The Hartree-Fock method

Ivo Cacelli
Dipartimento di Chimica e Chimica Industriale, Università di Pisa
Via Moruzzi 13, 56124 Pisa , Italy

Summary:
1. Introduction
2. The Slater determinant
3. The Hartree-Fock equations and the Brillouin theorem
4. The Fock operator
5. The canonical Hartree-Fock spin orbitals
6. The Hartree-Fock-Roothaan equation
7. Solution of the Hartree-Fock equations
8. The Fock operator as the molecular Hamiltonian
9. Orbital energies and Koopmans’theorem
20. Restricted and Unrestricted Hartree-Fock

Keywords: indistinguishability principle, Slater determinant, Hartree-Fock, Koopmans’

theoerm

Abstract
Starting from the general context of the time-independent Schrödinger equation, the Hartree-
Fock method is presented in a didactic way in order non experts readers can understand some
subtle problems arising along the development of the theory. Emphasis is given to the accor-
dance of Slater determinants with the …rst principles of quantum mechanics and to rationalize
in a intuitive way some properties of the Hartree-Fock wave function. The projected equations
and the Koopmans’theorem are discussed in some detail and the restricted and unrestricted
treatments are brie‡y presented.

0.1 Introduction
The exact solution of the Schrödinger equation for an atom or molecule with more than two
electrons is a rather formidable problem. Within the Born-Oppenheiner approximation, the
di¢ culty is mainly due to the two electron terms of the Hamiltonian which introduce a correla-
tion between the motion of the electrons, in the …elds of the attractive potential of the nuclei.
For this reason one has to look for approximate solutions, whose accuracy is determined by
the observable one wants to evaluate as well as by the computer resources at disposal. One
of the most popular approximation is the Hartree-Fock method in which the wave function is
as simple as possible, provided in accordance with the principles of quantum mechanics and in
particular with the indistinguishability principle of identical particles. This principle imposes
that the electronic wave function must be antisymmetric with respect to the exchange of any
two electrons i.e. it must change its sign when the position of two electrons is interchanged.
The most simple function that satis…es this requirement is the Slater determinant built on a
set of one-electron functions called spin orbitals. A spin orbital is a function of the three spa-
tial variables and one spin coordinate, which describes the spatial-spin distribution of just one

1
electron. A spin orbital is the wave function of the one-electron systems as for instance, the
hydrogen-like atoms. In the case of many-electron systems the motion of each electron can no
more described by a single particle function which does not depends explicitly on the motion
of the other electrons and the attempt to describe such systems through spin orbitals leads to
approximate wave functions. However, from a number of spin orbitals a Slater determinant can
be formed that represents a wave function which, hopefully, is able to capture the main physics
of the system. The Hartree-Fock problem is just the search of the ‘best’set of spin orbitals, or,
by invoking the variational theorem, the set of spin orbitals which minimizes the energy of the
corresponding Slater determinant.
Now let’s consider the general context of the Hartree-Fock theory. The molecular electronic
Hamiltonian in atomic units has the following expression
X 1 2 X ZA X 1
Hel = r + (1)
i
2 i iA
riA r
i<j ij

where the three terms are the kinetic energy, nuclear attraction and electron-electron repulsion
operators, respectively. The indices i; j run over the N electrons of the system and the index A
over the nuclei. The electrostatic potential depends on the electron-nucleus distances riA and
the electron-electron distances rij . By collecting the one-electron terms, the Hamiltonian can
be written in a more compact form
X X
Hel = h(ri ) + 1=rij
i i<j

where it is evident that h includes kinetic energy, nuclear attraction operators. The electronic
Hamiltonian arises from the separation of the nuclear and electronic motion within the Born-
Oppenheimer approximation. In fact in the electronic Hamiltonian (1) the nuclear positions
are not dynamical variables, rather they are …xed parameters which a¤ect the electron-nucleus
attractive potential. Thus the eigenvalue of the time-independent Schrödinger equation

Hel (x; R) (x; R) = E (R) (x; R) (2)

is just a point in the multi-dimensional potential energy at a given nuclear arrangement. In

the above equation x collects all the spatial-spin electronic coordinates whereas the parameter
R indicates a given nuclear conformation and a¤ects the nuclear attractive potential in the
Hamiltonian.

0.2 The Slater determinant

For one-electron systems the solution of the time-independent Schrödinger equation is just a
function of the three spatial coordinates x; y; z and the spin coordinate s, ' (x; y; z; s). As the
electrons are fermions with spin 1=2, the possible spin functions are well de…nite: (s) and
(s) which correspond respectively to eigenvalues 1/2 and -1/2 of the spin operator projected
onto the z axis. The spin variable s is a formal coordinate which allows to write the integrals
involving the spin functions in the same manner as the spatial integrals. The physically relevant
quantities are the and functions which correspond to di¤erent intrinsic angular moment of
the electron along the quantization axis. This type of function ' (x; y; z; s) suitable to describe
the motion of a single electron is called ’spin orbital’.

2
 In the case of two-electron systems, a reasonable approximation to the exact antisymmetric
wave function, could be an antisymmetrized product of two spin orbitals 'a and 'b
1 1 'a (x1 ) 'a (x1 )
(x1 ; x2 ) = p ['a (x1 ) 'b (x2 ) 'b (x1 ) 'a (x2 )] = p (3)
2 2 'a (x2 ) 'b (x2 )
where the symbol j:::j stands for the determinant of the enclosed matrix and the argument x1
and x2 indicates the spatial-spin coordinate of the two electrons. The above formula expresses
the wave function in terms of a Slater determinant (SD) in which each column includes a
di¤erent spin orbital as a function of the coordinates of electron 1 in the …rst row and of
electron 2 in the second row. So, it is never possible to say that 'a describes the …rst electron
and 'b the second electron. The terms ’…rst electron’and ’second electron’have no meaning as
the electrons are indistinguishable and it is never possible to put a label on one speci…c electron.
In fact the spin orbital 'a appears two times in the expression: in the …rst term with x1 as
argument and the second time with x2 as argument. Moreover the function (3) obeys to the
antisymmetry principle as the exchange of the position of the two electrons (i.e. x1 exchanged
with x2 ) originates a change of sign of the entire wave function [1].
In the case 'a = 'b the two electrons would be in the same spin orbitals and the matrix
in (3) has two identical columns, so that its determinant is null for all possible values of the
spatial-spin coordinates. Therefore the wave function is everywhere null. This is the famous
Pauli principle which descends from the antisymmetry principle and says that two electrons
can never occupy the same spin orbital in a Slater determinant. In addition, when x1 =x2 the
two rows of the matrix (3) are identical and again the SD vanishes in accordance with the
antisymmetry principle of fermions which establishes that two identical particles can never be
found in the same point of the spatial-spin coordinates [2,3].
A generic Slater determinant for N electrons is
'1 (x1 ) '2 (x1 ) ::: 'N (x1 )
1 '1 (x2 ) '2 (x2 ) ::: 'N (x2 )
(x1 ; :::::; xN ) = p (4)
N! ::: ::: ::: :::
'1 (xN ) '2 (xN ) ::: 'N (xN )
which again is consistent with the principle of indistinguishability of identical particles. The
normalization factor accounts for the N ! permutations terms arising from the determinant. As
in the previous case, each electron is associated with every spin orbital and therefore there is no
one-to-one mapping between electrons and spin orbitals, rather the correspondence is all-to-all.
A Slater determinant can be speci…ed by the indices of its spin orbitals, without indicating
explicitly the electronic coordinates. In the above formula the spin orbitals entering the SD go
from 1 to N so that the short notation
j i = j'1 '2 '3 :::'N i (5)
indicates the (ket) SD (4). In the case we have at disposal a number of spin orbitals greater
than N , many SDs can be built and each of them can be speci…ed by a sequence of the N
indices of the involved spin orbitals. The properties of the SD for two electrons deduced above,
are still true for a generic SD: the SD is null when two spin orbitals are identical and when two
electrons are put in the same spatial-spin coordinate. Thus for instance
j'1 '2 '2 '3 :::'N i = 0 (x1 ; x2 ; x3 ; x4 ; x3 ; :::::; xN ) = 0 (6)

