Chemical Hardness Ralph G. Pearson
Chemical Hardness Ralph G. Pearson
WILEY-VCH
Chemical Hardness. R. G. Pearson
Copyright© 1997 WILEY-VCH Verlag GmbH, Weinheim
ISBN: 3-527-29482-1
Chemical Hardness
Ralph G. Pearson
University of California,
Santa Barbara
WILEY-VCH
Weinheim • New York- Chichester • Brisbane • Singapore • Toronto
This book was carefully produced. Nevertheless, author and publisher do not warrant the information
contained therein to be free of errors. Readers are advised to keep in mind that statements, data,
illustrations, procedural details or other items may inadvertently be inaccurate.
John Wiley & Sons (Asia) Pte Ltd, 2 Clementi Loop #02-01,
Jin Xing Distripark, Singapore 0512
A catalogue record for this book is available from the Deutsche Bibliothek
A catalogue record for this book is available from the British Library
ISBN 3-527-29482-1
Preface vii
Abbreviations used in this book ix
3. Applications of DFT 59
Introduction 59
Rates of Reaction 68
Reactions of Free Radicals 74
Aromatic Electrophilic Substitution 78
Fukui Functions and Atomic Charges 84
Improving the Energy from an Approximate Wave Function 88
Solvation Effects 89
References 95
Index 197
Preface
TS transition state
UBER Universal Binding Energy Relation
UV ultraviolet
vis-UV visible-ultraviolet
INTRODUCTION
The cohesive energy of the oxides is always greater than that of the sulfides.
But the amount, A, by which it is greater can vary with the metal. Thus A is as
large as 54 kcal/mol for Mg and as small as 1 kcal/mol for Hg. It is easy to see
why Mg and Ca occur as carbonates, and Hg and Cd occur as sulfides. The
oxides, of course, would be converted to the carbonates by CO2 in the air.
More recent foundations for the concept of chemical hardness lie in the works
of Chatt2 and Schwarzenbach.3 Independently, they showed that metal ions
Table 1.1 Cohesive Energies of Some Binary Metal Oxides and Sulfides
Mg Ca Ni Zn Pb Ca Cd Hg
A [kcal/mol] 54 32 30 27 21 18 16 1
could be divided into two classes, (a) and (b), depending on the relative affinities
for ligands with various donor atoms, in aqueous solution.
Class (a) N > P > As > Sb Class (b) N < P > As > Sb
O > S > Se > Te O < S > Se - Te
F > Cl > Br > I F < Cl < Br < I
Edwards had done something similar even earlier.4 His classification, however,
depended on the proton basicity and ease of oxidation of various ligands. He also
made the important step of comparing rates of reaction of various substrates with
the same ligands. The existence of two classes of electrophiles was clearly shown.
Metal ions were simply one group of electrophiles. Class (a) metal ions reacted
most rapidly, and more strongly, with nucleophiles which were very basic to the
proton. Class (b) metal ions reacted most rapidly with nucleophiles that were
easly oxidized. Metal ions are also one class of Lewis acids. The ligands with
which they form complexes are simply Lewis (and Bronsted) bases.
Thus it would be possible to classify other Lewis acids as class (a) or (b).
Because of a shortage of information on formation constants in water, it is
necessary to use a variety of other experimental data. Putting the donor atoms of
bases in order of increasing electronegativity gives
As<P<Se<S~I~C<Br<Cl<N<O<F
The criterion used is that class (a) acids form more stable complexes with the
donor atoms to the right, and class (b) acids prefer donor atoms to the left. This
is essentially the same criterion used by Schwarzenbach and Chatt. Often the
existence or non-existence of related compounds, or complexes, could be used as
an indicator. Rates of reaction could also be used by considering the activated
complex to be an acid-base complex, A: B.
Table 1.2 shows the result of this classification, as presented initially.1 It also
shows that classes (a) and (b) were renamed as "hard" and "soft", respectively.
There are two reasons for this change in nomenclature. One is that it is
often useful to employ comparative terms for two acids, such as "Hg 2+ is softer
than Pb 2 + ". The other comes about as a result of thinking about the funda-
mental properties of a given acid which made it class (a) or (b). The acceptor
Introduction
atoms of the first class are usually of high positive charge and small size, with no
unpaired electrons in the valence shell. Class (b) acids have acceptor atoms of
low positive charge and large size, and often have unshared pairs of electrons in
the valence shell. These characteristics meant that class (a) acceptor atoms are
not very polarizable, whereas class (b) acceptor atoms are very polarizable.
Since polarizability means deformation of the electron cloud in an electric
field, and since things that are easily deformed are soft, this leads to the two
classes of acids being called hard and soft, respectively. What one really has in
mind is deformation in the presence of other atoms or groups to which bonding
was occurring. Thus optical polarizability, although a useful measure of softness,
is not quite the correct measure.
Looking at the list of donor atoms for bases given above, it is obvious that
polarizability is high on the left side and diminishes as one goes to the right. By the
same argument as before, bases which donor atoms such as As, P, Se, S or I are
called soft bases. Bases with F, O and N are hard bases. A hard base has an
electron cloud that was difficult to deform chemically. Electrons were held tightly,
so that loss of an electron is difficult, whereas a soft base is easily deformed and
even oxidized. Table 1.3 shows some typical examples.
With this new nomenclature it is possible to make a simple, general statement:
"Hard acids prefer to coordinate to hard bases, and soft acids prefer to coordinate
soft bases." This is the Principle of Hard and Soft Acids and Bases, or the
HSAB Principle.
The HSAB Principle
Hard Soft
H", R"
Borderline
(1.3)
where the a terms could be positive numbers for soft species and negative
numbers for hard species.
One could, of course, attempt to use Equation (1.4) by taking one set of data to
assign values of S&, S&, a\ and <JB to a number of acids and bases, and then using
another set of data to check them. Actually something similar to this has already
been done by Drago and Wayland, with their well-known four-parameter
equation.6
ORDER OF HARDNESS
Actually chemistry is usually not based on Reaction (1.1), but on the exchange
reaction
The following conclusion can then be drawn, provided A and A' are acids of the
same strength, or B and B' are bases of the same strength:
where h and s are read as the harder and softer of the two acids (bases).
Equation (1.7) is the result of applying Equation (1.5) to the four acid-base
complexes contained in reactions (1.6). The strength terms cancel out, and the a
terms give the value of AH°. Now it is not practical to demand that S\ = SA' or
SB = SBS but it is possible to ensure that they be comparable in magnitude. This
can be done by requiring that A and A' be acids of the same charge, and of
similar size, and the same for B and B'. This minimizes the influence of the SASB
terms, and enhances the CFA<JB terms.
Equation (1.7) is based on Equation (1.5) but does not require numbers for its
application, only the relative hardnesses of the reactants. But these can be
obtained by applying Equation (1.7) to a set of related data. Turning back to
6 The HSAB Principle
cohesive energies, we can assume that the metal ions are all acids of comparable
strength. The cohesive energies of all the oxides are larger than those of the
sulfides because the oxide ion is a stronger base, due to its smaller size.
Similarly, the value of AH° for MgO is larger than that for HgO because Mg 2+
is smaller than Hg 2+ . Nevertheless, the strengths are comparable, not grossly
different as they would be, say, for Al 3+ and K + . Accordingly, Equation (1.7)
enables us to write the order of increasing softness as
using the values of A and the reasonable assumption that the sulfidle ion is softer
than the oxide ion.
By comparing only acids of the same charge, and bases of the same charge, we
have the great advantage that the predictions of Equation (1.7) are not sensitive
to the environment.
Also we see from Table 1.2 that in aqueous solution Cu 2+ is softer than Ca 2+ .
Indeed, Table 1.2 is in agreement with the ordering above, but with less detail for
the various metal ions.
A further check on the hardness ordering can be made using a different set of
reference reactions. The bond dissociation energies of fluorides and iodides were
used in the earliest attempt to order the metal ions:9
Defining A as AH? - AHf, we find the same order of A as before, and assuming
that I" is softer than F~, the same order of increasing softness.
Note that even though we are considering the reactants to be M 2+ and F~ or
I~5 it is possible to use ordinary bond energies because only differences are
important. Table 1.4 gives the empirical hardness order for a number of singly
charged Lewis acids, based on the bond dissociation energies of fluorides and
iodides, Z>F and Df.10 We see that SiH^ is the hardest acid of this group,' followed
by CF^. In accordance with Table 1.2, Cu + , Ag + and OH + are very soft acids.
The same kind of analysis may be made to rank a series of bases in order of
increasing softness. Because of the large amount of data, it is convenient to use
H + and CHjj" as the reference acids, with H + being the harder. Table 1.5 gives the
hardness order for a number of singly charged anions, using A = Z>H — ^CH 3 as
the criterion. As expected, F " and OH~ are the hardest bases. Simple carbon
ions are much softer, e.g., CH^ and C 6 H 5 CH^.
Order of Hardness
SiH+ 150 71 79
CF+ 130 54 76
CH 3 CO + 120 50 70
HCO+ 122 52 70
H+ 136 71 65
C6H+ 124 64 60
C 2 H^ 119 63 56
t-C4H+ 108 50 58
i-C3H+ 107 53 54
C2H+ 107 53 54
CH+ 109 56 53
C 3 Ht 98 44 54
c-C3H+ 111 59 52
Li + 137 82 55
Na + 123 69 54
T1+ 105 64 41
CN+ 112 73 39
N0+ 56 20 36
Cs+ 118 82 36
1+ 67 36 31
Cu + 102 75 27
Ag + 87 61 26
HO+ 56 52 4
(a)
After Reference 10.
The cyanide ion is an ambident base. The carbon end is seen to be much softer
than the nitrogen end, as would be predicted. The greater strength of the H-CN
bond, compared with H-NC, means that the carbon end is much more basic
than the nitrogen end. This would be true even in solution, since the ions are the
same for both acids. As a result, binding to carbon will be more common than
binding to nitrogen, even for hard acids.
Tables 1.4 and 1.5 contain acids or bases with a wide range of acceptor or
donor atoms. In such cases, use of different references will not give identical
orderings. Usually the variations are not very great, being shifts of two soft, or
two hard, acids or bases with respect to each other.10'11
The most reliable (i.e., transferable) results are obtained if two similar acids or
bases are used as references. Thus OH~ and SH~ will give the same ordering as
F~ and I~, but H~ and F~ will give orders that are quite different: the alkali
metal cations will be much harder than in Table 1.4. Such changes in order are
evidence for special bonding effects in certain A: B combinations.
The HSAB Principle
(a)
After Reference 1
In the case of H~, the special effect is the total absence of 7r-bonding or 7r-anti-
bonding. Hydride ion is a pure a-donor, compared with iodide ion. Figure 1.1
shows the 7r-interaction of a d (or p) orbital on an acceptor atom, and a p (or d)
orbital on a donor atom. If the d orbital is filled, and the p orbital is empty, there
is a stabilizing effect. This would be the case in a soft-soft A: B complex. If the
Order of Hardness
^Lc—r>- M -«<--
Figure 1.1 A p orbital on a ligand atom and a d orbital on a metal atom. The plus and minus
signs refer to the mathematical signs, or phases, of the orbitals in different regions of space.
d orbital is empty, and the p orbital is filled, there would also be a stabilization,
as in a hard-hard combination. If both the d and p orbitals are filled, there would
be a destabilization, as in a hard-soft combination.
These 7r-bonding effects are part of the theory of the HSAB Principle. We can
also imagine that London dispersion energies between atoms or groups in an A: B
complex could stabilize it. Since these dispersion energies, or van der Waals
energies, depend on the product of the polarizabilities of the two groups, soft-soft
combinations would be stabilized in this way. The hydride ion is very polarizable,
and its softness depends on this factor, presumably.
Because of the absence of 7r-orbital effects, the bond strength of HX, where X is
any element, depends almost entirely on the clectronegativity of X. D° ranges from
42kcal for CsH to 136kcal for HF, for the Main-Group elements. For the
1100
900
DC 700
500
300
100
2
XP
Figure 1.2 Plot of force kR0 [kcal/mol A] against the Pauling electronegativity of X, for HX
molecules. Reprinted from R.G. Pearson, J. Mol. Struct. (Theochem.), 300, 519 (1993), with
kind permission of Elsevier Science - NL, Sara Burgerhartstraat 25, 1055 KV Amsterdam,
The Netherlands.
10 The HSAB Principle
Table 1.6 Comparison of Na + and Cu+ as Lewis Acids
transition metals, the range is more variable because of the necessity to promote to
the valence state (half-empty s orbital).13 Promotion energies are also important
for the hydrides of Groups 2 and 12.
Figure 1.2 shows kR0 plotted against the Pauling electronegativity, Xp> f° r all
the diatomic hydrides where the data is available. The quantity kR0, where k is
the force constant and Ro the internuclear distance, is simply the force that the
atom X exerts on the H atom.13 This force is a good measure of what the bond
energy would be without the interference of promotion energies.
As expected, H + is also a rather special Lewis acid because of the absence of
filled inner shells. Consider the reactions in Table 1.6. The results show that Cu +
is a "stronger" acid than Na + , though the "strengths" are comparable. This
accords with copper being more electronegative than sodium. The numbers also
show clearly that Cu + is softer than Na + . The proton is seen to be a very strong
acid because of its small size and ability to burrow into the electron cloud of
bases. There is no destabilization with the soft hydride ion in H 2 . Note that
hydride ion is a strong base, comparable with fluoride ion.
SOLVATION ENERGIES
(a)
Data from Reference 14.
For neutral molecules acting as bases, the effect of solvents would be similar to
that of ions, but smaller in magnitude.1 That is, a hard acid solvent like water
would deactivate other hard molecules more than related soft molecules. Table 1.7
lists the calculated bond dissociation energies for a number of complexes in the
gas phase
AH0 (1.12)
The numbers are theoretical ones, but the method of calculation used gives very
good agreement with experiment, where numerical results have been obtained.15
It might be thought that the slightly stronger bonding in Ag(NH 3 ) + than in
Ag(PH3)+ is a failure of the HSAB Principle, since Ag + is soft. But this
overlooks the requirement that data in aqueous solution are to be used. Since the
heats of hydration of NH 3 and PH 3 are —8.1 kcal/mol and —3.6kcal/mol, the
difference is enough to overcome the small difference in bond energies for Ag +
(2.8 kcal/mol), but not enough for Li + (19kcal-mol).
The HSAB Principle can also be tested by applying equation (1.7):
Ag(NH3)+(g) + Li(PH3)+(g)
This test works for all the examples in Table 1.2, and also for similar data for
Na + compared with Ag + . 16
Table 1.8 has some data on the proton affinities (PA) of a number of neutral
molecules:
-AH° = PA (1.14)
12 The HSAB Principle
Base PA [kcal/mol]
HF 117
H2O 167
NH 3 205
CH 4 121
HC1 135
H2S 170
PH 3 189
HBr 139
H2Se 171
AsH 3 179
(a)
Data from Reference 16.
The results for H 2 O, H2S and H2Se are surprising and would not have been
anticipated in 1963, when few data on PAs were known. The strength of bonding
increases with the size of the donor atom, even though the base strength is not
expected to increase, and the softness does increase. The opposite trend should
have been observed, as it is for NH 3 , PH 3 and AsH3. The hydrogen halides show
an even greater reversal.
Neither Equation (1.4) nor Equation (1.7) can explain these results. It is a gas-
phase phenomenon, since it vanishes in solution, where the expected orders are
found for H2O, H2S and H2Se.16 The simplest explanation is that the inverted
order is the result of the classical charging energy for a sphere. This energy (the
Born charging energy) is given by
AE = q2/2eR (1.15)
where q is the charge, R the radius of the sphere, and e the dielectric constant of
the medium.
In more detail, since the proton in Equation (1.14) bears a positive charge, the
energy increase results from the repulsion of the proton by the shielded nuclear
charges of B. These are largest for a short bond to the proton, as we have in
H 3 O + , and are lower for H 3 S + and H 3 Se + . Also the nuclear repulsions decrease
as charge is removed from the nucleus to the outer parts of the molecule, as in
going from H 2 F + to N H j . Since Equation (1.15) depends inversely on R, the
overall effect is to make small molecules reluctant to add protons in the gas phase.
This is unfortunate, since there is a great deal of data available on the methyl
cation affinities (MCA) of neutral molecules17
Complex formation between metal ions and ligands in aqueous solution has
always been of great interest. This interest is enhanced by the role of metal ions
in biology. The first use of the concept of chemical hardness was to explain
complex ion formation in water. However, it turns out that this is not so easy as
one might expect.
There is a great deal of data available, not only on equilibrium constants, but
also on heats of reaction between metal ions and various ligands.18 For various
practical reasons, it is difficult to find data that can be used even in a comparative
sense. Solvation energies are one obvious problem. Only for restricted cases can
we use Equation (1.7).
Nevertheless, some useful generalizations may be made. Consider the typical
reaction for the formation of a complex in water (charges are omitted for
simplicity):
There are four interactions in Equation (1.17) which are acid-base in character.
One is the interaction of MLW with H 2 O, the second is the MLW interaction with Y,
the third is the interaction of Y with the solvent, and the fourth is the interaction
of water with itself, a constant factor. The solvation energies of the large complex
ions, MLW(H2O) and MLWY, are governed primarily by the Born equation.
If ML« is hard and Y is soft, or vice versa, we can expect AH° to be positive, or
only a small negative number. Therefore, complexation will not be favorable.
Suppose MLM is hard and Y is hard. Then their interaction will be favorable. But
H2O is a hard molecule, both as an acid and a base. Therefore Y(aq) and
MLW(H2O) are also stabilized. Overall, only a small negative value for AH° can be
hoped for, and moderately stable complexes.
But if ML« is soft and Y is soft, then everything works in favor of a large
negative value for A//°, and very stable complexes. Ahrland has made a detailed
study of the available data and has found a remarkable agreement with the above
predictions.19 Hard acids rarely form complexes with soft bases, and hard bases
do not form very stable complexes with soft acids, except for strong bases such
as OH".
14 The HSAB Principle
Hard acids form only moderately stable complexes, even with hard bases, and
AH° is close to zero. Soft acids and soft bases usually form very stable complexes
in aqueous solution, and AH° is then a large negative number. For a soft reference
acid, such as Pd 2+ or CH 3 Hg + , the order of increasing values of -AH° is
F" <C Cl- < Br- ~ N j < I" < SCN" < RS~ < CN". 18 " 20 For neutral ligands
the order is H2O < NH 3 < thiourea < PR3. Note that the order for anions is not
exactly the same as in Table 1.5 because of different hydration energies.
There are not enough comparable data to put many metal ions in rank order,
but the following orders of decreasing values of -AH° can be established:
Hg 2+ > Cd 2+ > Zn 2+ ; Pt 2+ - Pd 2+ > Ni 2+ . Also the Irving-Williams order21
for formation constants of complexes, V 2+ < Cr2+ > Mn 2+ < Fe 2+ < Co 2+ <
Ni 2+ < Cu 2+ > Zn 2+ , follows the experimental values of rj for these metal ions.
As expected, the magnitude of the changes in log^ e q increase with the
polarizability of the ligand donor atoms.22
There is an excellent recent review of the role of metal ions in biology by
Hancock and Martell.23 The discussion is in terms of HSAB, but for quantitative
work the most accurate treatment is by an equation related to Equation (1.5):
K\ is the equilibrium constant (in H 2 O, at 25 °C) for Reaction (1.1), and the E1
and C parameters are analogous to those in Equation (1.5), but are empirical
numbers from data in water.24
The D parameters are necessary to account for steric effects. The last term in
Equation (1.19) can be neglected for large cations, but is important for small
cations. It emphasizes another important difference between reactions in solution
and in the gas phase. In solution the Lewis acid is multicoordinate, and has a
number of water molecules attached to it. The steric effects arise from clashes of
the ligand with these water molecules. A large number of E\ C and D
parameters are now available, and can be used to estimate formation constants
for complexes between metal ions and the common ligands.
In Table 1.3, BF3 is listed as a hard acid but BH3, which also contains B(III), is
considered a soft acid. For example,25
BH 3 : NH 3 + BF 3 : CO
- BH 3 : CO + BF 3 : NH 3 AH° = -12kcal/mol (1.19)
Symbiosis and Anti-symbiosis 15
Also, in Table 1.4, CFj is shown as a harder acid than CH3". These are examples
of a very general phenomenon, first noted by Jorgensen and called by him the
"symbiotic" effect.26 Soft bases attached to the same central acceptor atom make
it a soft acid, and hard bases make it a hard acid. In coordination chemistry,
symbiosis explains why some ligands, such as CN~ or phenanthroline, make a
metal ion form strong complexes with other soft ligands, whereas F~ and H2O
favor the bonding of other hard ligands.27
The symbiotic effect is also common in organic chemistry, but here it has been
called the clustering, anomeric, or geminal effect.28'29 Clustering refers to the
stabilization caused by adding several substitutents to the same carbon atom.
Some extreme examples are shown by Reactions (1.20) and (1.21), in which the
number of bonds of each kind is preserved.
In Reaction (1.20) the comparison is between F~ and H~, the hardest and softest
bases in Table 1.5, and the effect is at a maximum.
Elements other than carbon may be influenced, for example29'30
Comparing Reactions (1.22) with (1.21), it appears that silicon is less affected
than carbon by clustering. But consider the reaction
Clearly SiFjj" is much harder than CF^, just as SiHjj" is harder than CHj. An
extreme example of the difference between the two elements is given by
C(OCH3)4(g) + SiH4(g)
= CH4(g) + Si(OCH3)4(g) AH° = 144 kcal/mol (1.24)
This illustrates the great affinity of silicon for binding to oxygen donors.
Organic chemists usually explain the anomeric effect by double-bond-no-bond
resonance.29 This is consistent with the observation that F~, OR~ and NR^
bases give the largest effects. However, the HSAB principle provides a much
simpler explanation. The acceptor atom in BF3, CF3" and SiF^ is much more
positive than in BH3, CH 3 and SiH^. Since a high positive charge enhances
hardness (Table 1.3), we now have acids which bind better to hard bases, such
as F~, than to soft bases such as H~.
In spite of this simple and logical explanation, there are many cases of anti-
symbiotic behavior for the soft metal ions of the second and third transition
16 The HSAB Principle
series.31 That is, soft bases attached to a soft acid center, such as Ir(I), Hg(II),
Pt(II) and Au(III), can lower the affinity for another soft base. But this only
occurs for a coordination site trans to the original ligand. The rule is that two
soft ligands in mutual trans positions will have a destabilizing effect on each
other, when attached to a soft metal ion.31
This rule explains a host of experimental observations. For example,
(CH3)2Hg is readily cleaved by dilute acid to form linear CH 3 Hg(H 2 O) + . But
this cation resists cleavage, even by strong mineral acids. There is a simple
explanation based on the theory of the trans effect.32 This effect occurs only for
soft metal ions, and ligands of high trans effect are always soft ligands.
The theory of the effect postulates that the bonding in such cases is largely
covalent. Ligands in trans positions compete for the same orbitals to form
covalent bonds (a or TT). Therefore it is advantageous to have a hard ligand, such
as H2O or OH~, trans to each soft ligand already attached.33 Anti-symbiosis
should be minor in tetrahedral complexes, or when the acceptor atom is hard.
NUCLEOPHILIC REACTIVITY
Early evidence for the HSAB Principle came from studies of nucleophilic
reactivity series towards different substrates, or electrophiles.34 Some electro-
philes, such as H + , in proton transfer reactions or CH 3 CO + substitution
reactions of esters, reacted rapidly with bases that were strong bases towards the
proton. Other electrophiles, such as Pt(II) or RO + , reacted rapidly with
polarizable bases, and were indifferent to proton basicity.
It soon became clear that these two classes should be called hard and soft
electrophiles, respectively. Since the terms nucleophile and electrophile refer to
rates of reaction, by definition, the acid-base reaction involved is
The rate then depends on the stability of the activated complex B7 : A : B. This,
in turn, depends mainly on the compatiblity of B' and A, if relative rates for a
series of nucleophiles, B', are compared.
Figure 1.3 shows the rate constants for the hydrolysis of/?-nitrophenyl acetate,
catalyzed by the attacks of bases on the carbonyl carbon.35 There is a rough
proportionality between log k and pKa. The scatter is expected when a number of
bases of different natures are compared. While F~ is fairly reactive, Cl~, Br~ and
I~ show no measurable reactivity.
In contrast, Figure 1.4 shows rate data for many nucleophiles reacting with a
pyridine (Pyr) complex of Pt(II) in methanol at 25 °C.36
Figure 1.3 Rate constants for the reactions of nucleophiles in aqueous solution at 25 °C,
plotted against the basicities of the attacking reagents. After Reference 35.
n = log(k/ks) (1.27)
where ks is the rate constant for solvolysis. The plot is that of n against the pKa in
water, but using pKa values in methanol, when known, makes no difference.
Clearly there is no dependence of n on the normal proton basicity. In fact
CH3O~, the strongest base possible in methanol, has no detectable reaction with
trans-Pt(pyY)2Cl2. The same is true for F~, but Br~ and I~ are good reagents.
An important substrate for nucleophilic reactivity is methyl iodide. This serves
as a model for substitution reactions at tetrahedral carbon in general, the
SN2 reaction
B + CH3I A CH 3 B + + I (1.28)
10 -
9 - o cP
8 -
o
7 - 0
6 - o
o
5 o
o
4 - o o o
o o
3 o
2 -
? ? ° ?
1 1 1
-8 -6 -4 -2 0 2 4 6 8 10 12 14 16
Figure 1.4 Plot of npt, in CH 3 OH, against pKa in H 2 O. After Reference 35.
There are enough data available to calculate AG° for a large number of anion
bases, and a smaller number of neutral bases, B.37
The order of decreasing values of AG° is H2O > CH 3 OH > Br~ - NO^ > I" >
F~ > Cl- > SCN~ > CH 3 COO (CH3)2S C 6 H 5 O- (CH3)3As
NH 3 > C 6 H 5 S- > n-C 3 H 7 S- O- > (CH3)3P CN~ CH 2 CN-
H~ > CHj > C6H5". The range of equilibrium constants is 60 powers of 10. The
order is that of decreasing thermodynamic leaving-group ability. It is in good
agreement with the kinetic leaving-group ability, the nucleofugality, when known.
The next step is to compare n with AG°. That is, to look for a linear free
energy relationship,. This is usually expressed as38
where AG-f is the free energy of activation. Over a wide range of AG°, a need not
remain constant, but it must change slowly and continuously. Figure 1.5 shows a
Nucleophilic Reactivity 19
10r
3 •
1•
plot of n, which is proportional to AG-f, against AG° for the cases where n is
known. First-row donors (F, O, N, C) are shown as circles and higher-row
donors as crosses. The former all lie close to the smooth curve that has been
drawn. The latter all lie above it.
The curve acts as a reference line, but it also shows that the slope, a, in
Equation (1.30) has a value near one for AG° near zero, and approaches zero as
AG° becomes a large negative number. Such behavior is an expected consequence
of the Bell-Evans-Polanyi-Leffler-Hammond principle.39 It corresponds to a
late transition state for the more difficult reactions and a progressively earlier
transition state for more exergonic reactions. However, the curve appears to level
off at an n value near 8, which corresponds to a second-order rate constant of
about 1 M" 1 s~! at 25 °C, far from the diffusion-controlled limit.
Recent developments, both experimental40 and theoretical in nature,41 help
explain this behavior. The energy barrier to reaction consists of a part that exists
in the gas phase and in solution, and a solvation part found only in solution. The
20 The HSAB Principle
gas-phase barrier exists because five groups attached to carbon in the transition
state are at a higher energy than four. The solvation barrier results because the
nucleophile must be partly desolvated before it can react, and the transition state
(TS) must be solvated. For the reaction of Cl~ with CH3C1, the two parts are of
comparable magnitude in water, each being about 14kcal/mol. Desolvation of
Cl~ requires more energy than is returned in solvating the TS.
For neutral nucleophiles the situation is somewhat different. The gas-phase
energy barrier is very large, so that reactions such as (1.28) simply do not occur.
For example, when B is NH 3 we have
log k [M- ]
Base CH3F CH 3 I
that the ratio kx/kz should be constant, or nearly so.46 But we have just seen that
the ratio k\/kF can change by a factor of 200 when Y changes fdrom OH~ to I~. In
other cases 47 the ratio changes by as much as 106. The theory behind the Marcus
equation does not allow for interactions between entering and leaving groups.
OTHER APPLICATIONS
In one way or another, the HSAB Principle has found many applications in most
areas of chemistry.48 Usually these depend on the implicit use of Equation (1.7).
22 The HSAB Principle
What causes the large difference of 3.73 V, or 86kcal/mol? Let us break the
overall reaction into two parts, one requiring energy and the other giving back:
K Ag
M(s) = M+(g) + e" AHi [kcal/mol]: 120 243
M+(g) = M+(aq) AH2 [kcal/mol]: 79 116
~4T T27
We see that potassium has an energy shortfall of 41 kcal/mol, but silver has a
much larger one of 127 kcal/mol. The difference is equal to the 86 kcal/mol of free
energy given by AE°, since entropy effects nearly cancel. Thus silver is less
reactive than potassium partly because it is harder to ionize, due to greater
electronegativity, but also because it does not interact with the solvent water as
strongly as expected. Since water is a hard base and Ag + is a soft acid, such a
result is reasonable.
Looking at the electronegative series of the elements, we see that the noble
metals, Hg, Au, Pt, Ir, Pd, Os, Rh and Ru, all form metal ions which are
classified as soft. Their heats of hydration are all less than expected and their
inertness is thus explained.
The advantage of this viewpoint is that it focuses attention on the solvent. It
suggests that by changing the solvent, we might invert the reactivities of the
alkali metals, which form hard metal ions, and the Cu, Ag, Au triad. Even more
easily, of course, by adding soft bases such as CN~ to water we can increase the
reactivity of the noble metals.
The whole subject of solubility and solvation energies can be looked at in
HSAB terms. The rule "Similia similibus dissoluntur" can be replaced by "Hard
solvents dissolve hard solutes, and soft solvents dissolve soft solutes". This
Other Applications 23
The ionization reaction is also a typical reductive elimination, since the formal
oxidation state of the metal decreases by two units. We can predict that hard
ligands will stabilize Fe(II) in preference to Fe(0), and will reduce the acidity of
H2Fe(CO)6.
This kind of prediction has been amply verified for other cases.49 Consider the
rhodium(III) complexes
A good strategy is to have a hard ligand attached to a soft metal atom. The hard
ligand is often a solvent molecule such as H2O or CH3CN. Since these will be
labile, vacant coordination sites for the organic reactant are readily available.50
Soft ligands, present in the solution, will act as catalyst poisons,
Saville has developed rules for chemical reactions that can also be used to
select homogeneous catalysts.51 Consider the four-center reaction
C 2 H 5 OH A CH3CHO + H 2 (1.39)
metal
are readily accounted for. The metal surfaces remove hydride ions (and return
electrons to the substrate); the acid clay removes H + and OH~.
Pauling's bond energy equation is one of the best-known empirical equations
in chemistry:
Applied to the exchange reaction (Equation (1.16)), the Pauling equations gives
That is, the reaction will be exothermic if the products contain the least electro-
negative (EN) element combined with the most EN element. But this usually is
the opposite of what Equation (1.7) will predict. Take the example
SUMMARY
In retrospect it seems clear that the original concept of hardness and softness is
an example of "fuzzy logic." In spite of its name, fuzzy logic is a respected
branch of mathematics, which was essentially invented in 1965 by L. A. Zadeh. It
is a method of reasoning and making decisions when the available information is
not precise enough to use numbers, or definite statement such as "Yes" or "No."
On the other hand there is some information which is qualitative in nature and
insufficient to make definite statements.
Classical mathematics requires that a concept has a precise definition which
partitions a class of objects into two classes: (a) those that belong, and (b) those
that do not belong. But the real world is filled with concepts such as many, old,
slow, large. These do not have sharply defined boundaries, but still convey
information. Fuzzy logic deals with the application of such knowledge.
As might be expected, there was at first considerable criticism and rejection of
Zadeh's proposals. Today, however, there are many papers being published that
concern the development and application of fuzzy logic. There are a number of
26 The HSAB Principle
REFERENCES
25. V. Jonas, G. Frencking and M. T. Reetz, / . Am. Chem. Soc, 116, 8741 (1994).
26. C.K. Jorgensen, Inorg. Chem., 3, 1201 (1964).
27. J.M. Pratt and R.G. Thorp, / . Chem. Soc. A, 187 (1966).
28. J. Hine, / . Am. Chem. Soc, 85, 3239 (1963).
29. A.E. Reed and P.v.R. Schleyer, / . Am. Chem. Soc, 109, 7362 (1987).
30. H.B. Schlegel, / . Phys. Chem., 88, 6254 (1984).
31. R.G. Pearson, Inorg. Chem., 12, 712 (1972).
32. F. Basolo and R.G. Pearson, Mechanisms of Inorganic Reactions, 2nd edn., John Wiley,
New York, 1967, Chapter 5.
33. J. Chatt and B.T. Heaton, / . Chem. Soc A, 2745 (1968).
34. J.O. Edwards and R.G. Pearson, / . Am. Chem. Soc, 84, 16 (1962).
35. W.P. Jencks and J. Carriuolo, / . Am. Chem. Soc, 82, 1778 (1960).
36. R.G. Pearson, H. Sobel and J. Songstad, / . Am. Chem. Soc, 90, 319 (1968).
37. R.G. Pearson, / . Org. Chem., 52, 2131 (1987).
38. J.E. Leffler and E. Grunwald, Rates and Equilibria of Organic Reactions, John Wiley, New
York, 1963.
