Symmetry of Properties

 Various properties:
 Mechanical, Electrical, Magnetic, Thermal, etc.
 Tensorial nature of properties.
 Symmetry imposed restraints.
MATERIALS SCIENCE
Part of & A Learner’s Guide
 Neumann Principle ENGINEERING
 Curie Principle AN INTRODUCTORY E-BOOK

 Illustration of principles using Anandh Subramaniam & Kantesh Balani

Materials Science and Engineering (MSE)
Thermal, Electrical and Elastic properties Indian Institute of Technology, Kanpur- 208016
Email: anandh@iitk.ac.in, URL: home.iitk.ac.in/~anandh
http://home.iitk.ac.in/~anandh/E-book.htm
Advanced Reading
 Physical Properties of Crystals: Their Representation by Tensors and Matrices.
J.F. Nye
Clarendon Press, Oxford, 1985.
 Properties of Materials: Anisotropy, Symmetry, Structure.
Robert E Newnham
Oxford University Press, Oxford, 2005.
Overview of Properties For now we concern ourselves with properties of crystals
 For every stimulus there is a natural response. E.g.:
 when electric field is applied to a dielectric material it can lead to electric polarization,
 when stresses develop inside a material, it can lead to strains,
 application of magnetic field can lead to magnetization.
 In addition there can be cross-coupling terms, i.e. one kind of stimulus can lead to response*,
which is not naturally associated with it. E.g.:
 application of pressure[1] can lead to electric polarization (piezoelectricity),
 presence of temperature difference can cause electrical potential difference
(pyroelectricity),
 the electrical resistance is altered in the presence of a magnetic field (magnetoresistance).
 A stimulus can be connected to a response via a material property. E.g.:
 M = .H,   = E.,  P = p.T.
The material property has to be measured via experiments (or computed from more
fundamental properties).
 In reality the stimulus and response are tensors** (of various orders) and hence the material
property is also a tensor of some rank (i.e. many values may be required to describe the property). The equations
(as above) in tensorial form become:  Mi = ijHj,  ij = Eijklkl,  Pi = pi T.

[1] The correct phraseology should be: “application of forces/constraints”. The application of forces can lead to stresses in the material, which
can further cause electric polarization. Actually, pressure is a parameter better used in the context of fluids (and stress, a tensor, a better
quantity for solids). The hydrostatic component of stress is “like the pressure”.
* There could be more than one response. ** More on this soon.
Classification of material properties Overview of Properties (Cotd.)

Polar tensors (rank 0-4)

Axial tensors (rank 0-4)

Material Properties
Transport properties  Relate flow to a gradient

Properties with hysterisis  Involve domain wall motion

Examples of 1st to 4th ranked tensor properties The number of subscripts denotes the order of the tensor
Phenomena Cause Effect Coefficient Equation
Pyroelectricity T Pi pi Pi = pi T
(Temperature difference) (Electric polarization) (1st rank tensor)
Thermal Expansion T ij αij ij = αijT
(Strain) (2nd rank tensor)
Piezoelectricity ij Pi dijk Pi = dijkjk
rd
(Stress) (Electric polarization) (3 rank tensor)
Elasticity kl ij Eijkl ij = Eijklkl
(strain) (Stress) (4th rank tensor)
“Cross-coupling” effects & Heckmann diagrams
 A stimulus can lead to effects, which are not naturally associated with it.
 E.g. stress can lead to electric polarization (piezoelectricity).
 The “natural” effects and “cross-coupling” effects can be pictorially represented using
Heckmann diagrams (as below).
Scalar, Vector and Tensor Quantities
 To describe a property at a point inside the material we may require to specify just:  a
number (magnitude of that property like temperature or density),  a magnitude and direction
(i.e. 3 numbers in 3D like for electric field or pyroelectric coefficient) or  even more
numbers. The number of values required is given by 3n in 3D (2n in 2D); where ‘n’ is called
the rank of the tensor.
 To understand why tensors are required let us consider a force ‘F’ (with a F) applied along
the x-axis and ask the question*: “what will happen to the body?”.
 Clearly the information available is insufficient to answer unequivocally. If the forces are
applied as in Fig.1b the body will elongate, while if applied as in Fig.1c the body will shear.
 Hence, we need to specify the plane on which the force is acting. In case in Fig.1b the +F
force is acting on the +x-plane along the +x direction this is written as Fxx. In Fig.1c the
force +F is acting along the +y-plane along the +x direction written as Fyx.
 This implies (in this example of mechanical deformation) two directions are required to specify the force
(and hence determine its effect):
(i) the direction of the plane normal and (ii) the direction of the force.
 A combination of these forces Fxx & Fxy could also be acting on the body.

Note. For tension/compression: +F on +x F Fyx

F F F face is positive and similarly, F force on x
Fig.1a Fig.1b plane is also positive. For shear: shear Fig.1c
F Fxx causing clockwise rotation is positive.
F
* For now we will assume that other forces are present to give us force and moment balance (i.e. equilibrium condition).
 Q&A Give examples of axial and polar vectors.

 Polar vectors: Force, Electric field, Polarization.

 Axial vectors: Mechanical moment, Angular momentum, Curl of a polar vector.

Some basic points...

 A scalar does not require a coordinate system to define and hence is independent of the
coordinate axes chosen. We require just one number at each point for a scalar.
 A vector can be represented by its components along a set of coordinate axes.
We require three numbers in 3D at each point to specify a vector.
A change in the coordinate axes system (change in the angle between the basis vectors,
translation/rotation of the basis vectors) will change the components along the axes; but, will
the vector itself will remain unchanged.
 The transformation of a vector (rotation, inversion, mirror, etc.) can be carried out using a
transformation matrix.
 A second rank tensor can also can be represented by its components along a set of coordinate
axes. It can be visualized as a combination of two directions.
We require nine numbers in 3D at each point to specify a second rank tensor.
A change in the coordinate axes system will change the components along the axes; but, will
the tensor itself will remain unchanged.
 The transformation of a second rank tensor can be carried out using two transformation
matrices. We will repeat some of these concepts soon.
Tensors of various ranks
 Tensors can be used to describe:
(i) fields (field tensors) or (ii) properties (property tensors).
 These tensors can belong to various ranks: zeroth rank, first rank, second rank, etc.
 E.g. Temperature field is a scalar field, where each point in space is described by one
number the ‘T’ at that point (T(x,y,z)).
Scalar fields are tensor fields of rank-0. On the other hand some fields require more numbers
to be specified at each point in space.
 Electric field (polar vector) and Magnetic field (axial vector) require three numbers (in 3D)
to be specified at each point. These 3 numbers are the components along the coordinate axes
and give the direction and magnitude of the vector. Such a field is a vector field. Vectors are
tensors of rank-1.
 Some other fields require more numbers to be specified. E.g. to describe the state of stress at
a point, we need 9 numbers (in 3D) in general (stress being a symmetric tensor we actually need only 6 numbers).
Stress is a tensor of rank-2.

Q&A Why do we need coordinate transformations?

 We have already seen (Matrix representation of symmetry operators) that symmetry operators can be written as transformation matrices;
which involve operations like rotation (about a crystallographic axis), inversion (about an inversion centre usually the centre of the unit
cell) and mirror (about a plane usually passing through lattice points are exactly between lattice points).

