Hermitian operators

The operator P is defined as hermitian if its r,s matrix element has the property
P
r s

r
* P
s
d = (P
r
)*
s
d =
s
(P
r
)* d = [
s
*

(P
r
*
)]
*
d P
sr
*
In other words, the matrix elements related by the leading diagonal of P are complex
con!gates of each other"
#perators that are hermitian enoy certain properties" The Hamiltonian
(energy) operator is hermitian, and so are the $ario!s ang!lar moment!m operators"
In order to show this, first recall that the Hamiltonian is composed of a %inetic energy
part which is essentially
m
p
&
&
and a set of potential energy terms which in$ol$e the
distance coordinates x, y etc" If we can pro$e that the $ario!s terms comprising the
Hamiltonian are hermitian then the whole Hamiltonian is hermitian"
The distance coordinate 'operator( x
The coordinate x is an operator insofar as it can 'operate( on a f!nction f(x) to
prod!ce another f!nction x f(x), albeit !st a m!ltiple of the original one"
( )
* * * *
*
* * *
) (
sr r s s r s r rs
x dx x dx x dx x x

which shows that x is hermitian"
The moment!m operator p
The operator for the linear moment!m
)
in the x direction is
x i
p
x

" *et
!s s!ppose that o!r space is one+dimensional, along the x axis so that
x

can be
written
dx
d
" Its r,s matrix element is
[ ]

,
_

dx
dx
d
i
dx
dx
d
i
p
r s s r s r rs x
* * *
) (

)
,e ha$e already seen that the operator for %inetic energy is

,
_

+ +
&
&
&
&
&
& & &
&
&
& & & & z y x m m
p
m
p
m
p
z
y
x

, i"e" this is the %inetic energy hamiltonian for a
particle in free space"
)
The first term is -ero (wa$e f!nctions of finite systems $anish at t ) and the second
term is
*
*
1
]
1

dx
dx
d
i
r s

which is ( )
*
sr x
p " Therefore p
x
is a hermitian operator"
The Hamiltonian H
,e co!ld (b!t won(t.) extend the abo$e treatment for p
x
to show that its
s/!are p
x
&
is also hermitian so that the %inetic energy part of the Hamiltonian is
hermitian" If x is hermitian so also are y, z and r = (x
&
0 y
&
0 z
&
) and se$eral
f!nctions li%e
r
)
that depend on them and which appear as potential energy terms in
the Hamiltonian" 1ltho!gh this is not a complete arg!ment its res!lts strongly s!ggest
that the Hamiltonian operator is hermitian"
The bra and %et notation
It is con$enient now to !se the 2irac bra and %et notation to denote f!nctions"
To explain their form s!ppose we wanted to write down the wa$e f!nction for the &p
atomic orbital for which m
l
= )" Then instead of something li%e
&,),)
we write
) ) &
, which is called a ket" Then an eigen$al!e e/!ation in$ol$ing the f!nction

pa
with eigen$al!e pre$io!sly written as P
pa
=
pa
now becomes
P
pa
=
pa
The normalisation integral of the f!nction
pa
, i"e"
d
pa pa

3
*
is written
pa pa
where
pa
is the bra corresponding to the %et
pa
" The f!nction
represented by the bra is ta%en to ha$e been complex con!gated and associating the
bra and %et together in this way implies integration o$er all the whole range of the
$ariables" 4o the o$erlap integral S
ab

d
b a

3
*
is written as
b a
and the
matrix element V
ab

d V
b a

3
*
as
b V a
"
4!mmarising,

a

a
d
b a

3
*

b a
&
d V
b a

3
*

b V a
5roperties of hermitian operators
Their eigenvalues are real
To show this consider the eigen$al!e e/!ation in$ol$ing the hermitian
operator P
P
a
=
a
and m!ltiply from the left with the bra
a
6
a
P
a a a
beca!se
a a
= )" 7!t if P is hermitian the res!lt wo!ld be the same if we were to
interchange the bra and %et and ta%e the complex con!gate of the whole element6