3
 Because of the two electron terms in the Hamiltonian, it is clear that a Slater determinant
can never be the exact solution of the Schrödinger equation for N >1. Only in the case of
one electron Hamiltonian’s (typical of independent particle systems) the exact solutions can be
written as a product (or an antisymmetrized product) of spin orbitals, but this is not the case
for real systems like atoms and molecules.
The general energy formula for any normalized wave functions is

E = h jHel j i (7)

which in the case is a SD assumes the expression [4]

X
N
1X
N
E ['1 :::'N ] = h jHel j i = h'i jhj 'i i + [h'i 'j j'i 'j i h'i 'j j'j 'i i] (8)
i=1
2 i;j=1

or in compact notation
X 1X
E= hi jhj ii + [hijjiji hijjjii] (9)
i
2 ij

where the one-electron integrals are

Z
hi jhj ji = dx 'i (x) h (x) 'j (x) (10)

and the two-electron integrals in Dirac notation are de…ned as

Z Z
1
hijjkli = dx1 dx2 'i (x1 ) 'j (x2 ) 'k (x1 ) 'l (x2 ) (11)
r12

The two-electron contribution to the energy in the expression (9) includes two distinct types
of integrals: the coulomb hijjiji and the exchange hijjjii terms. To better rationalize these
energy contributions we introduce the one-body density matrix arising from a generic wave
function
Z Z
0
(x1 ; x1 ) = N dx2 ::: dxN (x1 ; x2 ; :::xN ) (x01 ; x2 ; :::xN ) (12)

which depends on two spatial-spin variables, and for this reason can be thought as a matrix with
continuum indices [5]. The density function (x1 ) (x1 ; x1 ) corresponding to the diagonal
elements x1 = x2 has an important physical meaning, namely (x1 ) dx1 is the probability of
…nding any of the N electrons in the in…nitesimal spatial-spin volume dx1 around x1 . Thus,
whereas the spin orbitals and the density matrix (for x1 6= x2 ) are mathematical objects, the
density function is a physical object. When the wave function is a SD formed by N spin orbitals,
the density matrix assumes the expression

X
N
(x1 ; x2 ) = 'i (x1 ) 'i (x2 ) (13)
i=1

4
which may be used to rationalize the coulomb and exchange elements in the energy expression
(9). Indeed the coulomb terms in (13) can be written in terms of the density function
Z Z
1X 1 (x1 ) (x2 )
Ecoul = hijjiji = dx1 dx2 (14)
2 ij 2 r12

and represents the electrostatic interaction between the density charge with itself. As it in-
cludes electrostatic interaction between charge of the same sign, it is a repulsive term and its
contribution to the total energy is always positive. This expression is the same of classical
physics and there is no apparent signature we are dealing with a quantum system. The second
term is the exchange contribution that can be expressed in terms of the density matrix (not by
the density function)
Z Z
1X 1 j (x1 ; x2 )j2
Eexch = hijjjii = dx1 dx2 (15)
2 ij 2 r12

Its origin is in the antisymmetrized property of the SD and for this reasons it has no counterpart
in classical physics. Notice that since the integral is always positive, the results is always
negative and therefore the exchange energy is attractive. Although it is not easy to rationalize
this result by simple physical intuition, is the indistinguishability principle that originates the
exchange terms which are attractive for fermions and repulsive for bosons, because of their
di¤erent statistics. One key feature of the coulomb and exchange terms is that in the case i = j
in (9) the coulomb and exchange terms cancel to each other, with the remarkable consequence
of avoiding the non physical electrostatic interaction of an electron with itself. This is not an
accidental cancellation but results from the accordance of SDs with the principles of quantum
mechanics, although they are a form of simpli…ed wave function.
A further consideration about SDs concerns with the electron correlation. The two-body
density function whose general de…nition is
Z Z
2 (x1 ; x2 ) = N (N 1) dx3 ::: dxN j (x1 ; x2 ; x3 ; :::xN )j2 (16)

represent the simultaneous density probability of …nding any two electrons at points x1 and x2 .
Thus the probability of any one electron being found in the in…nitesimal spatial-spin volume
dx1 around x1 and one other electron in the in…nitesimal volume dx2 around x2 is

2 (x1 ; x2 ) dx1 dx2 (17)

In the case the appearing in the expression (16) is a SD, the two-body density function
assumes the simpler expression

2 (x1 ; x2 ) = (x1 ) (x2 ) (x1 ; x2 ) (x2 ; x1 ) (18)

which shows that the simultaneous probability concerning two electrons, is written as a product
of two probabilities both referred to one electron, with the addition of a second corrective term.
The latter accounts for the antisymmetrized nature of the SD, as in the case x1 = x2 the
function 2 (x1 ; x2 ) is null, meaning that there is no chance two electrons occupy the same spin-
space point, in agreement with the indistinguishability principle of identical fermions. This

5
feature is generally known as the Fermi correlation; so we may say that the SD includes Fermi
correlation, which concerns pair of electrons with the same spin. However from (18), neglecting
for a moment the corrective term, the probability of the simultaneous event of …nding two
any electrons at x1 and x2 is the product of the probability of two single events, which means
that the occurrence of one event does not a¤ect the probability of the other. In other word
the two events are independent to each other, which translated in quantum mechanics means
that the probability of …nding any two electrons at x1 and x2 is the product of probability of
…nding one electron at x1 times the probability of …nding one electron at x2 . For this reason
the SD does not contain the electron correlation and the methods concerning SDs are said
independent particle methods or even mean …eld methods. However the second term in (18)
make this consideration not true, because, owing to the non-classical antisymmetrized nature
of the wave function, classical probability theory is not completely adequate to treat quantum
systems. What is certainly true is that for a SD the two-body density matrix can be expressed
by quantities concerning the one electron density matrix and that, due to (6), two identical
fermions can never be completely independent.