39. For references see S. Wolfe, D.J. Mitchell, H.B. Schlegel, / . Am. Chem. Soc, 103, 7692,
7694 (1981).
40. M. Pellerite and J.I. Brauman, / . Am. Chem. Soc, 102, 5993 (1980); ibid., 105, 2672
(1983).
41. J. Chandrasekhar, S.F. Smith and W.L. Jorgensen, / . Am. Chem. Soc, 108, 154 (1983).
42. R.G. Pearson and J. Songstad, / . Org. Chem., 32, 2899 (1967).
43. J. Koskikallio, Acta Chem. Scand., 26, 1201 (1972).
44. L.H. Sugemyr and J. Songstad, Acta Chem. Scand., 26, 4179 (1972); A.J. Parker, Chem.
Rev., 69, 1 (1969).
45. For a review see W.J. Albery, Annu. Rev. Phys. Chem., 31, 227 (1980).
46. E.S. Lewis, M.L. McLaughlin and T.A. Douglas, / . Am. Chem. Soc, 107, 6668 (1985).
47. R.G. Pearson and P.E. Figdore, J. Am. Chem. Soc, 102, 1541 (1980).
48. Y. Marcus, / . Phys. Chem., 91, 4422 (1987).
49. R.G. Pearson, Chem. Rev., 85, 41 (1985); idem, Bonding Energetics in Organometallic
Compounds, T. Marks, Ed., ACS Symposium Series 428, American Chemical Society,
Washington, DC, pp. 260-261.
50. W.J.A. Davies and F.R. Hartley, Chem. Rev., 81, 79 (1981); S.L. Randall et al,
Organometallics, 13, 5088 (1994).
51. B. Saville, Angew. Chem., Int. Ed. EngL, 6, 928 (1967).
52. L. Pauling, The Nature of the Chemical Bond, 3rd Edn., Cornell University Press, Ithaca,
1960, pp. 88-105.
53. R.G. Pearson, Chem. Commun., 65 (1968).
54. R.S. Drago, N. Wong and D.C. Ferris, J. Am. Chem. Soc, 113, 1970 (1991).
55. R.L. Matcha, / . Am. Chem. Soc, 105, 4859 (1983). D.W. Smith, / . Chem. Ed., 67, 911
(1990).
56. D.H. Rouvray (Ed.), Fuzzy Logic in Chemistry, Academic Press, New York, 1997.
2 Density Functional Theory
INTRODUCTION
Density functional theory (DFT) is a form of quantum mechanics which uses the
one-electron density function, p, instead of the more usual wave function, i/;, to
describe a chemical system.l Such a system is any collection of nuclei and electrons.
It may be an atom, a molecule, an ion, a radical or several molecules in a state of
interaction.
Hohenberg and Kohn proved in 1964 that the ground-state energy of a chemical
system is a functional of p only.2 A functional is a recipe for turning a function into
a number, just as a function is a recipe for turning a variable into a number. For
example, the energy is also a functional of the wave function. The variational
method is one recipe for turning ij) into a number, E.
reveals that all properties of the ground state are functions only of TV and v(r), the
potential due to the nuclei. But it can be proved that p{r) determines both TV and
v(r).2 Thus p(r) also determines ^o and all ground-state properties. It can also be
proved that a trial electron density, p(r), which is not the exact p(r), will give an
energy higher than the exact energy, EQ.
We can write the energy as
where all terms are explicit functional of the electron density. Fne is the nuclear-
electron potential energy, / is the classical part of the electron-electron repul-
sion energy, Exc is the so-called exchange-correlation energy and T is the kinetic
energy. Given the function p, we can readily calculate VnQ and /, but we do not
know the exact functional dependence of 7"and Exc on p. Hence we cannot calculate
the energy without approximations.
The usual method of proceeding is to solve for the Kohn-Sham orbitals, wz:7
f em (2.4)
DFT and the Concepts of Chemistry 31
where ve(r) is the classical potential due to the electrons and vxc is the potential
leading to Exc in Equation (2.3). Fortunately very good approximations are
already available for vxc.3 The kinetic energy is given by — ^V 2 , in atomic units.
Since we now have a one-electron problem, the Kohn-Sham equations (2.4)
can be solved in a self-consistent manner. We obtain a set of orbitals and their
energies, much as in HF theory. The density function, p(r), can be found as the
sum of the squares of the w,-, for the occupied orbitals. From p(r) the expectation
value of the energy can be found, as well as other one-electron properties. Just as
in the HF method, the total electronic energy is equal to the sum of the energies
of the occupied orbitals, minus a correction because the electron-electron
interactions have been counted twice.
Strictly speaking, the Kohn-Sham (KS) orbitals are fictitious entities, created
by a certain mathematical procedure. Of course, the same can be said about
molecular orbitals (MOs) in HF theory. But we know that MOs are given reality
by their successful use in many applications. The KS orbitals and the HF MOs
are not the same, but there is a one-to-one correspondence, and their orbital
energies are similar. It would appear that KS orbitals will one day be used in the
same way as MOs are now.
Since in principle DFT theory is exact and the KS equations are exact, if the
exact vxc is used in Equation (2.5) we can expect that the KS orbitals will
eventually prove superior for some applications. At least in one case, we know
that this is true. The orbital energy of the highest occupied KS orbital will
approach the negative of the first ionization potential exactly.9
The foregoing has emphasized the value of DFT in the accurate calculation of
molecular properties. But there is another aspect which has not been mentioned.
DFT is rich in conceptual content. Many of the basic concepts of chemistry,
including hardness and electronegativity (EN), appear simply and naturally.
Thus DFT is useful not only for calculations but also for understanding them.
Much of this is due to the pioneering work of R.G. Parr and his collaborators.
The quantity \x is the Lagrange multiplier that ensures that the integral of p over
the volume is equal to N. It follows that
6E[p] 6F[p]
(2.7)
Sp V
8p
where F[p] is the sum of the last three terms in Equation (2.3). The use of 6 in
Equation (2.7) implies that the terms are to be read as "the functional
dependence of E[p] and F[p] on p". We can also write the total differential of E
as
The quantity \i is called the electronic chemical potential. The name comes
from the thermodynamic equation
At zero pressure and temperature, we also have \ij — (dE/dN). In this case N
is the number of molecules in the system, and jij is the ordinary chemical
potential of thermodynamics. The electronic chemical potential of a single
molecule plays somewhat the same role. At equilibrium \i must be constant
everywhere, and p will be the correct electron density for the ground state.
The quantity x is called the absolute electronegativity, for reasons that will
become clear.8
The definition of fi in Equation (2.9) is much preferable to that in Equation
(2.7), which gives JJL in terms of the functional dependence of E(p) on p, which
is not known. But we do know something about the variation of E with N.
Figure 2.1 shows a plot of the energy of a chemical system as a function of the
number of electrons. The energies are all negative, with zero energy way up on
top. Experimentally we only know points on the curve for integral values of N,
from data such as ionization potentials, /, and electron affinities, A. If we
connect these points by a smooth curve, then (dE/dN) is simply the
instantaneous slope of the curve.
We do not know this instantaneous slope, nor is it straightforward to
calculate it, since the Schrodinger equation is defined only for integral values of
N. However, if we pick a starting point, such as the neutral species, we know
the mean slope from N to (TV— 1) electrons. It is the negative of the ionization
DFT and the Concepts of Chemistry 33
neutral
N
Figure 2.1 Plot of the total electronic energy vs. the number of electrons for a fixed collection
of nuclei.
potential, /. Also the mean slope from TV to (N+l) is the negative of the
electron affinity, A. By the method of finite differences we can estimate the
slope at N as
But this is simply the Mulliken EN, XM- In addition, there is a previous history
of calling (dE/dN) the electronegativity.
Actually, if we define the absolute EN, x, as (/+ A)/2, it is not quite the same
as Mulliken's EN. He was interested in a scale which could be used to estimate
the polarity of chemical bonds, just as Pauling was. Mulliken's / and A are for
supposed valence states of an atom or radical, such as it might have in a
molecule. The DFT values of / and A are for the ground state of any system,
atom, ion, radical or molecule. Also, since v is to be constant, they refer to
vertical values and not adiabatic ones.
Besides Equation (2.11), there is a second reason to call x the absolute EN.
If two systems, C and D, approach each other, with different values, /1Q and /x^,
for their electronic chemical potentials, there must be a flow of electrons from
one to the other until the potentials are equalized, i.e. /xc = Ato- This is the
condition for equilibrium. But this means that the ENs must also equalize:
Xc = XD- Thus we have a proof of EN equalization, originally proposed by
Sanderson as a postulate.12
34 Density Functional Theory
The quantity rj is called the absolute, or chemical, hardness.13 From the method
of finite differences we obtain the operational definition of rj:
v~(I-A)/2 (2.13)
(2.14)
(2.15)
(216)
Thus the difference in electronegativity drives the process, and the sum of the
hardness parameters acts as a resistance. Equation (2.16) is a chemical form of
Ohm's Law. It shows hat electrons will flow from the system of low EN to that
of high EN, as expected. It also shows that Equation (2.12) is a reasonable
definition of hardness, since the common meaning of hardness is resistance to
deformation or change.
The most immediate reason for Equation (2.12), however, was that it agreed
with the chemical observations that led to the concept of hard and soft acids and
bases. For example, consider Ca 2+ and Fe 2+ . The third ionization potential, / 3 ,
and the second, I2, would be /and A, respectively (Table 2.1). Accordingly, Fe 2+
is much softer than Ca 2+ , as expected. Also Ca 2+ is more EN than Fe 2+ , meaning
that it is much less likely to find Ca 3+ than Fe 3+ . Similar results are found or
almost all the metal ions.
Fortunately for the further development of hardness and DFT, a large
number of/and A values have become available. It turns out that most common
molecules have negative values for their electron affinity; that is, energy is
required to force an electron on to the molecule. The technique of electron
DFT and the Concepts of Chemistry 35
an
Table 2.1 Ionization Potentials, x d V f ° r Ca 2+ and Fe 2 +
Ca 2 + Fe 2 +
/(a)
Molecule A*> X V
(a)
Data from References 49 and 50, except as indicated.
(b)
G.L. Gutsev and T. Ziegler, Can. J. Chem., 69, 993 (1991).
(C)
M. Allan, J. Am. Chem. Soc, 115, 6418 (1993).
(d)
S. Moran and G.B. Ellison, Int. J. Mass Spectrom., 80, 83
(1988).
(e)
Singlet state.
ai
Table 2.3 Values of \ id rj for Various Monatomic Cations
Na+ Cu + Ca 2 + Fe 2 + Al 3 + T1 3+
There are still more reasons to believe that 77, as defined in Equation (2.12), is
indeed what is meant by chemical hardness. To understand this, it is necessary to
see whether the chemical concepts derived by DFT are compatible with mole-
cular orbital (MO) theory.15 This theory is certainly the most widely used by
chemists and is very successful in many areas. It is almost universally applied to
explain structure and bonding, visible-UV spectra, chemical reactivity and
detailed mechanisms of chemical reactions.
The best way to combine DFT and MO theory is to incorporate % and 77
into the commonly used orbital energy diagrams. Figure 2.2(a) shows such a
diagram. A typical case of a molecule where / = lOeV and A — — 2eV is taken.
Within the validity of Koopmans' theorem,16 the frontier orbital energies are
given by
Hard molecules have a large HOMO-LUMO gap, and soft molecules have a
small HOMO-LUMO gap.
Correlation with Molelcular Orbital Theory 39
(a) (b)
LUMO
• -x
E -5
-L -X
Figure 2.2 Orbital energy diagram for (a) a molecule and (b) a radical. HOMO = highest
occupied molecular orbital; LUMO = lowest unoccupied MO; SOMO = singly occupied MO
This statement agrees with the earlier, imprecise definition. For example, optical
polarizability in quantum theory results from a mixing of suitable excited-state
wave functions with the ground-state wave function. The mixing coefficient is
inversely proportional to the excitation energy from the ground to the excited
state. A small HOMO-LUMO gap automatically means small excitation
energies to the manifold of excited states. Therefore, soft molecules, with a small
gap, will be more polarizable than hard molecules. High polarizability was the
most characteristic property attributed to soft acids and bases.
A number of papers have appeared showing a correlation between polariz-
ability, a, and softness.17 Empirically it is found that a 1 / 3 is a linear function of
(/— A)~l, the softness. This is equivalent to the classical result for spheres of
radius R, that charging energies are proportional to 1/R, whereas polarizability is
proportional to R3. Calculations of a using DFT, and EN equalization, shows
that a is equal to (/— A)~l times a factor dependent on the size of the system.
In simple MO theory, the energy gap between the HOMO and the LUMO
defines the first excited state, or the energy of the first absorption band in the
visible-UV. A hard molecule would have a large value of hum2LX, where umax is the
vertical frequency of the lowest-energy transition where the multiplicity does no
change. For example, H 2 O, H2S, H2Se and H2Te have z/max at 1655, 1950, 1970
and 2000 A, respectively, showing increasing softness.
40 Density Functional Theory
- A) = - 2KnL (2.18)
where JUL and XHL are coulomb and exchange integrals for one electron in the
HOMO and in the LUMO.
Figure 2.3 shows a plot of ( / - A) for a number of molecules from Table 2.2
against hvmSiX for the same molecules in the gas phase. The slope of the best
straight line is 1.45. The scatter shows that molecules can have quite different
values of / and K. In addition, there are errors due to orbital relaxations, not
allowed for by Koopmans' theorem.
Table 2.4 gives the values for hvmdiX for a number of selected molecules in the
gas phase. Generally there is a good correlation between these energies^ and
25.0 r
20.0
15.0
<
i
10.0
5.0
0.0
0.0 20.0 40.0 60.0 80.0 100.0
(Frequency, kK)
Figure 2.3 Plot of (/ — A) against the vertical frequency of the first absorption band for a
number of gas-phase molecules. The multiplicity of the excited state is the same as that of the
ground state. From 55, reprinted with permission.
Correlation with Molelcular Orbital Theory 41
(a) l
Same multiplicity as ground state, data from Reference 18. 1000 cm
is one kiloKayser, kK.
(/— A) for a series of related molecules, but some exceptions are seen. It would
appear that CF4 is the hardest molecule known, since humax is 110.5kK,
compared with 78.2 for CH4. Neither / o r A is known accurately for CF4, but /is
about 16 eV, and A is about —8eV, making 77 ~ 12 eV, compared with 10.3 eV
for CH4.
The visible range is from 12.5 to 25 kK (K^cm" 1 ). Some free radicals are
included in Table 2.4, to remind us that many would be colored, if we could
42 Density Functional Theory
see them. While CH3 is colorless, it is shifted towards the visible, compared
with CH4. These results are consistent with the great reactivity of free
radicals.
(2.19)
(2.20)
Finally, for a radical attack, or any case where electrons flow in both directions,
we have
The orbitals are those of C, in all cases, but there would be complementary
equations for D.
These are just the rules for classical frontier orbital theory, as proposed by
Fukui21 - hence the name Fukui function. DFT has again been reconciled with
MO theory, this time for chemical reaction. The Fukui functions are reactivity
indices. Further analysis of the interaction of C and D shows that the value of
Chemical Reactivity 43
ifcfo) helps determine the net change in chemical potential.22 This implies that
good orbital overlap is also required.
The necessity for three reactivity indices, as shown in Equations (2.21)—(2.23),
reminds us that (dp/dN) is not a continuous function but has discontinuities at
integral values of N. The same is true for the energy, of course.23 If our system
consists of a single isolated molecule, then there is a constraint in that electrons
may be lost, but not gained. In such a case \i is simply equal to the orbital energy
of the HOMO. 24 The hardness is not defined.
Unlike the chemical potential, which must be the same everywhere in the
system at equilibrium, the hardness is not required to be constant, and has local
values, fj. From Equations (2.7) and (2.12) we find that25
2v=((d2F/dpdp')p/N) (2.24)
An average of these local values over the syste'm then gives the global value, 77.
The calculation of 77 is not easy, but Equation (2.24) is useful in telling us that the
hardness depends only on the way in which the kinetic energy and the inter-
electronic repulsion energy change with the number of electrons.
An easier function to deal with is the local softness, a. This is defined as26
Thus the local softness is readily found from the global value by multiplying by the
appropriate value of the Fukui function from equations (2.21)-(2.23). The local
softness measures how easy it is to change the electron density at different parts of
the molecule. As expected, the softness is different for accepting or losing
electrons. Unfortunately, 77 is not the simple inverse of a.27 However, it is reason
able to assume that in comparing two sites in a molecule with quite different
values of <r, that the fj values will also be quite different in the inverse sense.
As the positive charge on an atom increases, the electron density around it is
compacted. It becomes more difficult to change and the site becomes harder. The
increase in the classical Coulomb potential of the electrons is the main factor in
this hardening.28 The effect will be largest for points close to the nucleus, but will
be appreciable at bonding distances. Local hardness will be an important
property in determining bond strengths.
CHEMICAL REACTIVITY
The ultimate goal of any general theory of chemistry must be to give information
about the relative stabilities of molecules, and their tendencies to undergo
chemical change under specified conditions. DFT gives such information in
44 Density Functional Theory
detail when used to make ab-initio calculations, as described earlier. We will not
dwell on this application, but instead look for more qualitative information from
the concepts derived from DFT.
We can immediately draw important conclusions about molecular stability
from Figure 2.2, and the identification of 2ry with the HOMO-LUMO energy
gap. Soft molecules will be less stable than similar hard molecules. They will
dissociate or isomerize more readily. In the perturbation theory of such
reactions, change occurs by mixing in excited-state wave functions with the
ground-state wave function. If Q is the reaction coordinate,
(2.26)
(Eo - Ek)
The index k refers to the excited states. Only states of the same multiplicity
can mix.
There are important symmetry restrictions contained in Equation (2.26).29
These are not important at this time, but what is important is that an easy
change of ^, and hence of p, can only occur if the transition energies (EQ — Ek)
are small. There is a sum over all excited states in Equation (2.26), but only a few
will be important in driving the reaction. These will usually lie close to the first
excited state.
Figure 2.4(a) helps to explain this. It shows the ground state and some of the
excited states on an energy scale. There is a dense manifold of excited states,
some chemically important and others not, but hvmSLX sets a lower limit to all of
these. Also the difference (/z^max) will usually be much larger than the differences
between the excited states. Figure 2.4(b) shows an exception in which there are
hv
(a) (b)
Figure 2.4 (a) Stable molecule with large energy gap between ground and excited states,
(b) Stable molecule with small energy gap for excited states that are not chemically useful.
Chemical Reactivity 45
low-lying excited states which are not chemically important. An example would
be the colored complexes of the transition metals.
As predicted from the above, there is a close connection between dissociation
energies and hi/m3iX.ls Plots of the latter vs. bond dissociation energy Z>0 for
diatomic molecules are fairly linear, providing that covalent and ionic molecules
are separated. In view of Figure 2.3, we can expect similar results for (/ — A) vs. Do.
Usually, if we replace a light atom in a molecule with a heavier member of the
same family, we can expect two changes. The experimental value of ( / - A) will
become smaller,30 and the energy needed to break a bond to the replaced atom
will become less.
There is another way to test chemical stability hat can be applied to aromatic
organic compounds.31 Aromaticity is usually taken to mean high stability and
low reactivity. It can be quantified by calculating the resonance energy per
electron (REPE). These values are usually calculated by simple Hiickel theory,
but they correlate well with experimental measures of reactivity.32
Figure 2.5 shows experimental values of (/— A)/2 for a series of benzenoid
hydrocarbons plotted against their REPE. In the figure, benzene is the most
stable compound and has the largest energy gap. Tetracene is the least stable and
REPE
0.07
0.06
0.05
0.04
3.0 4.0 5.0 6.0 (I - A) / 2
Figure 2.5 Experimental values of (/ — A)/2 versus the calculated resonance energy per
electron (REPE) for several benzenoid hydrocarbons
46 Density Functional Theory
has the smallest gap. These results can be extended to many more examples, if we
use rj values also calculated from Hiickel theory.31
We next consider hardness and its role in the case of bimolecular reactions
where CD* is the transition state. We have already shown that there will be an
initial transfer of electrons from D to C
(216)
(T7C + m)
This will result in an energy lowering:
(XC-XD) (2>28)
Equations (2.16) and (2.28) are very appealing, since they give information on
reactivity using only data for the reactants, and a minimum number of
parameters. In fact they were used before the development of DFT, mainly to
estimate the polarity of bonds.33 But they are obviously incomplete. They are
based on C and D transferring electron density at distances that are so large that
other interactions between them are negligible. However, as they come to bonding
distance, there will be large changes in energy due to the added potentials of the
nuclei.34 Also, there will be changes due to the delocalization of electron density
corresponding to the formation of covalent bonds.35
The energy lowering in Equation (2.28) is typically only a few kilocalories per
mole, far short of bond energies. Bringing C and D closer together to form CD-I-
creates interactions which require consideration of orbitals in MO theory, or
Fukui functions in DFT. In the former case perturbation theory is generally
used, as pioneered by Dewar36 and Fukui and Fujimoto.37 Figure 2.6 shows the
MO-energy diagrams for two interacting molecules, which are of similar EN, so
that electrons will flow in both directions. One molecule, however, is harder than
the other.
Two kinds of interactions are shown:
1. Partial transfer of electrons from the HOMO of each molecule to the LUMO
of the other. This occurs by mixing of the orbitals.
2. There is a mixing of the filled MOs of each molecule with its own empty MOs.
Chemical Reactivity 47
hard soft
Figure 2.6 Partial transfer of electrons from HOMOs of each of two molecules to LUMOs
of the other (delocalization), and mixing of excited states within the same molecule
(polarization).
The first effect is called delocalization and is the mechanism whereby new bonds
are formed between the reactants and old bonds are broken. Other orbitals besides
the HOMOs and LUMOs may be involved to a lesser extent.
The second effect is called polarization. It has the effect of lowering the repulsive
energy between the two molecules as they approach each other. Polarization is
easiest when the energy gap is small for each molecule.
Considering only the four frontier orbitals, an approximate value for the
energy lowering due to delocalization is given by
2/3? (2.30)
AE =
(Ac-ID) (AD-IC)
The /3s are exchange integrals of the perturbation Hamiltonian over the
interacting MOs. One /3 could correspond to a bonding, and the other to TT
bonding, for example. The stabilization is greatest if A is large for both molecules
and / is small. This means that both energy gaps should be small or both
molecules should be soft.
Note that Equation (2.30) already requires two more parameters than Equation
(2.28) does. Also, to estimate (3\ and /%, we need to select definite occupied and
empty orbitals on both C and D. Finally, as C and D approach more closely, the
orbitals of the original reactants will become so changed as a result of mixing that
they will be useless for further calculations.
It is of interest that a different MO perturbation theory, using localized orbital
interactions, projects out the local softness and the Fukui functions directly.38
48 Density Functional Theory
Both of these can be evaluated for the different atoms in each molecule, and they
define the localizability of the interaction. Instead of the global EN, a local EN is
calculated for each atom and determines the amount of local electron transfer.
There have been several papers probing the further interaction of two molecules
using DFT. 22 ' 39 The focus is on the electron density of the combined system, and
how it changes as reaction proceeds. The guiding principle is the constancy of the
chemical potential over the system. As expected, some parts gain electron density,
and other parts lose it. The calculation of actual energy changes is not easy. The
importance of the Fukui function, or the equivalent local softness, <j, is confirmed.
Reaction usually takes place at the two atoms with the largest values of/. But this
is only true if there is considerable electron transfer between C and D. For
reactions between two hard molecules, reaction is favored at atoms where/has
minimal values.
All of these theoretical studies, whether by MO or DFT methods, provide
support for the HSAB Principle, "Hard likes hard, and soft likes soft". This is
easily seen in Equation (2.16). Obviously if one reactant easily loses electrons,
it is best if the other reactant easily gains electrons. Support for the HSAB
Principle also comes from ab-initio calculations of Hartree-Fock accuracy on
combinations of hard and soft metal ions with hard and soft neutral ligands.40
The energy changes can be broken down into contributions of electrostatics,
exchange repulsion, ligand polarization, ligand-to-cation charge transfer and
electron correlation effects. As expected, charge transfer is greatest in the
reaction of soft metal ions with soft bases, and least in the reaction of hard metal
ions with hard bases. The increased stability of a soft-soft combination is largely
due to increased correlation energy. This reduces electron-electron repulsion.
It is essentially the same as the polarization energy shown in Figure 2.6.
While all of these studies are in accord with the HSAB Principle, none of them
can be taken as a general proof. Indeed, a rigorous proof may be very difficult to
formulate. However, there is one noteworthy attempt.41 Write Equation (2.28) as
(2.31)
Assume that for a given (/x^ — /x£.) and TJD, AQQ is minimized with respect to rjc-
It follows that rjc — rjD. Minimizing A£7D will give the same result, so that
Aftc = Af2o, and (AQc + AOD) is also minimized.
The reason for this procedure is that the product molecule, CD, can be
considered an open system. That is, there is an exchange of both energy and
matter (electrons) between the parts, C and D. In ordinary thermodynamics, the
quantity that determines equilibrium in an open system is the grand potential,
Q = E — Nji. At equilibrium the grand potential is as negative as possible.42 But
Electronegativity Scales 49
the AVL values of Equation (2.32) are just the changes in their grand potentials.
This assumes that a single molecule, where N is the number of electrons, can be
considered in the same way as a collection of molecules, where TV is theif number.
But this is the same assumption that has already been made in the case of
Equations (2.9) and (2.12), which define \i and r\. Accordingly, we have proved
that if a given acceptor molecule reacts with a set of donor molecules, the most
stable product will be formed with the donor whose hardness is the same as
that of the acceptor. The HSAB Principle! However there is an important
restriction. The quantity (/xc — MD) must be constant, so that only donors of
the same electronic chemical potential can be compared. A similar conclusion
can be drawn for a series of acceptors.
This restriction is reminiscent of the restriction applied to Equation (1.7) of
Chapter 1, that only acids (or bases) of comparable strength can be considered.
It is not quite the same since, as explained earlier, orders of electronegativity
are not orders of strength. The main objection to the above proof of the HSAB
Principle is that it is based on Equation (2.28), which only gives the initial
interaction of C and D. The fact that it apparently leads to the right result
suggests that equations (2.28) and (2.16), while incomplete, may often be
harbingers of better values of AE and AN.
ELECTRONEGATIVITY SCALES
The concept of electronegativity (EN) is almost as old as chemistry itself.
Berzelius classified atoms as electronegative or electropositive. By the turn of
the 20th century it was understood that these terms referred to the electron-
attracting and -holding power of the atoms. During the 1920s the founders of
physical-organic chemistry extended the terms to include groups of atoms as
well as atoms. There was an approximate ordering of the EN of various atoms
and radicals.
In 1932 Pauling made a landmark contribution.43 He created an empirical
scale of EN based on heats of formation or, essentially, bond energies. A number
of other scales eventually appeared, such as the widely used Allred-Rochow
scale.44 These scales had two characteristics in common: one was that they were
calculated from properties of the free atoms of the elements; the other was that
they were tested by seeing if they agreed with the original Pauling scale. Failure
to do so would be a serious deficiency. In 1939, in the first edition of The Nature
of the Chemical Bond, Pauling gave his meaning of the word electronegativity:
"the power of an atom in a molecule to attract electrons to itself". Many would
accept this as a definition of the term.
It is clear that absolute EN differs substantially from Pauling EN. It applies to
molecules, ions and radicals, as well as to atoms. For the latter, it is a property of
a free atom in the ground state and not an atom in an excited valence state
suitable for its appearing in a molecule. As might be expected, applications of the
50 Density Functional Theory
two scales are quite different. The Pauling scale is useful for estimating bond
polarities and, to some degree, the strengths of bonds between different atoms.
The absolute scale is a measure of the chemical reactivity of an atom, radical, ion
or molecule.
The absolute EN does not conform to the Pauling definition of EN as a
property of an atom in a molecule, but the essential idea of EN is that of
attracting and holding electrons. There is no compelling reason to restrict this to
combined atoms. The extension of the concept of EN to molecules seems to be a
natural and useful step. Donor-acceptor interactions are at the very heart of
chemical bonding. The absolute EN is a measure of the intrinsic donor-acceptor
character of a species.
There is no inconsistency in the EN of a free atom being different from that of
an atom in a valence state. Scales such as Mulliken's and the recently developed
spectroscopic scale45 show that the absolute and Pauling-like scales can be
commensurable. Since the applications are so different, it is not a meaningful
question to ask which scale is more correct. Each scale is more correct in its own
area of use.
While Equation (2.16), in principle, can be used to calculate bond polarities, it
is not as reliable as methods using the Pauling scale. However, the absolute scale
of EN can be used in a unique way to probe bond polarity. For example,
consider a molecule X—Y, consisting of two atoms or radicals held together by a
bond. The polarity of the bond could determine whether the molecule behaves as
X + , Y~ or X~, Y + . The same question can be asked by looking at the reactions
X-(g) + Y+(g) = X-Y(g) = X+(g) + Y-(g) (2.33)
The difference in energy between the products on the right and those on the left is
easily found:
Atom XAW xp ( b )
(a)
R.G. Pearson, Inorg. Chem., 27, 730 (1988).
(b)
Reference 45.
more correct. This scale tells us that bonds between non-metallic elements and Be
or Mg are more ionic than bonds of B or Al. The absolute scale tells us that it is
easier to remove a 2p or 3p electron than a 2s or 3s electron. Both statements are
equally true.
Unfortunately the fact that the same label, electronegativity, is used for both
scales creates ample opportunity for confusion and misunderstanding. Since the
Pauling scale has the advantage of seniority and long-established usage, a
solution may be to find another term for the absolute scale. One alternative is to
use the name "electronic chemical potential", /i.
This presents some difficulties. The \i scale is a set of negative numbers, and it
is always more difficult to decide which of two negative numbers is the larger!
More serious is the conflict with the existing usage of \i as the ordinary
thermodynamic chemical potential /ix-
(2.35)
dN P,T
s,v
Electronegativity Scales 53
-E -Nfi -E -Nil
There is a relationship between \i and //T, but it is not a simple one. Recall that
\i depends on constraints on the changes in the number of electrons. If the system
is a single molecule, then an electron can be lost but not gained. In such a case
M = £HOMO> as already mentioned.24 The relationship between the energy of the
molecule, E, and /i is then given by8
where F is the sum of the kinetic energy and the electron-electron repulsion
energy, as before. It is an explicit, but unknown, functional of p. Vnn is the
nuclear-nuclear repulsion.
For all molecules except the smallest, E is much larger than N/i. Examples are
given in Table 2.6. Most of the energy is not relevant to chemistry since it comes
from the inner-shell electrons. For this reason, we subtract from E the energies of
the constituent atoms. This gives us the so-called electronic energy of statistical
thermodynamics, Ej, which is by far the largest part of the thermodynamic
chemical potential.46
Table 2.7 shows some energy changes on forming stable molecules from the
atoms. Now we see that changes in Nfi are comparable with changes in total
(a)
All quantities are the changes upon forming the molecule from
its atoms, and all are based on near Hartree-Fock calculations.
Energies in atomic units.
54 Density Functional Theory
energy, Ej, since they both involve the valence-shell electrons. Also included for
comparison are the changes in total repulsion energy, which are much larger.
Not shown are the large changes in F ne , the nuclear-electronic attraction which
lead to the formation of the molecule.
While Ej is always negative in forming a stable molecule, ANfi can be either
positive or negative. This means that \i cannot be used to predict bond energies
or equilibrium constants for chemical changes, even though it is a component
of Ej and /ix, which can be used make such predictions. Although [x is
constrained to be constant at equilibrium, it is not required by DFT to have a
minimum value.
In summary, unless a better name can be coined, it appears that the term
"absolute EN" will be around for some time. This means that chemists must be
aware of the difference between XP and XA, and make it clear which scale they
have in mind.
Considering the relationship between the two scales further leads to an
interesting result:47 XA changes to XP as a result of changing the electron density
about the atom due to chemical bond formation. But changes in p depend on the
hardness of the atom in question. Perhaps some mix of XA and 77 for the free
atom will lead to Pauling EN.
Figure 2.7 shows the result of searching for such a mix.48 The equation of the
straight line is
Pauling EN
Figure 2.7 Plot of a mixture of XA and 77 vs. XP (Pauling EN) for the representative elements.
After Reference 48.
Calculated \x and rj Values 55
Thus the Pauling scale is more closely related to hardness than to the absolute
EN. This is perhaps not too surprising, if we remember the original concept of a
hard acid or base as one; where the acceptor or donor atom, held on to its valence-
shell electrons tightly.
However it should be pointed out that Figure 2.7 shows only the representative
element atoms. The results would not be so good if the transition metals were
included. Also, calculated values of 77 are shown, and not the experimental ones.
Nevertheless, the close relationship of XP to 77 is undeniable. As might be expected,
there is a linear relationship between a" 1 / 3 and XP, 48 just as there is between a 1 / 3
and rj~l.
When experimental values of / and A are lacking, calculated results are often
satisfactory. A number of methods can be used, including both Hartree-Fock
and density functional theory.51 The best method would be an accurate
calculation of the energies of M, M + and M~, where M is a molecule. This
gives reliable values when M~ is a stable species, i.e., for positive electron
affinities.
Negative electron affinities are troublesome, because the variational method
will simply give A = 0. This is the adiabatic value with the electron not attached
to M. Fortunately, the self-consistent calculations, either by HF or KS methods,
give the orbital energies of the empty, or virtual, orbitals as well as the occupied
ones. The energies of the latter are for an electron in the field of (N — 1)
electrons, which is correct. The virtual orbital energies are for a fictitious electron
in the field of N electrons.