* Note: only one number needed at each point. If there is a to point to point variation of temperature a set of T have to be specified for each
(x,y,z) which gives rise to the temperature field.
 The number of values required to be specified can be thought of as the number of directions
involved. This is specified by adding subscripts to the parameter.
 A scalar involves zero directions and does not have a subscript (e.g. T)*.This implies a scalar
quantity is the same in all coordinate axes.
 A vector involves one direction and has a single subscript (Ei , with i= 1,2,3 in 3D, i.e. (E1, E2,
E3) or (Ex, Ey, Ez)).
 A tensor (rank-2) has two directions involved and has two subscripts (ij , with i,j = 1,2,3 in
3D).

Rank n (2D) n (3D) Field Tensors Property tensors

Polar Axial

0 1 1 Temperature Specific heat Rotatory power

1 2 3 Electric field, Pyroelectric Pyromagnetism

Magnetic field coefficient

2 4 9 Stress field, Thermal expansion, Magnetoelectricity

Strain field Permittivity

3 8 27 Piezoelectricity Piezomagnetism

4 16 81 Elastic Stiffness Piezogyrotropy

n- How many numbers need to be specified
 Vectors and their transformations
 The direction and magnitude of a vector (vector is a first rank tensor) is same in any
coordinate system. Let us consider a vector P in (x,y) coordinate axes with intercepts (Px,
Py). Let us consider rotation of axes by an angle  (using z-axis) as in the figure below.
 The new intercepts along the coordinate axes (Px’ ,Py’) can be related to the old ones (x,y) by
a transformation matrix (aij). Needless to say the magnitude and direction of the vector
remains unchanged during the rotation of the coordinate system only the intercepts (x,y)
along the coordinate axes change. The aij are nothing but the direction cosines between the
new and the old axes.
 For a certain specific orientation (=0) the component along y-axis (or for that matter the x-axis) will
be zero.
 a11 a12   Cos Cos(90   )   Px '   cos sin   Px 
aij     P    P 
 21 22  
a a Cos (90   ) Cos    y'    sin  cos   y 
Px '  Cos .Px  Sin Py  a11Px  a12 Py
Py '   Sin .Px  Cos Py  a21Px  a22 Py
These equations can be written as: Transformation
P Matrix

 Px '   a11 a12   Px  Using

P   P 
Einstein’s
summation
Pi '  aij Pj
 y '   21 22   y 
a a convention
Pi '  aij Pj In this equation ‘i’ is free index and ‘j’ is the dummy index (over which the sum is performed)
i=1,2 & j = 1,2 (here the ‘=‘ means ‘takes values of’ ). ‘x’, ‘y’ can also b replaced with indices ‘1’, ‘2’.

In 3D the aij will be a 3×3 matrix and the equation using direction cosines (as in table below) can be written as:
Direction cosines (3D)
 Px '   a11 a12 a13   Px  The condensed equation (in 3D)
P   a a23   Py  still remains the same except x1' x2' x3'
 y '   21 a22
 P  a a33   that i = 1,2,3 & j = 1,2,3 Pi  aij Pj
 z '   31 a32  z 
P x1 a11 a12 a13

The reverse transformation x2 a21 a22 a23

Sometimes we need to go from the old to the new coordinate system*. x3 a31 a32 a33
In this case the equations turn out to be:
P1  a11P1'  a21P2'  a31P3'
P2  a12 P1'  a22 P2'  a32 P3' Pi  a ji Pj' Note the position of the dummy index now (in aji)

P3  a13P1'  a23P2'  a33P3'

Q&A Which is the vector we are transforming here?

 The vector we have considered could be a field vector, like a stimulus or a response (E or P);
or could be a property vector like pyroelectric coefficient (p).

* Hey! Can’t I just call the new old and the old new?
Transformation of Tensors
 As mentioned before there are two kinds of tensors (here we are keeping our focus on 2nd and higher rank
tensors): (i) field tensors and (ii) property tensors. Many material/physical properties have a
tensorial nature (i.e. are tensors).
 The effect of symmetry on physical properties can be determined by how the tensor
transforms under a symmetry operation.
 The magnitude of a property in any arbitrary direction can be evaluated by transforming the
tensor.
 As expected, under the action of a symmetry operator of a crystal the Tensor properties do
not change.
 Scalar such as Temperature is same in all coordinates. So, if the temperature is ‘T’ at a point
in space, it is still same whatever coordinates we choose.
 We have already seen how vectors transform; now let us see how 2nd rank tensors transform
(due to coordinate transformation). Let us consider a vector (pi) related to another vector (qj)
via a tensor (Tij). [p & q could be fields as we shall see later]. Let p,q be polar vectors for now.

p1  T11q1  T12q2  T13q3  p1  T11 T12 T13  q1  Using

 p   T T T   q 
p2  T21q1  T22q2  T23q3 In matrix
form  2   21 22 23   2 
Einstein
summation
pi  Tij q j
p3  T31q1  T32q2  T33q3  p  T T T   q 
 3   31 32 33  3  convention

‘i’ is free index and ‘j’ is the dummy index (over which the sum is performed)
Continued…
 We have used only two concepts here:

(i) vector transformation from one coordinate axes to another.

(ii) Stimulus (qj) leading to response (pi) via a property tensor (Tij)
We have pi  Tij q j (1)
The vector pi (or pk) transforms to the new coordinate system as: pi'  aik pk (2)

In terms of the alternate indices (1) can be written as: pk  Tkl ql (3)

The vector qi (or qk) transforms to the old coordinate system from new as: ql  a jl q'j (4)
Putting together equations (2,3,4):
p i'  aik  pk   aik Tkl ql   aikTkl  ql   aikTkl  a jl q'j    aik a jlTkl  q'j
That is: p i'   aik a jlTkl  q'j
Now this is nothing but the relation between p and q in the new coordinate system (the ‘’’
system). This implies that the quantity in the brackets must be Tij’  The tensor in the primed
coordinate system.
Tij   aik a jl  Tkl This is the transformation of a second order property
tensor from ‘old’ to ‘new’ coordinate axes.

For the transformation of a vector we had one matrix here we have two.

‘i,j’ are free indices and ‘k,l’ are dummy indices (over which the sum is performed)
Alternate Matrix method for Tensor Transformation
We have noted that we can transform 2nd rank tensors using the relation Tij   aik a jlTkl 
A alternate matrix form of this equation is also very useful, which can be derived as below.

The relations we have already seen are:

Vectors from old to  p   a  p 

new coordinate system
      q    a   q
q   a q
1

One vector to another in old

coordinate system
 p   T  q   For a orthogonal matrix the transpose is the
inverse: (a)T = (a)1

One vector to another in new

coordinate system
 p  T  q
Next we substitute one by one these relations to identify [T’]

 p   a  p    a T  q    a T  a   q   a T  a   q  T  q

1 T

T    a T  a 
T Don’t get confused by the two Ts (one is for Tensor and other for transpose)

Noting that:  a T  a  T

 
  a  T  a 
T
 We get: T    a  T  a 
T
Notes on transformation matrices
 The new set of axes is related to the old set by three direction cosines (3 nos.). This implies that
not all the 9 terms in the transformation matrices (aij) are independent and only 3 are
independent. This implies that there must be 6 relations between the (aij).
 As each row in the (aij) matrix represents a direction of a vector with respect to 3 orthogonal
axes (i.e as direction cosines), we have three relations of the type: Cos2α+ Cos2β+ Cos2 = 1.

a112  a122  a132  1 2

a21  a22
2
 a23
2
 1 a31
2
 a32
2
 a33
2
1

 As any row represents one of the orthogonal directions, product of any two rows is zero.

a11a21  a12a22  a13a23  0 a11a31  a12a32  a13a33  0 a21a31  a22a32  a23a33  0

Determinant of the transformation matrices

a11 a12 a13 For a transformation that leaves the handedness of

+1 an axis unchanged (e.g. a rotation)
aij  a21 a22 a23
a31 a32 a33 1 For a transformation that changes the handedness
of an axis (e.g. a inversion or mirror)

 The determinant of any orthogonal matrix is either +1 or −1.