=
a
P
a
=
a
P
*
a
*

4o =
*
, i"e" the eigen$al!e is real"
4ince we ha$e shown that the Hamiltonian operator is hermitian, we ha$e the
important res!lt that all its energy eigen$al!es m!st be real" In fact the operators of
all physically meas!rable /!antities are hermitian, and therefore ha$e real
eigen$al!es"
Their eigenfunctions are orthogonal
8onsider the pair of eigen$al!e e/!ations for the hermitian operator P
P
a
=
a
P
b
=
b
and s!ppose that the eigen$al!es are non+degenerate ( )" 9ow m!ltiply (from
the left) the first e/!ation by
b
and the second by
a
6
a b a P b
b a b P a
7!t if P is hermitian
a P b
=
b P a *
and so
a b
=
*
*
b a
=
*
b a
beca!se is real"
4o
a b
=
a b
beca!se
*
b a
=
a b
:
7!t as we ha$e stated that the eigen$al!es and are !ne/!al, the only way the last
e/!ation can hold is if
a b
= 3
i"e" the eigenf!nctions are orthogonal (
3
3
*

d
b a )"
What if
a
and
b
are degenerate?
If the eigen$al!es and are e/!al the pair of e/!ations is
P
a
=
a
P
b
=
b
1nd we define a %et
A
as a linear combination of
a
and
b
6
A
=
b c a c
& )
+
"
1pplying the operator to
A
we get
P
A
=
A b c a c a c a c b P c a P c + + + ) (
& ) & ) & )
showing that a linear combination of the degenerate states
a
and
b
is also an
eigenf!nction of P, with the same eigen$al!e " In /!ant!m chemistry we often ta%e
combinations of degenerate f!nctions to prod!ce f!nctions which are more con$enient
for the p!rpose
&
" 1s we ha$e seen, in this case
a
and
b
are not necessarily
orthogonal, b!t combinations li%e
A
and
B
can be constr!cted !sing coefficients
c
)
, c
&
etc" that ens!re the orthogonality of
A
and
B
"
&
;ecall the <4 co!rse where we made two combinations of the energy+degenerate complex &p atomic
orbital f!nctions &p0) and &p
) by adding and by s!btracting them to prod!ce the real orbital f!nctions
&px and &py"
=
The comm!tation of operators
There are many sets of physical /!antities (energy, moment!m, position etc")
describing /!ant!m particles which may not be simultaneously meas!rable, i"e" it may
not be possible to specify sim!ltaneo!s $al!es to more than one of them at a time"
They are defined by the $ario!s e/!ations describing the >ncertainty 5rinciple, b!t
!st which pairs of /!antities 1 and 7 are related by A B are shown by the
operators concerned" If a state

has exactly determinable /!antities 5 and ? which

ha$e operators P and Q so that
P
p
and Q
q
then the following theorem can be pro$ed6
Theorem6 1 system that is specified by a set of states
, ,
& )

has
sim!ltaneo!sly and precisely specifiable /!antities P and if and only if t!eir
operators P and Q commute i"e" P Q

Q P

, for all the wa$e f!nctions

"
There are two statements to be pro$ed6
(a) For P and Q to have a common set of eigenfunctions they must commute.
*et one of the common eigenstates be represented by
m
in the eigen$al!e
e/!ations
P
m p m
and Q
m q m
#perate on the first e/!ation with Q and on the second with P6
Q P
p m
Q
m pq m
P Q
q m
P
m pq m
Therefore P and Q comm!te, i"e" P Q = Q P"
(b) If P and Q commute they must have a common set of eigenfunctions.
This is the more !sef!l statement" 4!ppose that
m
is an eigenf!nction of P with
eigen$al!e p6
P
m p m
@
#perate on this e/!ation with Q6
Q P
p m
Q
m
7eca!se of comm!tation we can write it as
P Q
p m
Q
m
i"e" P [Q|m] = p[Q
m
]
This e/!ation implies that Q
m
is an eigenf!nction of P with eigen$al!e p" 7!t
the only way that this can be the case is if
m
is an eigenf!nction of Q" Therefore
m
m!st be an eigenf!nction of both P and Q operators"
1pplication in /!ant!m chemistry A 4ymmetry operators
Here is a simple Hamiltonian" It is for the H
&
molec!le with two n!clei a and
b and with two electrons ) and &, b!t a Hamiltonian for any atom or molec!le wo!ld
ha$e the same sort of terms"
H =
( )