This makes the calculations poor for matching electronic spectra, but it is just
right for the electron affinity. There is still a problem, however, because the
LUMO energy is very dependent on the size of the basis set used, and its quality.
This means that results for both \i and 77 can be seriously in error. Fortunately, it
appears that relative values for a series of related molecules, or a series of
possible structures for a given molecule, are often quite reliable. The same basis
set, or its equivalent, must be used.
It is also possible to calculate \x and 77 in DFT by using the fundamental
definitions of Equations (2.9) and (2.12). The method used is the transition-state
procedure introduced by Slater.52 In essence, it requires the assumption that the
energy is a continuous function of TV, so that derivatives such as Equations (2.9)
and (2.12) have exact meanings. Although it is not true in reality, this is a
convenient mathematical device which gives useful results. Actually what is done
is to calculate both \i and /by the transition-state method, and then to find A and
hence 77, by Equations (2.11) and (2.13).
56 Density Functional Theory
(a)
E. Clementi and C. Roetti, At. Data Nucl. data Tables, 14, 177 (1974).
(b)
Q . Zhao, R.C. Morrison and R.G. Parr, Phys. Rev. A, 50, 2138 (1994).
(c)
K. Siegbahn et al. ESCA Applied to Free Molecules, North Holland/
American Elsevier, New York, 1969.
(d)
Average o f / = 1/2 and 3/2.
There are, in fact, two cases where a fractional value of N makes good sense.
One is a statistical mixture of molecules where the average value of N need not be
integral.23 The other is exemplified by Equation (2.16), where molecules C and D,
after their initial reaction, will have non-integral NQ and N& values. In molecules,
in general, chemists have long considered that a population analysis will give N
values for each atom that are not integers. This is the basis for the concept of
electronegativity, after all.53
Since all of the above calculations are strongly dependent on orbital energies,
it is worthwhile to close with a short comparison of orbital energies, as calculated
by HF and by DFT, and as measured experimentally by ESCA and photo-
electron spectroscopy. These are shown in Table 2.8. Both the HF and KS
orbital energies are quite close to the experimental ionization potentials. In
principle, the KS energy for the outermost orbital should equal the first
ionization potential, but this has not yet happened. It will be recalled that the KS
results depend on how well the exchange-correlation potential is represented.54
REFERENCES
1. General references on DFT: R.G. Parr and W. Yang, Density Functional Theory for Atoms
and Molecules, Oxford University Press, New York, 1989; N.H. March, Electron Density
Theory of Atoms and Molecules, Academic Press, New York, 1992; E.S. Kryachko and
E.V. Ludema, Density Functional Theory of Many Electron Systems, Kluwer Press,
Dordrecht, 1990.
2. P. Hohenberg and W. Kohn, Phys. Rev., 136, B864 (1964).
3. For excellent brief reviews of DFT, see T. Ziegler, Chem. Rev., 91, 651 (1991); W. Kohn,
A.D. Becke and R.G. Parr, J. Phys. Chem., 100, 12974 (1996).
4. J.C. Slater, Phys. Rev., 81, 385 (1951); J.C. Slater, Adv. Quantum Chem., 6, 1 (1972);
J.C. Slater and K.H. Johnson, Phys. Rev., 135, 544 (1972).
References 57
5. J.G. Snijders and EJ. Baerends, /. MoL Phys., 36, 1789 (1978).
6. L. Fan and T. Ziegler, /. Am. Chem. Soc, 114, 10890 (1992) N. Allinger and K.
Sakakibara, /. Phys. Chem., 99, 9603 (1995).
7. W. Kohn and L.J. Sham, Phys. Rev., 140, A1133 (1965).
8. R.G. Parr, R.A. Donnelly, M. Levy and W.E. Palke, /. Chem. Phys., 68, 3801 (1978).
9. M. Levy and J.P. Perdew, Phys. Rev. A, 32, 2010 (1985).
10. R.S. Mulliken, J. Chem. Phys., 2, 782 (1934).
11. R.P. Iczkowski and J.L. Margrave, /. Am. Chem. Soc, 83, 3547 (1961).
12. R.T. Sanderson, Science, 121, 207 (1955).
13. R.G. Parr and R.G. Pearson, J. Am. Chem. Soc, 105, 7512 (1983).
14. G.J. Schulz, Phys. Rev., 5, A1672 (1972); K.D. Jordan and P.D. Burrow, Chem. Rev., 87,
557 (1987).
15. R.G. Pearson, Proc Nat. Acad. Sci. USA, 83, 8440 (1986).
16. T. Koopmans, Physica, 1, 104 (1934).
17. L. Komorowski, Chem. Phys., 114, 55 (1987); T.K. Ghanty and S.K. Ghosh, /. Phys.
Chem., 97, 4951 (1993); U. Dinur, ibid., 7894; S. Hati and D. Datta, /. Phys. Chem., 98,
1436 (1994).
18. R.G. Pearson, /. Am. Chem. Soc, 110, 2092 (1988).
19. R. Nalewajski and R.G. Parr, /. Chem. Phys., 77, 399 (1982).
20. R.G. Parr and W. Yang, /. Am. Chem. Soc, 106, 4049 (1984).
21. K. Fukui, Theory of Orientation and Stereoselection, Springer-Verlag, Berlin, 1972;
I. Fleming, Frontier Orbitals and Organic Chemical Reactivity, John Wiley,
New York, 1976.
22. M. Berkowitz, /. Am. Chem. Soc, 109, 4823 (1987).
23. J. P. Perdew, R.G. Parr, M. Levy and J.L. Balduz, Jr., Phys. Rev. Lett., 49, 1691
(1982).
24. J.A. Alonso and N.H. March, J. Chem. Phys., 78, 1382 (1983).
25. S.K. Ghosh and M. Berkowitz, /. Chem. Phys., 83, 2976 (1985).
26. W. Yang and R.G. Parr, Proc. Nat. Acad. Sci. USA, 82, 6723 (1985).
27. M. Berkowitz and R.G. Parr, /. Chem. Phys., 88, 2554 (1988).
28. M. Berkowitz, S.K. Ghosh and R.G. Parr, J. Am. Chem. Soc, 107, 6811 (1985).
29. This is called the second-order Jahn-Teller (SOJT) method; R.G. Pearson, Symmetry
Rules for Chemical Reactions, John Wiley, New York, 1976.
30. W. Yang, C. Lee and S.K. Ghosh, /. Phys. Chem., 89, 5412 (1985).
31. Z. Zhou and R.G. Parr, J. Am. Chem. Soc, 111, 7371 (1989).
32. Z. Zhou, Int. Rev. Phys. Org. Chem., 11, 243 (1992).
33. J. Hinze, M.A. Whitehead and H.H. Jaffe, /. Am. Chem. Soc, 85, 148 (1963); R.S. Evans
and J.E. Huheey, /. Inorg. Nucl. Chem., 32, 373 (1970).
34. R.F. Nalewajski, /. Am. Chem. Soc, 106, 944 (1984).
35. R.G. Pearson, /. Am. Chem. Soc, 107, 6801 (1985).
36. M.J.S. Dewar, /. Am. Chem. Soc, 74, 3341, 3357 (1952).
37. K. Fukui and H. Fujimoto, Bull. Chem. Soc. Japan, 41, 1989 (1968); idem, ibid., 42, 3399
(1969).
38. H. Fujimoto and S. Satoh, /. Phys. Chem,, 98, 1436 (1994).
39. A. Tachibana and R.G. Parr, Int. J. Quantum Chem., 41, 527 (1992); J.L. Gazquez and
F. Mendez, /. Phys. Chem., 98, 4591 (1994); Y. Li and J.N.S. Evans, /. Am. Chem. Soc,
111, 7756 (1995).
40. P.K. Chattaraj and P.v.R. Schleyer, /. Am. Chem. Soc, 116, 1067 (1994); D.R. Garmer
and N. Gresh, ibid., 3556.
41. P.K. Chattaraj, H. Lee and R.G. Parr, /. Am. Chem. Soc, 113, 1855 (1991).
58 Density Functional Theory
42. See properties of the grand canonical ensemble, for example, in T.L. Hill, Statistical
Mechanics, McGraw-Hill, New York, 1956, p. 72.
43. L. Pauling, / . Am. Chem. Soc, 54, 3570 (1932).
44. For a summary of scales see J. Mullay, in Structure and Bonding, 66 "Electronegativity", 1
(1987).
45. L.C. Men, / . Am. Chem. Soc, 111, 9003 (1989); L.C. Allen and E.T. Knight, J. Mol.
Struct. (Theochem.), 261, 313 (1992).
46. The zero point energy is usually added to Ej, making it less negative.
47. L. Komorowski, Chem. Phys. Lett., 103, 201 (1983); Z. Naturforsch, 42A, 767 (1987).
48. J.K. Nagle, / . Am. Chem. Soc, 112, 4741 (1990); J.K. Nagle, private communication.
49. R.G. Pearson, / . Orgn. Chem., 54, 1423 (1989).
50. S.G. Lias, L.F. Liebman and R.D. Levin, / . Phys. Chem. Ref. Data, 17, Suppl. No. 1
(1988).
51. See the papers in Structure and Bonding: (a) 66 "Electronegativity", K.D. Sen and C.K.
Jorgensen, Eds. (1987); (b) 80 "Chemical Hardness", K.D. Sen, Ed. (1993).
52. J.C. Slater, Quantum Theory of Molecules and Solids, McGraw-Hill, New York, 1974, Vol. 4.
53. For an informative review see D. Bergmann and J. Hinze, Angew. Chem., Int. Ed. EngL,
35, 150 (1996).
54. For a discussion see A.D. Becke, / . Chem. Phys., 98, 5648 (1993).
55. J.B. Maksic, (Editor) Theoretical Models of Chemical Bonding, Part 2. Springer-Verlag,
Heidelberg, 1990, p. 55.
3 Application of DFT
INTRODUCTION
(XC-XD)
This equation can be very useful, as we shall show. However, it is easy to over-
estimate its importance. Chemical reaction between two species depends on many
other factors besides the ease of electron transfer; the donating and accepting
orbitals are certainly among these other factors to be considered. But equally
important is the nature of the reactants, which determines whether there is a
reasonable reaction path to reach stable products. This will be illustrated by
considering reaction between Cl2 and both Mg and (CH3)2O in detail.
Calculation with Equation (3.1) gives the results
which does not explain why Equation (3.2) represents an energetic reaction
leading to MgCl2, whereas reaction (3.3) is only mildly energetic, leading to the
charge transfer complex (CH3)2O:C12. Consideration of the orbitals involved is
of little help. In both cases the electrons come from an orbital which is non-
bonding, a b! orbital in the case of the ether, and an s orbital in the case of the
metal.1 The accepting orbital on Cl2 is an anti-bonding au.
Putting electron density into the au orbital helps break the Cl-Cl bond.2 But the
bonding in the ether remains intact and the reaction stops after the initial charge
- LUMO
-2
r • -x
r • -x
Mg
-10 HOMO
(CH 3 ) 2 O
-12
Cl2
eV
an
Figure 3.1 Orbital energy diagrams for several molecules, showing x d TJ.
The last step is a simple rotation. The ether does not follow the same path because
of the higher energy of both (CH 3 ) 2 OCl + ,Cr and (CH3)2OC12, containing
tetracovalent oxygen.
Introduction 61
Figure 3.2 shows the HOMO, the LUMO and x f° r several more molecules.
NH 3 will donate electrons to Pd, and Pd will donate electrons to CO. Experi-
mentally, Pd:NH 3 is an unstable compound, Pd:CO is more stable, and NH3:CO
shows no signs of existing, though NH 3 should donate electrons to CO more
than Pd does. In this case, consideration of the orbitals does help to understand
the results.
Figure 3.3(a) shows the accepting orbital (the LUMO) of CO. The donating
orbital (not shown) is a filled d orbital on Pd, or any transition metal, M. Clearly
there is good overlap between the two orbitals. Also not shown are the HOMO
of CO, a weakly anti-bonding a orbital concentrated on carbon, and the LUMO
of Pd, the 5s orbital. These also overlap well and lead to the synergistic effect of
+5 eV
LUMO
-X
-X
-5
-X
Pd
-10 eV
- O - O - HOMO
CO
Figure 3.2 Orbital energy diagram showing HOMO, LUMO and x f° r several species.
62 Application of DFT
M
0
(a) (b)
Figure 3.3 The TT* orbitals of (a) CO (b) N 2 , showing greater overlap of CO with a d orbital
on the transition metal atom, M.
(TT + cr)-bonding. NH 3 , being a pure a-donor, cannot interact in this way with
either Pd or CO. The empty n orbitals of NH 3 are very high in energy, even
higher than the LUMO shown.
The conclusion that 7r-bonding in metal carbonyls is more extensive than
cr-bonding was at first quite surprising.4 Looking at the formula M(CO)«, most
chemists would consider the metal atom center as the acceptor and the CO
ligands as the electron donors. In fact CO was earlier listed as a soft base. It now
must be considered an acid, and a hard one at that!
Earlier ab-initio calculations on metal carbonyls were contradictory on the
question of whether 7r-bonding or cr-bonding was dominant. However the latest,
and best, calculations using DFT with relativistic corrections show clearly that
7r-bonding is the more important.5 There is a wealth of information available on
bond strengths and stabilities of transition metal carbonyls. Since we are dealing
with orbitals on CO and the metal that are rather constant, we can check to see if
Equation (3.1) is consistent with the bond energy data for various metals.
There is a problem in that the ground state of the metal atoms are all high-
spin, except for Pd, which is (4d)10. The metal carbonyls are all low-spin with all
electrons paired up in d orbitals, as far as possible. Fortunately, in most cases it
is possible to correct / and A for this change in configuration, using spectroscopic
data.4 Although DFT is not really valid for these excited states, we can assume
that a low-spin metal atom, M, is a good model for a low-spin M(CO)«_i, which
then reacts with CO to form M(CO)«.
Table 3.1 gives a list of x a n d rj values calculated for the valence state of the
transition metals where the data are available. The valence state is (nd)k in all
cases, except Cu, Ag and Au, where the ground state and valence state are both
(nd)l0(n + l)s, but ionization is from the d shell and not the s. The values of
AN calculated from Equation (3.1) are also given, and the values of AH$ for
the reaction
M X
(a)
[eV] r/ a) [eV] AN L\H$ [kcal/mol] (b)
(a)
For low-spin valence state.
(b)
Activation enthalpy for loss of first CO in known carbonyls. See
References 4 and 5 for sources.
(C)
V(CO) 6 is stable to dissociation.
(d)
ForMn2(CO)10.
(e)
ForCo2(CO)8.
(f)
Theoretical values (Reference 5).
Various experimental criteria are used for this rating, especially IR spectra, but
the order is conirmed by DFT calculations in a number of cases.5 Bond energies
also fall roughly in the same order, both from experiment and theory. However,
the variation from one end to the other is quite small, about 10 kcal/mol. This
64 Application of DFT
results from the best <j-donors being at the end of the list, offsetting their poor
7r-accepting characters.
Taking a low-spin nickel atom as our metal, we next calculate AN for the
ligands. The following is the order of decreasing AN
The last five compounds have AN negative, meaning they are net <j-donors.
The calculated ordering in AN is similar to that given by experiment, but there
are discrepancies. These are due, for the most part, to orbital effects. For
example, Figure 3.3(b) shows the TT* orbital of N 2 , compared with that of CO in
Figure 3.3(a). Clearly N 2 overlaps more poorly with a d orbital than CO does.
Also the cr-orbital of CO, concentrated on carbon, is a better donor than that of
N 2 . An examination of the frontier MOs of pyridine and acetonitrile reveals that
they also are poorly placed to give good overlap.1
Some variation of the orbitals can be tolerated, if the changes are due to
substituents on a common substrate. In a series of olefins, the reactive orbitals
are always the TT and TT*, with perturbations due to the attached groups. An
example of the Equation (3.1) would be the bonding of various olefins to Ni(0).7
The data are the equilibrium constants in benzene for
NiL 3 + olefin <± NiL2(olefin) + L Keq (3.6)
where L is a phosphite ligand.
Table 3.2 shows values of Keq for various olefins, together with their x and 77
values. Assuming that x and 77 for NiL2 are the same as for low-spin Ni, the values
of AN have also been calculated by using Equation (3.1) and are shown in
Table 3.2. These are for
NiL2 + olefin <± NiL2(olefin) (3.7)
which is related to Equation (3.6) by a constant term.
The calculated AN values correlate very well with the equilibrium constants.
Large positive values mean strong bonding, with vr-bonding from metal to olefin
dominating. Negative values mean that cr-bonding to the metal is greater than
7r-bonding. Clearly cr-bonding is less effective than 7r-bonding.
Ethylene has more TT- than cr-bonding, in agreement with theoretical calcu-
lations for the reaction of Ni(PH3)2 with C2H4.8 The calculated values of AN for
the reactions of low-spin Pt and Pd are -0.0495 and + 0.005, compared with that
for Ni, +0.053. This agrees with the theoretical results for binding energies,
Ni > Pt > Pd, especially if 7r-bonding is better than cr-bonding.
When we are primarily interested in the relative values of AN for a related
series, approximate values of x and 77 for the common reactant are usually
adequate. But we can do a little better by referring back to Equation (2.29) of
Introduction 65
Table 3.2 Electron Transfer in Reactions of Olefins with Low-Spin Nickel Atoms (a)
Chapter 2. A version of this for the cases where we have unequal stoihiometries
of the two reactants, as in the reaction9
(3.8)
(/WM + nfiLaL) _
X
-
where \x is the common value for MLrt after equilibrium is reached.
Thus we can calculate the change (JJL - //M) due to the attached ligands.
Unfortunately, we know that Equation (3.9) is incomplete, since it gives only the
initial effect. Further changes in fi will occur due to covalent and ionic bonding
changes, which change p.9 For example, take the reaction of two identical atoms:
2H = H2
(3.10)
v }
6.7 eV
Here Equation (3.9) predicts no change.
Nevertheless, Equation (3.9) should give some idea of the effect of substituents
on a reactive center. This is a problem of major importance in chemistry. Ziegler
has used DFT to calculate the bond dissociation energies for a number of
Ru(CO4)L complexes,10
Table 3.3 Values for RuL 4 Complexes Estimated from Equation (3.9)
To estimate \i for Ru(CO)4, we again start with a low-spin Ru atom and use
Equation (3.9) to calculate a better value. The same may be done for substituents
other than CO, such as the examples in Table 3.3. These results are very
reasonable. By removing electron density from the metal atom, CO should make
Ru more positive, and a poorer electron donor. Phosphine does not remove
electron density so much as CO, and has a smaller effect. The result with NH 3
may be surprising at first, but it is also reasonable, since ammonia has no
7r-bonding tendency, and does give electron density to the metal by cr-bonding.
The comparison between PH 3 and NH 3 is in agreement with theoretical
calculations.8'11 The 7r-bonding to PH 3 does not use d orbitals on P, but instead
the anti-bonding <x* orbitals of the P-H bond.
The predicted lower EN of Ru(NH 3 ) 2 compared with Ru(PH3)2 means that
the former will be more reactive in cases where electron donation is required.
An example would be the bonding of ethylene. The dissociation energy of
Ru(NH3)2C2H4 is 64kcal/mol, compared with 30kcal/mol for Ru(PH3)2C2H4.8
Of course Ru(NH 3 ) 2 would be much more difficult to prepare than Ru(PH3)2.
Analogs of Ru(PH3)2C2H4 do exist, but not those of Ru(NH3)2C2H4.
In any case, there is a definite prediction about the effect of a neutral ligand on
the reactivity of a transition metal. A ligand with a small value of x will be acti-
vating for electron donation. This is almost the same as saying that a hard ligand
will be activating. Hardness and small EN usually go together because the
electron affinity has a large negative value.
There is ample evidence to support this prediction, since it simply says that hard
ligands will favor a higher positive oxidation state for the central metal. We saw
an example of this in Chapter 1, when the acidity of transition metal hydrides was
discussed. It is also an example of the HSAB Principle, or the symbyiotic effect.
To obtain a value of \ adequate for more quantitative comparisons, we can
also consider the further changes in \i that take place as the interacting fragments
approach each other more closely. There is plenty of experimental evidence on
this point, since we simply look at fi calculated from Equation (3.9), and the
experimental value of fi for various ML«.9 In this case we need L = CO, and there
are data for Cr(CO)6 and Fe(CO)5 (Table 3.4).12
It appears that covalent bonding leads to an increase in the electronic chemical
potential.9 As the interacting orbitals of two fragments get closer, the bonding
ones go down in energy and the anti-bonding ones go up even more. Thus / is
increased and A becomes more negative, so that ( / + A) gets smaller.
Introduction 67
With the estimated value of x — 4.0 eV for Ru(CO)4, we can calculate AN for
reaction (3.11) for a series of ligands, L. We still use r/ — 2.3 eV, the value for low-
spin Ru, since we do not know how to calculate the change in r\ due to the ligands
in any easy manner. As we shall see later, r\ undoubtedly increases, Table 3.5
compares AN with Do for a series of paired ligands, where O is replaced by S.
We find that AN correlates with Ah if we compare the paired ligands with
each other: CO < CS, CH2O < CH2S; CO2 < CS2. Larger AN means stronger
bonding, but not for O2 > S2, where the opposite occurs. The inversion for the
last case is due to the promotion energy. The ground state for O2 and S2 is a
triplet, whereas the valence state needed is a singlet with two electrons paired in a
single 7T* orbital. The pairing energy is 253kJ/mol for O2, but only 122kJ/mol
for S2. This reduces the net bonding energy for oxygen.
Table 3.5 contains more information on the yr-bonding in each case. There is
an estimated value of the energy lowering, AE^ and an estimate of the number
of electrons transferred from the metal to the ligand, nw. The values of AN and
AEK are correlated now in all cases. While AN and nn are correlated, they are not
equal. A major correction would be na, the number of electrons donated from the
ligand to the metal, but there are other factors as well.
Do values have also been calculated for CSe and CTe, as well as for CO and CS
binding to Ru(CO)4, and for Se2 and Te2 as well as O2 and S2.10 The bonding for
Se and Te is about the same as for S, and O is the one that differs. This is an
illustration of a general phenomenon in chemistry. The behavior of the first-row
(a)
See Reference 10.
(b)
Z>o, dissociation energy.
(c)
Number of electrons transferred from metal to ligand.
68 Application of DFT
! • -o-o-
Ru(CO)4
cs
-O~O-
co
Figure 3.4 Frontier orbital energies on a relative scale for Ru(CO) 4 , CO, CS, CSe and CTe.
elements (Li to F) is always different from that of the later rows, which do not
differ much from each other.13 This is particularly true for the non-metallic
elements. Fluorine, with / = 17.42 eV, differs markedly from chlorine, bromine
and iodine, where / = 12.97, 11.81 and 10.45 eV. Also, the energy difference
between ns and np is large for n — 2 and small for n = 3, 4, 5 and 6.
The result of these differences is illustrated in Figure 3.4, which shows the
relative frontier orbital energies for CO, CS, CSe and CTe, compared with those
of Ru(CO)4. Carbon monoxide is much harder than any of the others. Therefore
electron transfer, in both directions, will be much easier for CS, CSe and CTe.
This leads to a larger value of DQ in all cases. The stronger bonding of CS,
compared with CO, and transition metals is well known.
These successful applications of Equation (3.1) to estimate bond energies are
exceptions, rather than the rule. If AN is too large, so that ionic bonds are
formed, then size factors will dominate the bonding. We want the bonding to be
mainly due to electron transfer in one direction, but limited in extent. The best
examples will be those where a coordinate covalent bond is formed. Charge
transfer complexes should usually qualify, but only if similar molecules (and
orbitals) are compared.14
RATES OF REACTION
Take a ground-state iron atom as representative of MLW. Then Table 3.6 shows
the values of AN calculated from Equation (3.1) or the reaction with a number of
common X - Y molecules. The results are in remarkable agreement with
observations. Large positive values of AN are found for facile, fast reactions,
and negative values for slow or unobservable reactions. Thse statements are based
on studies of isolated iron atoms in low-temperature matrices, and on rate studies
of various low-valent metal complexes reacting with these X—Y molecules.16
Reaction (3.12) proceeds by electrons being transferred from MLW to an empty
orbital which is anti-bonding for X and Y, usually the LUMO of X - Y . 2 The
unreactive molecules in Table 3.6 have high-energy LUMOs, large negative
X-Y AN X-Y AN
F2 0.26 HF 0.030
Cl 2 0.17 HC1 0.027
Br2 0.15 H2S 0.007
h 0.12 C 2 H-H 0.004
H2 0.11 C 6 H 5 -H 0.002
o2 0.11 CH3-CI -0.013
CCI3-CI 0.10 C 6 H 5 O-H -0.015
NO3-H 0.094 CH 3 -F -0.035
HI 0.072 H2O -0.036
H-CN 0.066 CH 3 -H -0.060
CH3-I 0.048 NH 2 -H -0.065
SiH 3 -H 0.035 CH 3 O-H -0.066
70 Application of DFT
values for A, and small values for x- Hence AN is negative, indicating a positive
contribution to the activation energy barrier, making it higher.
Naturally, other factors that influence rates of reaction cannot be ignored. For
example, the rate of addition of HX molecules is some 107 times faster than the
addition of CH3X. This refers to the concerted addition of molecular HX to
IrCl(CO)[P(C6H5)3]2 in solvents such as toluene.17 Clearly steric hindrance is
much greater in CH3X than in HX and accounts for the slower reaction.
In agreement with Table 3.6, oxidative addition of SiH4 is much more facile
than that of CH4. An ab-initio MO calculation has been made of the reactions of
both SiH4 and CH 4 with Pt(PH3)2.18 The activation barrier for CH 4 is calculated
to be 28 kcal/mol, and there is no barrier for SiH4. As expected, there is an early
TS for SiH4 and a late one for CH4. Experimentally it is found that phenols react
readily with Pd(PR3)3 in toluene, and alcohols do not react.19 This is in agreement
with AN being much more negative for CH3OH than for C6H5OH in Table 3.6.
In addition to changing the substrate, we can also change the metal atom.
Metals of lower electronegativity than iron should be more reactive, and metals
of higher % should be less reactive, up to a point. If x for the metal becomes
comparable with x f° r the atoms or radicals X and Y, then electron transfer in
one direction is not required, and Equation (3.1) is no longer a good criterion.
There is limited evidence available on free transition metal atoms in low-
temperature matrices. The most reactive atoms are Sc, Ti and V, as expected.16
As we shall see in Chapter 5, there is reason to believe that there is a parallelism
between the reactivity of isolated metal atoms and the atoms in the bulk metal.
Thus the free atoms of the noble metals, such as Os, Ir, Pt and Au, are expected
to be slow to react.
But there is an opposing effect. The strength of the bonds to the noble metals
are usually the strongest, when comparing the metals in a given triad. The order
most often found is 5d ~ 3d > 4d. Particularly when the bond to the metal is a
pure a-bond, as is the case for H or CH3, the most EN metals form the strongest
bonds.20 When vr-bonding is dominant, as in the metal carbonyls, the situation
is reversed.
There is a great deal of information available on rates of oxidative addition for
metal complexes,21 but we can only make meaningful comparisons between
metals if the ligands and oxidation state of the metal are held constant. This
again means comparing metals in a triad. The same result is found for the rates
of reaction as for the bond strengths, 5d~3d^>4d. This suggests that for
reaction (3.12) we usually have a late transition state resembling the products.
The final energy, including the bond strengths of M—X and M—Y, determines
the height of the energy barrier.
A clear prediction is made about the effect of the auxiliary ligands, L, on the
ease of oxidative addition for a given M. Ligands with a small value of x will
increase the reactivity. That is, hard ligands will make the complex a better
reducing agent, or a better electron donor. Equally we can say that hard ligands
will increase the strength of M - X and M - Y bonds.
Rates of Reaction 71
ML + H2 = MLH 2 AE (3.14)
so that a negative AE means MLH 2 is stable. The results found for neutral
ligands are listed in Table 3.7. The large stabilizing effect of the hard ligand,
H2O, stands out. On the other hand, the bond energy between Ni and H2O, is
very low, 5kcal/mol, compared to 14 to 40kcal/mol for the other ligands.
While Equation (3.1) predicts that H2O will donate electron density to the
nickel atom, this is not what actually happens at the bonding distance of 2.33 A.
The bonding is van der Waals in character, being half dipole-induced dipole
and half correlation energy.23 There is little, if any, electron transfer in either
direction. This result is not a violation of DFT since the chemical potential of
Ni-H 2 O is made constant everywhere by the changes in electron density which
do occur.
The weak bonding of H2O and NH 3 to neutral atoms means that it is not easy
to make complexes conaining both a metal atom in a low oxidation state and
hard ligands. An attempt to make Rh(NH3)4~ in aqueous solution would
certainly lead to the formation of Rh(NH 3 ) 4 HOH + instead. The Rh(III) hydride
would result from the oxidative cleavage of water.24 A Rh(I) complex containing
four nitrogen donor atoms has been made, Rh(C2DOBF2), where C2DOBF2 is a
complicated tetradentate ligand.25 It is very reactive towards oxidative addition.
In a more straightforward example, a detailed study has been made of the
reaction
The order of rates found for different X~ was F~ > N j > CL" > Br~ > NCO~ >
I" >NCS", with hard F~ reacting 100 times faster than soft (S-bonded)
AE [kcal/mol]
L Ni Pd
None -6 -9
C2F4 4-9 -
C2H4 +2 + 13
CO +3 +25
N2 -9 +3
PH 3 -9 +4
H2O -17 -19
72 Application of DFT
NCS~. 26 Results like these, and the others cited above, are rather unexpected in
more classical thinking. Soft ligands were thought to put negative charge on the
metal atom, making it a better electron donor.
Ethylene, with x = 4.4 eV and rj = 6.2 eV, is a typical organic molecule in being
intermediate in EN so that reaction with both electrophiles and nucleophiles is
possible. For substituted olefins, \ ranges from 3.0 eV for (CH3)2 = C(CH3)2 to
7.3 eV for (NC)2C = C(CN)2. More EN olefins react best with electron donors, or
nucleophiles, as we have already seen in Table 3.2. The least EN olefins react best
with reagents like Br2 and H 3 O + . In the case of Br2, with x = 6.6 eV, there is good
agreement between the rate constants and AN calculated from Equation (3.1), as
shown in Table 3.8.27
In the Diels-Alder reaction, butadiene (x = 4.3eV) is less reactive towards
olefins than cyclopentadiene (x = 3.8 eV). This suggests that the diene is the
electron donor to the olefin
In that case the most EN olefins should react the fastest. Table 3.9 shows that
this is the case, with near-perfect agreement between the ordering of AN and k,
the second-order rate constant. The very low reactivity of ethylene itself is not
explained, however. Cyclopentadiene reacting with itself as an olefin is included
in the table, and it may be the one in the table that is out of line.
An unusual application of Equation (3.1) was made in a study of anionic
polymerization of olefins as a function of the olefin and of the initiator, AB.28
Table 3.8 Values of AN and Relative Rates of Olefin Reactions with Bromine in Methanol
(a)
After Reference 27.
(b)
Estimated from rates of hydration.
Rates of Reaction 73
Table 3.9 Calculated values of AN and Rate Constants for Reactions of Olefins with
1,3-cyclopentadiene
(a)
At 20°C in dioxane. See Reference 29.
(b)
Estimated from gas-phase data.
AB + C H 2 - C H R -> B - C H 2 - C H R - + A+ (3.17)
The role of A + and the terminating steps are not well understood. Since the reac-
tions are run in solvents such as dioxane or tetrahydrofuran, ion-pairing and
aggregation are occurring.
Eleven olefins and ten initiators were considered in all. The initiators were
NaCH 3 , LiCH3, n-C4H9Li, t-C4H9Li, CH3MgCl, NaOCH3, LiOCH3, C5H5N,
(CH3)3N and H2O, in order of decreasing efficiency. Since there were no
experimental values of / or A available for the first eight of these, they were
all calculated by ab-initio MO methods. The energies of AB, AB + and AB~ were
calculated separately to find / and A for all the initiators. To have consistency,
the same was done for the 11 olefins. Comparison of the x values found in this
way with those that were known from experiment showed that the right ordering
was obtained, though the theoretical values were about 0.8 eV lower, except for
(CH3)3N and H2O.
The values of AN were calculated from Equation (3.1) for all 110 possible
combinations. Table 3.10 gives some representative results. The olefins are in
order of increasing reactivity, as found experimenally, with the most reactive at
the bottom. The values of AN calculated for all ten initiators put all the olefins in
the right order of reactivity, the largest value of AN being for CH 2 =C(CN) 2 .
74 Application of DFT
Initiator
(a)
After Reference 28.
Placing the initiators in the right order was not successful. AN should have
decreased, or become negative, going from left to right in the table for any one
monomer. This is only roughly true, CH3MgCl having AN too small, and all the
molecular initiators having AN too large. Considering the complex nature of a
Grignard solution, and the fact that molecular initiators must have different
mechanism, these failures are not surprising. The success with the monomer
ordering shows that the transfer of negative charge to the olefin must always play
a key role.
(a)
Data from reference 32.
(b)
Vertical value. O. Edquist, E. Lindholm, L.E. Selin, H. Sjogren and
L. Astorink, Phys. Scr., 7, 172 (1970)
react the fastest. Electrons are being removed from the n orbital of ethylene,
converting the double bond to a single one:
R + C2H4 A R - C H 2 - C H 2 (3.19)
76 Application of DFT
Table 3.12 Values of AN for Reaction of Free Radicals with Ethylene (a)
(a)
Gas-phase reactions at 437 K. Data as given in Reference 27, except for
NCO, Reference 31.