 All the matrices associated with symmetry operations (rotations, inversion, mirror) are orthogonal.
 The value of the determinant is 1, as these transformations should not change the magnitude of the
vector.
Transformation of Tensors of all ranks (polar)
 We can generalize the transformation law to any rank tensor as in the Table-T below.
 These transformation laws serve as the very definition of these tensors.
 If these tensors represent physical quantities, it becomes clear that the components of the
tensor change (from one coordinate system to another) by the physical quantity itself
remains unchanged. E.g. the electric field vector may be represented by a sent of 3 different
numbers in two different coordinate systems, however (as obvious) the electric field strength
itself remains the same.
 Confusion may arise in the case of 2nd rank tensors in that they resemble a transformation
matrix (both are specified by 9 numbers and can be written as a 3×3 matrix). However, they
represent different quantities. Tensors (2nd rank) can be transformed to different axes using
these transformation matrices; but (clearly) the transformation matrices themselves cannot be
transformed (from one axes set to another).

Table-T Values required Transformation law

Rank 2D 3D Old  New New  Old
0 (scalar) 1 1 ' =   = '
1 (vector) 2 3 pi   aij  p j pi   a ji  pj
2 (*) 4 9 Tij   aik a jl  Tkl Tij   aki alj  Tkl
* Causally called a tensor (i.e. when
3 8 27 Tijk   ail a jm akn  Tlmn Tijk   ali amj ank  Tlmn
 the word tensor is used without
specifying the rank, this implies a
4 16 81    aima jn ako alp  Tmnop
Tijkl Tijkl   ami anj aok a pl  Tmnop
 rank-2 tensor).
Polar and axial tensor properties of rank 0, 1, 2, 3 and 4
Symmetry of property & symmetry of a crystal
 In a amorphous material there is no symmetry at the atomic level. This implies that there is
no preferred direction in a glass  i.e. glasses are isotropic* (any property measured along
any direction will have the same value).
 On the other hand crystals can** be anisotropic (i.e. properties can be direction dependent).
 It is perhaps obvious that the symmetry of a property should have$ the symmetry of the
crystal ($as we shall see soon this is actually ‘at least the symmetry of the crystal’). E.g. for a crystal with 4mm
symmetry a property measured along +x will have the same value as that along +y. E.g. if a
field is applied along +x and the current is measured along +x (Jx) this will be equal to the
current measured along +y (if field is applied along +y).
 We expect from 4-fold symmetry and inversion centre (mirrors) present:
J+x = J+y = Jx = Jy

Ey ** Will it really be anisotropic or not will

depend on other factors (including the rank
of the tensor and the symmetry of the
Jy crystal) and will be considered later.

Jx
Ex
* It is important to note that glasses with no symmetry at the atomic level have the highest
symmetry with respect to properties!
Neumann principle*
 We have already noted that the symmetry of a property should be equal to that of the crystal.
 However, a property may have higher symmetry. So the correct statement is:
the symmetry of a property can be equal to that or greater than that of the crystal.

Other statements of the Neumann principle.

 Symmetry elements of a physical property of a crystal must include all the symmetry
elements of its point group (all its rotational axes, mirror planes, etc.).

Examples of a property having a higher symmetry (than the point group symmetry).
We will understand the details soon.
 All properties that can be represented by tensors of rank up to 2 are isotropic for cubic
crystals. E.g. electrical conductivity of cubic crystals is isotropic.

* A closely related principle is the Curie Principle (Curie Principle).

How to use Symmetry & Neumann principle to determine the number of independent constants
in a property tensor?

1) Start with the property tensor (say in matrix form).

2) Reduce the number of independent components based on any constrains out the symmetry of
the crystal (e.g. symmetry of the stimulus, response, energy considerations, etc.).
List out the independent components at the end of all these considerations.
3) Use the appropriate transformation law based on the rank of the tensor (Table-T). Use
alternate matrix based transformation rules where applicable (and easier).
4) Apply the appropriate transformation matrices corresponding to the symmetry operators of
the crystal (only the bare minimum symmetry operators need to be used (the generator matrices)).
 In some cases where direct application of transformation principles are possible, the same
can be used (e.g. a 3-fold along [111] in cubic crystals takes: xy, yz & zx).
5) After the application of each transformation, apply Neumann’s principle. I.e. the transformed
components must be equal to the original components. Based on this determine the ‘equal’
components of the property tensor and the zero components.

These steps will become clear once we take up some actual solved examples in this chapter.
How to understand anisotropy?
 Let us start with a toy model to ‘visualize’ anisotropy in crystals. If the glass ball is pulled as
shown then it will not move along the direction of pull, but will move along the direction as
marked. I.e. even if the stimulus is along ‘P’, the response is along ‘A’.

P
Tension Shear

 We can think of two alternate ways to understand anisotropy:

 The direction of the response is different from the direction of the stimulus (if electric
field is applied along ‘S’ direction of a crystal the current may flow along ‘R’).
 The response may involve other terms than the ‘natural’ expectation due to the stimulus
(e.g. if a crystal is stretched along ‘F’, it may shear in addition to being elongated).
Funda Check What can happen if we pull, bend or twist an anisotropic crystal?

 If we pull the crystal it may elongate and shear.

 If we bend the crystal using pure bending moments, it may bend and twist.
 If we twist the crystal using pure torsional moments, it may twist and bend.

 The words in red are not the natural responses to the applied load.
 In an isotropic material only the green responses will exist.
How is that a block of Copper is isotropic (say with respect to its elastic properties)?
Origin of isotropy at the level of the microstructure and partial anisotropy in crystalline materials
 At the level of the crystal structure (in a single crystal), a given property may be isotropic
due to Neumann’s principle (e.g. cubic crystals like NaCl are optically isotropic). However, isotropy may also
arise due to spatial averaging of properties at the level of the microstructure.
 A single crystal of Cu will be anisotropic with respect to its elastic properties (we will see
later that 3 independent elastic constants are required: E11, E12, E44).
 However, a block of Cu is polycrystalline with multiple grains oriented randomly: thus there
is no preferred direction for the Cu block and its elastic properties are isotropic* (we require
only two independent moduli: E (Young’s modulus) and  (Poisson’s ratio) to describe the its
elastic response).
 In a wire of Cu, the grains may not be oriented randomly and such a material is said to
possess crystallographic texture. I.e. in a textured material there is (some degree of)
preferential orientation of grains. Texture can develop during rolling, extrusion, wire-
drawing, thin film growth, etc.
 In textured materials, the anisotropy present at the level of the single crystal is partially
recovered.
Atomic structure level
Origin of Isotropy Single crystals
Due to Neumann’s principle
* Assuming that the block has many many grains
and hence all orientations are represented. Microstructure level Also for textured materials
Stimulus and Response
 We can apply various kinds of stimuli on a material (like Force, Electric Field,
(As stated before)