,
_

+ + + +
ab b b a a
r r r r r r
e
m
) ) ) ) ) )
= &
)& & ) & ) 3
&
&
&
&
)
&

Bo! recall that the *aplacian operator is

&
)
&
&
)
&
&
)
&
z y x

for the first electron and

has a similar form for the second, with & replacing )" 9ow switching the n!clei a and
b and the electrons ) and & means performing a symmetry operation (call it P) on the
molec!le" 7!t notice that the effect of these interchanges lea$es H !naffected" 4o
if the prod!ct P H is applied to a wa$e f!nction we can write the operation as
P (H )= H (P)
with P bypassing H to act directly on " 7!t this means that P and H commute, and so
sho!ld ha$e common eigenf!nctions" ,hat does this meanC
4taying with o!r H
&
molec!le recall that if there were no symmetry we wo!ld
write its molec!lar orbital wa$e f!nction as
& & ) )
c c +
b!t when the $al!es of c
)
and c
&
are fo!nd that ma%e this an eigenf!nction of H by
applying the $ariation principle we get the pair
D
( )
& ) )
&
)
+
and
( )
& ) &
&
)

1s eigenf!nctions of the hermitian operator P we ha$e seen that
)
and
&
are
orthogonal" If yo! ha$e any do!bts on this, chec% it o!t6
E
)
P
&
F =
)
* P
&
d = G(
)
0
&
)* (
&

)
) d
= G[
)

&
d
&

&
] d
)
*
&
d 0
&
*
)
d ]
= G[ ) A ) A3 0 3]
= 3
7!t since
)
and
&
are also eigenf!nctions of H, matrix elements "
rs
that they
form with H will similarly be -ero6
E
)
H
&
F = 3
This is a $ery important res!lt A one which deser$es to be expressed $erbally6
If eigenf!nctions of H
)
and
&
are fo!nd which are sim!ltaneo!sly
eigenf!nctions of a comm!tinmg operator P that correspond to different ei#en$alues
of P then the matrix elements that they form with H are -ero"
This is the basis of the powerf!l methods of symmetry theory to red!ce the
labo!r of /!ant!m calc!lations and to symmetry+classify molec!lar wa$e f!nctions
and states" In the simple example ill!strated abo$e we don(t need to calc!late the
coefficients c
)
and c
&
by sol$ing a second+order

determinant, beca!se they ha$e been
obtained by ma%ing an eigenf!nction of P" Hor the same reason, the application of
symmetry arg!ments to the I#s of pentalene 8
J
H
D
in o!r <4 co!rse enabled !s to
a$oid sol$ing an J
th
order determinantal e/!ation and instead to sol$e separately one
:
rd
order, two &
nd
order and a )
st
order e/!ation (: 0 & 0 & 0 ) = J)"
8onsider what wo!ld happen if yo! tried to form matrix elements of H !sing

)
and
&
" <!st as & )
P
= & )
" )
which is -ero beca!se of the
orthogonality of
)
and
&
, so also & )
"
=
( )
& )
"
= 3 and the
Hamiltonian (energy) matrix is
K