R-f-O2^R-O2 (3.20)
Table 3.13 compares AN with the experimental second-order rate constant for a
number of radicals. Certainly the most nucleophilic radicals now react the
fastest, though the ordering is not perfect. The phenoxide and methoxide radicals
react more slowly than predicted.
To understand this, recall that only an early transition state should show
a correlation between AN and k. A late TS is more likely to correlate k with
the exothermicity of the reaction, or the R—O2 bond strength.34 Since O—O
bonds are notoriously weak, a slow reaction is not unexpected for R O radicals.
It is likely that the same explanation serves for the non-reactivity of CH3S,
and part of the inertness of HS, though we usually think of the S-O bond
as strong.
This S—O bond, however, is not strong since addition of RS- to O2 will lead
to the formation of a derivative of sulfenic acid, RSOR'. Such derivatives
are very unstable. Stable S-O bonds are the ones present in the isomeric
sulfoxides, RR'SO, which are coordinate covalent bonds. In HSAB terms,
RSOR' is a combination of a soft acid, RS + , with a hard base, R'O~, while
sulfoxide is a combination of a soft acid, oxene (the ! D oxygen atom), with a soft
base, RSR'.
Reactions of Free Radicals 77
(a)
Data rom Reference 33, except where otherwise indicated.
(b)
At298K.
(c)
Reference 35.
(d)
Reference 36.
(e)
Reference 37.
(f)
Reference 38.
Table 3.14 Values of A N and Rate Constants for Reaction of O H Radical with Olefins (a)
(a)
Data from Reference 39.
(b)
A t 2 9 8 K , gas phase.
(c)
Cis and trans both the same.
78 Application of DFT
XP XA
Another variation of Table 3.12 is to change the olefin, while keeping the
radical constant. Table 3.14 shows some results for the reaction with OH radical,
X = 7.5 eV and 77 = 5.7 eV. As expected, the least electrophilic olefins react the
fastest. As the large values of AN suggest, there is no energy barrier for most of
these reactions. Instead, electron transfer must be increasing the frequency of
collision, or the duration of the collision. Only tetrachloroethylene has an
appreciable energy barrier of 2.6kcal/mol. This shows up as a reduction in the
rate due to steric hindrance.
It is worth nothing that the use of both / and A for both reactants gives the
best correlation with the rate constants. The more common procedure in frontier
orbital theory is to consider only / for the donor and A for the acceptor.40 This
works for the alkyl olefins, but fails badly for the chloro olefins. If a nuleophilic
radical, such as (CH3)3C, is used as the common reagent, then the order in
Table 3.14 is completely inverted.41 The chloro olefins now react very rapidly,
and the alkyl olefins are slow.
Pauling electronegativities (xP) are also available for many radicals or groups.
These are obtained, not from the free radicals, but from properties of molecules
containing the groups. Various properties have been used, and this leads to some
fluctuations in the values calculated. Some typical results42 (Table 3.15) may be
compared with the absolute values (XA) of Table 3.11. While the units are not the
same, the fact that the orders are reversed suggests correctly that the Pauling
ENs will not be useful in discussing free-radical reactions.
E H
(3.21)
2
•LUMO
o
-2
-4
-6
-8
O^OHOMO
-10 C6H6
C 6 H 5 NO 2
-12
eV Cl 2
an
Figure 3.5 Orbital energies for several molcules, showing \ d V-
80 Application of DFT
(a)
Reference 27.
Aromatic Electrophilic Substitution 81
Table 3.17 Values of AN and krei for Reaction of Cl2 with Benzene
and Heteroaromatic Compounds
Other condensed aromatic ring compounds will also fall in line, since as the
softness increases, both AN and reactivity will increase. We also see that multiple
substitution by NO 2 or CN groups will increase x- This will make nucleophilic
substitution possible, if a suitable leaving group is also present.
None of the above results addresses the question of the position of reaction in
the aromatic molecule. In DFT this is done by considering the Fukui function,/.
It is interesting that this orientation problem was also the topic of the first paper
on frontier orbital theory.47 Reaction was predicted to occur at the position of
highest frontier orbital (FO) electron density. The frontier orbital in electrophilic
substitution would be the HOMO. If this orbital were written as the usual linear
combination of atomic orbitals, then the density at each atom would simply be
the square of the coefficient in the LCAO, or c? where / indicates the atom. Since
this is also one of the ways of approximating /, the success of the FO method
may also be claimed for DFT. However, the details of the Fukui function
application will be postponed briefly to look at a method unique to density
functional theory, and using the concept of hardness.48
We can see from Equation (3.20) that the intermediate C is a different 7r-system
from the reactant benzene. In terms of the 7r-electrons only, it is a linear
pentadienyl cation, C5H5". Thus there is a change in the ?r-electron energy. Let us
assume that the activation energy for electrophilic substitution comes only from
this 7r-energy change. We can readily calculate this from simple HMO theory.
Figure 3.6 shows the resulting orbital energies in terms of a and /?, the usual
coulomb and exchange integrals.
The energy of the 7r-electrons of C6H6 is 6a + 8/3, and that of C5H^ is
4a + 5.46/?. The change of 2a is added to the energy change in the cr-system,
which is then ignored by assuming it to be constant for various similar molecules.
The energy change of —2.54/3 was called the cation localization energy, L + , by
Wheland.44 Changes in L+ from one aromatic system to the next were then
responsible for their differing reactivities.
This assumption was very successful when applied to the problem of the
orientation of substitution in a molecule where there was more than one kind of
carbon atom. For example, naphthalene can react at either the 1- or 2-position.
Reaction occurs primarily at the 1-position, and L | is indeed smaller than L j .
The use of L + was not always successful for changes in the reacting molecule,
presumably because differences in the cr-energies became important.
82 Application of DFT
-2(5
-V3p
-p -P
C6H6 C5H5
Figure 3.6 Hiickel energies for the 7r-electrons of benzene and the pentadienyl cation. The
zero of energy is a.
Figure 3.6 can also be used to calculate /x and 77 for the reactant and the
intermediate. The orbital energies in HMO theory are given by
where the Xi are numbers which are fixed for the particular conjugated molecule,
once the connectedness is given. From the definitions of \i and 77, we have
For C6H6 we have // = a and 77 = -/3, remembering that both a and (3 are
negative. For the intermediae C, ji = (a + (3/2) and 77 = —(3/2.
The question to be answered is whether changes in 77, or /x, on going from
reactant to transition state, offer any clue as to the magnitude of the activation
energy. The complex C is assumed to be close in energy to the transition state.
If we define the activation hardness as
then in the case at hand, A77T is —0.50/3. Similar calculations can be made for
electrophilic substitution at any of the positions in other aromatic molecules.
Aromatic Electrophilic Substitution 83
Table 3.18 Reactivity Indices for Substituted Benzenes
C6H5F 2 0.462 6
3 0.492 0.5
4 0.435 87
C6H5C1 2 0.480 15
3 0.494 0
4 0.462 70
C 6 H 5 Br 2 0.483 19
3 0.494 0
4 0.463 62
C 6 H 5 OH 2 0.421 20
3 0.486 0
4 0.363 60
C 6 H 5 NH 2 2 0.391
3 0.484 Ortho, para directing
4 0.307
C 6 H 5 CH 3 2 0.392 28.5
3 0.485 1.5
4 0.339 40
C 6 H 5 CHO 2 0.269 9.5
3 0.139 36
4 0.276 9
C 6 H 5 CO 2 H 2 0.322 9.3
3 0.222 40.1
4 0.325 1.3
(a)
In units of -/?.
(b)
For nitration.
(c)
For references, see Reference 48.
What Zhou and Parr found was that the smaller the activation hardness is,
the faster is the reaction.48 Thus Arp- is a reactivity index. Table 3.18 shows the
results for the amounts of ortho-, para- and meta-substitution in the nitration of
substituted benzenes. Similar good results were found for the site selectivity in a
large number of condensed-ring hydrocarbons and heterocyclic molecules.
It can be seen in Table 3.18 that there is no correlation between reactivity and
the activation hardness, if the molecule is changed. Thus benzoic acid has a smaller
Arfi than benzene, but is much less reactive. Again, changes in the cr-bonding have
become important, similarly to the case of the cation localization energy. Actually
both L + and Arp do correlate with reactivity, if only the condensed-ring
hydrocarbons are compared with each other.48
It is difficult to give a proof as to what the exact relationship between the
activation hardness and the energy barrier should be, or even that a relationship
Application of DFT
LUMO
LUMO
HOMO
HOMO
Figure 3.7 Typical changes in hardness (HOMO-LUMO gap) for a chemical reaction. The
harder the TS, the more stable it is, and the easier it is to reach. After Reference 48.
should exist. Figure 3.7, however, shows that there are good reasons to believe that
a small value of Arp should correspond to a low activation energy. The transition
state is shown to be softer than the reactant in Figure 3.7. This follows from
the general properties of the activated complex, which is on a maximum in the
potential energy-reaction coordinate plot. Therefore it decomposes immediately
upon formation.
From the perturbation theory of chemical reaction, this means that there must
be low-lying excited states for the activated complex, or that the HOMO-LUMO
gap is small.49 From the same theory we can assume that the softer the transition
state, the more unstable it is, and the more difficult it will be to form. A hard TS
will be more stable, and hence easier to form (low energy barrier). We can see in
Figure 3.7 that increasing the hardness of the TS will mean lowering the HOMO
of the activated complex, and reducing the energy difference between it and
the reactant.
The natural DFT quantity to use in probing site selectivity within a molecule is
the Fukui function, f(r). The definition is
/* = 4 (3.29)
The quantity/^r is called the condensed Fukui function.50 It has a single value
for each atom, k, in the molecule, and is not otherwise a function of position. The
<7£S are net charges on the atoms. In the last equation c\ is simply the square of
the atom coefficient in the HOMO. It is also the frontier orbital density in FMO
theory. It is the easiest to calculate, since we only need the wave function for the
HOMO, which can often be found, at least roughly, from HMO theory.
Equation (3.28) requires all the filled orbitals, followed by a Mulliken (or
other) population analysis, for both the reactant molecule and its cation. It has
the advanage that the charge, qk(N), can also be useful. It will be recalled that
only soft-soft interactions between two reactants are controlled by the Fukui
function. That is, electron transfer, or covalency, is dominant. For hard-hard
interactions, the charges on each atom dictate where reaction will occur.51
The function/~(r) requires a complete calculation, but it can be approximated
from the HOMO, which is simpler. It gives more detailed information about
the stereochemistry of the reaction path. For example, in a theoretical study
of the reactions of HCHO with both electrophiles and nucleophiles, it was found
that a base would approach the carbon atom from a direction perpendicular
to the plane of the formaldehyde molecule.52 An acid such as H + would attach
itself to the oxygen atom, as a result of the net charge being more negative on
that atom.
An ab-initio calculation of the MOs of the NCS" ion gives the wave function
for the HOMO as53
where ip is a suitable valence shell atomic orbital. Squaring 0, we see that PHOMO is
much larger on S, than on N. Soft electrophiles will react at sulfur. From the total
electron density, however, N is more negative than S, — 0.68e~ vs. — 0.25e~. Hard
electrophiles, whose reactions are controlled by electrostatics, will react at N.
Sometimes even a simple diagram of the frontier orbitals will give us a great
deal of chemical information. For example, consider the malonaldehyde anion, a
model for the important /3-diketone anions
86 Application of DFT
HOMO LUMO
Figure 3.8 The HOMO and the LUMO of malonaldehyde. The shading shows the phases, or
signs, and the diameter of the spheres shows the magnitude of the coefficient in the MO.
Reprinted with permission from R.C. Haddon, J. Am. Chem. Soc, 102, 1807 (1980). © 1980
American Chemical Society.
The HOMO and the LUMO are both ?r-type orbitals for which an HMO
calculation has been made.54 Figure 3.8 shows a diagram commonly used to
show such orbitals. The diameters of the spheres represent the coefficients of
each atomic orbital in the MO. We readily see that carbon atom 3 is the reactive
site for reaction with soft electrophiles. We can assume that the oxygen atoms
will be more negative than any of the carbon atoms, and will react with hard
electrophiles.
The LUMO diagram, which gives/J, shows that carbon atoms 2 and 4 will be
the reactive sites. Even hard bases, such as OH", will react at these positions,
since these are carbonyl carbons and will be more positive than the central
carbon atom. Such qualitative conclusions are supported by detailed calculations
on maleimide, a cyclic molecule related to the malonaldehyde ion.55 In this case
/ + was calculated from
The function/ + (r), shown as density contour lines, PLUMOOO a n d / £ all gave the
same predictions about reaction sites.
By simply multiplying/ + , o r / ~ , by the global softness, a, one obtains the
local softness, a. Since softness is an index of molecular reactivity, one might
hope to rank-order a series of similar molecules in terms of reactivity to a given
reagent. This has been attempted in the case of electrophilic substitution of
substituted benzenes, C6H5X.56 It was found that/~(r),/^ and PHOMOW all gave
correct answers for the relative ortho,para-meta reactivity, except for X —CN
and NO2 This is not too surprising since c|, i.e., frontier orbital theory, also fails
in these cases. It is necessary to include the next-highest MO in order to get the
correct answer.57
More importantly, it was found that a did not give the correct order of
intermolecular activity. Table 3.16 shows why this is the case. The relative values
Fukui Functions and Atomic Charges 87
of 77 do not correlate with reactivity, but the values of x do. In frontier orbital
theory, it is the ionization potential, or the energy of the HOMO, which is most
closely related to overall reactivity in reactions such as aromatic electrophilic
substitution.57 In other cases it is the variations in electron affinity which explain
the experimental results. Generally speaking, it is usually better to use both /and
A, as required by Equation (3.1).
It is worth noting that another function, the molecular electrostatic potential58
(MEP) was tested as an index of intermolecular and intramolecular activity and
found to give the correct order in all cases.56 It is the sum of the potentials due to
the nuclei and all the electrons at each point in the molecule. It has been widely
used as an index of reactivity, but requires considerable effort to calculate.
A condensed form of the MEP could be used, if we knew the net charge on each
atom with certainty. But finding these charges, qk is a difficult, if not impossible,
task since is is not well defined. The commonly used Mulliken population analysis
has obvious deficiencies. Several dozen alternative methods, both theoretical and
experimental, have been proposed.59 Density functional theory also offers a
method, which has some advantages.
One approach would be the calculation of AN, which for a diatomic molecule
gives qk = ±A7V.60 But this can only be considered a zero-order approximation; it
is necessary to take into account the further changes that occur as the interacting
systems approach each other more closely. This has been done and presented
as the electronegativity equalization method, or EEM.61 The key equation is
given by
xl and rjl are the EN and hardness of the free atom; Axk and A ^ are the
changes due to changing size and shape of the atom in the molecule; qi/Rki is
the potential of the shielded atom / on the fcth atom. We also have
An approximate wave function for a given system will not give the correct electron
density, nor will it give the correct electronic chemical potential, constant
everywhere. Instead it will give a potential, /i, which has positive and negative
deviations from the average value, /2, at various points. An improved density and
energy can be had by transferring density from regions where // is too positive to
regions where fi is too negative.
The key equations are familiar
4(r?c
so that the polarizability, a, is calculated as 3.31 a.u., compared with the exact
value of 4.50 a.u. We can lower the energy further by transferring charge from
larger values of z to smaller values in region D, and the reverse for region C. This
improves a to 3.97 a.u.
However, there is a problem with this kind of calculation. It does not seem
reasonable to use the approximation rj = (I - A)/2 for problems of this kind. We
are not adding or removing charge in the sense of adding or removing electrons,
but merely rearranging a constant amount of charge. There is no change in
electron-electron repulsion, obviously, so only the change in kinetic energy with
p should be considered. But this functional dependence is not easily found.63
Sohation Effects 89
The same approach as the above has been applied to atoms and some very
simple molecules.64 The difference in fi is approximated by (e — e), where e is the
local orbital energy. It was found that good results were obtained if it was
assumed that rj = ((t+ ve)), where t is the local one-electron kinetic energy and ve
is the electron repulsion potential. Such an assumption can only be partially
justified.
The cases of slightly perturbed particles in boxes and harmonic oscillators has
also been treated.65 Good results were obtained if the hardness was simply
equated to the kinetic energy. It is of some interest that DFT can be applied to
particles other than electrons. The similarity of Equation (3.35) to the energy
correction of second-order perturbation theory is also worth noting
Hr is the perturbation and E is the average excitation energy between the ground
state, ipo, and the various excited states, ^ . Only the lowest one or two states are
important, in many cases.
For heavy atoms, and for complex systems such as solids, it is common practice
to replace the inner shells by an effective potential, the pseudopotential (PP),
which has the same influence on the outer parts of the atom as the real potential.
The PP is used in a region with a surface boundary beyond which the real potenial
is used. Since we know something about the behavior of wave functions and
orbital energies at large distances from the atom, the results for the outer region
are known to some extent.
Accordingly, it is important to match the orbitals and energies generated by the
PP in the pseudo-atom region to the corresponding properties in the true atom
region. It has been found that considerable improvement occurs if the hardness is
one of the conserved properties.66 That is, the total energy of the atom and
pseudo-atom must be the same up to the second order for small changes in the
frontier orbital occupancy. The improvement lies mainly in the transferability of
the PP. The same PP can be used for a given atom in a variety of environments.
SOLVATION EFFECTS
So far the concepts derived from DFT have been applicable to isolated chemical
systems, i.e., to gas-phase molecules, atoms and ions. Actually some of the
applications discussed were based on results in solution, with only minor
comments on the effects. Since solvation energies will always be important for
chemical reactions, it is time to examine solvation in more detail. There are two
possible procedures: one is to use the gas-phase theory first, and then to modify
90 Application of DFT
It is convenient to define I' and A' as the potentials of the standard Gibbs free
energy changes in Equations (3.39) and (3.40). The free energy of the electron at
rest in the gas is set equal to zero.
The direct measurement of V and A' is not possible, but they can be calculated
in several independent ways. Electrochemical methods depend on knowing the
absolute potential of the hydrogen electrode.
There is now general agreement on a value near 4.5 V at 25 °C.68 This value
enables us to find the free energy of hydration of the proton (-260 kcal/mol), and
from this the free energies of hydration of many other ions.67'69 Combined with
gas-phase data on / and A, many values of V and A! can be calculated.67
Since a thermochemical cycle is used, these would be adiabatic values. It is .of
interest that vertical values can be found experimentally by the technique of photo-
electron emission spectroscopy.70 These are 1 or 2 V higher because they do not
include the reorganization energy of the solvent around the product.
By combining elecrochemical methods with other data, it has been possible to
obtain values of/' and A' for large numbers of both inorganic71'72 and organic73
molecules. The determination of half-wave potentials, £"1/2, by cyclic voltam-
metry has been particularly useful for organic molecules.74 Even when "£1/2" *s
irreversible, the variation for a series of related molecules is about the same as for
the reversible values.75 These results for organic molecules are usually obtained
in solvents such as acetonitrile or dimethyl sulfoxide. Fortunately there is a large
body of information on the free energies, or enthalpies, of transfer of ions from
one polar solvent to another.76
Table 3.19 gives a small sample of the results which may be obtained. For
various practical reasons, it is usually difficult to measure both / ' and A1 for the
same species, so that // and rf are not known for most systems. However, this
is not as serious a problem as might be supposed. It may be seen that values of
/ ' are 2-4eV less than the gas-phase values /for the same molecule. Also A' values
are 2-4 eV greater than the gas phase A values. The larger the neutral species,
Solvation Effects 91
Table 3.19 / ' and A' for Some Sample Systems, Water Solution, 25 °C (a)
C 6 H 5 NH 2 4.53 CH 3 3.73
(CH 3 ) 3 N 5.21 PH 2 4.18
(CH 3 ) 2 S 6.11 C6H5 4.27
(CH 3 ) 3 P 6.34 o2 4.31
C6H6 6.56 C6H5S 5.36
(C 2 H 5 ) 2 O 6.72 C6H5O 5.42
(CH 3 O) 3 P 6.75 NH 2 4.72
CH 3 OH 7.25 I 5.64
H2O 7.94 OH 6.22
CH 3 NO 2 8.65 CN 7.02
CH 3 CN 9.35 F 7.88
(a)
See Reference 67 for data. Also see A. Bagno and G. Scorrano, J. Am. Chem.
Soc, 110, 4577 (1988) for new data on pA:as.
the smaller is the difference. For example, for OH and C6H5S radicals, using
Tables 3.11 and 3.19, we find {A! — A) equal to 4.39 and 2.79 eV, respectively.
Obviously the differences between / and / ' and A and A! are almost entirely the
result of strong solvation of the ions, M + or M". It is easier to form ions in any
solvent, than in the gas. Figure 3.9 now shows the effect of this on /jf and rj' in an
energy diagram. On going from the gas to solution, A becomes greater and /
becomes less so that the difference between them, the energy gap or 2rf9 becomes
less. For large moleculs the change from / t o V will be equal and opposite to the
change from A to A'. Therefore the numerical value of JJ! (or xO will be the same
as that of \i (or %)• This will not be the case, however, for small molecules,
particularly the hydrogen atom.77
Therefore, for a possible transfer of an electron from C to D when both are large,
we can use either xc or Xo compared with XD or xi> t° decide which molecule is
the donor and which is the acceptor. I' and A1 also give the thermodynamic one-
electron oxidation potential and reduction potential of the molecule.
-A
- o o - -»'
-o-o- -•
(a) (a)
Figure 3.9 (a) Orbital energies of HOMO and LUMO in the gas phase; (b) apparent orbital
energies in solution, see text.
that of the LUMO. The orbital energies, in fact, are relatively unchanged. The
solvent lowers the energy of all orbitals a small amount for M, and a much larger
amount for M + or M~.
The evidence for this can be readily seen by measuring the shifts in the vis-UV
spectra of molecules on going from the gas phase to solution. Such shifts are
usually quite small. Even for charge transfer bands, where the polarity of the
molecule changes, the shifts are less than 1 eV.79 Therefore rf correctly predicts
that a molecule is soft in reactions such as
Table 3.20 Experimental A//het, Ai/homo and AGET for Organic Free Radicals
9-Phenyl-xanthyl 42 16 27
Benzyl 112 62 50
t-Butyl 120 72 48
dissolve in soft solvents, and hard solutes in hard solvents. But there has been
little attempt to quantify this in terms of DFT.
As mentioned, there are cases where // and rf are the correct measures for
reactions in solution. Let R be an organic free radical containing trivalent carbon.
Then we have81
where AGET is the sum of the free energies for the one-electron oxidation of R,
and the one-electron reduction of R. The quantity TASET is known to be small,
and AGET is the difference (/' - A1). Some experimental results for different R
are given in Table 3.20.
The 9-phenylxanthyl radical is a resonance-stabilized triphenylmethyl analog.
The corresponding carbonium ion and carbanion are also stabilized and can be
prepared in sulfolane, so that A//het can be directly measured.82 The data for
benzyl and t-butyl are obtained by measuring the reduction and oxidation
potentials of the radicals in acetonitrile.83 The results show that C6H5CH2 and
(CH3)3C are much harder than the 9-phenylxanthyl radical (the latter is just one
of several studied with similar properties82). The solution hardnesses are then
responsible for the difficulty in forming the ions in the benzyl and t-butyl cases,
and the stability of the ions in the resonance-stabilized cases. The effect of the
small hardness in the latter cases also is evident in the small bond energy for
homolytic dissociation.
Equations (3.44) and (3.45) are applicable to the symmetrical case shown in
reaction (3.43), but they also apply to the general case
where X is any radical. For a pair of radicals, such as R and X, the effective
hardness is given by81
40.0
30.0
20.0
10.0
0.0
0.0 5.0 10.0 15.0 20.0 25.0
AGET [kcal/mol]
Figure 3.10 Correlation of Ai/het of carbon-carbon bonds with A GET for the trityl cation
reacting with substituted fluorenide anions. Energies are in kcal/mol. After Reference 81.
In the example given, /R < Ix and Ax > ^4R, SO that the heterolytic bond
breaking leads to R + and X~, rather than R~ and X + .
Equation (3.45) has been tested for the breaking of different carbon-carbon,
carbon-oxygen, carbon-nitrogen and carbon-sulfur bonds.81 For a given R, X
was varied. Both R + and X~ were cases of stable ions, by resonance or otherwise.
The test consisted of plotting experimental values of A/fhet against AGET in
sulfolane. Figure 3.10 shows the results for carbon-carbon bond breaking. The
cation is the trityl ion, (C 6 H 5 ) 3 C + , and the anions are substituted fluorenides.
The good linear relationship shown means that AH^omo either varies as A//het
does, or is constant. The latter turns out to be the case. The main point is that the
effective hardness is a good measure of the energy needed to form the ions. This is
most useful, since the customary practice in organic chemistry is to use the
stabilities of R + , R~ and R as criteria for ranking the reactivity of their tetra-
covalent precursors.
For example, good linear relationships are found between Ai/het and
(p^Ta - pKR+) for related families of R and X.84 The pKa and p^R+ refer to
the equilibria
HX = H X (3.48)
REFERENCES
1. For enlightening diagrams of MOs, see W.L. Jorgensen and L. Salem, The Organic
Chemist's Book of Orbitals, Academic Press, New York, 1973.
2. R.G. Pearson, Ace. Chem. Res., 4, 152 (1971).
3. For example, see R.G. Pearson and H.B. Gray, Inorg. Chem., 2, 358 (1963).
4. R.G. Pearson, Inorg. Chem., 23, 4675 (1984).
5. T. Ziegler, V. Tschinke and C. Ursenbach, / . Am. Chem. Soc, 109, 4825 (1987); J. Li,
G. Schreckenbach and T. Ziegler, / . Phys. Chem., 98, 4838 (1994).
6. R.G. Pearson in Bonding Energetics in Organometallic Compounds, T. Marks, Ed., ACS
Symposium Series 428, Washington, DC, American Chemical Society, 1990, pp. 251-262.
7. C.A. Tolman, / . Am. Chem. Soc, 96, 2780 (1974).
8. K. Kitaura, S. Sakaki and K. Morokuma, Inorg. Chem., 20, 2292 (1981); T. Ziegler, ibid.,
24, 1547(1985).
9. R.G. Pearson, Inorg. Chim. Acta, 198-200, 781 (1992).
10. T. Ziegler, Inorg. Chem., 25, 2721 (1986).
11. D. Marynick, / . Am. Chem. Soc, 106, 4064 (1984).
12. M. Guerra, D. Jones, G. Distefano, A. Foffani and A. Modelli, / . Am. Chem. Soc, 110,
375 (1988).
13. W. Kutzelnigg, Angew. Chem., Int. Ed. EngL, 23, 272 (1984).
14. R.G. Pearson, / . Chem. Ed., 64, 561 (1987).
15. J.P. Collman and L.S. Hegedus, Principles and Applications of Organotransition Metal
Chemistry, University Science Books, Mill Valley, CA, 1980, Chapter 4.
16. R.G. Pearson, Inorg. Chem., 27, 734 (1988).
17. W. Walper and H. Keim, Z. Phys. Chem. (Munich), 113, 207 (1978).
18. S. Sakaki and M. Ieki, / . Am. Chem. Soc, 113, 5068 (1991).
19. C. DiBugno, M. Pasquale, P. Leoni, P. Sabatino and D. Braga, Inorg. Chem., 28, 1390
(1989).
20. For theoretical calculations see M.R.A. Blomberg and P.E.M. Siegbahn, / . Chem. Phys.,
78, 986, 5682 (1983); J.J. Low and W.A. Goddard, Organometallics, 5, 609 (1986);
T. Ziegler, V. Tschinke and A. Becke, J. Am. Chem. Soc, 109, 1351 (1987).
21. R.G. Pearson and P.E. Figdore, / . Am. Chem. Soc, 102, 1541 (1980); R.G. Pearson in
Nucleophilicity, J.M. Harris and S.P. McManus, Eds., ACS Symposium Series 215,
American Chemical Society Washington, DC, 1987, pp. 233-246.
22. M.R.A. Blomberg, J. Schule and P.E.M. Siegbahn, / . Am. Chem. Soc, 111, 6156 (1989).
23. C.W. Bauschlicher, Jr., Chem. Phys. Lett., 142, 71 (1988).
24. R.D. Gillard, J.A. Osborn and G. Wilkinson, / . Chem. Soc, 4107 (1965).
25. J.P. Collman and M.R. Laury, / . Am. Chem. Soc, 96, 3019 (1974).
26. M. Kubota, Inorg. Chim. Acta, 1, 195 (1973).
27. R.G. Pearson, J. Org. Chem., 54, 1423 (1989).
28. A. Tachibana and K. Namamura, / . Am. Chem. Soc, 117, 3605 (1995).
29. J. Sauer, H. Wiest and A. Mielert, Chem. Ber., 97, 3183 (1964).
30. J.M. Tedder, Angew. Chem., Int. Ed. Ebgl, 21, 401 (1982); B. Giese, ibid., 22, 753 (1983).
96 Application of DFT
31. K.H. Becker, R. Kurtenbach and P. Wiesen, /. Phys. Chem., 99, 5986 (1995).
32. S.G. Lias, J.E. Bartmess, J.F. Liebman, J.L. Holmes, R.D. Levin and W.G. Mallard,
/. Phys. Chem. Ref Data, 17, Suppl. No. 1 (1988).
33. R.G. Pearson, /. Mol Struct. (Theochem.), 255, 261 (1992).
34. K. Heberger and A. Lopata, /. Chem. Soc, Perkin Trans., 2, 91 (1995).
35. R.W. Quandt and J.F. Hershberger, Chem. Phys. Lett., 206, 355 (1993).
36. R.J. Balla, H.H. Nelson, and J.R. McDonald, Chem. Phys., 109, 101 (1986).
37. J.J. Russell, J.A. Seetula, D. Gutman, F. Danis, F. Caralp, P.D. Lightfoot, R. Lesclaux,
C.F. Melius and S.M. Senkan, J. Phys. Chem., 94, 3277 (1990).
38. A. Masaki, S. Tsunashima and N. Washida, J. Phys. Chem., 99, 13126 (1995).
39. R. Atkinson, Chem. Rev., 86, 69 (1986); J.P.D. Abbott and J.G. Anderson, /. Phys.
Chem., 95, 2382 (1991).
40. G. Klopman, /. Am. Chem. Soc, 90, 223 (1968).
41. K. Munger and H. Fischer, Int. J. Chem. Kinet., 17, 809 (1985).
42. L.C. Allen, E.T. Egolf and C. Liang, /. Phys. Chem., 94, 5602 (1990).
43. A. Streitwieser, Jr., Molecular Orbital Theory for Organic Chemists, John Wiley, New
York, 1961, Chapter 11.
44. G.W. Wheland, J. Am. Chem. Soc, 64, 900 (1942).
45. N.S. Isaacs, Physical Organic Chemistry, Longman, Harlow, UK, 1987, Chapter 10.
46. S. Clementi, F. Genel and G. Marino, Chem. Commun., 498 (1967).
47. K. Fukui, T. Yonezawa and H. Shingu, /. Chem. Phys., 20, 722 (1952).
48. Z. Zhou and R.G. Parr, J. Am. Chem. Soc, 111, 5720 (1990).
49. R.F.W. Bader, Can. J. Chem., 40, 1164 (1962).
50. W. Yang and W.J. Mortier, J. Am. Chem. Soc, 108, 5708 (1986).
51. For a review of orbital control and charge control, see G. Klopman, Chemical Reactivity
and Reaction Paths, John Wiley, New York, 1974, Chapter 4.
52. C. Lee, W. Yang and R.G. Parr, /. Mol. Struct. (Theochem), 163, 305 (1988).
53. K.A. Jorgensen and S. Lawesson, /. Am. Chem. Soc, 106, 4687 (1984).
54. R.C. Haddon, /. Am. Chem. Soc, 102, 1807 (1980).
55. F. Mendez, M. Galvan, A. Garritz, A. Vela and J. Gasquez, /. Mol. Struct. (Theochem.),
277, 81 (1992).
56. W. Langenecker, K. Demel and P. Geerlings, /. Mol. Struct. (Theochem.), 234, 329
(1991); idem, ibid., 259, 317 (1992).
57. J. Fleming, Frontier Orbitals and Organic Chemical Reactivity, John Wiley, New York,
1976, Chapter 3.
58. R. Bonaccorsi, E. Scrocco and J. Tomasi, J. Chem. Phys., 52, 5270 (1970).
59. For a review, see J. Meister and W.H.E. Schwarz, /. Phys. Chem., 98, 8245 (1994). See also
J. Cioslowski and S.T. Mixon, J. Am. Chem. Soc, 115, 1084 (1993).
60. D. Bergman and J. Hinze, Structure and Bonding, 66, 145 (1987).
61. W.J. Mortier, S.K. Ghosh and SJ. Shankar, J. Am. Chem. Soc, 108, 4315 (1986);
K.A. Van Genechten, W.J. Mortier and P. Geerlings, /. Chem. Phys., 86, 5063 (1987).
62. B.G. Baekelandt, W.J. Mortier, J.L. Lievens and R.A. Schoonheydt, /. Am. Chem. Soc,
113, 6730 (1991).
63. R.G. Parr and W. Yang, Density Functional Theory of Atoms and Molecules, Oxford
University Press, New York, 1989, Chapter 6.
64. R.G. Pearson and W.E. Palke, Int. J. Quantum Chem., 38, 103 (1990).=
65. R.G. Pearson, Theor. Chim. Acta, 78, 281 (1991).
66. M. Teter, Phys. Rev. B, 48, 5031 (1993).
67. R.G. Pearson, /. Am. Chem. Soc, 108, 6109 (1986).
References 97
INTRODUCTION
From what has been said already, it is obvious that the hardness of a chemical
system plays a major role in determining its stability or reactivity. Alternatively,
we can use the HOMO-LUMO energy gap as a criterion. If stability is desired,
then it is advantageous to have a large energy gap, or a high value of the hardness.
