NIELINIOWA OPTYKA

MOLEKULARNA
chapter 1
by Stanisław Kielich

translated by:Tadeusz Bancewicz

http://zon8.physd.amu.edu.pl/~tbancewi

Poznan,luty 2008
ELEMENTS OF THE VECTOR AND TENSOR ANALYSIS

Reference frames

The spatial properties of material systems usually are described

three-fold Euclidean space. In order to describe the physical
phenomena going in space the appropriate reference frame must be
chosen. In principle this choice is arbitrary and we select it in such
a way that the physical process is described univocally and the
most conveniently. Generally it is convenient, in Euclidean space,
to deal with the rectangular Cartesian reference frame. The
Cartesian reference frame Oxyz connected with the quiescent
material system we call the nieruchomy? reference frame and/or
the laboratory frame whereas the frame M123 connected to the
microsystem (atom or molecule) of this material system we call the
moving system and/or the molecular system. We will use only the
right angle reference systems (the direction of Oz in connection to
directions Ox and Oy defines regula sruby prawoskretnej.) The
components of a point in the quiescent frame Oxyz we denote by
xi where latin index i called the free index runs through x, y, z.
Similarly in the molecular frame M123 the components of a point
read x1 , x2 , x3 or briefly xα , where the greek symbol α runs over
the values 1,2,3.
Rysunek 1: The laboratory reference frame XYZ versus the
molecular reference frame 123
Vector calculus

Physical properties which magnitude is given completely by the real

number we named the scalar quantities or shortly scalars. The
examples of scalar values are: mass, temperature, volume,
potential, energy.
In physics occur also the vector values.
Rysunek 2: A spherical reference frame
Rysunek 3: A cylindrical reference frame
Rysunek 4: The angle between vectors r1 and r2
Rysunek 5: The Eulers’s angles
Cartesian tensors

In physics beside scalars and vectors we use tensor quantities. The

scalar is the zero order (rank) a tensor. We denote it by T (0) .
Previously described vector quantities we name the first rank
tensors T (1) and as we have shown they are fully described by their
components Tx , Ty , Tz along axes Ox, Oy, Oz of a selected
Cartesian reference frame. In index (or tensor) notation the first
rank tensor we denote by Ti , where the free index runs over x, y, z.
In the matrix notation the first rank tensor we write in the form of

a
We use expression tensor’s rank instead of order since the word order we will
use to subsequent terms of the perturbation calculus
one column (or one row) symbol
 
 Tx 
 

(Ti ) =  T  = Tx Ty Tz (1)
 y 
 
Tz
Rysunek 6: Polar and axial vectors
Rysunek 7: Colinear and anticolinear vectors
The second rank tensor T (2) we denote by the symbol Tij with two
lower subscripts i, j. In the matrix form it reads
 
 Txx Txy Txz 

 
Tij = 
 Tyx Tyy Tyz 
 (2)
 
Tzx Tzy Tzz
Rysunek 8: The sum of vectors
Rysunek 9: The projection of vector A on vector B
Rysunek 10: Multiplication of vectors A and B
Rysunek 11: The mixed multiplication of vectors bf A, B and C
 Rysunek 12: The vector A in Cartesian reference frame

The examples of the second rank tensors are: respectively the

electric and magnetic primitivity tensors ij and µij , the linear
electric polarizability tensor aij . Among the third rank tensors we
have: piezoelectric tensor tijk , the nonlinear polarizability
(hyperpolarizability) tensor Bijk etc.

Properties of second rank tensors

We discuss the transformation properties of tensor components

between two different reference frames. We define the first rank
tensor (the polar vector) as the quantity having, in respect to the
frame Oxi three components Ti , Eq.(1), transforming between
laboratory and molecular reference frames according to

Ti = ciα Tα (3)
 Rysunek 13: Transformation of vectors

In a similar way we define the second rank tensor as the quantity

having, in respect to the reference frame xi , 9 components, Eq.(2),
transforming in the following way

Tij = ciα cjβ Tαβ (4)

Let us remind that the indices i and j are free ones whereas the
indices α and β are the connected ones and we perform, according
to the Einstein convention, summation over the connected indices.
Then we transform Eq.(4) into its expanded form

Tij = ci1 cj1 T11 + ci1 cj2 T12 + ci1 cj3 T13 + ci2 cj1 T21
+ ci2 cj2 T22 + ci2 cj3 T23 + ci3 cj1 T31 + ci3 cj2 T32 + ci3 cj