Magnetic Field, Temperature difference, etc.). These fields may give rise to natural
responses (like elongation, electric current or polarization, magnetization, heat flow, etc.).
 However, we may also get responses which are usually associated with other kinds of
stimuli. E.g. the application of pressure may lead to the polarization of the crystal (Piezo-
electric effect) or the application of a magnetic field may lead to strain in the material
(magnetostriction).
 A material/physical property connects the stimulus to the response.
 The stimulus response plot may be linear, non-linear or may even show hysteresis.
 The physical property connecting a stimulus of rank-p to a response of rank-q will be of
rank-(p+q). E.g. in the table below stimulus jk is of rank-2, response Pi is of rank-1 and the
property dijk is of rank-3.

Phenomena Cause Effect Coefficient Equation

Pyroelectricity T Pi pi Pi = pi T
(Temperature difference) (Electric polarization) st
(1 rank tensor)
Thermal Expansion T ij αij ij = αijT
(Strain) (2nd rank tensor)
Piezoelectricity ij Pi dijk Pi = dijkjk
(Stress) (Electric polarization) (3rd rank tensor)
Elasticity kl ij Eijkl ij = Eijklkl
th
(strain) (Stress) (4 rank tensor)
Crystallographic axes vs Property Axes
 The crystallographic axis (a,b,c) may or may NOT coincide with the property axis (Z1, Z2,
Z3).
 The property axes are always orthogonal, while the crystallographic axes may or may not be
orthogonal. The standard settings are as in the table below.

Property axis
Z1 Z2 Z3
Cubic a b c
Hexagonal a  (Z1, Z3) c (6-fold)
Tetragonal a b c (4-fold)
Trigonal a  (Z1, Z3) c
(hexagonal setting) (3-fold, [001])
Orthorhombic a b c
(c < a < b)
Monoclinic  (Z2, Z3) b [010] c
2-fold or  m
Triclinic  (Z2, Z3)  (010) c
On (010)
Electrical Properties
 The response of an electric field on a material can be:
(i) motion of charges (electrons or ions), (ii) polarization.
In metals electrons flow on the application of an electric field, while in materials like solid
electrolytes ions carry the current.
 Materials which lack free charges are called di-electrics (should have been called dia-electrics, which
would have made it consistent with dia-magnetics). In dielectrics, the internal field weakly opposes the
external field (like in diamagnetic materials the internal magnetic field weakly opposes the
external field).
 In some crystals the centre of mass of the positive charges does not coincide with the centre
of mass of negative charges, leading to a net electric dipole. These crystals are spontaneously
polarized in the absence of an external field  polar materials. Such crystals lack a centre of
symmetry (inversion centre).
 Polar materials come in two types:
 (i) those in which the reversal in field direction can reverse the direction of spontaneous
polarization (called Ferro-electrics) and
 (ii) those in which this does not happen.
Ferroelectric materials show a hysteresis in the electric field-polarization plot.
 In general, materials which show hyteresis in stimulus-response plots are called Ferroics.
 In certain crystals pressure (mechanical stress) can lead polarization (or change in
polarization if the crystal already has spontaneous polarization). These materials are termed
as Piezoelectric. It is also possible that electric field can lead to mechanical strain in the
material this is termed as Inverse Piezoelectric effect. Both of these effects are “cross-
coupling terms” as we had seen when we discussed Hartmann diagrams.
 In the next slide we consider the effect of symmetry on piezo-, pyro- and ferro- electric
properties.
 We will see that out of the 32 point groups only 20 have the potential to show
piezoelectric properties. The point groups with centre of symmetry (11 nos.) cannot show piezoelectric effect, as centre of
mass of the +ve and ve ions will coincide in these cases.
 Only 10 of these 20 show the pyroelectric effect.
 A subset of these 10 point groups show the ferroelectric effect.
Symmetry and Pyroelectric, Piezoelectric and Ferroelectric Effects
32 point groups 4 2
3
2
3
6 2 2 6
m m m mmm m
Crystal Symmetry Point Groups
Do not show Pyro-, Piezo-
or Ferroelectric effects
21 point groups 11 point groups 3
2
3
Noncentrosymmetric Groups Centrosymmetric Groups m
4 2 2 4
mmm m
1 point group 432 20 point groups 2 2 2 2
1, 2, m, 222, mm2, 4, 4, 422, 1
Possessing Piezoelectric Effect 4mm, 42m, 3, 32, 3m, 6, 6,
mmm m

Other Symmetry 622, 6mm, 62m, 23, 43m

(Polarized under
Elements (Nonpolar) Mechanical Stress)
Out of the 20 PG 10 show pyroelectricity

Polarized in the absence 10 point groups 1, 2, m, mm2, 3, 3m, 4, 4mm, 6, 6mm

of an electric field Polar crystals Pyroelectric Effect
(Spontaneously Polarized)
Out of the 10 PG a subset show Ferroelectric effect

Subgroup of the point groups

Ferroelectric Effect
Polar crystals
(Spontaneously Polarized
With Reversible Polarization)
Pyroelectricity
 On heating/cooling in some materials (like tourmaline*, gallium nitride (GaN) based materials,
caesium nitrate (CsNO3), Lithium tantalate (LiTaO3, LiNbO3)) voltage develops, this phenomenon
is called pyroelectricity. On heating/cooling charge develops on the opposite faces of crystals
lacking centre of inversion.
 Heating leads to shift in atomic positions so as to lead to a voltage.
 As temperature change (stimulus) does NOT have a directionality, the response
(polarization) must correspond to the symmetry of the crystal.
 All pyroelectric materials are also piezoelectric, but the vice-versa is not true.
 The temperature change (T) is related to the polarization (P) by:

 Pi → polarization vector [C/m2]

Pi  pi T  pi → pyro-electric coefficient (polar first rank tensor/vector property) [C/m2/K]

 Pi is a polar vector, which transforms as: Pi  aij Pj

'

* Common black tourmaline has a nominal formula NaFe2+3Al6Si6O18(BO3)3(OH)3(OH)

 Crystals with centre of symmetry do not show pyroelectric effect. Let the property axes be
the orthogonal set (Z1, Z2, Z3) and let the components of the pyroelectric coefficient be: (p1,
p2, p3). The matrix operator for centre of inversion is:  1 0 0 
0
 1 0 