If reactivity is desired, then a small gap or hardness is desirable.
Examples have already been given in Chapter 2 showing that soft molecules are
more reactive than similar, but harder, molecules. There is much more evidence
consistent with the statement that increasing hardness goes with increasing
stability. For example, theoretical calculations almost always indicate that the
most stable structure has the largest HOMO-LUMO gap. A simple case is shown
in Figure 4.1, where the MO energy diagram is given for CH4 in both its stable
tetrahedral structure and an unstable planar form.
The smaller gap in the latter case arises because the p z orbital of carbon is
removed from bonding while still occupied. It rises in energy and becomes the
HOMO. At the same time a linear combination of hydrogen Is orbitals is
removed from an anti-bonding MO of tetrahedral methane and becomes non-
bonding. It goes down in energy and becomes the LUMO. In fact, since we know
that the tetrahedral structure of CH 4 has the maximum overlap of the valence
orbitals of C and H, we can assume that any change in structure will result in
energy effects similar to Figure 4.1. The hardness will get less and the stability
will decrease.
The same results are found in all of the cases where structures are predicted
by simple rules, such as the valence-shell electron-pair repulsion theory.2 There
are also many examples involving more complicated molecules. Table 4.1 shows
the results of some calculations by the extended Hlickel method for the closo-
borane anions.3 For B6H^~ and BvH2" the HOMO-LUMO gap is given for
both the stable structure and unstable ones. The stable structures, octahedral
for BgH^" and pentagonal bipyramidal for B7H2~, have by far the largest gaps.
For B8Hg~ the gap has been calculated for three structures, for any of which
no strong preference is seen. In agreement with this, the ion has been found to
be fluxional.
t
t -o*o-
1
Figure 4.1 MO energy diagram for methane in its stable tetrahedral form and in the unstable
planar form.
(a)
From Reference 3.
Introduction 101
s
LUMO LUMO
\ LUMO
\
LUMO
-(Kh v
HOMO x
N
-o-o-
HOMO
-O-O-N -O-O-N
HOMO
HOMO
(a) (b)
Figure 4.2 Energy changes during the approach of two molecules: (a) HOMO-LUMO
interaction; (b) HOMO-second filled orbital interaction.
net decrease in energy. The HOMO will be an atomic orbital on the anion, and
the LUMO will be an orbital of the cation, in the usual case. As the ions
approach, the potential of the cation will lower the orbital energy of the HOMO,
and the potential of the anion will raise the orbital energy of the LUMO. The
HOMO and LUMO will move apart, just as in Figure 4.2(a).
These conclusions are drawn for the case where a bond is being formed
between two subsystems, but they also apply more generally, as is easily seen in
the case of ionic bonding. The HOMO will be a linear combination of anion
orbitals, and the LUMO a linear combination of cation orbitals. The largest
HOMO-LUMO gap will be found if the arrangement of the ions gives the
largest potential at each of the ions, which, in turn, means the minimum energy
for the system. Covalent bonding gives the same result, except that orbital
overlap is the dominant feature.
Thus covalent and ionic bonding give similar results: the energy decreases
when the HOMO-LUMO gap increases. But this conclusion is based on orbital
energies, or electronic energy, only. The nuclear repulsion must also be included
to get the total energy. This is a major reason why the interaction shown in
Figure 4.2 does not continue beyond a certain point.
There is another test that can be applied. Since experimental values of r\ are
available for many systems, we can check to see if r\ does increase on going from
atoms or radicals to stable molecules. Table 4.2 shows some sample results. As
expected, the hardness increases on forming the stable product from the unstable
reactants. These are not isolated cases; an examination of a large number of
reactions shows similar behavior.5
Unfortunately there does not seem to be a simple relation between the
magnitude of the change in r\ and the change in energy for the reaction. Partly this
results from the neglect of the nuclear repulsion energy, as mentioned. But the
102 The Principle of Maximum Hardness
(a)
F r o m Reference 5.
(b)
Value for reactants
(c)
Value for products.
(d)
Vertical values for / and A.
numbers in Table 4.2 show that A77 is greater for ionic bonding than for covalent
bonding. In the latter case it is sometimes necessary to make the correction to the
vertical values of / and A, to obtain an increase in 77 as predicted.
The increase in the HOMO-LUMO gap due to pure ionic bonding is easily
calculated.5 For NaCl, at the equilibrium distance of 2.36 A for the diatomic
molecule, the electrostatic contribution to 77 is 6.1 eV. The observed value of
4.8 eV reflects some of the same factors that decrease the gap for covalent
bonding. It is interesting to note that ionic bonding should not change the
electronic chemical potential, \i. The HOMO is lowered by 6.1 eV, and the LUMO
is raised by 6.1 eV, in the case of NaCl. Therefore the midpoint is unchanged.
For covalent bonding, the LUMO is raised more than the HOMO is lowered,
as shown in Figure 4.2(a). Hence there is a small positive increase in /i. The
results for fj, in Table 4.2 are in general agreement with this, but there is no
correlation between A// and the AE of the reaction.
A question may arise as to the experimental values of fi and 77 for mixtures,
such as the reactants in Table 4.1. Consider the case of NaCl again, where the
reactants are an equimolar mixture of Na and Cl atoms, which are momentarily
prevented from reacting in some way. Removal of an electron from the
metastable mixture would certainly cause the formation of Na + , rather than Cl + .
Introduction 103
15
10
10 20 30 40 50 60 90
Figure 4.3 Atomic hardness [eV] plotted against the atomic number Z. Shells and sub-shells
are determined by the local maxima in hardness. Reproduced with permission from Reference 7.
104 The Principle of Maximum Hardness
that 77 for the atoms is a measure of stability, with the noble gases showing
maximum values of 77 and maximum stability in the sense of chemical inertness.
The special stability offilledsub-shells, or half-filled sub-shells, is also clearly
seen. Compare Be with Li and B, or N with C or O. The sub-shell effects are
obscured by other factors at large atomic numbers, but are still visible if experi-
mental values are used.
Molecules are also examples of the special stability offilledshells. The simple
molecular orbital rule is that a molecule is thermodynamically stable if all the
bonding MOs are filled, and all the anti-bonding ones are empty.9 Other
examples are the Octet Rule of Lewis and Langmuir and the 18-Electron Rule of
Sidgwick. These relate hardness to stability just as Figure 4.1 does. So do the
famous Huckel rule (4n + 2), and the Wade-Mingos rule (2n + 2).10
Parr and Zhou have pointed out that the relationship between hardness and
stability can be extended to atomic nuclei.7 The hardness is defined as half the
difference between the highest filled nuclear energy level and the lowest empty
one. This difference is greatest for closed-shell nuclei compared with the neigh-
boring open-shell nuclei.11 Shells can be identified by the "magic numbers" 2, 8,
20, 28, 50, 82 and 126. Hardness seems to be a general measure of stability. Local
maxima can be used to locate shell and sub-shell structure.7
In quantum mechanical calculations on atoms and molcules, we usually have
only approximate electron density functions available. What is the relationship
of hardness, 77, to pi Will it increase to a maximum value as a set of trial basis
functions becomes better and better, approaching the true electron density?4
Some conclusions can be drawn, if we restrict ourselves to LCAO-MO theory.
The eigenfunctions, <pm, are given by
~~ •mirl'i (4.2)
where the ?/>s are atomic orbitals. The orbital energies are found from the
determinant
|#i/-%| = 0 (4.3)
The best values of the Cm/S define the best wave function, and the best value of
p, that can be obtained from the selected basis set of AOs. Any change from the
best values will cause the HOMO to rise in energy, or be unchanged, and the
LUMO will fall in energy, or be unchanged. Thus the energy gap between them is
a maximum for the best values of the coefficients, or the "best" electron density
function. Usually, of course, this will not be the true density function.
The conclusion that the hardness has a maximum value for the lowest-energy
solution can be readily verified in simple cases, such as the Hiickel MO theory.
Inclusion or omission of overlap integrals has no effect. Examples where all
atoms are the same and cases where the atoms and the orbitals are different give
the same result. In fact, in simple Hiickel theory it can be proved that the most
negative value of the 7r-electron energy is always accompanied by the largest
value of the hardness, since both are proportional to the exchange integral, /3.12
However, the proof requires that [i be kept constant.
Provided the self-consistent field (SCF) condition is met, calculations at the
Hartree-Fock level also obey the mathematics of Equations (4.3) and (4.4).
Therefore the HOMO-LUMO gap should also be a maximum in these cases.
Because the solutions are normalized and orthogonal, and because the atomic
orbitals are conserved, the coefficients for different MOs are not independent.
Therefore wrong coefficients in one orbital will usually lead to wrong coefficients
in all orbitals.
We can also change the trial wave function for a system by enlarging the basis
set of atomic orbitals. If this is done, what usually happens is that the HOMO is
changed only slightly, to a more negative energy, but the LUMO is decreased in
energy much more, so that the HOMO-LUMO gap is smaller, even though the
total energy is decreased. At the same time, the value of /i becomes more
negative, because of the decrease in £LUMO- This turns out to be important, as
will be seen in the next section.
From much circumstantial evidence of the kind given in the preceding section, it
was concluded that "There seems to be a rule of nature that molecules arrange
themselves to be as hard as possible".1 But none of the evidence is rigorous enough
or general enough to be conclusive. Fortunately Parr and Chattaraj have given a
rigorous and widely general proof of the "Principle of Maximum Hardness",
or PMH.13
Their proof is based on a combination of statistical mechanics and the
fluctuation-dissipation theorem. Since it is not easy to follow, some preliminary
discussion is needed. The statistical mechanics part depends on the properties of a
grand canonical ensemble. Such an ensemble consists of a large number of
106 The Principle of Maximum Hardness
systems, each identical with the system of interest. These can exchange energy and
particles with each other. Therefore the various members of the ensemble can
have different numbers of particles and different values of the total energy. These
quantities will fluctuate.
The Hohenberg and Kohn theorem applies to ground states at the absolute
zero of temperature. Fortunately there is a finite-temperature version of DFT,
first proved by Mermin.14 The equilibrium properties of a grand canonical
ensemble are determined by the grand potential, il, which is defined as follows:
n = E-fiN-TS (4.5)
where ft — 1/kT, (N) is the average of N over the emsemble, which is constant,
and N is the value for each member of the ensemble, which can vary.
For an ensemble of molecular systems we would have15
ensemble be F. The latter is the equilibrium distribution function for the grand
potential ft + AfL However, at time t = 0 the perturbation is turned off in some
way. Then F will relax to /, usually in a first-order process characterized by a
relaxation time, r. From the relation between the grand potential and the prob-
ability distribution function, we know that
(4 8)
'
F=(A)~lAf (4.9)
In the equilibrium state, the average value is independent of time. Therefore the
average value of the observable, (A), is a constant, as is/. Both F and A relax to
the final values of/and (A). At / = 0 the value of A can be written as A(0), where
the overbar indicates an average over the now non-equilibrium ensemble.
All averaging must be done with the /i, v and T of the equilibrium ensemble.
Therefore we are comparing equilibrium and non-equilibrium systems of constant
/i, v and T. Looking at equations (4.5) and (4.8), we can see that, if/x is constant,
the energy of the non-equilibrium ensembles will be higher than that of the equi-
librium one. At least this will be the case when changes in entropy are small.
Otherwise, it is the free energy which is higher.
As time goes on, the non-equilibrium ensembles will relax towards equilibrium.
The excess energy will be degraded to heat, by a mechanism of molecular
collision. This means that the fluctuation-dissipation theorem will apply. This
theorem dates back to early work by Nyquist16 and Onsager17 and has been
generalized by Callen and Welton.18 In essence, it says that small deviations from
equilibrium have the same relaxation times, whether they are spontaneous or
induced. This follows because the mechanism is the same.
The theorem applies to any observable, A which relaxes from the non-
equilbrium value A, averaged over the ensemble, to the equilibrium value (A).
The relaxation of a spontaneous fluctuation in an equilibrium system is given by
the time correlation function, C{i)
~ (A))
_ c =
C(0) (A(0) -
This establishes that (A(t) — (A)) is directly proportional to C(t), but it does not
give the sign of the constant of propotionality. Some observables in a non-
equilibrium system are less than the equilibrium value, and some are greater. For
example, consider concentrations of reactants and products in a chemical
equilibrium.
To find the constant of proportionality we follow Parr and Chattaraj,13 who
followed Chandler.15a We will show that
The integration is to be over all of phase space. Next replace F(0;rNpN) by its
equal from (Equation 4.9) and carry out the integration.
This result inserted into Equation (4.12) will give C(t) equal to its value in
equation (4.10).
Accordingly, it follows that
( A ( 0 ) - ( A » = ((j-<<7»>0 (4.16)
Thus the equilibrium ensemble has the minimum softness, or the maximum
hardness. This result is valid for all nearly non-equilibrium distributions which
obey Equations (4.8) and (4.7). Presumably it includes all cases of interest, since
these are common assumptions in linear response theory.
The restriction to constant /i and v is severe, and greatly limits the usefulness
of the proof. However, it is likely that these restrictions can be relaxed somewhat.
The Maximum Hardness Principle 109
The most important feature seems to be that the non-equilibrium system has a
higher energy than the equilibrium one, so that the fluctuation-dissipation
theorem can be applied. The result of interest in Equations (4.14) and (4.15) is
that the constant of proportionality is a quantity that is always positive. Its exact
value of (A) is not actually used.
An interesting feature of the Parr-Chattaraj proof of the Principle of Maximum
Hardness, is that the specific example of chemical softness is not introduced until
the last step. The proof should then be valid for many other observables, provided
that certain restrictions are met. One requirement is that the observable always
has a positive value (or in some cases always a negative one).
Equation (4.7) suggests that the fluctuations in an equilibrium system will offer
many examples of the minimum (maximum) principle. These fluctuations are
usualy presented as the variance, or second central moment, since the average, or
first moment, is zero.19 The variance is always positive, as required, and its
magnitude is usually taken as an inverse measure of goodness. That is, we want
the variance of a measured or calculated variable to be as small as possible. It
seems entirely reasonable that the equilibrium system would have the smallest
variance.
For example, the variance of the local energy, ^loc = Hijj/^, has long been
used to gauge the goodness of an approximate wave function, ^. 20 The variance,
(£foc — ^i20CX wiH approach zero as the true wave function is approached, since
then the local energy will be a constant equal to the exact energy. Note that the
exact energy will also be a minimum compared with the other E\oc.
The energy of an electromagnetic field, confined to a hohlraum, is given by
E = ~(Fi + W) (4.17)
where .Fis the electric field strength, and B is the magnetic field. Since F = B = 0,
the minimum energy is wherejhe variance of F and B is minimum. In Brownian
motion, the mean values of p2 and p, the momemtum, are given by
where M is the molecular weight. The smallest value of the variance again goes
with the minimum value of the energy.
New information can be obtained by applying the principle of minimum
variance. The fluctuations of the energy are given by12
CvT=f3((E-(E))2) (4.19)
As mentioned earlier, the restriction to constant /J, and v, if absolute, would be very
severe. As chemists we normally wish to examine changes in the nuclear positions,
leading to chemical change. Fortunately, there are cases where this can be done,
keeping ji and v fixed.21 Start with a molecule in its equilibrium geometry and
calculate its obital energies at, or near, the HF level. Now distort the molecule a
small distance along directions given by the vibrational symmetry coordinates and
re-calculate the orbital energies. By using the full set of symmetry coordinates,
the hardness can be probed for all possible changes in the equilibrium geometry.
Table 4.3 shows the results of such calculations for the CO2 molecule.22 In
order to understand the results, Figure 4.4 shows the normal modes of vibration
of CO2. The asymmetric modes, IIU and S u , differ from the symmetric mode, S g ,
in several respects. First of all, they destroy some elements of symmetry,
changing the point group to C2V and Coov, respectively.
Secondly, positive deviations from equilibrium produce a configuration which
gives the same average nuclear potential as negative deviations. Therefore,
an
£HOMO? £LUMO> M d V must be the same for both. If we let Q represent a
symmetry coordinate, then both (dfx/dQ) and (dr]/dQ) must be equal to 0 at the
equilibrium go-
(a)
F r o m Reference; 2 2 .
(b)
Equilibrium values.
Tests of the PMH 111
OR, C R2 O
nu
• •
Also, if we expand the energy as a power series in AQ, the linear term must
vanish, and the quadratic term is the first non-vanishing one. Symmetry
arguments can be used to show that (dvn/dQ) and (dve/dQ) are separately equal
to 0 when averaged.17 Here vn and ve refer to the potentials of the nuclear
repulsion and the nuclear-electron attraction, respectively. Hence, for the non-
totally symmetric distortions, we have met the restrictions of Parr and Chattaraj.
Accordingly, the hardness should be a maximum at <2o- Table 4.3 shows that this
is the case. Note that /i can either decrease or increase upon distortion from
equilibrium.
The totally symmetric mode gives different results. The hardness and fi both
increase steadily as the nuclei approach each other. If the nuclei coalesced, fi
would be a maximum. This does not happen because at 2o we have the condition
correct point group of a molecule, but not the equilibrium bond angles or bond
distances when these are changed by symmetric distortions.
Datta was the first to point out an interesting corollary of the PMH. The
transition state of a chemical reaction, which has a maximum energy, by
definition, should have a minimum hardness as well.25 Figure 4.5 shows the
variation of both the total energy and the hardness for the inversion reaction of
NH 3 . The reaction coordinate is the angle between an N—H bond and the three-
fold axis, so that 90° is the planar form. It is also the transition state (TS) for the
inversion reaction of ammonia. The reaction coordinate has AJ( symmetry at the
TS, but is A! thereafter. The double potential well is at a = 65° and 115°. The
hardness is clearly a minimum at a = 90°, as expected, just as the energy is a
local maximum. The energy in this case is the total energy. But symmetry
arguments show that, for a non-symmetric reaction mode, all quantities such as
^nn? Ee\ and E must be extrema at the TS.26 Since F nn , the nuclear repulsion
energy, is a minimum at this point, EQ\ must be a maximum.
That the hardness would be a minimum at the TS is expected on other
grounds. Long ago Bader pointed out that an activated complex which exists at
the TS must have low-lying excited states, or a very small HOMO-LUMO gap.27
This follows from the second-order Jahn-Teller effect since activated complexes,
8.50 -247.00
I CD
CD
7.50 -249.00
6.50 -251.00
30.00 70.00 110.00 150.00
Reaction coordinate, a [deg]
Figure 4.5 Inversion of ammonia. Variation of the total energy (broken line) and the
hardness (solid line) as the angle, a, is changed. Hardness is in eV, and total energy in au.
From Reference 25.
Tests of the PMH 113
This says that, if \i is constant, as E = EQ\ decreases, the hardness will increase,
and vice versa. A maximum in E means a minimum in Jf, and a minimum in E
means a maximum in Jf: this is the PMH. Any state that is an extremum for
both EQ\ and Vnn will be an extremum for the hardness.
The correctness of these conclusions is borne out by a number of studies made
on the variations of \i and E in rotational isomerization reactions.29 These have
the interesting feature of having double-well potential energies with an inter-
vening barrier, or a single well with two barriers.
As an example of the latter case, consider H2C>2. This molecule is unstable in
both the cis and trans forms, and is stable at a = 70° setting a = 0° for the trans
form. Figure 4.6 shows the calculated variations of both the potential energy and
the hardness as a function of a.30 The mirror-image relationship between the two
is what is predicted by Equation (4.21). At both 0° and 180°, r\ is a minimum and
V is a maximum. This is required since the reaction coordinate is asymmetric at
these points. The point groups are C2h for the trans isomer and C^ for the cis.
The equilibrium point group is C2, at a = 60°, somewhat off the experimental
value of 70°.
In all these applications of the grand potential, the energy which appears in
Equation (4.5) is the electronic energy. The nuclear-nuclear repulsion energy
does not appear explicitly, since it is supposedly constant. Also, the proof for the
PMH is valid for small departures from equilibrium only. Therefore it cannot be
assumed that the relative stabilities of isomers can be linked to the relative sizes
of the HOMO-LUMO gap.
For example, some results for several possible structures of Si4, calculated by
DFT methods, are given in Table 4.4.31 The most stable form is a rhombus, or
diamond shape. This also has the greatest hardness. All three structures are
readily interconvertible by small distortions from the stable structure.
Compare this case with that of possible structures for the (unknown) mole-
cule P 6 . Warren and Gimarc have made accurate SCF MO calculations on five of
the possible valence isomers.32 These were selected by looking at the five most
stable valence isomers of C6H6, since P 6 is valence iso-electronic with C6H6.
Replace each CH unit by a P atom in the structures 1-5 in Table 4.5. The energy
values shown are for the P 6 isomers. There is no relationship between the relative
energies and either the gap size or the chemical potential. The most stable isomer
has the benzvalene structure, and almost the smallest gap.
114 The Principle of Maximum Hardness
10
V HOOH
\
y- "v _
4
8 _- \
t \ /
/ \ / \
/
2
\ /
/ \ /
6 -
/
0
\\ JI
\ I
4 - -2
A - -4
J| |
2 - /
-6
0 -- -^ >_
-8
i i i
-2 -10
-30 30 60 90 120 150 180 210
a
Figure 4.6 Calculated variation for H 2 O 2 of the potential energy, V (solid line), and the
hardness, 77 (broken line), as a function of the angle of rotation, a. The trans form is taken as
the reference so that a, V and 77 are all zero. Energies are in kcal/mol. Reproduced with
permission from G.I. Cardenas-Jiron and A. Toro-Labbe, / . Phys. Chem., 99, 5325 (1995).
© 1995 American Chemical Society.
V = - £HOMO)/2 (4.22)
Table 4.4 Calculated Energy Values for Possible Si4 Structures [eV]
Structure
D2h £>4h ?d
Structure
r=\
Dewar
Benzene Prismane benzene Benzvalene Bicyclopropenyl
1 2 3 4 5
The orbital energies, e, also have been calculated by methods of quite variable
sophistication and accuracy. Yet the results always seem to support the concept
that "molecules tend to be as hard as possible". Even a method as primitive as
Hiickel MO theory can be used with good results.12
A study has been made of the effect of basis set selection on ab-initio SCF
calculations of molecular hardness.34 Not surprisingly, the value of the LUMO
energy, which is set equal to —A, is the term which changes most. This means
that the hardness calculated from Equation (4.22) will be very sensitive to the
quality of the calculations. Of course, \i will also vary greatly. But as long as the
same basis set, or its equivalent, is used to compare different structures, or
similar molecules, relative values of rj and \x are found which are usually reliable.
A test of the PMH which is quite different, and more general, has recently been
given.35 It follows earlier work by Gyftopoulos and Hatsopoulos, who used a
grand canonical ensemble with a limited number of discrete energy levels, so that
the distribution function was known.36 Those authors then calculated the
electronic chemical potential, /i, which was found to be /x° = (1+ A)/2. The
ensemble was a collection of systems containing the three species M°, M + , M~
and with energy levels E°, (E° + /) and (E° - A).
The ensemble average electronic energy can then be calculated in the usual
way, and the softness from Equation (4.7):
(4.23)
l+2exp(-/3r7°)
(4.24)
l+2exp(-/3r?°)
116 The Principle of Maximum Hardness
These are the equilibrium values. The softness (cr) is not the same as a° — I/770
where rf = (/ — A)/2. It may be thought of as an additional softness due to the
finite temperature. At zero Kelvin, (cr) = 0.
For two different equilibrium systems, Equations (4.23) and (4.24) tell us at
once that the system of higher energy has the greater softness. The restrictions
are constant temperature and constant (N) = N°. The restriction on /i° has been
removed. This proof of the PMH would apply to small changes in the
internuclear distances, since each set of nuclear positions corresponds to an
equilibrium value of the electronic energy.
A more general case is that of a non-equilibrium distribution that is close to
an equilibrium distribution. The parameters fi, rj, E, /, v and A are all allowed
to change, but (N) = N° is maintained. It is then found that a — (a) increases
as E — (E) increases, where the overbars indicate averages over the non-
equilibrium ensemble. An interesting restriction is that AA > A/, i.e., the
energy of the LUMO must change more than the energy of the HOMO. This
is usually the cases for non-equilibrium electron distributions, as well as non-
equilibrium geometries.
Adding more energy levels, such as the energies of M 2+ and M 2 ", has no
practical effect on these results. However, they do depend on ignoring the contri-
bution of TS in Equation (4.5). But the entropy in question is only the entropy
associated with the different electronic degeneracies of M°, M + , M~. At any
reasonable temperature, the changes in TS can be ignored. The model calcula-
tion then strongly supports the PMH, as derived by Parr and Chattaraj on very
different grounds.13 It also supports the essential requirement that the non-
equilibrium ensembles have an energy higher than equilibrium, and that small
variations in /J, and v are permissible.
all lower states. A case for which this difficulty disappears is when the excited
state is the lowest state of a given symmetry species.37 Then the energy is
determined by p for that state.
The search for a workable DFT of excited states is an important and active
field at the present time.38 There has been little done, however, on the concepts
of chemical potential and hardness, which is our main concern. If we assume
that Equation (4.1) is valid, we can draw some conclusions about fi and 77 for
excited states.
Figure 4.7 displays the MO-energy diagrams for a ground-state molecule M°,
its first excited state, M*, and the states which give the minimum / and the
maximum Ay M + and M~. These are the same whether we start with either the
ground or excited state. But the energy required to go from M° to M + is clearly
larger than that for M* to M + . The difference is just the excitation energy, E*.
Similarly the energy gained in going from M° to M" is less than that for M* to
M~ by the same amount, E*. We have
(/* + v4*)/2 = (/° + A°)/2 = -//*- = -ii° (4.25)
(/* - A*)/2 = (1° - A°)/2 - E* = 77* (4.26)
There is no change in the electronic chemical potential, but the hardness of the
excited state is less than that of the ground state.
This conclusion can be drawn in another way. Write the reactions
r 1— -r>>
Energy ApxAx
(*2 i Y/2
Translational nx —r- /Z/2TT
\12 2n
iJ
Rotational (J{J + 1))1/2/Z/2TT
Vibrational (2v + 1)1/2/Z/4TT
Electronic - 1 +- O - ^ — } h 2n
3 [ 2 \ n2 \
compared with the total energies of such system. Quantities of chemical interest,
such as bond energies and activation energies, are calculated as small differences
between very large numbers. This greatly magnifies the errors of calculation. If,
instead, we could calculate changes in /i or 77 and convert them to changes in
energy, in principle greater accuracy could be achieved.
The preceding sections have shown that it is 77, and not //, which is more closely
related to energy. Also, the relationship is an inverse one in that +AE means
—A77. An annoying restriction is that /i should be constant, or change very little.
Also it is the electronic energy, EQ\, which appears in DFT-based equations such
as (4.5) and (4.21), where N is the number of electrons. Changes in nuclear-
nuclear repulsion are usually easy to calculate, however.
Changes in /a can be accommodated, if we assume that it is the grand potential,
Q, which is related to the hardness. At low temperature,
O = EQ\ — Nji =f(rj) (4.31)
The reason for including TV and TV2, both equal to one, is that the exact
definitions of 77 and 77 demand them, or their dimensional equivalent
0.010
0.005 "
0.000
-0.010 - - -0.010
-0.015 - - -0.015
i • i
-0.020 -0.020
4.9 5.1 5.3 5.5 5.7 5.9 6.1 6.3
AE
0.1 -
0.0
-0.1
-0.02 -0.01 0.00 0.01
(b)
Figure 4.8 (a) Changes in /J, and in rj for Li2 as the internuclear distance, R is changed from Ro,
indicated by the arrow, (b) Changes in the electronic energy, E, plotted against changes in 77.
Energies are in a.u. Reproduced with permission from J.L. Gasquez et al., J. Phys. Chem., 97,
4059 (1993). © 1993 American Chemical Society.
are known at several values of R in the region of the equilibrium value, Ro.
Figure 4.8 shows some results for Li2, which may be taken as fairly typical.
Figure 8(a) shows \L and rj as a function of R, in terms of changes from the
values at Ro. Figure 8(b) has Arj plotted against the change in the electronic
122 The Principle of Maximum Hardness
energy. While A/i, is small, A77 is a linear function of AEe\ with a negative slope,
as Equation (4.33) would predict, but the magnitude of the slope is wrong by a
factor of about four. The same results are found for H 2 , N 2 and HF, except that
the slopes are variable, being either more or less than N2, except for hydrogen.
Going back to equation (4.21),
we see that the slope depends on the changes in the hardness functional, Jf7. The
definition of Jf is given by28
where F[p] is the universal functional of DFT for the kinetic energy and the
electron-electron repulsion energy. Clearly J-f for H, H j and H 2 is just the
hardness, 77, of Equation (4.33) times N2. In other cases it is proportional to 77, at
least for a certain range of variation.
Unfortunately, we do not yet know how to calculate Jf with any accuracy, so
that the constants of proportionality are not known. But this is an area where
progress can be made, hopefully, by further study of Equation (4.35). Another
problem also exists, in attempting to relate changes in 77 to changes in energy.
The values of fi and 77 depend very much on the quality of the method used to
calculate them. While the relative values that have been calculated may be quite
good, as mentioned earlier, absolute values are not. The calculation of electron
affinities is especially difficult. On the positive side, if we know Ro, then we also
know (OEQI/OR) without any further quantum mechanical calculations. It is
simply equal to the force due to nuclear repulsion, ZAZB/^Q-
The nuclear-nuclear repulsions must be taken into account; often, but not
always, they account for relative stabilities. It is worthwhile looking at all the
relevant parameters in a few cases, taking theoretical results. Table 4.9 shows the
results for two pairs of isomers: naphthalene and azulene, N — 68; and cyanic
acid and isocyanic acid, N = 14. The most stable isomers, naphthalene and
(a)
R.J. Buenker and S.D. Peyerimhoff, Chem. Phys. Lett., 3, 37 (1969).
(b)
Reference 46.
Hardness and the Electronic Energy 123
HCN, cannot be predicted by F nn , EQh Np or fi. The only consistency is the size
of the HOMO-LUMO gap, used to calculate 77.
Similar results are found in detailed calculations on ten pairs of isomers of the
type HAB and HBA.46 However, in only eight of the ten cases does the most
stable isomer have the greater hardness. In any event, the PMH does not predict
that the hardness should correlate with the total energy, but with the electronic
energy, or ft. Of course, a larger gap going with greater stability is a reasonable
result, though not predicted.
The failure in three cases to correlate Ee\ with the energy gap means that going
from HCN to HNC, for example, is too large a change to meet the requirement
of PMH. It is found in all cases that the two isomers, in their most stable
geometry, are local maxima in hardness, as predicted. Also the transition state
for interconverting HAB and HBA is always much softer than either isomer.
Apparently, going from one stable structure to the TS is a smaller change than
going all the way to the other isomer, and the PMH holds.
REFERENCES
21. R.G. Pearson and W.E. Palke, /. Phys. Chem., 96, 3283 (1992).
22. W.E. Palke, unpublished calculations.
23. S. Pal, N. Vaval and R. Roy, /. Phys. Chem., 97, 4404 (1993).
24. P.K. Chattaraj, S. Nath and A.B. Sannigrahi, Chem. Phys. Lett., 212, 223 (1993).
25. D. Datta, /. Phys. Chem., 96, 2409 (1992).
26. R.G. Pearson, Symmetry Rules for Chemical Reactions, Wiley-Inter science, New York,
1976, Chapters 1 and 3. See also G. Makov, J. Phys. Chem., 99, 9337 (1995).
27. R.F.W. Bader, Can. J. Chem., 40, 1164 (1962).
28. R.G. Parr and J.L. Gazquez, J. Phys. Chem., 97, 3939 (1993).
29. P.K. Chattaraj, S. Nath and A.B. Sannigrahi, /. Phys. Chem., 98, 9143 (1994);
G.J. Cardenas-Jiron, J. Laksen and A. Toro-Labbe, /. Phys. Chem., 99, 5325 (1995).
30. G.I. Cardenas-Jiron and A. Toro-Labbe, /. Phys. Chem., 99, 12730 (1995).
31. M. Galvan, A. Dal Pino, Jr. and J.D. Joannopoulos, Phys. Rev. Lett., 70, 21 (1993).
32. D.S. Warren and B.M. Gimarc, /. Am. Chem. Soc, 114, 5378 (1992).
33. D.S. Warren and B.M. Gimarc, Int. J. Quantum Chem., 49, 207 (1994).
34. S. Nath, A.B. Sannigrahi and P.K. Chattaraj, /. Mol. Struct. (Theochem.), 306, 87 (1994).
35. P.K. Chattaraj, G.H. Liu and R.G. Parr, Chem. Phys. Lett., 237, 171 (1995).
36. E.P. Gyftopoulos and G.N. Hatsopoulos, Proc. Nat. Acad. Sci. USA, 60, 786 (1965).
37. O. Gunnarson and B.I. Lundquist, Phys. Rev. B, 13, 4274 (1976).
38. R.G. Parr and W. Yang, Density Functional Theory for Atoms and Molecules, Oxford
Press, New York, 1989, pp. 204-208.
39. Reference 38, p. 96.
40. P.K. Chattaraj, A. Cedillo and R.G. Parr, Chem. Phys., 204, 429 (1996).
41. For example, see I.N. Levine, Quantum Chemistry, 2nd Ed., Allyn and Bacon, Boston,
1974, p. 63.
42. H.S. Robertson, Statistical Thermophysics, Prentice-Hall, Englewood Cliffs, NJ, 1993,
Chapter 1.