Having in mind that for each pair of indices i, j both running over
x, y, z we obtain one equation, then finally we have 9 equations
with 9 components each. With this example we can see the
advantages of the tensor techniques being simultaneously simple,
compact and clear.
The inverse transformation of the second rank tensor from
molecular to laboratory reference system reads

Tαβ = cαi cβj Tij (6)

Tensor addition

We can add or substrate only tensors of the same rank and of the
same indices and defined in at the same point and the same
reference frame. By the sum two tensors Aij and Bij we mean the
third tensor Cij which components are the sums of the respective
adding tensors

Cij = Aij + Bij (7)

The tensor is symmetric when

Sij = Sji (8)

and is antisymmetic if

Aij = − Aji (9)

The symmetry and antisymmetry properties of the tensor are the

reference frame independent. Each tensor of the second rank Tij
we be univocally divided as a sum of the symmetric Sij and
antisymmetric Aij tensors

Tij = Sij + Aij (10)

where
1

Sij = Tij + Tij (11)
2
is the symmetric tensor, and
1

Aij = Tij − Tij (12)
2
stands for the antisymmetric tensor.
According to Eqs (11) and (12) the symmetric tensor in three
dimensional space posses 6 independent components: 3 diagonal
components Sxx , Syy , Szz and three off diagonal components
Sxy = Syx , Syz = Szy , Szx = Sxz . The Sij tensor can be written in
the form of the diagonally symmetric matrix
 
 Sxx Sxy Sxz 

 
Sij = 
 Sxy Syy Syz 
 (13)
 
Sxz Syz Szz

The Kronecker delta, δij , tensor serves as an example of the

symmetric tensor which we write in the form of the unit matrix
form
 
 1 0 0 

 
δij = 
 0 1 0 
 (14)
 
0 0 1

The antisymmetric tensor, in the Cartesian space, posses only 3 off

diagonal components: Axy = −Ayx , Ayz = −Azy , Azx = −Axz
since according to Eq(12) the diagonal components vanish. In the
matix form we write the antisymmetric tensor in the following form
 
 0 Axy Axz 

 
Aij = 
 −Axy 0 Ayz 
 (15)
 
−Axz − Ayz 0
Contraction of tensors

In tensor calculus we often use the operation of contraction of a

tensor. In this way we obtain a new tensor with its rank lowered by
two to the origin tensor rank. This operation is of a specific
tensorial character and has no equivalence in other arithmetic
operations. The contraction consists in the summation over the
pair of selected tensor indices. E.g. when within in the third rank
tensor Tijk we sum over the indices j and k the new first rank
tensor Tijj will result. In a similar way by contraction of its indices
from the second rank tensor Tij we obtain the zero rank tensor–in
other words scalar. In the case of the higher rank tensors
contraction can be done several times (manifolds?). When in the
n − th rank tensor the contraction is performed p–times as a result
we obtain the tensor of the rank lowered by 2p
(n) n−f old contraction
T −→ T (n−2) (16)

Now we are in a position to give the definition of the n-th rank

tensor in the d-fold space. By this tensor we understand the set of
dn components Tα1 ,...,αn , (αn = 1.2, 3, ..., d) connected to the
Cartesian reference sets in the way that by changing the set they
transform by the rule given by Eq(??).

Contraction of the product of tensors

With the use of the unit Kronecker tensor δij in its matrix form,
Eq.(??), we easily note that the one-fold contraction is equivalent
to multiplying of the tensor by the unit tensor δij

Tij δij = Tii (scalar)

Tijk δjk = Tijj (vector)
Tijkl δkl = Tijkk (second rank tensor)
Tijkl δij δkl = Tiijj (scalar) (17)

We carry on contraction not only on tensors but also on products

of tensors. E.g. let us take into account the second rank tensor Tij
and the third rank tensor Tijk ; the product Tij T klm is the fifth
rank tensor. When now we carry on the contraction over pair of
indices, let’s say j and m we obtain the third rank tensor

Tij Tklm δjm = Tij Tklj = ?Tikl (18)

The above process we call the contraction of the product of tensors
Tij and Tklm over the indices j and m.
In the general case when we have the product of the tensors of
ranks r and s T (r) T (s) = Ti1 ,...,ir Tk1 ,...,ks than by one–fold
contraction over arbitrary pair of indices we obtain the new tensor
of the rank r + s − 2. Generally a tensor of the rank r + s over the
p–fold contraction gives the new tensor of the rank r + s − 2 p
(s) p−f old contraction
T (r)
T −→ T (r+s−2 p) (19)

The product of the type of Eq.(18) based on the combination of

the external tensor product and contraction we sometimes call the
internal tensor product.
The trace of tensor

Using ono–fold, two–fold etc. contraction of the tensor of a given

rank leads to tensor invariants or tensor traces, e.g.