Point groups with inversion symmetry not showing pyroelectricity

0 0 1
Crystal Group with
 System Centre of
symmetry (i)
Cubic 4 2
The coefficients after the application of the inversion operator is: (2)
3
m m

 p1'   1 0 0   p1    p1   p1'    p1  2
3
 '       '  
m

 p2    0 1 0   p2     p2    p2     p2  Hexagonal 6 2 2

 p3'   0 0 1 p3    p3   p3'    p3  (2) m m m

        6
m
Trigonal 2
 If a component of a tensor changes sign after a transformation (2)
3
m

of coordinates, the component must vanish (= 0) for that crystal 3

Tetragonal 4 2 2
in accordance with Neumann’s principle. (2) m m m
4
m

By Neumann’s principle:  p1'   p1    p1  Orthorhombic 2 2 2

 p1   0  (1) m m m
 '      p    0
 p2    p2     p2    2  
Monoclinic
(1)
2

 p   0
m
 p3'   p3    p3   3  
    Triclinic (1) 1

This implies that in all crystals with centre of symmetry pyroelectricity is absent (i.e. if we heat
these type of crystals no charge/potential will develop):
(i) 11 of the 32 point groups &
(ii) 2 out of 7 Curie groups (/m m, m).
 We have noted that not all crystals which lack an inversion centre are pyroelectric. This
implies that not all piezo-electric crystals are pyroelectric. Quartz is a well known
piezoelectric crystal, which is not pyroelectric. Quartz has a point group symmetry 32 (2 ||
Z1 & 3 || Z3). This implies that the pyroelectric coefficients will remain the same under these
transformations. Note: all cubic crystals have 2-fold || <001> and 3-fold || <111>.

 p1'   1 2 3 2 0   p1    p1 2  3 p2 2 
 '     
 p2     3 2 1 2 0   p2    2 
 3 p1
2
p2
3-fold along Z3
 p3'   0
   0 1   p3   p3 

  
By Neumann’s principle:  p1    p1 2  3 p2 2   p1   0 

 p     3 p1  p2    3 p1  p2 Both these conditions  p  0 
 2  2  are satisfied only  2  
p  p 
2
 p     
 1 p  3 p if p1 = p2 = 0  3  3
 3  p3  2

Now we apply the other symmetry operation (2-fold along Z1 (after the 3-fold operation)
 p1'   1 0 0   0   0 
 '     
2-fold along Z1  p2    0  1 0   0    0 
 p3'   0 0 1 p3    p3 
    
 p1   0   p1   0 
p  0   p    0
By Neumann’s principle:  2      2  
 p  p   p   0
 3  3  3  
 It is to be noted that though we are deriving the effect of symmetry on the pyroelectric
coefficients, it is applicable to all 1st order tensor (vector) properties (as there is nothing specific to
pyroelectricity in the matrix operations). all vector properties for point group 32 vanish.
  Now let us look at the tourmaline crystal, which addition to being pyroelectric is
piezoelectric. Tourmaline has a point group symmetry 3m (m  Z1 & 3 || Z3). This implies
that the pyroelectric coefficients will remain the same under these transformations.
 p1'   1 0 0   p1    p1 
m  Z1  '     
 p2    0 1 0   p2    p2 
 p3'   0 0 1  p3   p3 
    
 p1   0 
By Neumann’s principle:  p1    p1  p p 

 p2    p2   2  2
p   p  p  p 
 3  3   3  3

 p1'   1 2 3 2 0   0   3 p2 2   3
   p2  0
 '   
3-fold along Z3 p
 2    3 2  1 2 0   2 
p   p2
2   
2
 p2 = 0

 p3   0
'
0 
1   p3   p3   p   p2
     2 2
 p1   0 
  p2    0  Since p3 survives tourmaline is pyroelectric
p  p 
 3  3
Polar axis
 The direction of polarization (polar axis) is determined by the symmetry of the crystal. If the
polarization can be visualized as an arrow mark (), the ends of the arrow cannot be related by
any symmetry operator of the crystal. The disawllowed operations are as below.

 As expected pyroelectricity always occurs along a polar axis, this does

not imply that all polar axis will show pyroelectricity. E.g. Z1 axis of
quartz (32 symmetry) is a polar axis (along ‘a’ direction in hexagonal
setting of the trigonal crystal), however, pyroelectric charges do not
accumulate along this direction. This happens due to the presence of the
3-fold orthogonal to the Z1 (which implies that there will be three
equivalent directions 120 apart which will lead to zero net
polarization).
How to locate the Polar axis?
 If vectors are drawn from the origin to each
of the equivalent points and the resultant
obtained, this will be along the polar
direction. Examples are shown for point
groups m and 2.

 If the vectoral sum leads to a zero

resultant then there will be no net
polarization and hence no polar
direction. E.g. for the point group 2/m.
 Pyroelectricity in cubic materials
 All cubic crystals have 2-fold || <001> and 3-fold || <111>. 3-fold along <111> will take
xy, yz & zx.

 p1'   0 1 0   p1   p2 
 '     
3-fold along [111]  p2    0 0 1   p2    p3 
 p3'   1 0 0  p3   p1 
    

 p1   p1 
By Neumann’s principle:  p1   p2  p p 

 p2    p3   2   1
p  p  p  p 
 3  1  3  1
Now we apply the other symmetry operation (2-fold along Z1 (after the 3-fold operation)

 p1'   1 0 0   p1    p1  By Neumann’s principle:  p1   0 

 '     
2-fold along Z1  p2    0 1 0   p1     p1  p  0 2  
 p3'   0 0 1  p1   p1   p   0
      3  

 It is to be noted that though we are deriving the effect of symmetry on the pyroelectric
coefficients, it is applicable to all 1st order tensor (vector) properties (as there is nothing specific to
pyroelectricity in the matrix operations). In cubic crystals first rank tensor properties vanish.
Pyroelectricity in textured materials
 Suppose we are not considering a single crystal but a material with texture with m
symmetry (a strong electric field along Z3, which aligns domains parallel to Z3). Will
pyrolectricity survive in the textured material?

 p1'   Cos Sin 0   p1   p1 Cos  p2 Sin 

 || Z3  ' 
 p2     Sin Cos 0   p2     p1 Sin  p2 Cos 
 p3'   0 1    
  0   p3   p3 
 p Cos  p2 Sin   p1  If equality in row-1 and  p1   0 
By Neumann’s principle:  1 p  p  0 
 p1 Sin  p2 Cos   2  row-2 are to be applicable   2  
p  p 
   p  for all  values  3  3
 p3   3

 p1'   1 0 0   0   0 
 '   0    0 
m  Z1 p
  
2  0 1 0    
 p3'   0 0 1  p3   p3 
    

 p1   0 
p  0 
  2   Since p3 survives the textured material with m symmetry is pyroelectric
p  p 
 3  3
Pyroelectric crystal classes
 We have noted that there are 10 points groups, which show pyroelectricity. Two Curie groups
also show pyroelectricity. These are listed below. The number of independent measurements
required is the number of non-zero coefficients in each case.