43. J.L. Gazquez, A. Martinez and F. Mendez, /. Phys. Chem., 97, 4059 (1993).
44. S. Pal, R. Roy and A.K. Chandra, /. Phys. Chem., 98, 2314 (1994).
45. A.P. Hitchock, M. Tronc and A. Modelli, /. Phys. Chem., 93, 3068 (1989).
46. T. Kar and S. Schneiner, /. Phys. Chem., 99, 8121 (1995).
5 The Solid State
INTRODUCTION
There are several cogent reasons to include a chapter on the solid state in a
treatise devoted to chemical hardness, and other concepts, derived from density
functional theory. One is that DFT has been the theoretical method of choice in
dealing with solid-state problems for a number of years.
This began with a landmark paper by Slater in which he proposed that the
effects of exchange in the wave function could be replaced by an exchange
potential, proportional to p 1 / 3 where p is the electron density function.1 This
followed earlier work by Dirac, in which he showed how to add the exchange
energy to the Thomas-Fermi theory of the atom.2 The exchange potential is also
dependent on a factor, a, which is allowed to vary somewhat from its value in a
uniform electron gas. The method is called the Xa method and involves solving a
series of one-electron wave equations in a self-consistent manner.
Slater was led to the Xa method by the difficulty of using the Hartree-Fock
method in the case of solids, which contain an enormous number of atoms and
electrons. The HF method would require a complicated superposition of deter-
minantal wave functions, and in addition was known to give some erroneous
results.3 The Xa approach immediately showed its usefulness in calculating the
physical properties of solids.4
At this point it should be noted that we are talking about crystalline solids
where there is strong bonding between all nearest neighbors. That is, molecular
solids are excluded. In the strong-bonding case, it is necessary to consider the
entire crystal as one giant molecule. In such cases, DFT has great advantages
over HF, as discussed in Chapter 2. In addition, very large molecules, such as
proteins, can be handled in DFT by the "divide-and-conquer" method.5 This
technique breaks the molecule down into its subunits and solves each of these in
a separate calculation. The results may then be joined together, obeying certain
restrictions.
The solid-state equivalent of the divide-and-conquer method has been known
for a long time. The natural subunits are the unit cells, which are repeated many
times to form the macroscopic crystal. The wave equations need to be solved
only for a single cell. The connection between the cells is made by multiplying by
the factor exp (ik-R) when one goes from any point in one unit cell to the
corresponding point in another unit cell. R is a vector marking the distance of
the second cell from the first and k is a wave vector in reciprocal space. This
and
H = {dE/N)s,v 2r/ = (82E/dN\v (5.1)
where TV is the number of electrons, are just simple thermodynamics. The second
equation is not generally used, nor is the interpretation of chemical hardness.
However the gap between the HOMO and the LUMO is an important one in
solids. It is called the energy gap or the band gap.
BONDS IN SOLIDS
We will begin our study of solids by considering the chemical bonds that must
exist between neighboring atoms. This will be done by the use of a semi-empirical
MO theory, which is simple and gives much insight into the nature of the
bonding.6 It is a localized MO theory, even though the idea of a single crystal
being a giant molecule calls for orbitals delocalized over the entire crystal.
These non-local orbitals are the canonical MOs of the giant molecule.7 But we
can always take linear combinations of the filled canonical MOs to produce an
equivalent set of filled MOs which are localized. The elecron density and the
energy are unchanged by this operation. The localized MOs are much easier to
picture and to comprehend. Afterwards we will see how the delocalized MOs
change the story.
The LCAO-MO model used here is very similar to models already developed
for bonding in single molecules.8 These are reasonably successful for predict-
ing, or rationalizing, bond energies. Actually the LCAO-MO semi-empirical
Bonds in Solids 127
approach has been used in solid-state theory for some time.9 It is called the tight-
binding model. The emphasis has been on understanding the electrical and
magnetic properties of solids, quite naturally. As a result these applications,
while very useful in some ways, have not been good at predicting the cohesive
energy of solids.10
From the chemist's point of view, this is probably the most important property
of a solid. It is defined as the energy required to dissociate one mole of solid into
its constituent atoms:
We will write the cohesive energy as AEexp since our immediate goal is to
understand the experimental results already available from heats of formation of
solids. We can ignore the small errors resulting from heats at 298 K rather than
at absolute zero, the reference temperature for most theories.
The cohesive energy tells us about the strength of the chemical bonds in the
solid. Its magnitude determines the stability and chemical reactivity of AB.
Eventually it is the quantity which determines the structure of AB, since different
possible structures will have different energies.
Another property of great interest is the ionicity of the bonding. To what extent
do the atoms of the solid resemble neutral atoms, held together by covalent
bonds, and to what extent are they like ions held together by electrostatic forces?
This is a difficult question. Even if we have an accurate X-ray picture of the
electron density of a compound, it is very hard to say whether atoms or ions are
being shown. The same is true for an electron density calculated by accurate
quantum mechanical methods.
Over the years numerous scales of "percent ionic character' have been proposed
for both simple molecules and solids.11 They may depend on some physical
property, such as dipole moments or refractive indices, or be extracted from some
theoretical analysis. The various scales usually agree on a rough ordering but the
actual numbers vary widely. It seems to be impossible to define exactly what is
meant by "percent ionic character". Therefore, the numbers in the various scales
have meaning only in terms of the model from which they were derived. Each
scale is useful for some experimental or theoretical property, but transfer to other
properties is hazardous.
The model used here gives a different approach to the question of ionicity. The
scale is based upon energy. The model can give a good approximation to the
energy for completely covalent bonding and for completely ionic bonding.
The relative values, AECOV and AE[on, give strong evidence for the nature of
the bonding, especially if one or the other is quite close to he experimental AEQxp.
The model also interpolates between the two limits in a way that is well grounded
in fundamental theory. This then allows an estimate to be made of the ionicity
that gives the maximum cohesive energy. The increase in this quantity from
or A^cov, or Aismix, is also important.
128 The Solid State
This method has only been possible for the alkali halides and the halides of
group 11 (Cu, Ag, Au), until recently. We will apply it to all the binary
compounds AB, often written as A"B8~", where A is a metal and B a non-metal.
The transition metals are not included, except for d10 and high-spin d5: otherwise
crystal field stabilization would have to be included.
The ionic model requires a knowledge of the higher electron affinities of atoms
like S and P. Only the first affinity (A\) is known for any of the elements. The
second and third (A2 and A?) cannot be measured experimentally, and are nearly
impossible to calculate theoretically.
However, very recently, a self-consistent set of values for A2 (and A3) has been
proposed for O, N, S, P, Se and As atoms.12 These come from a semi-empirical
method using Slater orbitals, but calibrating the orbital exponents with the
known values of / and A\. Table 5.1 presents these new values in the form of the
total charging energy, ^An. This is the energy evolved in forming the ion Bw~
from B. Estimates are also given for Te and Sb, but these cannot be found from
Slater orbitals and are only crude extrapolations. The Prewitt-Shannon ionic
radii R# are listed for the various anions, and the orbital exponents, £, for the
valence-shell Slater orbitals are also shown. These indicate that ions like P3~ and
As 3 " are more compact than might have been imagined.
The great majority of AB compounds have either a rock-salt structure (coor-
dination number 6) or a sphalerite-wurtzite structure (coodination number 4).
A good theory should predict the correct structure, or at least the correct
coordination number (CN). Pauling had originally done this by means of his
radius-ratio rules. However, it has been pointed out that these rules work only
50 percent of the time.13 The evidence is now strong that CN6 compounds
are highly ionic, and that CN4 comounds are more covalent.14 The usual
explanation for this is that CN6 has the higher Madelung constant, favoring
(a)
Energy evolved in the process B +we = B"
(b)
Ionic radii or CN6, experimental.
(c)
Orbital exponent for the valence-shell Slater orbital.
(d)
Estimated only.
(e)
Calculated, using data from ionic salts.
Details of the Model 129
ionic bonding, and that CN4 compounds can form four tetrahedral hybrid
orbitals at each atom, favoring covalent bonding. Some qualification is necessary
for the latter explanation.
The accepted criteria for the strength of covalent bonds are due to Mulliken.15
The exchange integral /? increases with the ionization potentials (or electro-
negativities) of the bonded atoms. It also increases with the magnitude of the
overlap integral S. Since a tetrahedral hybrid gives the maximum overlap with an
orbital of a given neighboring atom, it is assumed that hybrid orbitals form the
strongest bonds. But this is by no means true if there are several neighbors. In
CH4, for example, the total overlap is exactly the same for the symmetry-adapted
MOs as for the four tetrahdral hybrids. All that hybridization does is to
concentrate the orbital in a particular direction.
One might even believe that there is more covalent bonding with six nearest
neighbors than with four, since there are more orbitals to overlap, but in
extended systems such as the AB solids this is not true either. The total overlap is
exactly the same for one s orbital and three p orbitals on each atom, whether the
CN is 4 or 6. This results because the contribution of any AO to any MO is
diluted according to the number of bonded atoms. This exactly compensates for
the greater number of overlaps, providng the interatomic distances are the same,
this is also true for CN8, as in the CsCl structure.
It is the distance feature that favors covalent bonding for CN4. As is well
known, intferatomic distances are greater for CN6 than for CN4, by about 5
percent, when the same atoms are bonded. This is a consequence of the greater
number of nearest-neighbor repulsions. Since the overlap integral falls off
exponentially with distance, it follows that /30Ct is less than /?tet, leading to weaker
covalent bonding.
For the case of completely ionic bonding, the Born-Mayer equation will be
used:16
iv (5>3)
M71
(5.4)
130 The Solid State
Here m is the coordination number and M is the Madelung constant (1.748 for
m = 6 and 1.64 for m = 4; no distinction is made between sphalerite and wurtzite
structures).
The equilibrium interatomic distance Ro is assumed to be known, in addition
to the compressibility n and the molar volume Vo. Actually, only a limited
number of reliable values for K are known,17 but regularities in V0/K allow
reasonable estimates to be made in other cases.18 When Z is 2 or 3 small errors in
these estimates can lead to large errors in the equilibrium potential energy C/o.
The equation for the cohesive energy becomes
NA + NB = 2 (5.7)
AE 8 p 2 / 3 K 8 n)]l/2 (5 9)
=- [ 0 ( V ) ] = - - -
When n= 1, 2, 3 or 4 we have the 1-7, 2-6, 3-5 and 4-4 cases. Equation (5.9)
gives 2(3y/7, 2/3\/l2, 2(3y/T5 and 8/3, respectively. If AB were a diatomic molecule
with each atom contributing one electron to a single bond, we would have
AE = 2/?.
Details of the Model 131
(5.11)
The first term on the right-hand side in Equation (5.10) is the same as — AE in
Equation (5.9), but explicitly showing that (3 depends on R. The dependence
shown follows from the repulsion term, depending on the first power of the
overlap integral, rather than on the square of it. Equation (5.10) is the solid-state
version of the Morse equation.
The cohesive energy, AEC0V, is now equal to — Uo, which is equal to AE/2 in
Equation (5.9). Since /3 is to be fitted in any case, a knowledge of p is not so
critical for covalent bonding. For orientation, p is equal to 0.305 A for carbon
and 0.486 A for gray tin.
It is necessary to evaluate /? in Equation (5.9) by fitting to some experimental
data. One method has been to use the Mulliken criteria relating (3 to /and S. This
was first done by Wolfsberg and Helmholz.20 It requires separate consideration
of s and p orbitals, however.
The best overall results are obtained by calibrating with single-bond energies
for the elements, and using the arithmetic mean:
To be consistent with Equations (15.0) and (5.11), /?AA is the single-bond energy
for element A, and similarly for /3BB- Although assumptions are necessary to
obtain these energies, Bratsch has compiled a list of the required single-bond
energies for the elements.21 One change was made - the experimental data for N,
O and F were corrected for the lone-pair bond weakening effect. The corrections
were assigned on the basis of estimates given by Politzer.22 (The correction is not
unambiguous for O and N, because O - O single-bond energies vary from 51 kcal/
mol for H 2 O 2 to 30 kcal/mol for (CH3CO)2O2, with similar variations for N - N .
Average values were taken.) The corrected values, with Bratsch's other results,
are listed in Table 5.2, for convenience. A few new values have been added.
The /3AB values calculated from Table 5.2 are for CN4. Since /3 is a function of
distance, that for CN6 will be less. Looking at the difference in R for CN4 and 6,
and using Equation (5.10), it appears that (3oct « 0.80 /?tet- This correction has
been used in calculating AEC0Y for CN6.
We must also consider two factors that relate to the ionic model. One is
usually called the penetration error. The potential energy in Equation (5.3) is
calculated by using point charges for both ions. A more realistic picture has the
132 The Solid State
Li 25 As 43
Be 55 Se 48
B 73 Br 46
C 85 Rb 12
85 (b)
O Sr 23
85 (b)
N Y 52
9 0 (b)
F Ag 38
Na 17 Cd 14
Mg 29 In 31
Al 43 Sn 36
Si 55 Sb 34
P 48 Te 38
S 64 I 36
Cl 58 Cs 11
K 14 Ba 19
Ca 24 La 53
Sc 48 Au 52
Cu 46 Hg 10
2 3 (b)
Zn 17 Tl
Ga 34 Pb 28
39 (b)
Ge 45 Mn
(a)
From Reference 21.
(b)
Estimated.
cation as a point charge, but sitting in the electron cloud of the anion. The
potential energy is made smaller by this change. Fortunately it is easy to
calculate the error with the Slater orbitals used to construct Table 5.1.
The second factor is the radius-ratio effect. Only nearest-neighbor interactions
were included in the repulsion part of Equation (5.3). But there is one situation
where next-nearest neighbors must be considered: this is the case where one ion,
usually the anion, is so much bigger than the other that they are in contact, or
strongly overlapping. The values of RB in Table 5.1 and values of Ro may be used
to assess this possibility.
Table 5.3 shows the experimental values of the cohesive energies and the
semi-theoretical values, calculated for pure ionic and pure covalent bonding.
Table 5.3 includes most of the available data on AB solids. The omitted exam-
ples, mostly alkali halides and alkaline-earth chalcogenides, show no unexpected
features.
The ionic energies in Table 5.3 have not been corrected for the penetration
error. Table 5.4 gives some sample values of this error. It is negligible for 1-7
salts, and small for 2-6 compounds, except for BeO and BeS. All of the 3-5 cases
Details of the Model 133
A/7 (b)
AB *o ( a ) A£ c o v A£ i o n CN
[kcal/mol] [kcal/mol] [kcal/mol]
(a)
Data from Reference 23.
(b)
Data from Reference 24.
(c)
Cubic form; see Table 5.4 for penetration error.
(d)
Extrapolated.
(e)
Interpolated.
(a)
Difference between Madelung energies calculated from point charges for
anions and from charge clouds for anions.
Details of the Model 135
have appreciable errors, especially those with CN4. The compounds BN and BP,
.with the diamond structure, have very large errors. For BN the correction for
A£ion = 375kcal/mol reduces the value to AE{on — 150kcal/mol.
The distance between next-nearest neighbors (RBB or RAA) is readily found. It
is 1.414i?0 f° r CN6, and 1.633i?0 for CN4. This number may be compared with
2RB or 2RA, the distance between two ions when they are in contact. RB is the
anion radius found by X-ray diffraction and listed in Table 5.1. If 2RB is greater
than RBB, then the ions are overlapping and large repulsion energies result. It is
found that 2RB is comfortably greater than 2RB for all cases in Table 5.3 except
for BN, BP, BeS and BeO. These all have CN4 in agreement with the radius-ratio
rule. But also, except for BeO, the covalent cohesive energy is much greater than
the ionic energy.
Two important conclusions can now be drawn from the data in Table 5.3. The
first is that, with few exceptions, compounds with CN4 have AEC0Y much larger
than A£j on . The reverse is true for CN6. This agrees with expectations: covalent
bonding is favored by CN4, and ionic bonding by CN6. The second conclusion
is that, with few exceptions, the larger of the two theoretical values, AEcoy or
A£i on , agrees reasonably well with the actual cohesive energy, Aisexp- This is
particularly true for CN4, where the agreement is remarkable considering the
simplicity of the bonding model. Note that the 4-4 examples are the elements
C-C, Si-Si and Sn-Sn, as well as Si-C.
The bonding ability of the ionic model can be improved by varying the net
charges on the ions to give the minimum energy.6 It is possible to write the MO
energy equations in terms of the charge x, which can vary from 0 to n for A"B8"".
When this is done, it is found that x = 0 is always a local minimum in the energy,
and that x ~ 0.9n is another local minimum. That is, mixing of the ionic state
does not improve the energy of the purely covalent state, but a little mixing
improves the energy of the purely ionic state.
A AEm[x of 5-30 kcal/mol is calculated. It is small for compounds like NaCl
and CaO, and large for cases such as CuCl, BeO or MnS. With these corrections,
the ionic model gives very good agreement with Aisexp for all CN6 compounds,
except CdO. BeO is predicted to be 95 percent ionic. It has CN4 because of the
radius-ratio effect. This is also the case for MgTe.
To summarize, the MO bonding model predicts that most AB solids will fall
into two classes: very ionic with the rock-salt structure and 100 percent covalent
with a tetrahedral structure. The ionicities are a result of the simplified model
and should not be taken too literally. Nevertheless, there are differences in
bonding between the solid state and small molecules which are real and which the
model correctly identifies. In the solid state, compared with a diatomic molecule,
interaction of an ion with the rest of the lattice favors a large value of x. At x ~ 0
there is no stabilization from the lattice. Ionic mixing does not help. Also the
metallic element, A, contributes only one to three electrons to four orbitals.
Immobilizing these electrons on B to give an ionic state causes an exaggerated
loss of covalent bonding.
136 The Solid State
The reason for two different classes of AB componds is also clearly identified
by the model. It is a result of quite different charging energies for metals to the
left and to the right in the Periodic Table. Thus, I\ — lOOkcal/mol for K and
178kcal/mol for Cu; h +h = 415 kcal/mol for Ca and 631 kcal/mol for Zn;
I\+h + h - 1017kcal/mol for Sc and 1320kcal/mol for Ga. ScN, ScP, ScAs
and ScSb all have CN6; GaN, GaP, GaAs and GaSb all have CN4. The non-
metalic elements have no part in the structure-making decisions, unless the
relative ion sizes are such as to enforce CN4.
The first entries in Table 5.3 are arranged so that the AB compounds formed by
elements in the same row of the Periodc Table are grouped together. The values
of RQ for all the CN4 compounds from each row are seen to be nearly constant,
including the 4-4 case. Sn-Sn, InSb, CdTe and Agl al have Ro = 2.81 A, for
example. This is consistent with pure covalent bonding in all four cases. The
change to CN6 is accompanied by increases of Ro to 3.08, 3.17 and 3.50 A for
YSb, SrTe and Rbl respectively. This suggests a sudden change to ionic bonding.
The distances increase steadily because of the decrease of the force of attraction
as the charge on the ions becomes smaller.
In molecules the various methods of estimating percent ionic character give
results ranging smoothly from 0 percent to near 100 percent. This includes the
kind of semi-empirical model used above,8 as well as much more sophisticated
calculations.25 But in solids there appears to be a discontinuity in ionicity, just
as there is a discontinuity in structure. It may be more meaningful to classify
solids as either highly ionic or highly covalent, rather than as positioned in a
continuous scale.
A few AB solids have body-centered cubic structures, with a CN of 8. An
example is CsCl, which is calculated to be much more stable in the ionic model.
Again ionic bonding seems to favor a high CN, since the Madelung constant
increases, being 1.763 for bcc. This theory proves to be short-lived, however,
when we consider bonding in the metals. By definition, the bonding here must be
covalent, since identical atoms are bonded.
But the metals are most commonly found with close-packed structures, body-
centered cubic (bcc), face-centered cubic (fee) or hexagonal close-packed (hep).
The number of nearest neighbors, or the CN, is high, being 12 for the close-
packed structures and 8(+6) for bcc. So covalent bonding also benefits from high
CNs in some cases. At first it seems unhelpful for so many atoms to be within
bonding distance, because the non-transition metals, at least, do not have enough
valence electrons to bind even four neighbors.
If we try to apply our semi-empirical LCAO-MO method to metals, we get poor
results. Taking lithium metal as an example, and using two Li atoms as our unit
(just as for C, Si and Ge), we have only two electrons to fill up the four possible
MOs. The cohesive energy is only (3 instead of 4/3, the same as it is for the diatomic
molecule Li2. This gives A2?cov = 25 kcal/mol, using Table 5.2. The value of A^xp
is 77 kcal/mol. A localized bonding model is not useful for the metals. We must
exploit the new features that result because a solid is a very large molecule indeed.
Bands in Solids 137
BANDS IN SOLIDS26'27
An easy way to see the effect of size on quantized energies is to take the well-
known problem of a particle in a box. Start with a single atom in a box of atomic
dimensions, and then stack a large number of these boxes to give a macroscopic
crystal. Taking Li as an example, and assuming a cubic box of side a, the energy
levels for the atomic box are given by
t _*y,+•;+•»
8mr
For the ground state, nx = ny = nz = 1. The mass m is the mass of the valence-
shell electron, which is the particle. The rest of the atom simply creates a
background potential, f/o, constant in the box.
The energy of Equation (5.13) is kinetic energy, which is positive. The
potential energy C/o is negative and can lead to a bound state. To apply Equation
(5.13) it is convenient to use atomic units: h = 2TT, m = e = #o = 1, and 1 a.u. of
energy = 27.21 eV. The ionization potential of a Li atom is 5.39 eV. Equating this
to the kinetic energy gives an a value of 8.65ao in (Equation (5.13). This is a
reasonable result. The average distance of an electron in a 2s orbital of hydrogen
is 6.0a0, for comparison.
Next we take Avogadro's number (No) of Li-atom boxes and put them
together to form a cubic crystal of lithium metal. From the density we calculate
the side of the cube to be 2.35 cm. Dividing by No, we find that each atom is in a
box which has a side a — 5.26a$. If each electron was confined to a single atom,
the energy would increase to 14.56 eV, using Equation (5.13).
Since we need the energy to decrease, showing a cohesive energy, we must allow
the valence electrons to move throughout the box. They now move in a constant
potential due to the nuclei, the inner-shell electrons and the other valence-shell
electrons. This is the free-electron model for metals, due to Sommerfeld.28
By imposing a periodic boundary condition in the form
V<x + L, y + L, z + L) = *Kx, y* *) (5-14)
where L is the side of the crystal, we find that the solutions of Schrodinger's
equation are plane waves
We must add No electrons to our crystal, in accordance with the Pauli Exclusion
Principle. Each level of Equation (5.16) can hold two electrons of opposite spin.
The volume element (27r/a)3 defines a primitive unit cell in k space, each cell
contains one energy level. The ground state will fill all levels from k = 0 to a
limiting value, kip. The No electrons will need NQ/2 unit cells, or the number lying
in a sphere of radius kF.
The highest energy is that at the surface of the sphere, the Fermi energy, £p-
h2k2F h2
The molar volume, Fo, is simply NQCP. Also, the average energy per electron can
be shown to be equal to 3/5 of e-p, so we can write
* - » (MO
40ma2
For lithium metal this is 2.82 eV, which is to be compared with the average
energy of the separate Li atoms of 5.39 eV.
There has been a reduction in energy of 2.57 eV, or 59.3kcal/mol. For two
atoms this means a cohesive energy of 119 kcal/mol, which is to be compared with
A£"exp = 77 kcal/mol. Of course we must also consider the change in UQ to obtain
a more complete story, but the essential point has been established. Putting a large
number of electrons in a large box produces a reduction in energy, or a delocaliza-
tion energy.
Whereas physicists are comfortable with plane waves, chemists will usually
prefer an explanation based on LCAO theory. This is the tight-binding approach,
in solid-state terminology. The result may be anticipated by our general know-
ledge of valence theory. Figure 5.1 shows the building up of energy levels as we
increase the number of atoms bonded together. For one valence orbital per atom,
one more MO is formed for each atom added. For a linear chain, the lowest-
energy MO is pushed down and the highest MO is pushed up with each addition
of an atom. Also the levels are more closely packed, so that for an infinite
number of atoms a continuous band of levels is formed.
If there is one electon per valence orbital, then only the bottom half of the
band will be filled, since we can put two electrons of opposite spin into each level.
Thus all the bonding orbitals will be filled and all the anti-bonding orbitals will
be empty. The result is a net stabilization of the system, or a cohesive energy. If
there are several valence orbitals on each atom with different energies, then
several bands will be formed, which may or may not overlap.
This picture can be made more quantitative, at least at the level of simple
Hiickel MO theory. Take a linear chain of H atoms, or a linear chain of CH
units, with a constant spacing, a, between the units. The energy of the levels in
such a case is given by
sj = a + 2/3cos(2irj/(N+ 1)) (5.19)
Bands in Solids 139
where the index j refers to the MO, or level, and N is the number of atoms,
defining a wave vector, k — Inj/Na, the energies are given by
= a + 2(3 cos ka (5.20)
when N is large.
The wave vector forms a reciprocal space as before and has continuous values
in the range — n/a < k < n/a. These limits define the (first) Brillouin zone of a
crystal. The point k = 0 is the zone center and k = ±ir/a are the zone edges. The
variation of the energy with k is called the dispersion of the band. Figure 5.2
shows the dispersion given by Equation (5.20). Only the range from 0 to ir/a
need be shown. The range from 0 to -ir/a is just the mirror image. Figure 5.2
also shows the Fermi level, k^ = rr/2a, which is the highest filled level if the levels
are doubly occupied.
The energy per electron ranges from e0 = a + 2(3 to e¥ = OL. We can find the
average energy by simply counting the energies of the occupied levels:
2 C l
t (V\A1r ^ ^ (5.21)
• = — e(k)ak = ha
ft Jo ft
Since the levels are nearly continuous, this can be done by integration. This
energy should be compared with the energy of an electron in the two-atom case,
e = a + (3. There is a delocalization energy of 0.27 ^/electron. Note that in
Hiickel theory, there is no repulsion energy term. The Hxickel (3 is one-half of the
(3 values in Table 5.2.
This analysis was for a one-dimensional solid. The extension to three dimen-
sions is not so easy for real structures, but can be done for a simplified model of a
metal. Imagine the case of a simple cubic lattice such as is shown in Figure 5.3.
140 The Solid State
a-2p
n/2a = kF n/a
Figure 5.2 Dispersion of the one-dimensional energy band formed by overlap of the s or p?r
orbitals of an infinite chain of atoms
Figure 5.3 A simple cubic structure for a solid which can form chains in the x, y and z directions
The coordination number is only 6, but if the valence orbital is an s orbital, then
we have three infinite chains for each atom, running in the x, y and z directions.
The energies are29
e(k) = a + 2/3(cos kxa + cos kya + cos kza) (5.22)
For two lithium atoms, the cohesive energy is 7.64/3, compared with 2/3 for an Li2
molecule. This can account for the difference found experimentally: 25 kcal/mol
compared with 77 kcal/mol. One should also consider internuclear separations,
2.67 A in the diatomic molecule and 3.03 A in the metal, so that (3 in the metal is
less than in the molecule.
The dispersion curve for the e{k) of Equation (5.22) is more complicated than
Figure 5.2, since kx, ky and kz can have different values. The convention is to
show the energies at various selected points in the Brillouin zone. These are
labeled as F, M, K and X, and are called symmetry points.
Returning to the one-dimensonal case again, the wave functions may be
written as
N
cj>(k) = {NY1'2 Y, expiikRvM - Rv) (5.24)
P=\
Rp is the vector which translates a point in one unit cell to the corresponding
point in another. The index p identifies the unit cell, or the atom, ip is an atomic
orbital, which is the same for all atoms. The <fi(k)s are called Bloch functions and
are simply symmetry-adapted linear combinations of the orbitals ij).
In three dimensions the exponential in Equation (5.24) must be written as a
dot product, exp(iA>/?p). Also, in real solids the atoms of the unit cell usually
contribute more than one valence orbital. Then one has a set of Bloch functions
and the crystal orbitals are linear combinations of the Bloch functions. Also, the
atoms of the unit cell are usually not the same. At least in the case of the AB
solids discussed in the previous section, we can show how this affects the results.
Take the one-dimensional case as a start. There is an infinite chain of alter-
nating A and B atoms, with a constant spacing a. The unit cell must now contain
two atoms. The energy levels are given by
I -1 J
£ =
(k) ~^~^—~ ± 9 V (aB - <^A)2 + 16/32 cos A:a (5.15)
For any value of k there are two roots, one bonding and one anti-bonding.
Figure 5.4(a) shows the two energies plotted as a function of k, for typical values
of as, «A and (3.
When aB ^> aA, Equation (5.25) may be expanded as29
4(32 cos ka
£\ = (*B + 7 v (5-26)
(a B - a A )
4/?2c
- - °S*a (5.27)
142 The Solid State
E2
n/2a n/2a
(a) (b)
Figure 5.4 (a) Dispersion of the bonding (lower curve) and anti-bonding (upper curve) levels
of an AB solid; (b) the same when a A = « B
Important new features are that there are now two bands, and that a band gap
exists at k = 7r/2a. The gap size is |QB — C*A|- Of course when A = B, as in C - C
or Si-Si, the gap must disappear, as shown in Figure 5.4(b). Although this looks
different from Figure 5.2, it is the same actually, as may be seen if one folds
Figure 5.2 back on itself in the middle. The difference exists because there are
now two atoms in the unit cell instead of one.
We also can see that, if (5 is very small, the gap would be constant,
independently of k. But it is a dependence of e on k that produces the band
width, or dispersion. Compounds which are highly ionic will have narrow bands.
The ground state will have the electrons concentrated on B, the non-metallic
atom. Excited states result from the transfer of an electron from B~ to A + ,
forming A,B.
In three dimensions there will be two sets of four overlapping bands for the
compounds in Table 5.2. In the rock-salt structure linear chains exist for the p x ,
py and pz orbitals and Equation (5.26) and (5.27) are appropriate. The s orbital
would give somewhat more delocalization energy, as in Equation (5.22). All four
of the stable bands will be filled completely by the eight valence electrons of the
AB unit. These are called the valence bands. All four of the unstable bands are
completely empty. They are called the conduction bands, for historical reasons.
In the limit where j3 = 0, we have the completely ionic case. The average
energies of the filled bands are given by Equations (5.3) and (5.4). In terms of
the model leading to Equation (5.26), the energy per electron is equal to QB, which
Bands in Solids 143
then must include the terms in Equations (5.4) and (5.6), assuming that a repul-
sive potential has been added. The average energy of the empty, or conduction,
band will be equal to QA, as in Equation (5.27). This must also include terms in / n
and An, as well as coulombic energies appropriate for an A,B unit in a matrix of
A + , B~.
In the wurtzite or zinc blende structure, we do not have linear chains of
nearest neighbors. Usually there is strong mixing of the s and p bands. The actual
calculation of the band structure of, say, diamond, is very complicated,30 but a
simple model gives the essential features. Refer back to Equation (5.9), showing
the energy lowering for eight electrons in four bonding orbitals as equal to |8/3|.
Not shown is the fact that there are also four anti-bonding orbitals, with an
energy increase of —/? per electron.
Just as we argued earlier that we could take linear combinations of the crystal
orbitals to form localized orbitals, we can now reverse the argument and take
linear combinations of the localized orbitals to form the crystal orbitals. If we
only combine orbitals which are completely filled, or completely empty, we will
not change the average energy. Thus in diamond we will generate a band of filled
orbitals with an average energy of (a -+- /?), and a band of empty orbitals with an
average energy of (a - /3) per electron.
Estimating the width of the bands is more difficult. However, we can obtain
some idea by looking again at Figure 5.2. Treating the filled and empty levels as
the valence band and the conduction band, we see that the average energy
difference in 8/3/TT, or 2.5/3. The width of each band is 2/3. Therefore, for covalent
bonding, the band width is of the same order as the average energy gap.
The conclusion is that band widths will be appreciable, but usually not great
enough to close the gap between the full and empty bands. Figure 5.5 shows sche-
matically the three situations that finally emerge. They are classified according to
their electrical resistance. Only the highest filled band and the lowest empty band
are shown. The vertical scale is energy, but the horizontal scale has no meaning.
Conductors, such as the metals, are characterized by a partially filled band, so
that the highest filled level and the lowest empty level are essentially at the same
energy, the Fermi energy. Insulators have a large residual gap between the valence
and conduction bands. Examples are ionic compounds, but also some covalent
compounds such as diamond. Semiconductors have a small gap between the
bands. Most of the covalent compounds in Table 5.3 fall into this class.
The electrical conductivity of electrons in a solid depends on the ability of an
electron to move to a higher energy level when accelerated by an electric field.
The energy change is very small, so that only partially filled bands can conduct.
In semiconductors thermal energy will promote a few valence-band electrons
into the conduction band. These electrons can now move in the field. So can the
electrons in the valence band whose energies are just below the levels of the
promoted electrons.
Figure 5.5 enables us to find the quantities /x and 77, as defined in Equation
(5.1), and as approximated by / and A. In the solid the ionization potential is
144 The Solid State
Figure 5.5 Schematic of the band structures for conductors (metals), semiconductors and
insulators. The shaded areas show the occupied levels. The lower band is the valence band,
and the upper is the conduction band
(£c~£y)
F°
til (5.28)
2
(5.29)
Bands in Solids 145
\i = M - 0 (5.30)
where <f> is the electrostatic potential in the bulk solid. The work function is
given by33
The added term is the potential difference between a point just inside the surface
and one just outside the surface of the solid.