0.3 The Hartree-Fock equations and the Brillouin theorem

For a Slater determinant the energy is expressed by integrals involving all the spin orbitals
which form the SD. At this point it is evident that the energy depends on the form of the
spin orbitals. But what spin orbitals? How can we …nd a criterion to choose the ’best’ spin
orbitals? Fortunately, the variational theorem comes in handy. It states that the mean value
of the energy of any physically acceptable wave function, as just the expression (7), is surely
higher than the exact energy of the system. Therefore the ’best’ spin orbitals will be those
that minimize the energy. Thus if we think the energy as a functional of the spin orbitals,
we require that the energy is stationary with respect to arbitrary variation of the spin orbitals
and that it increases for simultaneous arbitrary variation of any pair of spin orbitals. As the
second condition is rather cumbersome, the Hartree-Fock equations will be determined by the
condition of stationarity of the energy, although this does not guarantee that the energy is at
a minimum.
Thus we want to …nd N spin orbitals to form the ground state SD 0 which should have
the property that the energy is stationary with respect to arbitrary changes in the orbitals
'i ! 'i + 'i . Such a stationary condition must be looked for by imposing that the spin
orbitals remain orthonormal, i.e.
h'i j'j i = ij (19)
A stationary condition with some constraints can be conveniently treated with the method of
the undetermined Lagrange multipliers, by forming an additional functional de…ned as
X
['1 :::'N ] = E ['1 :::'N ] ji [h'i j'j i ij ] (20)
ij

where the ij are indeterminate Lagrange multipliers. It is evident that we have added to E a
function always null, in the case the spin orbitals are orthonormal. The change of the functional
1
with respect to arbitrary variations of the spin orbital 'k , which we suppose to be a real
1
A functional F of a function Rg(x) is p
a recipe to get a number from an integral which include g(x). For
instance in the expression F [g] = dx x2 g (x) we say that F is a functional of the function g (x) :

6
function, is
1X
= h 'k jhj 'k i + [h 'k 'j j'k 'j i h 'k 'j j'j 'k i] (21)
2 j
1X X
+ [h'i 'k j'i 'k i h'i 'k j'k 'i i] + jk h 'k j'j i + c:c: (22)
2 i j

where the variations of the bra spin orbitals are considered explicitly whereas those of the ket
spin orbitals are contained in the c:c:(complex conjugate) term which can be neglected since
it does not include additional information. The two sums arise from the case k=i and k=j. As
habjcdi = hbajdci, the two sums over two-electron integrals are identical and can be collected
by eliminating the 1=2factor
X X
= h 'k jhj 'k i + [h 'k 'j j'k 'j i h 'k 'j j'j 'k i] jk h 'k j'j i (23)
j j

For 'k completely arbitrary, the condition = 0 is satis…ed by the spin orbital 'k which
obeys to the following equation
XZ
h (x1 ) 'k (x1 ) + dx2 'j (x2 ) 'j (x2 ) 'k (x1 ) =r12 'j (x2 ) 'j (x1 ) 'k (x2 ) =r12 (24)
j
X
= jk 'j (x1 ) (25)
j

By de…ning the coulomb Jm and exchange Km operators

Z
Jm 'k (x1 ) = dx2 'm (x2 ) 'm (x2 ) 'k (x1 ) =r12 (26)
Z
Km 'k (x1 ) = dx2 'm (x2 ) 'k (x2 ) 'm (x1 ) =r12 (27)

this equation can be rewritten in a more compact form

" #
X X
h+ (Jm Km ) 'k (x) = jk 'j (x) (28)
m j

We notice that the left hand side of the above equation is just an operator which act on a spin
orbital, This operator is called the Fock operator
X
F =h+ (Jm Km ) (29)
m

and has the very advantageous feature that it does not depend on the spin orbital on which
it acts. It is just an one-electron operator that includes kinetic energy, nuclear attraction and
electron repulsion. A very important feature of the Fock operator is that it is an hermitian
operator, then its eigenfunctions form an orthogonal set and its eigenvalues are real. We will
discuss later of this operator in more details. Now we want to complete the derivation of the

7
Hartree-Fock equations. In order the functional is stationary for arbitrary variations, this
equation must hold for every spin orbital

X
N
F 'k = jk 'j 8k; j N (30)
j=1

where the Lagrange multipliers are deduced to be the matrix elements of the Fock operator

jk = h'j jF j 'k i (31)

which is hermitian [4]. This equation has to be read as: the stationary conditions are satis…ed
if the action of the Fock operator on each spin orbital yields a linear combination of the
same spin orbitals forming the Slater determinant. At …rst sight this does not seems a very
remarkable condition, but it is not so. Let’s assume to work in a functional space formed by
M orthonormal spin orbitals (with M >N ) and the …rst N form the optimal SD. For sake of
clarity in the following we use the indices i; j; k; l; m (going from 1 to N ) for the spin orbitals
that enter the Slater determinant and with a; b; c; d (going from N +1 to M ) the remaining spin
orbitals. The former may also be denoted as occupied spin orbitals whereas the latter as empty
or virtual spin orbitals. The indices p; q; r; s run over the full space and therefore go from 1 to
M . In this whole space the action of a generic one-electron hermitian operator T on a generic
spin orbital (included or not in the SD) yields a linear combination of the spin orbitals of the
whole space

X
M X
N X
M
T 'k = Trk 'r = Tik 'i + Tak 'a (32)
r=1 i=1 a=N +1

where Tr k = h'r jT j 'k i is a matrix element of the T operator. By comparing (30) with (32) it
is apparent that (30) expresses a clear condition on the occupied spin orbitals: the action of the
Fock operator on any occupied spin orbitals gives rise to a linear combination of spin orbitals
which must not include virtual spin orbitals. Therefore the stationarity condition implies that
the Lagrange multipliers between an occupied and a virtual spin orbital must vanish. Then the
condition to be satis…ed is

h'a jF j 'k i = 0 8k N 8a j N < a M (33)