Point groups Curie Groups Pyroelectric coefficients

 p1 
p  Number of
1  2 (3 Nos.) independent
p 
 3 values required
0
p 
2  2 (1 No)
0
 
 p1 
0
m   (2 Nos.)
p 
 3
mm2 (OR),
0
3, 3m (Trigonal), 0
, m   (1 No)
4, 4mm (Tetragonal) p 
 3
6, 6mm (Hexagonal)
Ferroelectricity With special reference to Perovskites

 BaTiO3 (a model material w.r.t to ferroelectricity) has a cubic structure with centre of
symmetry above 120C and hence has no permanent dipole moment. Below 120C the
crystal structure is tetragonal, with the ‘Ti’ shifted from the centroid of the octahedra formed
by ‘O’. This is an example of a Perovskite structure, which have a general formula ABO3
(other examples are: SrTiO3, KTaO3, SrSnO3, and many more).
 In fact BaTiO3 [(Ba+2Ti+4) (O2)3] can exist in hexagonal, cubic, tetragonal, orthorhombic,
and rhombohedral crystal structures. All of these phases exhibit the ferroelectric effect
except the cubic phase (with an inversion centre).
 Not all Perovskites are ferroelectric, SrTiO3 (at RT) with centre of symmetry is cubic (cP5,
Pm3m) and is not a ferroelectric material. In BaTiO3 the Ti+4 ion is slightly smaller than the
space given by the octahedral position and hence the ion slightly shifts along one of the
<001> directions; thus lowering the symmetry of the crystal to tetragonal. This further
introduces a permanent dipole to the structure. Note: SrTiO3 becomes tetragonal at low T (below 168C).
 Regions within the material where the electric dipoles are parallely oriented (say along
[001]) form a ferroelectric domain. In neighbouring domains the dipole may be oriented
along one of the other members of the <001> family (e.g. [010]).
 Different Perovskites can exhibit a variety of properties like they can be: dielectric (CaTiO3),
ferroelectric (BaTiO3), piezoelectric (Pb(Zr,Ti)O3), semiconducting ((Ba,La)TiO3),
superconducting, etc. (they can even show GMR effect). It is to be noted that some of these compounds have
alloying in one of the sublattices and some depend on off-stoichiometry for their properties.
BaTiO3
 The structure below 120 is tetragonal (next slide for figures and data) with the Ti+4 ion
slightly displaced towards one of the <001> directions.
BaTiO3
Ti not at the centre of
the cell (slightly
displaced up)

Ti

Ba

BaTiO3 Wyckoff Site

x y z Occ
position Symmetry
Lattice parameter(s) a = 3.99 Å, c = 4.033 Å
Ba 1a 4mm 0 0 0 1
Space Group P4mm (99)
Strukturbericht notation Ti 1b 4mm ½ ½ 0.515 1

Pearson symbol tP5 O1 1b 4mm ½ ½ 0.025 1

Other examples with this O2 2c 2mm. 0 ½ 0.483 1
structure
The displacement of Ti
gives BaTiO3 a net
electric dipole
SrTiO3 Ti not at the centre of the
cell (unlike in BaTiO3)

Ti

Sr

SrTiO3 Wyckoff Site

x y z Occ
position Symmetry
Lattice parameter(s) a = 3.90 Å
Ti 1a m3m 0 0 0 1
Space Group Pm3m (221)
Strukturbericht notation Sr 1b m3m ½ ½ ½ 1

Pearson symbol cP5 O 3d 4/mm.m ½ 0 0 1

Other examples with this
structure
Piezoelectricity
Thermal Expansion
 When a material is heated it usually expands* (conversely a material will contract on cooling). The strains
introduced are stress free strains (i.e. the body will be stress free in the expanded state, in the
absence of any external constraints).
 Since the stimulus (T) has no directions involved, the response (thermal strains) must
correspond to the symmetry of the crystal. This implies that none of the symmetry elements
of a crystal can be destroyed during free thermal expansion.
 Thermal expansion (strains) can be related to the temperature change by:

2nd rank tensor (field)  ij   ij T Scalar

Thermal expansion coefficient 2nd rank tensor (material property)

 11 12 13  11 12 13 

Written in full          .T
 21 22 23   21 22 23 
 
 31  32  33  31 32 33 
 As strain is a symmetric tensor**, so is the tensor representing the coefficient of thermal
expansion. Note that the strain components are scaled-up versions of the coefficients of thermal expansion (with appropriate change
in units).

* Examples to the contrary exist (in some materials coefficient of thermal expansion may be negative in some temperature regimes).
** In usual materials.
Thermal strains along principal directions
 The directions of principal strains will coincide with the principal directions of αij. Along the
principal directions the strains are:

1  1 T  2   2 T  3   3 T
 This implies that a sphere in the crystal will become an ellipsoid (the strain ellipsoid) on
heating/cooling. This is illustrated in the 2D example below, wherein the new lengths along x1
& x2 are: (1+α1T) & (1+α2T).

heat

 The coefficient of bulk expansion is: (α11 + α22 + α33) = (α1 + α2 + α3) is an invariant.
Isotropic versus anisotropic thermal expansion
 In a isotropic material: α11 = α22 = α33 (& α12 = α23 = α31 = 0)*. To understand the effect of
isotropic expansion (in 2D) let us consider a unit square which is randomly oriented w.r.t to the
principal axes (x1, x2). Before and after thermal expansion (as the green circle goes to the red
circle) the square (yellow) remains a square (pink).
Fig.1: Isotropic expansion
Fig.2: Anisotropic crystal

 In the case of an anisotropic crystal (in Circular hole

remains circular
2D), α11  α22 & hence 11  22. In such a
crystal on thermal expansion the green Isotropic
circle becomes the red ellipse and a
square (yellow in Fig.2) becomes a
parallelogram (pink) after thermal Anisotropic
expansion. Circular hole
becomes ellipse

* We will soon see “why?”.

 Symmetry imposed restrictions on the number of independent coefficients

T    a  T  a 
T
We have noted before that axes transformations can be carried out by:

Let us consider a cubic crystal with (4/m 3 2/m) symmetry (highest available for cubic crystals)
11' 12' 13'   0 1 0  11 12 13   0 1 0 
4-fold  ' '   1 0 0     1 0 0 
 21  '
22  23 
   21  22  23  
along Z3
 31  32  33   0 0 1   31  32  33   0 0 1 
' ' '
 
Like in tetragonal crystals
 0 1 0   12 11 13    22  21  23 
  1 0 0   22  21  23    12 11 13 
  
 0 0 1   
   32 31 33  32 31 33 
By Neumann’s 11 12 13    22  21  23  11 12 13  11 0 0 
principle:    22  23    12 11 13 
     0  0 
 21  21 22 23   11 
 31 32 33   32 31 33   31  32  33   0 0  33 
 If a crystal had only a 4-fold axis then, two independent coefficients would survive: α11 &
α33. This implies that all tetragonal crystals have only two independent coefficients (out of a
possible 6).
 It is to be noted that though we are deriving the effect of symmetry on the thermal expansion
coefficients, it is applicable to all 2nd order tensor properties (as there is nothing specific to thermal expansion
coefficient in the matrix operations).
Now we apply the other symmetry operation m  Z3 (after the 4-fold operation)
11' 12' 13'   1 0 0  11 0 0 1 0 0 
m  Z3  ' '  0 1 0  0  0 1 0 
 21  '
22  23 
   11 0
 
 
 31  32  33   0 0 1  0
' ' '  
0  33   0 0 1
 
 1 0 0  11 0 0  11 0 0  We got no reduction in the number of
  0 1 0   0 11 0  0 11 0  coefficients based on the mirror plane.
   