The term ((/> — (j)') is also the potential due to the surface dipole layer. It is a
function of the condition of the surface and is not a constant. The problem in
elctrochemistry is that the emphasis is on measuring electrical potentials. In
addition to the potential due to different values of \i for two electrodes, the
1.0
<—T=0
f(E) 0.5
U—T>0
measured voltages will include surface potentials, which are often substantial.
However, in terms of energies they are negligible and are usually ignored in solid-
state physics. In the same way, a slight charging of an electrode can produce a
measurable change in voltage, but will have little effect on the total energy, or the
average energy of an electron in the bulk.
When two solids are put in contact with each other, there will be a contact
potential, called the Volta potential, V. It is given by
F=($i-$2) (5.32)
the difference in work functions. The potential is due to the transfer of electrons
between the two solids. The condition of equilibrium is that the electrochemical
potentials be equal, i.e.,
Mi = M2 = Mi - <t>\ = ^2 - 02 (5.33)
it is obvious that there is no simple relationship between /' and any of the
quantities discussed above. Reaction (5.34) does correspond to a definite change
in energy, or free energy; at standard temperature. Attempts to measure it, or
calculate it, will always run into problems, such as the existence of surface
potentials.33
the band gaps, which are always too small.35 This is attributed to errors in the
conduction-band energy levels, which are mainly unoccupied and similar to the
LUMOs of molecules.
Returning to a consideration of /i and 77, Equation (5.28) came from the
assumption by Parr that the chemical potential of an electron in a molecule was
of the same nature as the chemical potential of an electron in a solid. This was
accompanied by a finite difference approximation to a derivative and the use of
Koopmans' theory, relating orbital energies to / and A.
In view of these uncertainties, it is fortunate that an independent proof of
Equation (5.28) exists in solid-state physics.36 Equation (5.29), the Fermi
distribution function, is used to count the number of electrons in the valence
band and in the conduction band. The result is that
<535)
where Nv and Nc are called the effective number of states per unit volume in the
valence and conduction bands, respectively. Their ratio is not very large, and
since kT is only 0.02 eV at room temperature, the second term in equation (5.35)
can usually be neglected.
Actually, a similar slight dependence of /i on the temperature is found for
molecules.37 It involves the ratio of the degeneracies of M + and M~. Looking at
equation (5.29) again, we see that if the total number of electrons is to remain
constant, then \i must change as the temperature changes. This shows up in
Equation (5.35). In the case of semiconductors and insulators, the convention is
used that e? = M, at all temperatures. For metals, the convention is that the
Fermi energy is always the value at 0 K.
The chemical potential or its equivalent, the work function, has long been used
to measure the electron-donating and -accepting power of a solid. For two solids
in contact, the difference determines the flow of electrons. This is an important
property for semiconductor devices.38 The interaction between a solid and a mol-
ecule also is influenced by the differences in their electronic chemical potentials.
However, such interactions almost always occur at a surface, which brings in
important new effects. A brief discussion of surfaces will be given later.
It seems natural to consider E\ (i.e., twice the chemical hardness), if transfer of
electrons is the critical property. Certainly the fact that the gap is zero for metals
accords with the great electrical conductivity of these solids. Also, the size of the
gap determines the conductivity of semiconductors and insulators. In general,
the conductivity, C, is given by
The quantities rae and rah are the mobilities of the electrons and the holes.
They are proportional to the time between scattering events and inversely
proportional to the so-called effective masses. The holes move in an electric field
by neighboring electrons jumping into the hole, and creating a new hole. As the
temperature increases, the mobilities decrease, but n and p increase. At room
temperature the increase in n and p for metals is less than the decrease in me and
mh, and C decreases with increasing temperature.
For insulators and semiconductors the situation is reversed, and the con-
ductivity increases with temperature. The concentrations of the charge carriers
are given by
The value of the energy gap may be found by plotting the log of C against \/T.
The minimum energy gap is also the important factor for other properties of
a solid which depend on the electrons in the conduction band. These include
the Pauli spin paramagnetism, and the (small) contribution of the electrons to
thermal conductivity. All of these properties are due to extremely small con-
centrations of free electrons. Thus for silicon, where E°g = 1.1 eV, the number of
conduction electrons is only 2 x 1010/cm3, compared with an atom concentration
of 5 x 1022/cm3. This is for a sample where impurity concentrations have been
reduced to 1 part in 1012 by zone refining.
But there are many other properties of a solid which are little affected by such
small numbers. These are the properties where the response to a change in
conditions requires a change in all of the valence electrons, not just the few near
the Fermi level. Certainly this would be the case for the cohesive energy. The
appropriate energy gap in such cases would be that between the average energy in
the valence band and the average energy in the conduction band, so in place of
Equation (5.28) we would have
(5.39)
Table 5.5 shows experimental values of Eg, El and the cohesive energies (from
Table 5.3) for a number of AB compounds. The average energy gap results are
those calculated from experimental data on high-frequency dielectric constants
for the crystals. Later we will compare these values of Eg with those calculated
from our earlier bonding models.
Insulators and Semiconductors 149
(a)
References39band45.
(b)
See Table 5.3.
(c)
References 43, 45 and 46.
150 The Solid State
A reason for selecting the dielectric constant as the experimental basis for Eg is
that the high-frequency value, e^, is due entirely to the polarization of the
electrons. Also e^ is equal to the square of the refractive index and is easily
measured. There is a simple relationship between the average energy gap and e^ 39
where N is the number of valence electrons per unit volume and m is the electron
mass.
Another reason for using s^ as the source of Eg is that the second term in
Equation (5.40) is the contribution due to the polarization of the valence
electrons, i.e., the response of all these electrons to the perturbation of a weak
electric field. In perturbation theory this response is inversely proportional to the
differences in energy between the various excited states and the ground state. It is
common practice to replace the various energy differences by a single average
value, our Eg. An electron at the bottom of the valence band will contribute
almost as much of the dielectric constant as an electron at the top. Looking at
Equation (5.38), such an electron contributes nothing to the conductivity at
room temperature.
An early use of Equation (5.40) was by Phillips and van Vechten.40 They applied
it to derive a scale of percent ionic character. Their assumptions were such that a
continuous scale was generated. While such a scale may be questioned for solids,
there was also an important conclusion: the structures of the AB compounds
depended entirely on the ionicity. Low ionicity led to CN4, and high ionicity to
CN6 or 8. This is consistent with our earlier analysis based on cohesive energies.
Our immediate concern is whether Eg, or possibly E°g, serves as a suitable
measure of chemical hardness just as ( / - A) does for molecules. Examination of
the data in Table 5.5 shows that it does. There is a good correlation btween Eg
and the cohesive energy, as long as related solids are compared. That is, the 4-4
compounds show Eg falling just as A2?COh does. The alkali halides also are
correlated with each other, but not with the 4-4 cases. In the 2-6 examples, we
can compare the CN6 compounds with each other, but not with the CN4 cases,
which have their own relationship. The 3-5 solids form their own family for
CN4, but there are no Eg data for the ionic 3-5 cases, which probably belong to a
different family.
Actually, the same correlations are found for the values of E°g and Ai?coh5 but
it is not as good, simply because of the smaller range of E°g. This behavior is not
unexpected. Note that Eg is always much greater than E°g, as it must be. As an
additional test of the direct relationship between hardness and stability, it has
been shown that, when different structures are possible for a solid, the more
stable structure has the largest energy gap.41
We can also obtain a set of numbers for Eg by using the simplified bonding
model which led to Table 5.3. To do this, we will simply calculate the change in
energy in promoting one electron from the localized bonding orbital to the
Insulators and Semiconductors 151
1.748
AE = 3.62 + 14.39 - 1.5 = 11.1 eV = / (5.42)
2.81
/ A
i i k
k
-E o
o f
i Ec
r-° ^
9
1IIP
Figure 5.7 Relevant energy levels for a semiconductor or insulator. The size of the arrows
gives the magnitude of energy differences
152 The Solid State
where es is the static dielectric constant of silicon. The ionic radius, RQ, is taken
to be the same as the interatomic spacing in silicon for both Si+ and Si". This
gives A£soiv = —2.81 eV for both cases.
Table 5.6 Comparison of Eg from Dielectric Constant with Theoretical Spectroscopic Gap
CN6 CN4
This value may be compared to the photoelectric thereshold, 7°, which is 5.1 eV.
I is greater, as it must be, because of the band width broadening.
The second change occurs in the calculation of A. The value of a in Equation
(5.46) was taken as the ionization potential of a free silicon atom, 8.15eV, in the
calculation of ev. But for ec we must take a as 1.39 eV, the electron affinity of a Si
atom.
This may be compared with A° = 4.0 eV for Si. It is less positive, as it must be.
While we only need about 5 eV of band width to account for the differences, the
actual valence band width of silicon is very much larger, about 15 eV. This is a
result of the overlappiong of s and p bands.
(7— A) — 5.62 eV is not the same as the spectroscopic gap of 2/3 = 4.78 eV.
The former involves the formation of ions and their solvation, whereas the latter
does not create any change in polarity. The spectroscopic gap, which we will
call (7— A)s is the one listed in Table 5.6. It is more closely related to Eg than
(I-A) is.
Calculation of ev and ec for the 3-5, 2-6 and 1-7 covalent cases is a little more
complicated. From Equation (5.8), we would have
n (8-/i) [n(S-n)]1'2
£v = -aA-\ — aB H (3 (5.51)
o o 4
This is the orbital energy of a bonding MO, (j>\9 with the composition
\ (5 ,2)
where I/JA and ipj& are atomic orbitals on the more metallic element A and the less
metallic B, respectively. The anti-bonding MO, 02, must be
1/2
4>\ keeps the atoms neutral, but <p2 will give an electron distribution so that A is
negative and B is positive. We have
(8 - n) n [«(8 - «)] „ , ^
L J !
^- «A + O«B-L A P (5-54)
Insulators and Semiconductors 155
The solvation nergy, 1.55eV, is calculated for Zn 025+ and S 075+ , using the Born
equation. With the use of equations (5.50) and (5.54), we calculate
The solvation energy is for ZnO75~ and S025~. The numbers are reasonable when
compared with 7° = 7.6 eV and A° = 4.0 eV.
Finally, we calculate (/— A)s by transferring an electron from </>\ to 02. This is
the spectroscopic gap, which should be equal to Eg.
The solvation energy is now calculated for Zn° 5 ~ and S 0 5 + , which are the net
charges after transfer of the electron. By analogous calculations, (7 - A)s has
been found for several more examples of covalent compounds. The results are
also listed in Table 5.6.
These calculations assume that the charges produced are spread over the entire
crystal. Excitons can also be formed, with adjacent charges; the binding energies
are much smaller, however. For example, in GaAs the exciton binding energy for
Ga O25 ~As O25+ would only be 0.33 eV. In the vis-UV spectrum, the exciton ab-
sorption is at 1.5 eV, the same as the threshold for conduction band absorption.47
In general, the agreement between (7 - A)s and Eg is quite good. The numbers
need not be the same, as already mentioned. What is important is that the same
range of energies is covered and trends are followed. The results for the ionic 2-6
compounds include polarization energies that are larger than for the 1-7 cases by
a factor of two or three, bcause the polarizability of O2~, for example, is two to
three times larger than for F~. Actually the polarizability of O2~ (and Se2~) is
not a constant, but varies with the cation. Note that the excited state of CaO
is Ca + O", in a matrix of Ca 2+ and O2~.
We can also calculate the electronic chemical potential, //, which is equal to the
Fermi energy. From Figure 5.7, we see that there is only one value for /r. that is,
_£+i2 (5.58)
The theoretical values for NaCl, Si, ZnS and GaAs are -4.9 eV, -4.8, -5.2 and
—4.4, to be compared with the experimental results of — 4.8 eV, —4.8, —5.8
and —5.1. Just as in the case of I' and A1 measured in solution (see Chapter 3),
the solvation effects nearly cancel in the sum of/and A. But they are additive for
the difference, and cause large changes.
156 The Solid State
Assuming that the polarization energies cancel each other, the absolute EN,
X = —M, for an ionic solid AB becomes
(/A + AB)
= XA B
XAB = 2 ' ( 5<59 )
This is just the result for a mixture of gas-phase A and B atoms before they react.
Also, for the 4-4 cases, we find
(J
X(s) = ^ = X(g) (5-60)
That is, the EN of the solid is the same as the EN of the gaseous atoms which
make up the solid!
For the 2-6 and 3-5 covalent solids, X(s) is a more complicated mixture of /
and A for both atoms, but the final result can be expressed empirically as48
All of these observations can be summed up by saying that there is very little
change in the chemical potential in the overall process
This is quite remarkable, considering the very large change in energy accompany-
ing reaction (5.61).
The hardness, rj, on the other hand, almost always increases for reaction
(5.62). But this is only true if we take Eg/2 = rj. Even so, the hardness decreases
as the solid is formed from the atoms in the case of Si, Ge and gray tin. The
reason for this is not clear.
The most important fact about the bonding in metals is that it results from a
reduction in kinetic energy due to delocalization of the electrons. This is favored
by a high coordination number, leading to a high density for most metals. The role
of the large number of nearest neighbors is to provide a more uniform potential
field in which the electrons move. This keeps the kinetic energy low, in contrast
with rapid variations of the potential. The strategy is then to fill the bands of levels
only about halfway, thus avoiding the anti-bonding crystal orbitals.
The number of valence orbitals on each atom should be greater than the
number of valence electrons, to give the best opportunity to fill only the stable
Some Properties of Metals 157
levels. Except in a few cases, it is better to doubly occupy the most stable levels,
even though this has an energy cost of increased electron-electron repulsion. The
valence atomic orbitals should be rather diffuse, since this will give the greatest
overlap, increasing the band width, which means higher delocalization energy.
Such diffuse orbitals are found on atoms of small x and small rj. Diffuse orbitals
also overlap at larger atomic distances. This, in turn, permits an increase in the
coordination number. Electron-electron repulsion is greatest when both electrons
are on the same atom; in this case diffuse orbitals minimize the repulsion energy.49
Table 5.7 gives some properties of the more common metallic elements. The
cohesive energies are included. These are for a single atom,
(a)
Data from Reference 24.
(b)
Data from S. Trassiti, J. Chem. Soc, Faraday Trans. I, 68, 229 (1972),
and Reference 48.
(c)
For the free atoms.
* E F ,|i
Figure 5.8 Energy levels in a metal; e and ef are the average energies of the filled and unfilled
levels, respectively
160 The Solid State
both the filled and unfilled levels of the conduction band. The average energy gap
between them cannot be determined by the use of Equation (5.40), since the
dielectric constant of a metal is infinite.
Also, for metals we cannot find (/ — A)s by examining the electronic spectrum.
Because of the band structure, all possible frequencies of the electromagnetic
spectrum can be absorbed. This is usually followed by the immediate re-emission
of the photon so that there is almost total reflectivity.51 In the visible, this
accounts for the appearance of metallic luster. Above the photoelectronic
threshold, electrons are emitted from the metal as well.
Even though the bonding in metals must be purely covalent, we cannot use the
simplified bonding model of the earlier section. That model is appropriate for
cases where the delocalized crystal orbitals can be replaced by average localized
orbitals. This is not possible for metals, or at least not easy. Actually the tight
binding theory at the Hiickel level of approximation has been used for metals in
several cases.52
It is necessary to distinguish between s, p and d orbitals, and to use different
exchange integrals for a-, ir- and ^-bonding. If this is done, one can successfully
account for differences in energy. For example, the choice of crystal structure
(fee, hep or bec) for different metals can be predicted.53 It is likely that cohesive
energies could also be calculated in this way, if values of (3 such as those in
Table 5.2 were used. Unfortunately, most of the (3 values listed for metals were
estimated from the experimental cohesive energies.
Assuming that the cohesive energy is due to the delocalization energy of partly
filled bands, we can conclude that Equations (5.20), (5.22) and (5.23) are valid.
We can also assume that the coulombic energy a is the same as for the free metal
atoms. This is what is done in the simple Hiickel theory used for other covalent
bonding. Then we can write
where N is the number of valence electrons per atom, and n is an unknown factor.
In the case of the group 1 and 11 metals we can use the (3 of Table 5.2. Then
Equation (5.64) gives n — 3.1 for the alkali metals and n — 5 for Cu, Ag and Au.
These are reasonable values in a band theory of metal bonding. For other metals,
we simply lump together the unknown n and (3. We can also use the properties of
a band to write
due to surface potentials. Table 5.7 shows the experimental results for the work
functions of the common metals. They are usually obtained for polycrystalline
samples, since the surface potentials vary according to the crystal face through
which the electron is emitted.
We see at once that <3> is not equal to the ionization potential of the free atom,
as we might have supposed. Instead it is very nearly equal to the electronic
chemical potential, //, of the atom. These data are also listed in Table 5.7 for
convenience. The agreement is really quite good. This should not surprise us,
since we know that the coulombic integral can have two values. One, for the
removal of an electron is —/. The other, for the addition of an electron, is — A.
Since the energy levels in the band theory must be continuous, the value
€f! = -(I+A)/2 is a reasonable result. We now have - / / = (7° + A°)/2 =
(I+A)/2, where 7° = A° = -ep and 7 and A are for the free atom. The fact that
fj, hardly changes in going from the atoms to the solid is the same result that we
found for the other classes of solids.
To find the spectroscopic gap, (7+ A)s, we can assume that a — a' since there
is no change in the charge. Then
This result is very similar to that for other covalent solids, and obviously gives us
a direct dependence of the cohesive energy on the effecive energy gap, or
hardness = Eg/2.
In accordance with the standard operational definition of the hardness, we
must have rj — (7° — A°)/2 = 0. This agrees with the ease of electron movement,
as evidenced by the high electrical and thermal conductivity. However, this
movement does not lead to the breaking of bonds. The energy gap, Eg, is
obviously the quantity which determines chemical reactivity, such as dissocia-
tion. The hardness decreases in going from the free atoms to the solid, if we take
(7° — A°)/2. But it increases if we take Eg/2 = AEQxp/N as an effective hardness.
The relation between the work function and the Mulliken electronegativity, as
shown in Table 5.7, has been known for some time.48'54 The early use of this
near-equality was to determine the unknown EN of some of the elements. Now
we would turn the procedure around and estimate $ from the usually well-
known data for the free atoms. In any case, the most interesting point is the
constancy of the electronic chemical potential.
It is not easy to find other data which show that Eg plays a role in the reactions
of solids. There is one case for metals which can be used as a test: the formation
of alloys, which represents, at least in a sense, the reaction of two solid metals
with each other. The stability of alloys of varying composition is influenced by
factors such as the relative atom sizes, and the number of valence electrons per
atom (the Hume-Rothery rules).
But it has long been known that differences in EN for two metals are favorable
for stable alloy formation. This has been put in a quantitative form by Miedema
162 The Solid State
and his co-workers.55 The heat of formation of a binary alloy can be given by the
empirical equation
4(ryc + m)
These two topics are treated together because they may both be considered as
dealing with solids that are incomplete in some way. Clusters may be as small
as three atoms or may contain many thousands of atoms. The study of clusters,
particularly metal clusters, is a very active field.56 The interest is two-fold: one
purpose is to learn how the properties of solids emerge as the number of units
increases to infinity; the other is because clusters are important in heterogeneous
catalysis. They have very large ratios of surface area to mass and are much more
reactive than large crystals.
For orientation, let us see how some important properties change as we go
from diatomic molecules to solids, for several kinds of systems. This is illustrated
in Table 5.8.
The largest change is in the cohesive energy, which increases in all cases. For
ionic solids, the increase is due to the larger Madelung constant. For metals, the
increase is due to the better delocalization energy with a higher coordination
number. For covalent solids, the increase occurs because all of the valence-shell
Clusters and Surfaces 163
(a)
For two atoms.
(b)
Calculated for Ga 2 As 2 . From Reference 57.
orbitals and electrons can be utilized. For ionic solids, there is little change in
/ o r A, because the band width is small. For metals and covalent solids, /becomes
smaller and A becomes larger because of the band width effect.
As the cluster size increases, two kinds of behavior are seen. In some cases
A/icoh, / and A all change rather smoothly as the number of units in the cluster
grows. An example would be nickel clusters, NiN, where N is the number of
atoms.58 The quantity to compare for energy is BE/TV where BE is the total
bonding energy. This number is 1.10 eV or Ni2, 2.70 eV for Ni13, 4.06 eV for Ni 147
and 4.44eV for nickel metal. At the same time, /decreases and A increases rather
regularly, so that they approach one another.
This behavior for / and A can be predicted on classical grounds. The work
function for bulk metal would be modified for small spherical samples by the Born
charging energy. The ionization potential would be increased and the electron
affinity would be decreased by the same amount
(5.70)
A = $-e2/2R (5.71)
where R is the radius of the sphere. As R goes to infinity, / and A become equal,
and equal to the work function. Equations similar to (5.70) and (5.71) have been
derived, both classically and quantum mechanically, except that values slightly
different from 1/2 are predicted.59
The other kind of general behavior observed for both metallic and non-metallic
clusters, is that A/scoh, / and A all change in a non-uniform way consistent with
the existence of shell structure within the cluster. That is, there is evidence for
extra stability for certain values of N, called magic numbers.60 The evidence
often comprises increased intensity in the mass spectra for the magic numbers.
This has been seen for the alkali metals and the noble metals of Group 11.
164 The Solid State
-0.5
Figure 5.9 Second energy differences, A2, for lithium clusters versus the number of atoms,
N Reprinted with permission from Reference 63
Clusters and Surfaces 165
Figure 5.10 Chemical hardness, I— A9 versus the number of atoms, N, for lithium clusters.
Reprinted with permission from Reference 63
Figure 5.9 shows that A2 is close to 0, except when N = 2, 8, 18, 20, 34, 40 and 58.
The peaks at these numbers show that the model is a good one. Figure 5.10
shows a plot of (/ - A) as a function of N. There is a general decrease of (/ - A)
as TV increases. This is consistent with Equations (5.70) and (5.71).
However, the most striking features of Figure 5.10 are the pronounced peaks at
N = 2, 8, 18, 20, 34, 40 and 58. At the magic numbers the hardness shows a local
maximum. Increased stability is accompanied by increased hardness. Identical
results have been calculated for sodium clusters.59 Similar, but not so spectacular,
resuls may be calculated for small silicon clusters.64
The case of carbon clusters is quite different.65 A model suitable for metals will
not work. Instead, a definite structure must exist using ordinary covalent bonds.
For TV < 30 the clusters exist in linear, cyclic and bicyclic forms.66 For N > 30
the structures are three-dimensional (3D) and evolve into the cage or soccer-
ball structure characterisic of C 60 and C70. For N < 30, the magic numbers are
given by N = An + 2, where n is an integer.61 This suggests an aromatic type of
stabilization.
For larger TV the magic numbers are 50, 60, 70, 90, 100.67 The criterion for
extra stability is that the carbon atoms can form a 3D structure in which the
MOs are such that all TV electrons are in bonding orbitals, and all non-bonding
orbitals are empty.68 In other words, there must be a large HOMO-LUMO gap.
166 The Solid State
x[eV] 77 [eV]
they will contribute less to the more stable crystal orbitals, but more to the
less stable ones.
Therefore the surface will develop a positive charge when the valence band is
less than half-full, but will become more negative as the band fills. The positive
charge will counteract the higher surface energy and make the work function
greater. Filled bands, however, should have a smaller work function. There is
experimental evidence for this variation.71
There are other problems with trying to predict the properties of surfaces. In an
effort to reduce the surface tension, the atoms of the surface will usually rearrange
themselves to some degree. The changes can be small, such as a variation in bond
lengths, or extensive, leading to a quite different structure for surface atoms. In
this case the surface is said to have been reconstructed. Moreover, real surfaces
are not the uniform flat planes that we visualize, but contain many steps, kinks
and defects.72
In an attempt to lower their surface energy, solids almost always will adsorb
small molecules, such as H2O or O2, on their surfaces. Molecules which form
molecular solids have much smaller surface tensions than other solids. The study
of surfaces is a difficult one, it can be seen. Fortunately a number of very special
experimental methods have been developed,72'73 and surfaces may be studied in
great detail. Such studies are useful in many important practical areas, such as
adhesion, lubrication, corrosion and adsorption. However, the most important
area is probably heterogeneous catalysis.
This subject is too vast to cover even superficially, but it is worthwhile to show
how the modern theory of heterogeneous catalysis is related to the topics in this
book.74 The key reaction is between the adsorbate molecule, acting as one reac-
tant, and the surface atoms of the catalyst, acting as the second reactant. There is
a transfer of electron density between the two reactants leading to chemisorption,
and bond-breaking or weakening in the adsorbate.71'75 The acid-base character
of the surface is matched to that of the adsorbate.76
Donating and accepting orbitals are identified and their overlaps considered.
The direction of electron flow is found by considering the work function of the
solid and / and A for the gas-phase adsorbate molecule. Until recently the DFT-
based concepts of absolute EN and hardness were not used. The older idea of
HSAB was used in cases where acid-base interaction seemed important.76
The most common theoretical tool used is calculation by the extended Hiickel
theory (EHT).77 For solids these lead to a density-of-states (DOS) diagram. This
is a picture of the number of energy levels per unit of energy, as a function of the
energy. A high density of states at the Fermi level leads to stronger adsorption of
a substrate. It may also be recalled that the DOS at the Fermi level is equal to the
local softness.78 In DFT a high value of the softness also leads to better interac-
tion with a substrate.
The examples of most interest are the reactions of the small molecules H2,
N 2 , CH4, O2 and CO on transition-metal surfaces. It was concluded in these
important cases that the main interaction was the donation of metal d-electron
168 The Solid State
density into the cr* and TT* orbitals of the substrate. It was also found that, in any
transition series, the early members such as Ti, Zn or Ta were better donors than
the late members, such as Ni, Pd or Pt. This does not mean that the former were
better catalysts, since too strong an interaction will reduce catalytic efficiency.
We can also consider these interactions from the viewpoint of DFT. Hopefully
the equation
- XD)
2(r/D + 77c) 2(77D + 77c)
will show a correlation between the amount of electron transfer and the strength
of bonding. For metals the work function will be used for XD and a value of ~ 0
for ?7D. The very large number of bulk atoms will serve as a reservoir for the few
surface atoms that actually react.
For the reactions of H 2 , N 2 , O2 and CO we would reach the same conclusions
as the extended Hiickel calculations. These are molecules of high EN. However,
for CH 4 the predicted direction of net electron flow would be from CH 4 to any of
the transition metals. In the extended Hiickel approach, there is electron transfer
in both directions, with a small preference for metal d to cr*. Note that the
reaction of methane with metal surfaces is much more difficult than for the other
substrates.
Equation (5.73) gives the same order of strength of bonding for the various
metals, since <£ and x become larger as we go from left to right in a transition
series, both for the free atoms and the bulk metals. The near-equality of x a n d 3>
in table 5.7 suggests that the relative reactivity of various substrates should be the
same for the atoms and their metals. The ordering for free atoms is given in
Table 3.6 of Chapter 3. It seems to be reasonable for metals also, though it is
hard to find comparable data.
The use of the DFT-based concepts in solid-state chemistry is in its infancy. But
enough has been done to show promise in a number of diverse areas. An impor-
tant application appears in the interpretation of scanning tunneling microscopy
(STM) signals.79 STM, invented in the early 1980s by Binnig and Rohrer, is one
of the ingenious new techniques for studying surfaces. It generates images of the
atoms on a surface by the tunneling of electrons between the atoms and a sharp
metal tip. The current can flow either way, depending on the bias voltage. The
brightness of the image depends on the magnitude of the current.
At low temperature and voltage, the current is directly proportional to the
local density of states at each point on the surface.80 But this DOS is just equal
Recent Applications of Concepts 169
to the local softness! Therefore it is possible to scan a surface to find the sites of
greatest softness, and hence of reactivity.79 Both the original surface and mol-
ecules adsorbed on the surface can be examined. By changing the bias voltage,
the local softness for both accepting and donating electrons can be measured.
The current and the brightness of the signal also depend on the work function
of the surface.80 Since this could change for various points on the surface, it may
also play a role. However, there is already some evidence pointing to a major role
for the local softness. The (111) plane of silicon contains two kinds of Si atoms,
one kind being softer for accepting electrons. Soft molecules are adsorbed on
the softer atoms and hard molecules on the harder atoms.79 Organic molecules
adsorbed on surfaces show bright spots for aromatic rings but are dark for
aliphatic chains. The brightness of functional groups increases in the order
CH 3 ~ OH ~ Cl < NH 2 < Br < K SH.81
The HSAB principle has long been used to rationalize the adsorption of
various molecules on various surfaces (see Chapter 1). This correlation is now
being done more quantitatively. For example, the adsorption of C6H5SH and
C6H5SO2H, and the non-adsorption of C6H5SO3H on gold, have been explained
in terms of the increasing HOMO-LUMO gap.82 The sites of adsorption of Na,
Al, As and Cl on a gallium arsenide surface agree with predictions based on local
softness.57 The HOMO-LUMO gap and its role in the adsorption of organic
polymers on metals has also been discussed.83 This is important in understanding
the operation of adhesives.
There is a theory explaining the stronger adsorption of H 2 on the early transi-
tion metals of each series, and the non-adsorption on Au, Ag and Cu.84 As the
H 2 molecule approaches the metal surface, the electron in the s orbital of
the metal is pushed into an empty d orbital to avoid repulsion. This detailed
mechanism has been confirmed by DFT calculations.85 The local DOS as well as
orbital symmetry are the determinants. These also determine the surface sites
where a hydrogen atom will be bound.
A rather remarkable paper has appeared on catalytic reactions in zeolites.86
It combines ab-initio calculations on zeolites with the results of the following
reaction, catalyzed by the same zeolites
Both ortho- and /?ara-xylenes are formed as products. In toluene, the ortho
position is more negative than the para, but the local softness is greater as the
para position. If coulombic effects are dominant, we expect a low p/o ratio. If
electron transfer, or orbital control, is dominant, we should have a high p/o ratio.
Both ab-initio and semi-empirical MO calculations were made for model
clusters of zeolites, in which Al/Si and Al/Ga/B ratios were changed. The
HOMO-LUMO gap was found in each case. The experimental pjo ratio, indeed,
was found to vary inversely with the size of the gap. Softer zeolites favored orbital
control, leading to more para-xylGne, as would be predicted.
170 The Solid State
There have been several studies in which certain properties of crystalline com-
pounds have been calculated. These include charges on the atoms, acidity, basicity
and reactivity of atoms and bonds. The approaches are based on absolute EN
and local softness, primarily. The objectives are to estimate the strength of
various bonds, reactivity of sites to electron donors and acceptors, and so on. It
is to be hoped that such information can be useful in synthesis of complex
substances and their stabiliy to various internal disproportionation changes.
Both electronegativity equalization methods (EEM)87 and EHT calculations88
have been used.
Most real solids are polycrystalline, consisting of grains that are more or less
cemented together at their boundaries. Impurities may or may not segregate at
these grain boundaries. This in turn has a large effect on materials used in
electronic devices. The HSAB and Maximum Hardness Principles have recently
been used to study segregation of As and Ga in germanium crystalites.89
Ab-initio calculations were made on combinations of Ge and As, with As in
the bulk and with As on the surface. The local softness, both for accepting
electrons and donating electrons, was calculated. It was found that arsenic, a soft
impurity (shallow donor) did congregate at the surface. A hard impurity (deep
donor) was predicted not to. Also it was predicted that gallium (a soft acceptor)
would not segregate at the grain boundaries. The PMH was obeyed for each kind
of behavior.
An interesting example of the hardness concept has been given by a study of
the charge capacity of TiS2 intercalated with lithium.90 It may be recalled that
Huheey had originally called (I - A)~l the charge capacitance of an atom or
group, K,.91 This can be written in terms of the electronic chemical potential and
the charge, Q, transferred to the group,92
Q
«=« , O , (5.75)
2f] (JJL° - fj)
REFERENCES
4. J.C. Slater, The Self-Consistent Field Method for Molecules and Solids, McGraw-Hill, New
York, Vol. 4, 1974.
5. W. Yang, Phys. Rev. Lett., 66, 1438 (1991); idem, Phys. Rev. A, 44, 7823 (1991).
6. R.G. Pearson, J. Mol Struct. (Theochem.), 260, 11 (1992).
7. C.A. Coulson, Valence, 1st Edn., Clarendon Press, Oxford, 1952, p.262ff.
8. R.G. Pearson and H.B. Gray, Inorg. Chem., 2, 358 (1963); G. Klopman, / . Am. Chem.
Soc, 86, 4550 (1964); R.L. Flurry, / . Phys. Chem., 69, 1927 (1965); idem, ibid., 73, 2111
(1969).
9. For introductory reviews, see M. Kertesz, Int. Rev. Phys. Chem., 4, 125 (1985); J.A. Duffy,
Bonding, Energy Level and Bands, Longmans, New York, 1990.
10. W.A. Harrison, Electronic Structures and Properties of Solids, Freeman, San Francisco,
1980.
11. For reviews and analyses, see P.G. Nelson, / . Chem. Ed., 71, 24 (1994); J. Meister and
W.H.E. Schwarz, / . Phys. Chem., 98, 8245 (1994).
12. R.G. Pearson, Inorg. Chem., 30, 2856 (1991).
13. J.K. Burdett, / . Am. Chem. Soc, 102, 450 (1980).
14. W.B. Pearson, J. Phys. Chem. Solids, 23, 103 (196?); J.C. Phillips, Rev. Mod. Phys., 42,
317 (1970).
15. R.S. Mulliken, / . Am. Chem. Soc, 72, 4493 (1950); idem, / . Phys. Chem., 56, 295
(1952).
16. N. Born and K. Huang, Dynamical Theory of Crystal Lattices, Oxford University Press,
London, 1954.