This is the …rst form to express the Brillouin theorem. The spin orbitals entering the best
SD obtained with the Hartree-Fock method have to satisfy the Brillouin theorem. Notice that
condition (33) says nothing of de…nite about each virtual spin orbital, rather it states that
the action of the Fock operator on any element of the space spanned by the occupied spin
orbitals must remain whithin itself. The second form of the Brillouin theorem can be obtained
by noticing that in the full spin orbital space it is possible to build a large number of Slater
determinants other than 0 , simply by choosing a sequence of N spin orbitals out of the M
included in the space. The set of SDs obtained by substituting one occupied spin orbital 'k
with a virtual spin orbital 'a forms the class of single excited determinants and are denoted by
a a
k . It turns out that the Hamiltonian matrix element between 0 and k

a
h 0 jHel j ki = h'k jF j 'a i = 0 (34)

8
is just the Fock matrix element between an occupied and a virtual spin orbital, which is null
in the case the 'k are the Hartree-Fock spin orbitals. Therefore the second way to express the
Brillouin theorem is: the Hamiltonian matrix element between the Fock Slater determinant
and any single excited determinant is null.

0.4 The Fock operator

The Fock operator is the key object in the Hartree-Fock method and it is worthwhile to analyze
it in some details. A very important feature is that it is an hermitian one-electron operator.
Besides the kinetic energy operator and the nuclear attraction potential (29), it includes a sort
of electron-electron potential with a sum over the spin orbitals entering the SD. The coulomb
term J does not depend on the antisymmetrized form of the wave function and has a rather
simple physical interpretation. Eq. (26) may be rewritten as

Jm 'k (x1 ) = vm (x1 ) 'k (x1 ) (35)

where
Z Z
' (x2 ) 'm (x2 ) m (x2 )
vm (x1 ) = dx2 m = dx2 (36)
jr1 r2 j jx1 x2 j

is the electrostatic potential in the point r1 arising from the electronic density connected with
the spin orbital 'm . By summing over all 'm we obtain the total averaged electron-electron
potential in the point x1 . We can de…ne the global coulomb operator acting on a spin orbital
N Z
X 'm (x2 ) 'm (x2 )
J'k (x1 ) = dx2 'k (x1 ) = v (x1 ) 'k (x1 ) (37)
m=1
jr1 r2 j
Z
(x2 ; x2 )
v (x1 ) = dx2 (38)
jr1 r2 j

where now is the total one-body density function (13) arising from the N electrons in the
occupied spin orbitals of the SD. This potential v (x) is repulsive as it is everywhere positive
and it is the analogous of the nuclei-electron potential, which, on the contrary, is attractive.
As the J operator is de…ned in every point of the coordinate space, J is a local operator. Few
rows above we have used the adjective averaged in referring to v (x). Why? There is a subtle
consideration about this. The term averaged arises from the fact that this potential does not
account for the speci…city of the spin orbital to which F operates. The potential at the point
r1 is …xed and is independent from the density charge in that point. In other words the charge
distribution is …xed and a direct correlation between pairs of electrons is absent. In the exact
wave function such a …xed potential can not exist, although we may imagine a sort of potential
in the point r1 but its value should depend on the electron density in r1 . This aspect is typical
of the uncorrelated wave functions like SDs, as we discussed in the …rst section about the
two-body density function of a SD.
The term K in eq.s (27,29) is the exchange operator. We can de…ne a full exchange operator

9
acting on a spin orbitals as
N Z
X 'm (x2 ) 'm (x1 )
K'k (x1 ) = dx2 'k (x2 ) (39)
m=1
jr1 r2 j
Z
(x1 ; x2 )
= dx2 'k (x2 ) (40)
jr1 r2 j
Z
= dx2 K (x1 ; x2 ) 'k (x2 ) (41)

where now is the density matrix that appears within the integral. In the last formula the
K (x1 ; x2 ) function is the analogous of the kernel function in integral equations. This operator
arises from the antisymmetrized nature of the SD and, unlike the coulomb operator (37), it is a
non-local operator since it does not generate a simple potential at each value of the coordinate.
When it operates on 'k the integral in (39) involves all the occupied spin orbitals (including
'k ), so that the result depends on the value of 'k at every point of the space weighted with the
kernel two-variable function. The exchange operator can be also written using the P12 operator
which interchanges electron 1 and electron 2
N Z
X 1
K'k (x1 ) = dx2 ' (x2 ) P12 'm (x2 ) 'k (x1 ) (42)
m=1
r12 m

In such a way, the Fock operator can be written with a de…nite argument x1 irrespectively of
the function on which it acts
XN Z
1
F (x1 ) = h (x1 ) + dx2 'm (x2 ) (1 P12 ) 'm (x2 ) (43)
m=1
r 12

A further very important feature of the Fock operator concerns the combined action of the
coulomb and exchange operators. Let’s consider the action on the spin orbital 'k and write
the Fock operator in the form (29)
X
F (x1 ) 'k (x1 ) = h (x1 ) 'k (x1 ) + [Jm (x1 ) Km (x1 )] 'k (x1 ) (44)
m

In the case the summation index m is equal to k, the J and K terms cancel to each other

[Jk (x1 ) Kk (x1 )] 'k (x1 ) = 0 (45)

This conclusion is quite important, as these two terms would represent the interaction of the
electron in 'k (x1 ) with itself. These non physical terms are called self interaction terms and are
not present neither in the Fock operator nor in the expectation value of the energy, according
to the fact that the Hartree-Fock method does not violate the …rst principles of quantum
mechanics. According to (45) the expectation value of the Fock operator for the 'k spin orbital
X
h'k jF j 'k i = h'k jhj 'k i + [h'k 'm j'k 'm i h'k 'm j'm 'k i] (46)
m

includes both the coulomb and exchange two-electron integrals, which cancel to each other
when m=k.