,
_

+
%
%


3
3

i"e" the sec!lar determinantal e/!ation is
3
3
3

+
%
%


&!e matrix elements of H bet'een different ei#enstates of P are zero( The Hamiltonian
matrix has been bro!ght to a diagonal form, or 'has been diagonalised("
;ecall the application of symmetry in the I# treatment of pentalene
(<4 symmetry co!rse)
The I#s of the pentalene molec!le transform according to symmetries 1
)
, 1
&
, 7
)

and 7
&
in point gro!p )
&$
, and we showed that they were
(A
)
) = c
*
G(
*
0
+
0
,
0
-
) 0 c
.
G(
.
/
0
) 0 c
1
G(
1
/
2
)
(A
&
) = c
*
3 G(
*
4
+
0
,
4
-
)
(B
)
) = c
*
L G(
*
0
+
4
,
4
-
) 0 c
.
L G(
.
4
0
)
(B
&
) = c
*
L(G(
*
4
+
4
,
/
-
) / c
1
L(G(
1
4
2
)
1 symmetry orbital li%e (A
&
) does not interact with an orbital with different
symmetry li%e (B
&
)" In other words the matrix element E (A
&
) H (B
&
)F is -ero, as
we(ll demonstrate" In recognising the interactions between atomic orbitals at the
HMc%el le$el remember that all parameters for non+nearest neighbo!rs are -ero"
E (A
&
) H (B
&
)F
= [c
*
( G(
*
4
+
0
,
4
-
) H c
*
L( G(
*
4
+
4
,
/
-
) / c
1
(LG(
1
4

2
)]d
= c
)
(c
1
( (
)
N
=
)[ 0 A A 0 3 0 3 " " " ] 0 (
)
N
&&
) c
)
( c
&
L( [ 0 A A 0 3 "
J
0 3 " " "]
= 3
4o the H matrix elements between f!nctions corresponding to different symmetries
(here A
&
and B
&
) are -ero" It is for this reason that we can treat the fo!r symmetries
A
1
, B
1
, A
2
and B
2
separately"
If an operator can be fo!nd which comm!tes with operator H (e"g" Hamiltonian)
then any pair of non+degenerate eigenf!nctions
)
and
&
of prod!ce -ero matrix
elements of H"
i"e" if H ! H
and 5
i
= a
i
and 5
5
= b
5
then
3
*

d "
5 i
Angular momentum
7y considering the form of the operator for the z+component of the ang!lar
moment!m operator

i
l
z

it is possible to show that
z
l
(b!t not sim!ltaneo!sly
x
l
or
y
l
) comm!tes with the Hamiltonian H" 4o eigenf!nctions of z
l
are also
eigenf!nctions of H" 1s a res!lt, !st as the matrix elements of z
l
formed from
different eigenstates
r
and
s
of are -ero, so also are the corresponding matrix
elements of H" (i"e"
s l r
z
= 3 and
s " r
= 3)" ,e shall !se this property later"
Iatrix form of the 4chrOdinger e/!ation
)" 4ol!tions of the 4chrOdinger e/!ation
4elect a sol!tion to H = % which can be expressed as a linear combination
of a complete set of basis f!nctions P
i
Q6
= c
)

)
0 c
&

&
0 c
:

:
0 " " " ())
R
4ince the set P
i
Q6is complete the set Pc
i
Qof coefficients can be chosen to render so
flexible that it can constit!te a sol!tion (the k
th
, say) of the 4chrOdinger e/!ation for a
partic!lar system" Inserted into the e/!ation the latter wo!ld read
H
k
= %
k

k
(&)
There are n independent ways of combining the n f!nctions according to e/n" ()), so

k
is a member of n sol!tions li%e those in e/n" (&)" Hor the moment let !s consider
!st the k
th
sol!tion6 then e/n" ()) needs to be written bearing a label on the left hand
side and on the set of coefficientsP c
i
Q6

k
= c
)k

)
0 c
&k

&
0 c
:k

:
0 " " " 0 c
nk

n
(:)
The P c
i
Q are calc!lated !sing the Sariation 5rinciple (energy minimisation), which
entails sol$ing the sec!lar e/!ations"
&" The Hamiltonian matrix
In order to sol$e for
k
and %
k
in e/n" (&) the energy (Hamiltonian) matrix H is
constr!cted, whose general element is gi$en by "
i5
T
6
*