 0 0 1  0 0 33   0 33  Now we try the 3-fold along [111].
  0
11' 12' 13'   0 0 1  11 0 0 0 1 0
 ' '  1 0 0   0 11 0   0 0 1 
3-fold || [111]
 21  '
22  23 
   
 31  32  33   0  0  33   1 0 0 
Which is present in all
cubic crystals

' ' '
 1 0   0

 0 0 1   0 11 0   33 0 0  By Neumann’s principle:

  1 0 0   0 0 11    0 11 0  11 12 13  11 0 0
     0  
 0 1 0   0   0 0 11   21   23  
0

   33 0 22 11

 31  32  33   0 0 11 

 This implies that for (4/m 3 2/m) crystals there is only one independent coefficient.
 We could have change the order of the operations (i.e. we could have operated the 3-fold
first) and this should not make any difference. Also, we could have operated the 3 instead
of the 3-fold (as this is available for the (4/m 3 2/m) crystals.
 Next, we ask the question: “how many independent coefficients will a second rank tensor
property have for cubic crystals”. This includes cubic crystals lower point group symmetry like ’23’.

All cubic crystals have a 3-fold || to <111> and 2-fold || <100>*

Now let us try the other option pointed out before; i.e. operate the 3-fold axis first.
11' 12' 13'   0 0 1  11 12 13   0 1 0 
3-fold || [111]  ' '   1 0 0   0 0 1

 21  '
22  23      21  22  23   
   
 31  32  33   0 1 0   31  32  33   1 0 0 
' ' '
 
 0 0 1  11 12 13  0 1 0   0 0 1  13 11 12   33  31  32 
  1 0 0   21  22  23   0 0 1    1 0 0   23  21  22   13 11 12 
     
 0 1 0      
   31 32  33  1 0 0   0 1 0   33  31  32   23  21  22 

By Neumann’s 11 12 13   33  31  32 

 α11 = α22 = α33
principle:
21  22 
 23   13 11 12  
α12 = α13 = α23
 31  32  33   23  21  22 
11 12 13  11 12 12 
 
 21
 22    
23   12
11  
12 

 31  32  33  12 12 11 

* Note: not all cubic crystals have 4-fold || <100>. In addition, cubic crystal may have 2-fold || <110>
Next we operate the 2-fold || [001]
11' 12' 13'   1 0 0  11 12 12   1 0 0 
 ' '   0 1 0     0 1 0 
2-fold || [001]
 21  '
22  23 
   12  11  12  
  
 31  32  33   0 0 1  12 12 11   0 0 1 
' ' ' 
 
 1 0 0   11 12 12   11 12 12 
  0 1 0  12 11 12    12 11 12 
  
 0 0 1   12 11  12 12 11 
   12
By Neumann’s 11 12 12   11 12 12 
principle:   12    12 11 12   α12 = 0
 12 11

12 12 11   12 12 11 

11 12 13  11 0 0

 
 21  22  23
 0 
  11 0

 31  32  33   0 0 11 

 All cubic crystals are isotropic with respect to second rank

tensor properties (require only one independent constant).
Centrosymmetry in 2nd rank tensor properties
 Let us consider a vector stimulus qj connected to a vector response pi via a tensor property
Tij. Let us reverse the direction of the stimulus, this will reverse the direction of the response
without affecting Tij.
 This implies that all second rank tensor properties are centrosymmetrical.

pi  Tij q j  Tij   q j   Tij  q j     pi 

 p1  T11 T12 T13  q1    p1  T11 T12 T13  q1 
 p   T T T q    p   T T   q 
  
2 21 22 23   2  Reversing stimulus T
 2   21 22 23   2 
 p  T T T   q    p  T T T   q 
 3   31 32 33  3   3   31 32 33  3 
The effect of crystal symmetry on properties represented by symmetrical second-rank tensors
Optical System Characteristic Nature of Number of
classifica symmetry[1] representation quadric independent Tensor[2]
tion and its orientation coefficients
Cubic Four 3-fold axes Sphere 1 S 0 0
Isotropic 0 S 0
(anaxial)  
 0 0 S 
Tetragonal One 4-fold axis Quadric of revolution 2  S1 0 0
Hexagonal One 6-fold axis
about the principal 0 S1 0 
Uniaxial symmetry axis  x3  z  
Trigonal One 3-fold axis  0 0 S3 
Orthorhombic Three mutually General quadric with axes 3  S1 0 0
perpendicular 2-fold  x1, x2 , x3  to the diad 0 S2 0 
axes; no axes of 
higher order axes  x, y, z   0 0 S3 
Monoclinic One 2-fold axis General quadric with one 4  S11 0 S31 
axis  x2  to the diad 0 S2 0 

axis  y   S31 0 S33 
Biaxial
Triclinic A centre of symmetry General quadric. No fixed 6  S11 S12 S31 
or no symmetry relation to S S22 S23 
crystallographic axes  12
 S31 S23 S33 

[1] Axes of symmetry may be rotation axes or inversion axes.

[2] The setting of the reference axes x1 , x2 , x3 in column 6 in relation to the crystallographic axes x, y, z and to the symmetry
elements is that shown in column 4.
4th Order Tensor Properties (mechanics)
 The stresses are related to strains by Hooke’s law via the 4th order tensor properties: Stiffness
(Eijkl) and Compliance (Sijkl). Note: Usually ‘C’ is the symbol for stiffness and the symbol for compliance is not ‘C’ and
confusing with ‘Stiffness’!

 ij  Eijkl  kl  ij  Sijkl  kl
 Being 4th order tensors there are 3×3×3×3 (= 81) components (in 3D). Not all these
components are independent. The number of independent components reduces 81 to 36
given that stress and strain are symmetric tensors. For diagonal terms:

11  S111212  S1121 21  is a symmetric tensor:  12   21 11  ( S1112  S1121 )12

 S1112 & S1121 always occur together and cannot be experimentally separated and keeping this
in view S1112 is assumed to be equal to S1121. For off-diagonal terms:

12  ( S1233 ) 33  21  ( S2133 ) 33 As 12 = 21 S2133  S1233 Thus not all components are
independent

 Eij kl As ‘ij’ & ‘kl’ refer to strain &

stress tensors which have 6
Eij kl
independent components (3D)
9×9 6×6
If all components are
independent
The single index notation for stress & strain Voigt’s notation
 Sometimes the indices are condensed (from 2 to 1) according to the notation as below. The
two index notation is the matrix notation and has to be converted back to tensor notation so
that transformations can be carried out (as usual for tensors).

 11  12  13   1  6  5  11 12 13   1 1

6 1
5 
  
2 2
      
 22 23   2 4

 22  23   2 1

2 4
Only one symmetric
half of the matrix is written
  33    3    33    3 
Tensor Matrix Tensor Matrix

 1   E11 E12 E13 E14 E15 E16   1 

   E26   2 
This will lead to condensed notation  2   E21 E22 E23 E24 E25
 
version of the stress-strain relation  3   E31 E32 E33 E34 E35 E36   3 
(matrix form)    
 4   E41 E42 E43 E44 E45 E46   4 
E.g. E43  E2333  5   E51 E52 E53 E54 E55 E56   5 
    
 6   E61 E62 E63 E64 E65 E66   6 

1 6 1 6 6
dW   i d  i   Eij j d  i
2 i 1 2 j 1 i 1
 Further reduction in the number of independent moduli is possible using the concept of
strain energy of a hyper-elastic material. If ‘W’ is the strain energy per unit volume:

1 3 3 1 3 3
dW   ij d  ij   Eijkl kl d  ij
2 i 1 i 1 2 i 1 i 1
 Let us consider the specific case where the only stress applied is 11 (= 1 in condensed
index notation) and the body is in plane strain condition (i.e. 13, 23 & 33 are zero). The
strain energy is given by (Taylor series expansion for small strain in a hyperelastic material):

      2  
1 W 1  W
W ( ij )  W (0)      ij       ij kl   ...