10
0.5 The canonical Hartree-Fock spin orbitals
Now we want to explore the possibility of transforming eq. (29) in a eigenvalue equation,
without altering the stationary conditions. Let’s suppose we have found a solution of eq. (29)
which we rewrite in compact form
F j'1 :::'N i = j'1 :::'N i (47)
where j'1 :::'M i is a row ket vector collecting the N occupied spin orbitals and is a N N
matrix. As is a symmetric matrix (for real spin orbitals), it can be diagonalized through an
orthogonal transformation matrix U of dimension N N
U = U" (48)
where " is the diagonal matrix containing the eigenvalues. The matrix can then be written
as
= U"U ~ (49)
which may be substituted in eq. (47)
~
F j'1 :::'N i = j'1 :::'N i U"U (50)
Multiplying on the right by U
F j'1 :::'N i U = j'1 :::'N i U" (51)
an eigenvalue equation is obtained. A new set of spin orbitals connected to the '’s by a unitary
transformation can be de…ned
j'01 :::'0N i = j'1 :::'N i U (52)
which diagonalize F
F j'01 :::'0N i = j'01 :::'0N i " (53)
Thus the canonical Hartree-Fock eigenvalue equation are obtained
F '0k = "k '0k 8k = 1:::N (54)
that is equivalent to the stationary condition (30) because the Brillouin theorem is satis…ed
both by the f'g and the f'0 g sets. As demonstrated before, the Fock operator does not change
in passing from the set f'g to the set f'0 g so that with a little change of notation we may
write the canonical Hartree-Fock equation as
F ' k = "k ' k 8k = 1:::N (55)
where the superscript has been omitted and the set f'g are the canonical Hartree-Fock orbitals
i.e. those orbitals which are eigenfunctions of the Fock operator. The canonical Hartree-Fock
orbitals are unique in contrast to the in…nite number of orbital sets connected to them by a
unitary transformation and that describe the same SD, although they do not diagonalize the
Fock operator. The eigenvalues "k are the diagonal elements of the Fock operator and are called
orbital energies
X
N
hi jF j ki = ik "k = hk jhj ki + [hkmjkmi hkmjmki] (56)
m=1
as they represent the kinetic, nuclear attraction an electron repulsion energy of one electron in
the spin orbital 'k . The orbital energies will be discussed in more details later on.

11
0.6 The Hartree-Fock-Roothaan equation
The stationary conditions for a SD lead to the Hartree-Fock equations (28,55) which must be
satis…ed by the N optimal spin orbitals. The exact solution of this integro-di¤erential equation
will give the "exact" Hartree-Fock spin orbitals. In practical cases we are forced to project the
spin orbitals in a …nite set of basis functions [4,5] and the integro-di¤erential equations translate
into a set of matrix equations. In the case the basis set is complete the solutions are identical,
otherwise, for …nite basis sets, the '’s are a more or less good approximation of the "exact"
Hartree-Fock spin orbitals. We start with a set of M (with M >N ) spin orbitals r which do
not satisfy eq. (28) and form a basis for the projection of the Hartree-Fock spin orbitals. The
basis may or may not be orthogonal. Therefore the goal is to …nd a linear combination of the M
spin orbitals, such that the …rst resulting N spin orbitals do satisfy the Hartree-Fock equations
X
M
'k = Crk r 8k = 1:::N (57)
r=1

In this way the stationariety condition is transferred to the C coe¢ cients. The above equation
may be extended to the remaining spin orbitals 'a (with a from N +1 to M )
X
M
'a = Cra r 8a = N +1:::M (58)
r=1

It is clear that such spin orbitals have no relevance in the stationarity condition, but they have
to be always considered as it will become clear in the following, when the iterative solution of the
Hartree-Fock equations will be discussed. In order to preserve orthonormality the C coe¢ cients
have to form a unitary matrix i.e. we are dealing with a linear unitary transformation in the
space spanned by the basis of spin orbitals r . The above equations can be rewritten in a more
compact form using the Dirac bra-ket notation as
j'1 :::'M i = j 1 ::: Mi C (59)
where j'1 :::'M i is a row vector collecting the spin orbitals, j 1 ::: M i collects the basis functions
and C is the matrix of the coe¢ cients. The k-th column of the C matrix contains the coe¢ cients
Crk which give the spin orbital 'k from the basis function r , Crk = h r j'k i. Therefore the
goal of the Hartree-Fock method is to …nd the …rst N columns of the C matrix for such a
transformation.
The matrix form of the Hartree-Fock equations is obtained by projecting the eigenvalue
problem onto the f g basis. The starting equation is
F j'1 :::'M i = j'1 :::'M i " (60)
and after substitution of projection (59) we obtain
F j 1 ::: Mi C = j 1 ::: M i C" (61)
By taking the scalar product with the bra column vector formed by the f g basis
h 1j h 1j
::: F j 1 ::: Mi C = ::: j 1 ::: M i C" (62)
h Mj h Mj

12
we obtain the matrix equation

FC = SC" (63)

where Frs = h r jF j s i is the projected M M Fock matrix. The metric matrix S whose
elements are Srs = h r j s i is the overlap matrix between the basis functions. The projected
form of the equation is called Hartree-Fock-Roothaan equation.
In the case of non orthonormal basis sets the S matrix is di¤erent from the identity matrix
and complicates the eigenvalue equation. The pseudoeigenvalue problem (29) can be e¢ ciently
faced by orthogonalizing the basis set through a non unitary matrix X which transforms the
non orthogonal basis f g into an orthonormal basis f 0 g

j 0i = j i X

where the bold face j i and j 0 i collect the row of and 0 functions, respectively. Each column
of the X matrix represents a new functions 0 as linear combination of the original functions
. The orthonormality condition on the new basis set is
~ h j i X = XSX
h 0j 0i = X ~ =I (64)

A solution for the transformation matrix X can be found by working on the metric matrix S,
which is symmetric and can be diagonalized with an orthogonal matrix X

SV = VK (65)

where K is a diagonal matrix. Using the spectral decomposition theorem, the matrix S can be
expressed through its eiegensolutions
~
S = VKV (66)

Since for any reasonable basis set, S is positive de…nite [4], its eigenvalues are positive and the
matrix K can be expressed by

K = K1=2 K1=2 (67)

where the elements of the K1=2 matrix are just the square root of the corresponding elements
of the K matrix i.e. K 1=2 ii = (Kii )1=2 . We can also de…ne the inverse of this matrix, K 1=2
which contains as diagonal elements the (Kii ) 1=2 values. The two currently employed choices
for the X matrix are the following. In the Löwdin symmetric orthogonalization X = VK 1=2 V, ~
whereas in the canonical orthogonalization X = VK 1=2 . In both cases the relation (64) is
satis…ed and, most important, the overlap matrix can be factorized as the product of two
~ 1 =S, as can easily be veri…ed. By substituting the last expression in eq. (63)
matrices X 1 X
and applying some simple algebra we can obtain the desired result
~ 1 C"
FC = X 1 X (68)
~ 1 C"
XFC = X (69)
1
~ X
XFX ~ C=X ~ 1 C" (70)
F0 C0 = C0 " (71)

13
where the primed matrices refers to the orthonormal basis f 0 g and are de…ned as

F0 = XFX~ transformed Fock matrix (72)

~ 1C
C0 = X transformed eigenvectors (73)

The form (71) allows direct diagonalization of the transformed Fock matrix and the eigenvectors
C can be retrieved by
~ 0
C = XC (74)

Equation (71) is completely equivalent to eq. (63) (they have the same eigenvalues) and can
be solved by standard diagonalization methods.