H
5
d" #b$io!sly the
elements of H depend on the basis f!nctions P
i
Q !sed to b!ild it" 1ny other set of
f!nctions, e$en if complete, wo!ld gi$e rise to different matrix elements "
i5
7 and
therefore to a different matrix H", altho!gh the energy eigen$al!es obtained from H
and H" wo!ld be identical"
%xample8 &!e allyl radical at "9ckel le$el
>sing the carbon &p
z
atomic orbitals as basis f!nctions, the energy matrix for
the allyl radical,
) & :
H
&
88H8H
&
is
is H

,
_

3
3
(@)
The U I# f!nctions and energy eigen$al!es for allyl are gi$en in the table6
k

k
%
k
) G (
)
&
&
0
:
) &
&
)
N
&
(
)

:
)
)3
:
G (
)
0 &
&
0
:
) 0 &
9ow instead of !sing the atomic orbital set (
)
,
&
,
:
) as basis f!nctions what if we
had !sed the molecular orbital set to constr!ct HC *et(s try6
"
))
= T
)
*

H
)
d = V T(
)
&
&
0
:
)* H(
)
&
&
0
:
) d
= V[ 0 & 0 & & & &]
= &
"
&&
= T
&
*

H
&
d = V T(
)

:
)* H(
)

:
) d
= G [ 0 ]
=
4imilarly,
"
::
= 0 &
"
)&
=
[ ] & &
& &
)
+
= 3
4imilarly,
"
)&
= 3 and "
&:
= 3
4o the H matrix generated by the I# basis set P
k
Q is

,
_

& 3 3
3 3
3 3 & +
H

(D)
1 comparison of the matrices defined in e/ns" (@) and (D) shows that the latter differs
from that in (@) by the fact that the non+-ero elements are confined to the diagonal
positions "
ii
, while "
i5
= 3 for i 5" This H is a diagonal matrix " >sing H gi$en by
(D) to form the sec!lar determinantal e/!ation prod!ces
3
& 3 3
3 3
3 3 & +

%
%
%

))
which shows that the diagonal elements in H are already the energy eigen$al!es %
)
, %
&
and %
:
that were listed in the table" 1 process which transforms a matrix li%e the one
in e/n" (@) into a diagonal form as in (D) is said to dia#onalise the matrix, thereby
re$ealing the eigen$al!es of the matrix in its diagonal positions"
:" Sector notation
The linear combination in e/n" (:) may be written as a scalar product of two
$ectors c
k
and where the !nderlining means that the /!antity is a $ector" If the basis
set defining
k
in e/n" (:) has n terms, the $ectors c
k
and are each rows or col!mns
of n elements, and allow an alternati$e expression for
k
as

k
= c
k

= c
)k

)
0 c
&k

&
0 c
:k

:
0 " " " 0 c
nk

n

To get the second of these e/!ations from the first we form the scalar prod!ct of the
$ectors c
k
and by s!pposing c
k
to be a ro' $ector and a column $ector, and then
!sing the r!les of matrix m!ltiplication" Hor clarity we(ll adopt the con$ention that a
$ector with elements (x
)
, x
&
, " " " ) will be written as x if it is a col!mn $ector, b!t as x
W
if it is a row $ector" 2oing this transposes a col!mn $ector x into a row $ector x
W
, and
transposing x
W
again to x
WW
(= x) restores the col!mn $ector" 9ote howe$er, that as
well as t!rning the col!mn into a row format, con$erting x into x
W
also replaces the
components x
i
by their complex con!gates x
i
*" ,e also define the transpose A
W
of a
matrix A as one in which the rows ha$e been swapped by the col!mns and $ice $ersa,
and also complex+con!gated" In other words if the general element of A is A
i5
, it
appears in A
W
as A
5i
*"
>sing this notation the abo$e k
th
state f!nction becomes

k
= c
k
W
(K)
=" Iatrix form of the eigen$al!e e/!ation" 5art )6 the eigen$ector matrix"
The energy %
k
of the state
k
is normally calc!lated $ia the sec!lar e/!ations,
which f!rnish both these /!antities, b!t if
k
is %nown, %
k
may be extracted from it by
calc!lating
%
k
= T
k
* H
k
d
)&
=
d c c
5 i
i 5
5k ik
H