17. W. Yang and R.G. Parr, Phys. Chem. Miner., 15, 191 (1987); G. Simmons and H. Wang,
Single Crystal Elastic Constants, MIT Press, Cambridge, MA, 1971.
18. H. Wang, Phys. Chem. Miner., 3, 251 (1978).
19. C.A. Coulson, L.B. Redel and D. Stocker, Proc. R. Soc. London, 270, 357 (1962).
20. M. Wolfsberg and L.J. Helmholz, / . Chem. Phys., 20, 837 (1952).
21. S.G. Bratsch, Polyhedron, 7, 1677 (1988).
22. P. Politzer, / . Am. Chem. Soc, 91, 6235 (1969); idem, Inorg. Chem., 16, 3350 (1977).
23. R.W.G. Wyckoff, Crystal Structures, 2nd Edn., Vol. 1, Interscience, New York, 1963;
P. Villars and J.D. Calvert (Eds.), Pearson's Handbook for Crystal Data, American Society
of Metals, Metals Park, OH, 1985.
24. D.D. Wagman, W.H. Evans, V.B. Parker, R.H. Schumm, I. Halo, S.M. Bailey, K.L.
Churney and R.L. Nuttall, "The NBS tables of chemical thermodynamic properties",
J. Phys. Chem. Ref Data, 11, Suppl. 2 (1982).
25. J. Simons and A.I. Moldyrev, / . Phys. Chem., 100, 8023 (1996).
26. For a simple introduction, see R. Hoffmann, Solids and Surfaces, VCH Weinheim, 1988.
27. For a more detailed account, see J.K. Burdett, Chemical Bonding in Solids, Oxford
University Press, New York, 1995.
28. For example, see J.R. Hook and H.E. Hall, Solid State Physics, 2nd Edn., John Wiley and
Sons, New York, 1991, Chapter 3.
29. T.A. Albright, J.K. Burdett and M.H. Whangbo, Orbital Interactions in Chemistry, Wiley-
Interscience, New York, 1985.
30. Reference 27, p. 96ff.
31. N.W. Ashcroft and N.D. Mermin, Solid State Physics, Saunders, New York, 1976.
32. For example, see J. Goodisman, Electrochemistry: Theoretical Foundations, Wiley-
Interscience, New York, 1987.
33. H. Reiss and A. Heller, / . Phys. Chem., 89, 4207 (1985).
34. For a brief review see Reference 3.
172 The Solid State
35. For example see M. Causa and A. Zupan, Chem. Phys. Lett., 220, 145 (1994); idem, Int. J.
Quantum Chem., 56, 337 (1995).
36. J.R. Christman, Fundamentals of Solid State Physics, John Wiley and Sons, New York,
1988, Chapter 8.
37. R.G. Parr and W. Yang, Density Functional Theory of Atoms and Molecules, Oxford
University Press, New York, 1989, p. 77.
38. For example, see Reference 28, Chapter 6, and Reference 36, Chapter 14.
39. (a) D.R. Penn, Phys. Rev., 1311, 5179 (1978); (b) J.A. van Vechten, Phys. Rev., 182, 891
(1969); idem, ibid., 187, 1007 (1969).
40. J.C. Phillips and J.A. van Vechten, Phys. Rev. Lett., 22, 705 (1969); J.C. Phillips, Rev.
Mod. Phys., 42, 317 (1971).
41. J.K. Burdett, B.A. Coddens and G.V. Kulkarni, Inorg. Chem., 27, 3259 (1988).
42. N.F. Mott and M.J. Littleton, Trans. Faraday Soc, 34, 485 (1938).
43. R.T. Poole, J.G. Jenkins, J. Leisegang and R.C.G. Leckey, Phys. Rev., 1311, 5179 (1975).
44. See Reference 4, Chapter 6.
45. E. Burstein, A. Pinczuk and R.F. Wallis, The Physics of Semimetals, D.L. Carter and
R.T. Bate, Eds., Pergaon Press, New York, 1971, p. 251.
46. M. Aniya, / . Chem. Phys., 96, 2054 (1992).
47. M.D. Sturge, Phys. Rev., 127, 768 (1962).
48. E.C.M. Chen, W.E. Wentworth and J.A. Ayala, / . Chem. Phys., 67, 2842 (1977).
49. For a discussion of the boundary between metals and nonmetals, both in the elements and
in compounds, see P.P. Edwards and M.J. Sienko, Ace. Chem. Res., 15, 87 (1982).
50. K.S. Pitzer, Ace. Chem. Res., 12, 271 (1979); P. Pykko and P. Declaux, ibid., 276.
51. P.A. Cox, The Electronic Structure and Chemistry of Solids, Oxford University Press,
Oxford, 1987, Chapters 2 and 7.
52. See Reference 27, Chapters 2 and 4, for some examples.
53. J.K. Burdett and S. Lee, / . Am. Chem. Soc, 108, 3050, 3063 (1985).
54. R.T. Poole, D.R. Williams, J.D. Ridley, J.G. Jenkins, J. Liesegang and R.C.G. Leckey,
Chem. Phys. Lett., 36, 401 (1975).
55. A.R. Miedema, F.R. DeBoer, / . Phys. F., 3, 1558 (1973); R. Boom, F.R. DeBoer and
A.R. Miedema, J. Less Common Metals, 46, 271 (1976).
56. For reviews, see J. Koutecky and P. Fantucci, Chem. Rev., 86, 539 (1986); J.C. Phillips,
ibid., 86 619 (1986). M.D.Morse, ibid., 86, 1049 (1986).
57. P. Piquini, A. Fazzio and A. DalPino, Jr., Surf. Sci., 313, 41 (1994).
58. G. Pacchioni, S.C. Chung, S. Kruger and N. Rosch, Chem. Physics, 184, 125 (1994).
59. For a discussion, see J.A. Alonso and L.C. Balbas, Structure and Bonding, 80, 229 (1993).
60. W.A. de Heer, W.D. Knight, M.Y. Chou and M.L. Cohen, Solid State Physics, 40, 93
(1987).
61. S.B.H. Bach and J.R. Eyler, / . Chem. Phys., 92, 358 (1990).
62. T.P. Martin, T. Bergmann, H. Gohlich and T. Lange, J. Phys. Chem., 95, 6421 (1991).
63. M.K. Harbola, Proc. Natl. Acad. Sci. USA, 89, 1036 (1992).
64. K. Rhagavachari and C M . Rohlfing, / . Chem. Phys., 94, 3670 (1991); W. von Niessen
and V.G. Zahrzewski, ibid., 98, 1275 (1993).
65. For a review, see G.S. Hammond and V.J. Kuck, (Eds.), Fullerenes. Synthesis, Properties
and Chemistry, American Chemical Society, Washington, DC, 1992.
66. G. von Helding, M.T. Hsu, P.R. Kemper and M.T. Bowers, / . Chem. Phys., 95, 3835
(1991).
67. J.A. Zimmerman, J.R. Eyler, S.B.H. Bach and S.W. McElvany, / . Chem. Phys., 94, 3556
(1991).
References 173
68. N. Kurita, K. Kobayashi, H. Kumahora, K. Tago and K. Ozawa, Chem. Phys. Lett., 188,
181 (1992).
69. R.S. Becker and W.E Wentworth, J. Am. Chem. Soc, 85, 2210 (1963); R.S. Becker and
E. Chen, / . Chem. Phys., 45, 2403 (1966).
70. C.A. Coulson and R. Taylor, Proc. Phys. Soc A, 65, 815 (1952).
71. (a) E. Shustorovich and R.C. Baetzold, / . Am. Chem. Soc, 102, 5989 (1980); (b) J.Y.
Saillard and R. Hoffmann, ibid., 106, 2006 (1984).
72. For an excellent review of surface phenomena of all kinds, see G. A. Somorjai, Chemistry
in Two Dimensions, Cornell University Press, Ithaca, New York, 1981.
73. R.J. Hamers, / . Phys. Chem., 100, 13103 (1996).
74. For introductions, see R. Hoffmann, Solids and Surfaces, VCH, New York, 1988; G.A.
Somorjaj, Introduction to Surface Chemistry and Catalysis, Wiley, New York, 1994; D.W.
Goodman, / . Phys. Chem., 100, 13090 (1996).
75. (a) R.C. Baetzold, / . Am. Chem. Soc, 105, 4271 (1983); H.S.S. Sung and R. Hoffmann,
ibid., 107, 578 (1985).
76. P.C. Stair, / . Am. Chem. Soc, 104, 4044 (1982).
77. R. Hoffmann, / . Chem. Phys., 39, 1397 (1963).
78. Reference 37, p. 102.
79. M. Galvan, A. Dal Pino, Jr., J. Wang and J.D. Joannapoulos, J. Phys. Chem., 97, 783
(1993).
80. J. Tersoff and D.R. Hamann, Phys. Rev., 1331, 805 (1985).
81. B. Venkataraman, G.W. Flynn, J.L. Wilbur, J.P. Folkers and G.M. Whitesides, J. Phys.
Chem., 99, 8684 (1995); D.M. Cyr, B. Venkataraman, G.W. Flynn, A. Black and G.M.
Whitesides, ibid., 100, 13747 (1996).
82. J.E. Chadwick, D.C. Myles and R.L. Garrett, J. Am. Chem. Soc, 115, 10364 (1993).
83. L.H. Lee, New Trends in Physics and Physical Chemistry of Polymers, Plenum, New York,
1989.
84. J. Harris and S. Anderson, Phys. Rev. Lett., 55, 1583 (1985).
85. M.L. Cohen, M.V. Ganduglia-Pirovano and J. Kudrnovsky, Phys. Rev. Lett., 72, 3222
(1994).
86. A. Corma, F. Llakis and C. Zicovich-Wilson, / . Am. Chem. Soc, 116, 134 (1994).
87. See Chapter 3 for a discussion of EEM.
88. R. Dronskowski, Inorg. Chem., 31, 3107 (1992); / . Am. Chem. Soc, 113, 6730 (1992).
89. A. Dal Pino, Jr., M. Galvan, T.A. Arias and J.D. Joannopoulos, / . Phys. Chem., 98, 1606
(1993).
90. F. Mendizabal, R. Contreras and A. Aizman, Int. J. Quantum Chem., 56, 819 (1995).
91. J.E. Huheey, J. Org. Chem., 36, 204 (1971).
92. P. Politzer, / . Chem. Phys., 86, 1072 (1987).
6 Physical Hardness
INTRODUCTION
many defects, necessarily. The important defects for gliding motion are edge and
screw dislocations. These are the front (back) and side (left/right) edges of a
spreading region that is locally sheared.
A shearing stress will cause the dislocations to move, increasing the size of the
locally sheared region. For the front edge the way in which this happens is
illustrated in Figure 6.1. In projection, planes of atoms are shown as lines. Atom 2
in Figure 6.1 is at the center of the dislocation. By breaking the 3-6 bond and
forming a 2-6 bond, the dislocation line has moved to the right. The actual
movement occurs at "kinks" along the otherwise smooth line. Of course, it is
accompanied by small changes in atomic positions.
Minute amounts of added material can change the strength greatly. For
example, added carbon atoms in iron can act as dislocation traps and halt the
gliding motion. Work-hardening of metals is a process whereby many of the dis-
locations intersect and collide with one another, thereby becoming partially
immobilized. The movement of dislocations can be studied at various tempera-
tures and the activation energy found. At 1500°C plastic flow can be seen, even
in diamond.
Gilman has called attention to the similarity of kink movement to a simple
substitution reaction,3
A - B + C = ( A - B - Q * -> A + B - C (6.1)
3 - 6 + 2 = (3 - 6 - 2)* - + 3 + 6 - 2
(a) (b)
Figure 6.1 Movement of an edge dislocation, or kink, under the action of a shearing force
(indicated by arrows), (a) Original bonding; (b) bonding after kink movement. After Reference 2.
Introduction 177
LUMO LUMO
HOMO HOMO
Figure 6.2 Orbital correlation diagram for kink movement, viewed as a simple substitution
reaction. The width of the kink is w, and the reaction coordinate is in units of w. After
Reference 3.
Taking the 4-4 covalent solids as examples, Figure 6.3 shows a graph of the
glide activation energy plotted against the energy gap, E°g, or the HOMO-
LUMO gap. The correlation is remarkable and explains neatly why C and Si are
hard and brittle, whereas a-tin is almost malleable. The slope is exactly 2.00,
which is explained by realizing that the activation energy must include the energy
of formation of new kinks. These are formed in pairs and then must move apart.
The total energy needed is 4 x E°g/2 = 2E°g.3
Since we are taling about a solid withfilledvalence bands and empty conduction
bands, the disappearance of the gap means that the kink system has become
metallized by the shearing stress. Indeed, it is well known that many substances
become metallic at high pressures.4 When the atoms of a semiconductor, for
example, are brought closer together, the orbital overlap increases, the band
widths increase and the band gaps become smaller. If the band gaps become zero,
the substance becomes a metal.
The usual criterion is that the molar volume, V, must become equal to the
molar refractivity, R = ^nNoa, where a is the gas-phase polarizability. Then
the Clausius-Mosotti equation becomes
£ = OO (6.2)
e+2 V
The dielectric constant, e, is infinite for a metal.
The transition pressure at which semiconductors become metals has been
measured for a number of 4-4, 3-5 and 6-2 solids. Gilman showed that these
numbers are a linear function of the hardness numbers measured by the Vickers
test.5 This test measures the resistance to a compressional force, acting in one
178 Physical Hardness
CD
CD
CO
CD
2 4 6 8
LUMO-HOMO gap (eV)
Figure 6.3 Plot of the glide activation energy against the band gap, E°g, for the Group 14
elements. Reprinted with permission from J.J. Gilman, Science, 261, 1436 (1993). © 1993
American Association for the Advancement of Science.
direction only. This leads to a rather different view of the metallization process.
Instead of increased overlap being due to a shortening of bond lengths, it results
from a change in bond angles.
How this happens is shown in Figure 6.4.6 The semiconductors to which the
above remarks apply all have the four-coordinate tetrahedral structures char-
acteristic of covalent bonding. Applying a compressional force along the z-axis
will cause a tetragonal distortion, as shown. There is a volume decrease of about
25 percent, but the nearest-neighbor distances are almost unchanged. Next-
nearest neighbors are brought closer, however. Crystallographic data show that
0O = 109.5°, but that 6X = 149° and 62 = 94°.
Although we could say that there is effectively more overlap because of an
increased number of neighbors, there is an even more satisfying explanation.
Consider again the change from the tetrahedral structure of CH 4 to a planar one,
discussed at the beginning of Chapter 4. The energy of the HOMO increased and
the energy of the LUMO decreased, as we went from tetrahedral to planar. The
same changes in a tetrahedral solid, as shown in Figure 6.4, will cause E°g to
decrease. Simple Hiickel theory leads to a prediction that 9\ should equal 180° and
Introduction 179
(a)
(b)
Figure 6.4 Schematic drawing of the change from (a) the diamond structure of a-im to
(b) the structure of metallic f3-im. For clarity, five of the atoms in the unit cell have been drawn
larger than the others. After Reference 6.
62 should be 90° for the gap to become zero. But this does not include the effects of
the band width, which will make the required changes smaller, as observed.
The energies required to change bond angles and bond lengths in solids can be
estimated from force constants for these changes. The force constants come from
measurement of the elastic constants.7 As expected, it is easier to change bond
angles than bond lengths. Therefore metallization by bond-angle changes is
easier than by bond-length changes. The stresses needed, called shearing stresses,
are less than compressional (dilational) stresses. A portion of a solid that is
metallic is a region where local bonding has vanished, or can fluctuate readily.
Chemically, it is a highly reactive zone.
This leads to a number of interesing possibilities for new applications. One is the
interpretation of the detonations of solid explosives.8 These usually decompose by
180 Physical Hardness
the very rapid propagation of a detonation front. The reactions are too fast to be
thermally activated. Instead the high pressure in a front can cause local
metallization, allowing rapid chemical reaction. This can occur if Equation
(6.2) is satisfied, or if the HOMO-LUMO gap becomes zero by bond-bending.
A number of explosives fit the latter mechanism. These include lead azide,
ammonium nitrate and pentaerythritol nitrate.8'9 The critical strains needed to
initiate explosion may be estimated. As expected, the sensitivity of explosives to
shock can also be rationalized by estimating their initial HOMO-LUMO gaps,
or 2?g. For solids such as Pb(N3)2, the energy needed to excite an electron from
the valence band to the conduction band can be found by spectroscopy. A small
gap means high sensitivity.10
For organic solids the individual molecular properties are used. For explosives
containing aromatic rings, there is a correlation between the sensitivity and the
resonance energy.11 The smaller the resonance energy per electron (REPE), the
more sensitive is the compound. All these examples show the expected
relationship: the softer the solid (in chemical terms), the more reactive it is.
There are many other examples of chemical reactions being induced by shearing
stresses.9 A mechanism involving metallization seems plausible. Areas of appli-
cation include photochemistry, degradation of polymers, friction and wear,
mechanical alloying and cutting processes.
= BV
ait) =4r » (6-4>
dNJTV N2K
where Vo is the molar volume and TV is the number of molecules. The factor of
N2 is needed for dimensional purposes. B has the units of pressure and BVo
has the units of energy. If Vo is calculated per mole of atoms, then TV2 may
be ignored.
Because of Equation (6.4) Yang et al. have proposed that BV0 be called the
physical hardness, HP It has the units of energy, the same as chemical hardness.
Its reciprocal is the softness, proportional to K, as desired. Their proposal was
strongly reinforced by showing that BV0 for a number of substances followed
much the same ordering as the Moh hardness of those substances.
Table 6.1 contains experimental values of H = BVo for solids of simple type,
with cubic structures. For comparison, the values of the cohesive energy, Ai^oh,
are also included. The hardness numbers are of the same magnitude as the
energies of atomization. There is also a definite trend for H to increase as AjE^h
increases. However it is not very regular. For ionic solids, H is almost always less
than the cohesive energy. For covalent solids, including metals, H is always much
greater than AEC0^ except for Li.
We also see that W and Pt, and a few other noble metals, have H greater than
carbon. On this scale they are harder than diamond, which means that H no
longer matches the scratch test for hardness. This does not invalidate BVo as a
legitimate scale for hardness. It is well defined, has a thermodynamic basis, and
measures the resistance to well-defined changes. But for very hard substances, it
is not equivalent to the Moh scale, which measures plastic hardness. Actually B,
rather than BV0, matches the scratch test best.14
Though not apparent from Table 6.1, there is a relationship between B and
(6.5)
Li 38 38 Zn 128 31
Na 39 26 Ag 246 68
Cs 35 19 Cd 137 27
Mg 113 35 In 147 58
Al 173 78 Pb 180 47
Ca 107 43 W 790 202
Si 100 39 Pt 651 135
Cr 332 95 C 478 171
Mn 250 67 Si 282 109
Fe 280 99 Ge 270 90
Cu 235 81 Sn(0) 285 73
(a)
Data from References 13 and 21. Ai?COh for compounds should be divided by two to
compare with elements.
(6.6)
R
MZ2
(l-p/Ro) (6.7)
MZ2 ,
(6.8)
A Definition of Physical Hardness 183
U0 = -CQ~R^2P (6.10a)
= U0R20/\Sp2 (6.10b)
Equations (6.8) and (6.10(b)) show that there is a connection between H and
the cohesive energy. Unfortunately we cannot readily calculate H from A i ? ^
because we do not know p, except as calculated from Equations (6,8) and
(6.10(b)). The values of p found in this way increase as Ro increases, and are
about five to seven times smaller than Ro. The cohesive energy for covalent
bonding is equal to C/Q, but for ionic bonding we must correct C/Q by the energy
released when the separated ions change to atoms.
Equations (6.8) and (6.10(b)) come from the definition of B in Equation (6.3),
and the relations
The curvature (d2 U/dR2)T may be identified with the force constant,/, when U
is the potential energy function for a diatomic molecule. It has also been identified
as the single most important property determining the scratch hardness of a
solid.14 The relationship between the hardness, H, and the cohesive energy now
seems to be the same as that between the force constant of a diatomic molecule and
its dissociation energy, Do. When one increases, the other usually does also. But
there is no simple relation between them.16 Force constants depend more strongly
on the repulsive part of the potential energy function than bond energies do.
Equation (6.12) suggests that we try to calculate BV0 from the point of view of
force constants for the bonds undergoing compression. This has already been
done for cubic crystals by Pauling and Waser.17 Their procedure was to equate
the pressure-volume work done to the energy of compressing the bonds. Their
result was
(6.13)
184 Physical Hardness
10 5 / /
/ i s in millidynes/cm.
where n — m/2, the number of bonds per atom. The coordnation number m is
divided by two because each bond is shared by two atoms.
Equation (6.13) is completely consistent with Equation (6.12), since the total
change in energy depends on the number of bonds, as well as their force
constants. VQ is the volume per mole of atoms, as before. The force constants for
a number of solids have been calculated from Equation (6.13).17 Comparing
them with the force constants for the related diatomic molecules gives the results
in Table 6.2.
The force constant for a solid is only a fraction of that for the molecule. The
reason, of course, is that the average bond energy for the solid is only a fraction
of that for the diatomic case. For metals there are too many bonds and not
enough electrons. For ionic solids, the charges tend to cancel each other. Only
for the covalent 4-4 solids, such as carbon (diamond) and silicon, are the bonds
in the solid equivalent to those in small molecules. Actually C2 has a double
bond. The force constant for a single C - C bond is about 5 mdyn/cm, which is to
be compared with / for the solid. The 3-5, 2-6 and 1-7 covalent cases give
reduced bonding compared with the 4-4, as discussed in Chapter 5.
where /3 — XjkT. Since nj Vo = 1/7/, we can say that crystals which are physically
soft have large fluctuations in N. In this case the systems of the ensemble are
crytals of identical volume, but with varying numbers of component atoms. The
average value (N) is a constant, N°.
Using the proof of Chattaraj and Parr, as was done in Chapter 4, we can
conclude that the mechanical softness is a minimum, and the hardness is a
maximum, for the equilibrium state. This would apply to any condensed system,
solid or liquid. This is a very reasonable result. Chemical hardness is the
resistance to change in the electron distribution, and physical hardness measures
the resistance to change of the nuclear positions. An equilibrium system should
have the greatest resistance to change for both of these properties.
To test this prediction for solids, we must show that BV0 is a maximum for the
equilibrium state of the crystal. That is, 8(BVo) = 0, subject to certain
restrictions, such as constant volume and temperature. Another restriction is
made: to consider only cubic crystals containing a single element, or the binary
AB compounds. This is so that we can use the information in Equations (6.11)
and (6.12). We wish to prove that the equilibrium crystal has V= Vo, the
experimental volume per mole of atoms.
Differentiate i?F with respect to V, using the definition of B in Equation (6.3)
and the definition of P in Equation (6.11).
With the relation between V and R in Equation (6.11) this can be written as
-2*5
variation of the bulk modulus with pressure, (dB/dP)T. The experimental results
in the latter case are given either by a power-series equation
a2P (6.20)
\9P.
fdB\' fib
(622)
apj-w *'"*"
Experiments show that the modulus is indeed a linear function of the pressure,
over the range of pressures available. The slope is given by Equation (6.22). The
intercept is B$.
Pauling and Waser interpreted both BQ and B\ in terms of force constants.17 In
Equation (6.13) it is BQ that appears and gives a value for/. B\ then gives the
value of g, again equating pressure-volume work to the energy of compressing
the bonds. They found that20
g = (l-b/a2)f/2R0 (6.23)
Using the result from the PMPH, g = -f/Ro, we find (b/a2) = 3 and B\ = 5.0.
Therefore, for cubic solids, the pressure derivative of the modulus is equal to a
constant and dimensionless number, 5.0.
The value of a, the compressibility at very low pressure, is well known for many
solids. However, b is difficult to determine. Older, static methods are not reliable.
Better results are obtained by ultrasonic pulse methods on single crystals, but
even here different investigators can differ by 10-20 percent. Table 6.3 shows
The Principle of Maximum Physical Hardness, PMPH 187
(a)
See Reference 18 for sources of data.
(b)
At220A.
The average displacement, x = (R- Ro), would be zero. But the term in x3,
Equation (6.17), makes a positive value of x more probable than a negative
value. This is based on the Boltzmann distribution law, and a negative value
for g. A calculation gives
^'"'=S> (6-25)
In the last equality, we have used the maximum hardness result, —/=
188 Physical Hardness
3a Vp
7 = CyK, (6.26)
Table 6.4 Griineisen Constants for Some Covalent Solids at Room Temperature(£
Table 6.5 Griineisen Constants for Some Ionic Solids at Room Temperature(a)
(a)
D a t a from Reference 18.
a
f —I (6.27)
Ja l
~ ON
Fa is the force exerted on nucleus a by the electron cloud. The reactivity index is
also a nuclear Fukui function. Although it is completely general, Equation (6.27)
has been applied mainly to surface atoms.
The Hardness of Molecules 191
N,T
A large local value of h(r) means a large resistance to change of the shape of the
local electron density. It involves fluctuations in nuclear positions, e.g. in vibra-
tions. The averaging in Equation (6.28) is over an ensemble. Since the covariance
of // and v does not necessarily have a fixed sign, we cannot say that h{r) is a
maximum or minimum.
Equilibrium for a solid would also stipulate the shape, since the surface free
energy should be a minimum. For an isotropic crystal this shape would be that
of a sphere, but this is hardly ever a factor. The loss in energy for a surface
atom is about one-half of the cohesive energy per atom. But the number of atoms
on a typical surface is only about 1016, or 10~8 mol, so the surface energy is very
small. There is one observable effect, however; a collection of small crystals will
cohere to form larger crystals, if a mechanism, such as digestion of a precipitate,
is provided.
If we use Equation (6.17), and if the PMPH is valud, then (V/V) = -f/3g and
we can write
Note that this definition is also true for solids, if/ = -gRo-
Equation (6.30) is a reasonable result for molecular hardness. Certainly H'
should increase as the force constant increases, and decrease as \g\ increases.
Remember g, being negative, is a softening factor. While we cannot test Equation
(6.30) in the same way as for solids, we can calculate H1 for a number of diatomic
molecules, to see if they are in some way informative.
In place of equation (6.17), the potential energy of a diatomic molecule is
usually given in the Dunham formulation:29
The hardness can be written as H' = AfRl/9a\. The values of ao, ax and a2 can be
found from a detailed analysis of the rotation-vibration spectra of the molecule.
The experimental results are presented as u e , the vibrational frequency, BQ, the
rotational constant, ueXe, the anharmonicity constant, and ae the rotation-
vibration coupling constant. The subscript e refers to the ground-state or
equilibrium value.30
Formulas are available to convert these constants into ao, a\ and a2.31 We have
ao = / R Q / 2 , a\ = 2gRo/f, and a2 — 2hR^/f. The necessary constants have been
obtained for a very large number of diatomic gas-phase molecules, both stable
and unstable.32 Table 6.6 gives some typical results for H'.
Table 6.6 gives the values of Do, the gas-phase dissociation energies. It can be
seen that H' is very similar to Do: the correlation is quite good, though far from
perfect. Ionic rtiolecules have Do values larger than / / ' , and covalent molecules
have Do less than Hf, as a rule. Comparison of BeO and BeS with CaO and CaS
suggests that the former are much more covalent than the latter.
We may also compare the relationships between Do and H\ between Do and/,
and between Do and 77. Whereas there is a rough correlation in the latter cases,
that between the dissociation energy and the physical hardness is the best, at least
for covalent molecules. For example, F 2 and I 2 have nearly the same dissociation
energies, but the force constant for F 2 is 2.75 times as large as that for I2. Also, rj
for F 2 is 7.3 eV and that for I 2 is 3.8 eV, taking the vertical values.
The Hardness of Molecules 193
Ionic molecules
LiF 54 136 Csl 33 82
Lil 48 82 BeO 114 106
LiH 46 56 BeS 104 88
NaCl 41 98 CaO 81 110
KBr 35 90 CaS 82 80
Covalent molecules
H2+ 38 61 B2 50 70
H2 79 103 c2 141 143
BeH 66 54 N2 243 225
BH 65 79 o2 122 118
CH 74 81 F2 45 37
NH 79 86 ci2 73 57
OH 91 102 h 47 36
HF 103 136 Na 2 25 17
HI 78 70 HF+ 56 79
(a)
Data from Reference 32.
Compared with solids, we now have accurate values for the constant h in
Equation (6.17). The range of ax and a2 data is demonstrated by the examples in
Table 6.7. In the case of solids, ax would have the constant value of -2.00.
Since we know h, we can at least test to see if BV0 is a maximum or a
minimum. If(dH'/dV) = 0, then (d2H'/dV2) < 0, if we are at a maximum of H'
at V= VQ. This works out to require that
This is true in all the examples in Table 6.7, and appears to be true in general.
However, it should be borne in mind that there is no direct proof for the
assumption of maximum physical hardness for molecules.
Nevertheless, it seems very natural that, just as for solids, molecules should
have maximum chemical hardness and maximum physical hardness, at equi-
librium. Polyatomic molecules offer a more difficult problem, since there are
different force constants for the different normal modes. Also there are different
anharmonicity constants, which need not be negative.
A natural choice for the physical hardness of a moelcule would be the totally
symmetric breathing mode. This would correspond to the selection of the bulk
modulus for solids. In Chapter 4 it was shown that the Principle of Maximum
Chemical Hardness applied to antisymmetric vibrational modes of a molecule, but
not to the symmetric modes. It would be of interest to see whether the Principle of
Maximum Physical Hardness governed the symmetric vibrations of molecules.
194 Physical Hardness
REFERENCES
1. For an introduction, see any elementary physics textbook. For more detail relevant to the
present work see J.H. Wernick, Treatise on Solid State Chemistry, N.B. Hannay, Ed.,
Plenum Press, New York, 1975, Vol. 1, Chapter 4; E. Nembach, ibid., Vol. 2, Chapter 7.
2. For a good discussion of defects and their motion, see A.R West, Basic Solid State
Chemistry, John Wiley, New York, 1984.
3. J.J. Gilman, Science, 261, 1436 (1993); J.K. Burdett and S.L. Price, Phys. Rev. B, 25, 5778
(1982).
4. P.P. Edwards and M J . Sienko, Ace. Chem. Res., 15, 87 (1982).
5. J.J. Gilman, Czech, J. Phys., 45, 913 (1995).
6. J.J. Gilman, Phil. Mag. B, 67, 207 (1993).
7. W.A. Harrison, Electronic Structure and the Properties of Solids, W.H. Freeman, San
Francisco, 1980, p. 193ff.
8. J.J. Gilman, Phi. Mag. B, 71, 1057 (1995).
9. J.J. Gilman, in Metal-Insulator Transitions Revisited, P.P. Edwards and C.N.R. Rao,
Eds., Taylor and Francis, London, 1995, p. 269ff.
10. W.L. Faust, Science, 245, 37 (1989).
11. A.V. Belik, V.A. Potemkin and N.S. Zefirov, Dokl. Akad. Nauk. SSSR, 308, 882 (1989).
12. For example, see D.A. McQuarrie, Statisical Mechanics, Harper and Row, New York,
1976, p. 67.
13. W. Yang, R.G. Parr and T. Uytterhoeven, Phys. Chem. Mineral, 15, 191 (1987).
14. R.Y. Goble and S.D. Scott, Can. Mineral, 23, 273 (1985).
15. J.N. Plendl, S.S. Mitra and P J . Gielisse, Phys. Stat. Sol., 12, 367 (1965).
16. R.G. Pearson, / . Mol. Struct. (Theochem.), 300, 519 (1993).
17. L. Pauling and J. Waser, / . Chem. Phys., 18, 747 (1950).
18. R.G. Pearson, J. Phys. Chem., 98, 1989 (1994).
19. N. Dass and M. Kumari, Phys. Stat. Sol., Ill, 103 (1985).
20. In the notation of Pauling and Waser, / = k and —6g = k!.
21. G. Simmons and H. Wang, Single Crystal Elastic Constants, MIT Press, Cambridge, MA,
1971.
22. C.F. Kittel, Introduction to Solid State Physics, 3rd Edn., John Wiley and Sons, New
York, 1967, p. 183ff.
23. R.G. Pearson, Int. J. Quantum Chem., 56, 211 (1995).
24. I.C. Sanchez, J. Chu and W.J. Chen, Macromolecules, 26, 4234 (1993).
25. I.C. Sanchez, J. Phys. Chem., 97, 6120 (1993).
26. M.H. Cohen, M.V. Ganduglia-Pirovano and J. Kudrnovsky, / . Chem. Phys., 101, 8988
(1994); idem, ibid., 103, 3543 (1995).
27. B.G. Baekelandt, G.O.A. Janssens, W.J. Mortier, H. Toufar and R.A. Schoonheydt,
J. Phys. Chem., 99, 9784 (1995).
28. A. Cedilli, B.G. Baekelandt and R.G. Parr, / . Chem. Phys., 103, 8548 (1995).
29. J.L. Dunham, Phys. Rev., 41, 713, 721 (1932).
30. For a good discussion, see I.N. Levine, Molecular Spectroscopy, Wiley-Interscience, New
York, 1975, Chapter 4.
31. D. Steele, E.R. Lippincott and J.T. Vanderslice, Rev. Mod. Phys., 34, 239 (1962).
32. K.P. Huber and G. Herzberg, Constants of Diatomic Molecules, Van Nostrand Reinhold,
New York, 1979.
33. J.H. Rose, J.R. Smith and J. Ferrante, Phys. Rev. B, 28, 1835 (1983); A. Banerjea and J.R.
Smith, ibid., 37, 6632 (1988).
34. J.L. Graves and R.G. Parr, Phys. Rev. A, 31, 1 (1985).
Index