0.7 Solution of the Hartree-Fock equations

Once a method to face the possible non orthogonality of the basis set has been found, we
can pass to the problem of solving the Hartree-Fock equations. The solution of the Hartree-
Fock equations require to determinate the C matrix in eq. (63) to be used in (59) to obtain the
Hartree-Fock orbitals projected in the basis f g. To be precise only the N vectors corresponding
to the occupied spin orbitals are necessary, but there is no real advantage in exploiting this
change and the whole set of eigensolutions of eq. (71) can be determined. The main di¢ culty
arises from the dependence of the coulomb and exchange terms from the density matrix, which
in turn is formed by the occupied spin orbitals. Thus the Fock operator depends on its solutions
f'g and the resulting non linear equation must be solved by iterative procedures.
n o
(0) (0)
Let’s suppose we are able to …nd a reasonable set of guess spin orbitals '1 :::'n using
our chemical intuition or performing a calculation with a certain semiempirical method, or
even only using the one-electron terms of the Hamiltonian instead of the Fock operator. The
corresponding density matrix allows to determine the coulomb and exchange matrix elements,
so the …rst Fock matrix can be built
(0)
F (75)

where (0) is the guess density matrix, built with the guess spin orbitals. By the matrix
transformation (72) with the S 1=2 matrix the F0 (0) matrix can be obtained for the …rst
iteration. Once diagonalization of F0 (0) and transformation (73) are accomplished, a new
set of M spin orbitals '(1) is obtained. These spin orbitals do not satisfy Brillouin theorem
(34) nor the stationarity conditions (30), so they are not yet the Hartree-Fock spin orbitals.
The reason is that they are eigenvectors of a Fock matrix built with di¤erent spin orbitals (the
guess spin orbitals)

'(1)
r F (0)
'(1)
s = "r rs (76)

but do not satisfy the same equation with the Fock operator they originate

'(1)
r F (1)
'(1)
s 6= "r rs (77)

Nevertheless the spin orbitals '(1) corresponding to the lowest N eigenvalues " can be used to
form a new density matrix (1) , which in turn allows to set up a new Fock operator F (1) .

14
This procedure can be repeated forming a new Fock matrix from the density matrix of the
previous step: F (n 1) . At each step n the …rst N spin orbitals '(n) will be a mixing of
occupied and virtual spin orbitals '(n 1) of the previous step. The procedure will continue
until convergence is reached i.e. when the density matrices at two consecutive steps di¤ers by
less than a given small threshold t
(n) (n 1)
<t (78)

and, correspondingly, the eigenvalue equation

'(n)
r F (n)
'(n)
s = "r rs (79)

is satis…ed within the same threshold t. When this occurs the Brillouin theorem is also satis…ed,
the current spin orbitals are the Hartree-Fock orbitals and the SD is the Hartree-Fock wave
function. The corresponding energy will be the lowest possible for a SD, within the given basis
set.

0.8 The Fock operator as the molecular Hamiltonian

The SD is the most crude approximation to the exact wave function which retains consistency
with the principles of quantum mechanics. We have seen that the spin orbitals to be used to
build the best SD are the lowest eigenstates of the Fock operator. The Hartree-Fock method
belongs to the class of the so called independent particle methods, which are characterized by
the fact that the spin orbitals are eigensolution of an e¤ective one-electron Hamiltonian. The
term ’e¤ective’means that each electron feels the nuclear attraction and a kind of e¤ective …eld
due to the remaining electrons of the molecule. In the Hartree-Fock method this one-electron
operator is the Fock operator, which has the rewarding property of being the same for all spin
orbitals. Its ’e¤ective’ form includes all the electrostatic Hamiltonian terms, adapted to the
case where the wave function is a SD. The Fock operator (43) is written just for one electron
but, as any physical operator, it may be written for N electrons

X
N
FN (x1 ; x2 :::xN ) = F (x1 ) (80)
i=1

and it is sometimes called the Fock Hamiltonian. Just as for all one-electron operators, it may
be demonstrated that each SD built with spin orbitals which are eigenfunctions of F (x1 ), is an
exact eigenstate of the operator (80). Indeed, if the orbital energies are arranged in increasing
order "i "i+1 , the eigenvalue of the Hartree-Fock SD is the sum of the orbital energies of the
…rst N spin orbitals
!
X
N
FN (x1 ; x2 :::xN ) 0 (x1 ; x2 :::xN ) = "i 0 (x1 ; x2 :::xN ) (81)
i=1

that is a general feature of all one-electron operators. Thus the Fock operator FN (80) can be
considered as the molecular Hamiltonian which is the best one-electron approximation of the
true Hamiltonian, for a system of N electrons in the ground state [6]. Moreover it shares with
the Hamiltonian the feature of possessing an in…nite number of eigenfunctions, that is a key

15
characteristic to be physically consistent. It should be noted that the energy of the system is
not simply the expectation value of FN since the eigenvalue of eq (81) are di¤erent from the
expectation value of the Hamiltonian
!
XN XN X
N
h jFN j i = "i = hi jhj ii + [himjimi himjmii] (82)
i=1 i=1 m=1
!
XN X
N
1
h jHel j i = hi jhj ii + [himjimi himjmii] (83)
i=1
2 m=1

The di¤erence is in the 1/2 factor, which appears in the correct energy mean value and avoids
the double counting of the electron-electron interaction, since the double sum of the orbital
energies includes the interaction between di¤erent spin orbitals twice. As the two-electron
repulsion term is always positive, the expectation value of the Hamiltonian is always lower
than the expectation value of the Fock operator. We conclude that the eigenvalue of the N -
electron SD has no relevant meaning being the energy the only relevant observable. Nevertheless
the molecular Fock operator is the best approximation of the true Hamiltonian.