* *
=
i5 5k
i 5
ik
" c c

*
4o H
i5

d
5 i
H

*
is the (i, 5)
th
element of a matrix H" 7!t !sing $ectorNmatrix
notation the same can also be written
%
k
= c
k
W
H c
k
(J)
%
k
=
8hec% that this expression follows the r!les of matrix m!ltiplication of the three
factors on the right of (J)6 the dimensions of the first (a row $ector) are ) n, the
second is a n n matrix and the third is a col!mn $ector with dimensions n )" The
res!lt is therefore the scalar (single n!mber) %
k
"
>ntil now we ha$e considered !st one sol!tion of the 4chrOdinger e/!ation
the eigen$al!e %
k
and the eigen$ector c
k
" 7!t can the whole set of sol!tions be
handled sim!ltaneo!slyC *et(s assemble all the eigen$ectors li%e c
k
into a matrix c,
whose order will be n n6

,
_

nn n n
n
n
n
c c c
c c c
c c c
c c c
" " "
" " " " " "
" " " " " "
" " "
" " "
" " "
& )
: :& :)
& && &)
) )& ))
c
(R)
The columns of c are the set Pc
k
Q, which are the eigen$ectors associated with the
energy eigen$al!es %
)
, %
&
, etc" The matrix c is called the ei#en$ector matrix, or
matrix of eigen$ectors, and it has some special feat!res"
@" 5roperties of the eigen$ector matrix
The eigen$ector matrix is an example of a type of matrix described as unitary"
1 !nitary matrix A has the following important properties6
)" The s!m of the s/!ares of the elements of any row or col!mn is one,
i"e"
)
&

5
i5
A
and
)
&

5
5i
A

&" The scalar prod!ct of any two rows or of any two col!mns is -ero,
):
i"e"
3
*

5
k5 i5
A A
and
3
*

5
5k 5i
A A
:" The transpose of the matrix is e/!al to its reciprocal,
i"e" A
W
= A
)
so that A
W
A = I ()3)
where I is the unit matrix which is a matrix of the same dimension as A, with all its
diagonal elements e/!al to !nity and all off+diagonal elements e/!al to -ero"

,
_

) 3 " " "

3 ) " " "
" " " " "
" " 3 ) 3
" " 3 3 )
I
1 !nit matrix I m!ltiplying any other matrix lea$es it !nchanged (I = )" If yo!
thin% of A (or c) as a matrix of *81# coefficients, properties ) and & follow from the
normalisation of the I# wa$e f!nctions and from the ort!o#onality of a pair of I#
f!nctions, respecti$ely"
=" The eigen$ector matrix c is s!ch a !nitary matrix" 1n important property
of c is its ability to diagonalise a matrix" This means that if the hamiltonian matrix H
is s!bected to a similarity transformation with its eigen$ector matrix c as follows6
c
)
H c = c
#
H c ! $
d
the res!lting matrix $
d
is dia#onal, i"e" its off+diagonal elements are -ero" Those
which are on the diagonal are the ei#en$alues %
)
, %
&
, %
:
, ( ( ( of H,
i"e"

,
_

&
&
)
3 3
3
3
3 3
%
%
%
%
The proced!re of finding the eigen$al!es of a H matrix is often referred to as
'diagonali-ing( it by finding a s!itable matrix c to perform the similarity
transformation c
#
H c.
)=