 1   ij    2   ij  kl  
  ij  0    ij  0 

W ( ij )  W (0)   ij   ij   T 3  ... Ignoring higher order terms and observing that the
quantity in blue font is Eijkl
Zero if no Zero if no
residual strain residual stress
Noting that the order of differentiation does not matter
  2     2  
 
1
W ( ij )   
W

 2   ij  kl 
 
ij kl
 
 2
1
 ijkl  ij kl  2     ij kl  2  Eklij   ij kl
E     1 W
    
1

  ij  0    kl ij ij 0 

 E   E 
ijkl klij
Eijkl is a symmetric tensor
Effect of symmetry elements on Eijkl
 We have noted that even for crystals without any symmetry, only 21 of the 81 possible
elastic constants are independent. For crystals with symmetry this number is further reduced.
Let us consider a tetragonal crystal with 4 symmetry.
 The 4-fold will take: (a)(b), (b)(a), (c)(c). Matrix of surviving coefficients is shaded
blue only 7 independent terms survive: E1111, E3333, E1122, E1133, E1313, E1212, E1112.
 1111 1122 1133 1123 1131 1112   2222 2211 2233 2213 2232 2221

 2222 2233 2223 2231 2212   1111 1133 1113 1132 1121

 3333 3323 3331 3312   3333 3313 3332 3321
   
 2323 2331 2312   1313 1332 1321 
 3131 3112   3232 3221
   
 1212   2121 

 1111 1122 1133 0 0 1112 

1131=1113=2223=2232=1113  

 all these coefficients are zero  1111 1133 0 0 1112 
4  fold
 3333 0 0 0 
Eijkl  
2213=1123=1132=2231=2213  1313 0 0 
 all these coefficients are zero  1313 0 
 
 1212 
3331=3332=3323=3313=3331
 all these coefficients are zero 3312=3321=3312
 all these coefficients are zero
2312=3221=1321=3112=3221
2331=1332=3213=2331
 all these coefficients are zero
 all these coefficients are zero
 2 4 2
Components of Eijkl for cubic crystals Cubic point groups  23, 43m, 3, 432, 3
m m m
 Cubic crystals belong to one of the 5 point groups as above. All the point groups have at
least a 3-fold symmetry (along <111>) along with at least a 2-fold along <100>.
  The 2-fold (along <100>) will take: (a)(a), (b)(b), (c)(c).
 The 3-fold (along <111>) will take: (a)(b), (b)(c), (c)(a).
 In the end we have only 3 independent terms: E1111 , E1122 , E2323. In 2-index notation: E11, E12, E44.
 1111 1122 1133 1123 1131 1112   1111 1122 1133 1123 1131 1112 
 2222 2233 2223 2231 2212 

 2222 2233 2223 2231 2212 

 3333 3323 3331 3312   3333 3323 3331 3312 
  Action of 2-fold 
2323 2331 2312 

 2323 2331 2312  
 3131 3112   3131 3112 
   
 1212   1212 
 1111 1122 1133 0 0 1112   2222 2233 2211 0 0 2223 
 2222 2233 0 0 2212   3323 
   3333 2233 0 0
 3333 0 0 3312  Action of 3-fold  1111 0 0 1123 
   
 2323 2331 0  3131 3112 0 

 3131 0   1212 0 
   
 1212   2323 

 1111 1122 1122 0 0 0 

We see that some of the terms are equal to one another. We 
 1111 1122 0 0 0 
have already noted that some of these terms are zero.
 1111 0 0 0 
Hence, the surviving terms (in the symmetric half) are :  
 23 23 0 0 
Note: the total number of non-zero terms is 12 out of  2323 0 
 
possible 36 (24 zero terms). But, only 3 are independent.  2323 
Pathway for reduction of number of independent Eijkl

Due to symmetry of stress Due to strain energy Due to symmetry present in

81 and strain tensors 36 considerations 21 cubic crystals 3
Curie Principle Click here for introduction to Curie groups

 When certain stimuli lead to certain responses, the symmetry elements of the stimuli should
be seen in the response.
 A crystal subjected to certain stimulus will display only those symmetry elements, which are
common to the crystal without the stimulus and the stimulus without the crystal.
 In terms of groups this can be written as:

R crystal with stimulus   G cystal symmetry   S Stimulus 

 The symmetry of the crystal in the presence of the stimulus is the intersection
group of the symmetry of the crystal (in the absence of the stimulus) with the
symmetry of the stimulus (in the absence of the crystal).
 R, G, S are point groups (and not space groups).

 Let us consider a cubic crystal with 4/m 3 2/m symmetry, which is loaded along [111]
direction. The symmetry of the stimulus is m (‘conical’ symmetry). The symmetry of the
strained crystal is 3m (the common symmetry to both the crystal and the stimulus).
 To illustrate the Curie principle, let us consider the example of Ammonium Dihydrogen
Phosphate (ADP, 42m) subjected to electric filed along the 4 direction of crystal (along
[001], with a pure rotation axis of 2-fold). The stimulus (field) has a symmetry of m. The
resultant symmetry of the crystal in the presence of the electric field (stimulus) is 2mm.
 This can be understood as follows: 2-fold is a subgroup of 4. This is common to that
between 4 and  and survives. The horizontal 2-fold present in 42m is lost as this is not
present in m. The surviving symmetry is 2mm as shown and belongs to the orthorhombic
class (can also be written as mm2).
42m m 2mm

 

4
[001]crystal
 Let us now impose the electric field along other directions and see it effect.
42m [100]crystal
42m [uvw]crystal
2
m
  
 m along a
m || 2 random direction

[110]crystal
42m
 1
 

m || m m
 2W  11  11  ( E11kl kl ) 11  ( E111111  E111212  E1121 21  E1122 22 ) 11
1
W   1111
2
W  2W
Taking partial derivatives (1st wrt to 22 then wrt 11) 2  ( E1122 ) 11 2  E1122 1
 22 11 22

Now reversing the order of the partial derivatives (1st wrt to 11 then wrt 22 )
W  2W
2  (2 E111111  E111212  E1121 21  E1122 22 ) 2  E1122 2
11  2211
Since the order of differentiation should not matter as far as the energy density (W) is
concerned, (1) = (2). This implies E1122 = E2211.
 Now we apply m  Z3 (after the 3-fold operation)

11' 12' 13'   1 0 0  11 12 12   1 0 0 

 ' '   0 1 0   0 1 0 
m  Z3 
 21  '
22  23      12  11  12   
   
 31  32  33   0 0 1 12 12 11   0 0 1
' ' '
 
1 0 0  11 12 12  1 0 0   1 0 0  11 12 12 
0 1 0  12 11 12   0 1 0    0 1 0  12 11 12 
   

0 0 1 12 12 11    
 0 0 1  0 0 1 12 12 11 

incomplete

 11' 12' 13'   1 0 0   11 12 12   1 0 0 

 ' '   0 1 0    0 1 0 
m  Z3 
 21  '
22  23      12  11  12   
  
  31  32  33   0 0 1  12 12 11  0 0 1
' ' ' 
    