0.9 Orbital energies and Koopmans’theorem

The eigenvalues of the Fock operator for the one-electron case are the orbital energies whose
expression has been given in eq. (56) for the occupied spin orbitals entering the Hartree-Fock
SD. Such an expression can be specialized for occupied and virtual spin orbitals
X
"k = hk jhj ki + [hkmjkmi hkmjmki] (84)
m
X
"a = ha jhj ai + [hamjami hamjmai] (85)
m

In the case of the occupied spin orbital 'k the term of the sum with m=k is null as the coulomb
and exchange integrals cancel to each other. This is relevant from a physical point of view since
it would represent the self interaction of one electron in a given spin orbital. Thus the orbital
energies of the occupied spin orbitals can be rewritten in a more physically consistent way
X
"k = hk jhj ki + [hkmjkmi hkmjmki] (86)
m6=k

where the self interaction term is omitted. The energy of one electron in an occupied spin
orbitals accounts for the one electron terms plus the coulomb and exchange interaction with
the N -1 electrons occupying the remaining spin orbitals. Conversely, no cancellation occurs
for the orbital energies of the virtual spin orbitals. They are determined in the …eld of all the
N electrons of the SD and feel a less attractive potential with respect to the corresponding
occupied spin orbitals.
Further insight in the orbital energies may be obtained by writing the coulomb and exchange

16
terms with explicit factorization of the spatial and spin parts.
Z Z Z Z
1
hkmjkmi = dr1 ds1 dr2 ds2 'k (r1 ) k (s1 ) 'm (r2 ) m (s2 ) (87)
r12
'k (r1 ) k (s1 ) 'm (r2 ) m (s2 ) (88)
Z Z Z Z
1
hkmjmki = dr1 ds1 dr2 ds2 'k (r1 ) k (s1 ) 'm (r2 ) m (s2 ) (89)
r12
'm (r1 ) m (s1 ) 'k (r2 ) k (s2 ) (90)
The functions can be or and due to their orthonormality, the integrals on the formal
spin variables are 0 or 1. The spin integration is irrelevant for the coulomb integral, whereas
the exchange vanishes in the case k 6= m . Thus we deduce a very general rule: the exchange
interaction only occurs between electrons with parallel spin i.e. with the same spin function.
Conversely the coulomb interaction is irrespective of the spin of the involved electrons. Thus we
may modify the comment about the orbital energy of an occupied spin orbital as follows: the
energy of one electron in an occupied spin orbitals accounts for the one electron terms plus the
coulomb interaction with the N -1 electrons occupying the remaining spin orbitals and minus
the exchange interaction with all the remaining occupied spin orbitals with the same spin.
As the Fock operator is an e¤ective operator whose relevance stands in its capability of
…nd the spin orbitals entering the best SD, it is not clear what is the physical meaning of its
eigenvalues i.e. the orbital energies. Moreover the possibility of performing unitary transfor-
mations among the occupied spin orbitals without a¤ecting the SD (except for a phase factor)
could make these quantities just as mathematical objects, as they are de…ned for canonical spin
orbitals. Nevertheless the Koopmans’ theorem give a rather interesting physical meaning to
the orbital energies.
Let’s consider a SD of N -1 electrons obtained from the Hartree-Fock SD by removing one
electron from the canonical spin orbital 'k without altering the spatial function of the spin
orbitals. The energy of these two SDs are
EN = N
0 H
N N
0 (91)
EkN 1
= N 1
k H N 1 N 1
k (92)
where we have added a superscript N or N -1 to indicate the number of electrons of the system.
Depending on the choice of the spin orbital 'k , the SD N k
1
may or may not represent the
ground state of the ionized system. In any case the energy of the ionized system can be obtained
by the general formula (8) by eliminating the k index in all the sums
X
N
1 XX
N N
EkN 1 = hm jhj mi + [hmijmii hmijimi] (93)
m6=k
2 m6=k j6=k

whereas the energy of the N electron SD can be written by expliciting the terms involving the
k-th spin orbital
X
N
1 XX
N N
N
E = hm jhj mi + hk jhj ki + [hmijmii hmijimi] (94)
m6=k
2 m6=k i6=k

1X 1X
N N
+ [hmkjmki hmkjkmi] + [hkijkii hkijiki] (95)
2 m6=k 2 i6=k

17
where the two-electron term with m=j=k has been omitted since it is null. The last two sums
are identical for real spin orbitals and can be collected in a single sum by eliminating the 1=2
factor Taking the di¤erence between these two energies

X
N
EkN 1
EN = hk jhj ki [hmkjmki hmkjkmi] (96)
m6=k

it is apparent that the right expression is just the orbital energy of the system with N electrons
and the left expression is an ionization potential of the N electron system

IPk = EkN 1
EN = "k (97)

This is the Koopmans’ theorem which establishes that the ionization potential of the system
obtained by removing one electron from the spin orbital 'k is the orbital energy of the same spin
orbital with the opposite sign. The orbital energies of the occupied spin orbitals are generally
negative so that the corresponding ionization potential is positive, in agreement with what we
expect at least for neutral molecules, where in the ionization process

A ! A+ + e (98)

requires some energy to bring one electron to an in…nite distance from the molecular ion.
It is clear that the Koopmans’theorem is an approximation to the exact ionization potential
since it includes several approximations. Let’s consider the usual case in which the N electron
system corresponds to a neutral molecule and the N -1 system corresponds to the cation. The
…rst approximation is that the spin orbitals used to build the ionized SD are the same used for
the neutral molecule (frozen orbital approximation) i.e they are not the optimal spin orbital
for the positive ion. Since the Hartree-Fock spin orbitals of the ion feel the repulsive …eld of
N -2 electrons and the attractive …eld of N protons, they are in general less di¤use than the
corresponding HF spin orbitals of the neutral species. In other words the Koopmans’theorem
neglects the relaxation e¤ects of the spin orbitals in the positive ion, so that "k is higher than
the di¤erence between the best SD of the ion and the best SD of the neutral system. Hence
the inclusion of relaxation would lead to a decrease of the ionization potentials. On the other
hand Hartree-Fock theory neglects correlation e¤ects other than Fermi correlation and, since
the correlation energy is larger for the neutral system, such correction would lead to an increase
of the ionization potentials. Therefore the corrections due to relaxation and correlation e¤ects
tend to cancel to each other and the Koopmans’estimate is reasonable, at least for the outer
shell spin orbitals. For inner shell spin orbitals the relaxation e¤ects are much higher and the
above cancellation does not yet occur. In these cases a reasonable estimate of the ionization
potential can be obtained by performing separate Hartree-Fock calculations of the ionized and
neutral species [4].
Finally we mention that by a similar procedure, a number of N +1 SDs can be formed by
putting one electron in a spin orbital which is empty in the N electron SD. In the case the
N electron SD represents a neutral system, these N +1 electron states represent the possible
anions. The corresponding electron a¢ nity of the system is the orbital energy of the involved
virtual orbital 'a with opposite sign

EAa = E N EaN +1 = "a (99)

18
where

EaN +1 = N +1
a H N +1 N +1
a (100)

In this case the frozen approximation and the neglecting of correlation e¤ects make the Koop-
mans’results rather inaccurate and the above formula is of little use.