Tr(Tij ) = Tij δij = Tii

Tr(Tijkl ) = Tiijj etc (20)

In particular the trace of the unit Kronecker tensor is 3.

Tr(δij ) = δij δij = δ11 + δ22 + δ33 (21)

Let’s focus of our attention on another property of the unit

Kronecker tensor δij , which we call the property of replacement
(substitution) of indices

Ti δij = Tj , Tij δjk = Tik

Tijk δkl = Tijl
δij δjk = δik , etc. (22)

where in the process of calculating the product of Ti and δij the

index i has been substituted by the index j.
The symmetric second rank tensor, Eq.(11), can be written in the
form of a sum

Sij = Kij + Dij (23)

where
1
Kij = Skk δij (24)
3
is the spherical tensor or the isotropic tensor given by the trace of
the symmetric tensor Skk = rmT r(Skl ) and the unit tensor δij ,
whereas the tensor
1
Dij Sij − Skk δij (25)
3
we call the deviator or anisotropic tensor. We can easily prove that
the trace of the anisotropic tensor vanishes

Tr Dij = Dii = 0 (26)

since the trace Tr δij = δii = 3.

Symmetrization of tensors

Previously we have shown that the symmetric part of the second

rank tensor can be written in the form of Eq.(11). We note that it
is the arithmetic mean value of the tensors Tij and Tji resultintg
from two possible interchange of the indices i and j. This
symmetrization operation we can extend to the tensor of an
arbitrary rank and in respect to all its indices as well as to some
selected group of indices. When we intend to symmetrize the
tensor in respect to s its indices then we perform s! their
permutations and next we take the arthmetic average of all the
tensors obtained in this way. When the indices under
symmetrization form a compact group then the result of the
symmetrization we write in the round parenthesis, e.g. (ijk). In
the other case we can use the symbol (ij|kl|m), where the indices
between the vertical lines are exempted from the symmetrization.
In the case of the second rank tensor, s = 2, the symmetrization is
given by Eq.(11), whereas in the case of the third rank tensor the
symmetrization process in respect to the indices j and k reads
1  
Ti(jk) = Tijk + Tikj (27)
2
When we intend to symmetrize the third rank tensor over its all
three indices, s = 3, then we have 3! = 6 permutations and then
instead of Eq.(27) we obtian
1  
T(ijk) = Tijk + Tjki + Tkij + Tjik + Tikj + Tkji (28)
6
In a similar way we perform the symmetrization over arbitrary
number of indices in result obtaining the symmetric tensor over the
indices active in a process of symmetrization.

We call the tensor the totally symmetric or just symmetric

according to some number of its indices when it does not change
within these indices interchange.

Antisymmetrization (alternation) of tensors

In the case of the second rank tensor the result of its

antisymmetrization is given by Eq.(12). The process of
antisymmetrization of a tensor in respect to its s indices goes in
the following way: we perform s! permutations on these indices
and in the case of even (cyclic) permutations we leave the sign of
the tensor unchanged whereas in the case of odd permutations we
change its sign. Finally we calculate the mean arthmetic value over
all s! terms. Usually the indices involed in the antisymmetrization
process we write in the square parenthesis [...]. For s = 3 the
antisymmetrization of the third rank tensor Tijk gives
1  
T[ijk] = Tijk + Tjki + Tkij − Tjik − Tkji − Tikj (29)
6
We call the tensor antisymmetric in respect to some indices when
it changes the sign after the interchange of an arbitrary pair of
these indices (then it changes the sign with respect to the odd
number of their permutations and leaves the sign unchanged to
even number of permutations). According to these properties for
the antisymmetric second rank tensor we have

Aij = − Aji (30)

whereas for the third rand tensor we obtain

Aijk = − Ajik = Ajki = − Akij = Akij = − Aikj (31)

It is easy to note the the trace of the second rank tensor vanishs

Tr (Aij ) = Aij δij = Aii = 0 (32)

We can separate the tensor on its symmetric and antisymmetric
part, e.g. for the second rank tensor we have

Tij = Sij + Aij = T(ij) + T[ij] (33)

In a similar way we separate the third rank tensor

2 2
Tijk = T(ijk) + T[ijk] + (T[ij]k + T[kj]i ) + (T(ij)k − Tk(ji) )(34
3 3