0.10 Restricted and Unrestricted Hartree-Fock

In the above discussion we have de…ned the spin orbitals as functions able to describe the
motion of one electron, and consequently, they have as argument both the spatial and spin
coordinates of one electron. Since the spin functions are simple mathematical objects and the
nonrelativistic electronic Hamiltonian does not include spin operators, it is rather convenient
both from physical and computational points of views, to perform the integrals involving the
spin functions and write the equations in terms of spatial orbitals. Let’s consider the restricted
case characterized by the assumption that for each spin orbital with spin function there is a
corresponding spin orbital with spin function and with the same spatial function. Then the
spin orbitals can be grouped into pairs: each pair is formed by a spatial orbital times the and
the spin function. Such an approximation is suitable for closed-shell molecules characterized
by an even number of electrons distributed in pairs in the orbitals, so that each spatial orbital
contains zero or two electrons. Therefore if the spin orbital 'k (r) (s) is occupied by one
electron then it is also 'k (r) (s) and the same holds for unoccupied pairs. According to the
de…nition (5), the closed-shell restricted Hartree-Fock SD is

j i = '1 '1 '2 '2 :::'N=2 'N=2 (101)

where now the spin orbitals are written with explicit reference to their spin function and the
…rst N=2 spatial orbitals are doubly occupied. The restricted constraints are expressed by

'k (r) = 'k (r) (102)

It can be demonstrated that such a SD is eigenstate of the spin operators with null eigenvalue
i.e. it is a singlet state. This method is called restricted Hartree-Fock (RHF).
The general spin orbital Hartree-Fock equation can be converted to a spatial orbital equation
by performing integration on the spin coordinates. Let’s start with the Fock matrix elements
between two generic spin orbitals which are now written by explicit reference to the spatial and
spin functions
* +
X
N
h p p jF j q q i = p p h+ (Jm Km ) q q (103)
m=1

It is evident that since h has no reference with the spin, the …rst term can be factorized in two
distinct integrals over the spatial and spin coordinates
Z Z
h p p jhj q q i = dr p h (r) q (r) ds p (s) q (s) (104)

=h p jhj q i h pj q i (105)

19
where it is understood that the …rst two brakets have spatial and spin coordinate, respectively,
as integration variable. Owing to the orthonormality of the spin functions we obtain

hp ;q = hp ;q =h p jhj qi hp ;q = hp ;q =0 (106)

Each two electron integral can be factorized in three distinct integrals over the spatial and spin
coordinates and in particular for the coulomb integrals

h p p jJm j q qi = h p p 'm m j q q 'm m i (107)

Z Z
1
= dr1 dr2 p (r1 ) 'm (r2 ) q (r1 ) 'm (r2 ) (108)
r12
Z Z
ds1 p (s1 ) q (s1 ) ds2 m (s2 ) m (s2 ) (109)

= h p 'm j q 'm i h p j q i h mj mi (110)

It is apparent that the integral is null for p 6= q and that the spin of the occupied spin orbitals
'm is irrelevant i.e. the result is identical for 'm and for 'm . Therefore the total coulomb
potential is twice the coulomb potential arising form the (or ) occupied spin orbitals. This
con…rms that the coulomb interaction is not a¤ected by the spin and occurs for all pairs of
electrons. For the exchange terms the two-electron integral can be manipulated as before

h p p jKm j q qi = h p p 'm m j'm m q q i (111)

Z Z
1
= dr1 dr2 p (r1 ) 'm (r2 ) m (r1 ) 'q (r2 ) (112)
r12
Z Z
ds1 p (s1 ) m (s1 ) ds2 m (s2 ) q (s2 ) (113)

= h p 'm j'm q i h pj mi h mj q i (114)

and the results shows that the product between the two overlap integrals over the spin coordi-
nates is di¤erent from zero only in the case p = q = m . So we found again that the exchange
interaction occurs between electrons with parallel spin and that the exchange matrix elements
are null for p 6= q . Therefore in the case of restricted spin orbitals and closed shell systems,
the Fock matrix between two spatial orbitals is
* N=2
+
X
h p jF j qi = p h+ (2Jm Km ) q (115)
m=1

where it is understood that the integrals are on the spatial coordinates. There is no need to
specify the spin function of the orbitals p and q as the results in the same for both and
both . Thus the matrices involved in Hartree Fock eigenvalue problem (63) are formed by
two identical diagonal blocks; one of them is to be diagonalized and the eigenvectors are the
coe¢ cients of the spin orbitals irrespectively of their spin.
The energy expression can also be expressed in terms of spatial orbitals by performing spin

20
integration (we use here a more compact notation)

X
N
1X
N
E= hi jhj ii + (hijjiji hijjjii) (116)
i=1
2 ij=1
N=2 N=2 ; N=2 ;
X 1 XXXX
=2 hi jhj ii + (hi ; j ji ; j i hi ; j jj ; i i) (117)
i=1
2 i j
N=2 N=2 N=2
X X X
=2 hi jhj ii + (2 hijjiji hijjjii) (118)
i=1 i j

The last expression has been obtained by noticing that the coulomb terms are never null whereas
the exchange terms are di¤erent from zero only in the case = .
By removing the constraints (102) that each spin orbital has a partner with identical
spatial function, we obtain the unrestricted Hartree Fock (UHF) equations which are useful
for radicals, where the number of spin up and spin down electrons is di¤erent, or when the
molecule is placed in a magnetic …eld, which discriminates between spin up and spin down
electrons. In UHF the Fock matrix is always made by two diagonal blocks but, since they
are di¤erent, both of them have to be diagonalized. Therefore, from a computational point
of view, UHF method is more expensive than RHF. In the usual case where the Hartree-Fock
iterative procedure leads to the absolute energy minimum, the UHF energy is always lower or
equal than the RHF energy, since the former is subject to no constraint. An unpleasant feature
of the UHF wave function is that it is not an eigenfunction of the spin operators, since the
UHF SD contains some spin contaminants. However UHF is more ‡exible than RHF and it is
capable of describing in a qualitative way the dissociation of bonded atoms [4].

References
[1] P. O. Lowdin, Quantum Theory of Many-Particle Systems I, Phys, Rev. 97(1955) 1474
[2] G. C. Schaltz, M. A. Ratner, Quantum Mechanics in Chemistry Dover Publications,
Inc. Mineola, New York 2002
[3] C. Cohen-Tannoudji, B. Diu, F. Laloë, Quantum Mechanics vol.I Wiley, New York 1995
[4] A. Szabo, N. S. Ostlund, Modern Quantum Chemistry, Dover Publications, Inc. Mineola,
New York 1996
[5] C. C. J. Roothaan, New Developments in Molecular Orbital Theory, Rev. Mod. Phys.
23(1951)69
[6] R. McWeeny, Methods of Molecular Quantum Mechanics, Academic Press, London, 1992

21