Introduction to modern theory of polarization.

Bilal Hyder

May 2022
Introduction

Modern theory of polarization is concerned with the electric polarization of the dielectric

materials. There are two viewpoints when we consider the electric polarization.

• Macroscopic polarization: A vector quantity signifying electric dipole moment per

unit volume. It is one the most fundamental properties of a dielectric material and is

essential when experimentally dealing with them. It is mostly taken for granted i.e. we

just calculate it for a given material and use it for further calculations.When a dielectric

material is placed inside an electric field, we see a separation of charges in the material

and two poles are formed.

• Microscopic polarization: It refers to dealing with the polarization properties of

the material on molecular level, considering the effect of applied electric field on the

individual atoms and their electric clouds and understanding the reason behind the

phenomenon of polarization. We try to calculate the dipole moment on the molecular

level and then just average it over the entire length of the material to get the value of

polarization. If everything is done right, we should get a value similar to what we get

experimentally

In this report we will be taking baby steps to understand the calculations for microscopic

polarization and how it is scaled up to a macroscopic level.

Macroscopic polarization

Let’s start off with a brief refresh of our concepts on electric polarization. When an electric

field is applied to any material, the electrons tend to move in the direction opposite to the

direction of the applied electric field. In case of dielectrics, the electrons are not free to move

and they can not flow through the materials as is the case with metals. In dielectrics the

applied electric field causes a shift in the electron cloud of the atoms. When dealing with all

of it macroscopically, we consider the material as having two center of charges, one for the

1
negative charges and the other for positive charges. Lets consider 1D case to make things

easier to understand. We can calculate polarization using the following formula,

p⃗
P⃗ = . (1)
L

_q
+q

P⃗ here represents Polarization, p⃗ is electric dipole moment, and L is the length of the material

since we are considering the 1D case. Dipole moment can be calculated by the following

formula,

p⃗ = Σi (qi × r⃗i ). (2)

q here is the magnitude of charge on one of the centers and ⃗r is the distance between the

two centers. This concept of measuring the polarization will be useful when we actually

know the exact positions of the charges like in the high school examples, or when we have

done some sophisticated modeling of materials to determine the actual shift of charges when

an electric field is applied. While working in the lab, however, we need to come up with

something measurable, something we can detect with our measuring devices and that can

be used in our calculations. It is not possible to simply measure the charge at any point

on a piece of insulator, there is no measuring device for that. To cope with our inability

to measure charge at a point, we take another route to measure the polarization. We flip

the direction of the applied electric field and it flips the induced polarization. This is known

as polarization reversal. Due to this flip some charge will need to flow from one end to the

other. The change of polarization will be the amount of charge per unit area of electrode

2
that flows during this flip of polarization direction. To measure the charge that flows from

one end to the other after the polarization reversal we provide the charges with an external

path to flow through (a metal wire), and then trap the flowing charges inside a capacitor.

To calculate the amount of current that is stored inside the capacitor, we use the formula

Q = CV . The ”effective” polarization will be half of the change in polarization caused by

polarization reversal. Effective polarization is also known as spontaneous polarization and

signifies the polarization at equilibrium.

_ _ _ _ _ + + + + +
+ + + + _ _ _ _
_ _ _ _ + + + +
+ + + + + _ _ _ _ _
E _ _ _ _ _ E + + + + +
+ + + + _ _ _ _
_ _ _ _ + + + +
+ + + + + _ _ _ _ _

Microscopic polarization

Dealing with microscopic polarization is a bit more complex (in the simplest case). Before

we try to apply an electric field to the material and make things more intricate, let’s just go

through a simple thought experiment and see what is the situation at the molecular scale.

Consider an ionic solid just like NaCl. With no electric field on a macroscopic scale, we just

see a neutral body with no charged centers and no polarization. However, when we go to

the atomic scale, we see that there a are portions of different charges alternating one after

another i.e. a 1D chain of ions in an ionic solid.

3
Since its a thought experiment, lets define a unit cell, sit on it, and observe our surroundings.

_ + _ + _ + _ +

On our left we see identical unit cells repeating till (seemingly) infinity and same is the

situation on our right hand side. Each unit cell has a couple of opposite charges separated by

a constant distance a/2 and so it must have some polarization; let’s calculate it! Considering

the left most point of our unit cell as our 0 reference, the polarization will be

1 1 a 3a 1 2a 1
P⃗ = Σqi xi = (−1 × + 1 × ) = = . (3)
a a 4 4 a 4 2

We have got a non-zero value of polarization for this unit cell. Our solid is supposed to

be non-polar but we see these infinitely many identical unit cells with a non-zero polarization

in the chain, so should this be the polarization of the entire solid? At the end of the day are

no these unit cells that make up the entire solid as well? The answer is yes, they do, but

wait, what if we chose another unit cell by flipping the order of negative and positive ions?

_ + _ + _ + _ +

This is still a perfectly valid unit cell. Let’s calculate polarization for this arrangement.

1 1 a 3a 1 −2a −1
P⃗ = Σqi xi = (+1 × − 1 × ) = = . (4)
a a 4 4 a 4 2

Turns out that polarization is still non-zero but this time it has a different value; the mag-

nitude is same as before but the sigh in opposite. In fact, there are many more ways to

define a unit cell for this ionic chain under discussion and it turns out that every time we

get a new value of polarization. In this case, these values will be ± 12 , ± 32 , ± 52 , .... This list of

possible values of polarization is known as a polarization lattice. Notice that these values are

centrosymmetric; there are equal magnitudes on both sides of the zero polarization. These

4
polarizations will cancel each other out, and overall the system will be non-polar.

1D ionic chain in an electric field

We got away with the different values of polarization in the non-polar case by using the

centrosymmetric approach. The opposite values nullify each other and the chain gets zero

polarization in the end. But what if we had applied some electric field to the chain. The ions

in the chain would have gotten displaced from the current position and we would lose the

centrosymmetric polarization lattice. What value of polarization out of the entire lattice will

we choose to extrapolate the polarization to the macroscopic scale? Here we use a trick similar

to the one that we used when we experimentally measured the polarization of a ferroelectric.

instead of setting a reference point, we measure the difference in polarization of the unit cell

before and after the application of the electric field. The figure below shows how the ions

displace in response to the applied electric field.

_ + _ + _ + _ +

_ + _ + _ + _ +

We have already seen that the unit cells in the upper chain of figure above. Now lets check

what is the change in polarization for the two unit cells before and after the application of

the electric field. Polarization for the two unit cells after the application of electric field:

1 1 a 3a 1 −2a 1 d
i) P⃗ = Σqi xi = (+1 × − 1 × ( + d)) = = + . (5)
a a 4 4 a 4 2 a

1 1 a 3a 1 −2a −1 d
ii) P⃗ = Σqi xi = (+1 × ( + d) − 1 × ) = = + . (6)
a a 4 4 a 4 2 a

The differences in polarization for the two cases after the field is applied is then given by

5
 1 d 1 d
i) δ P⃗ = ( + ) − = (7)
2 a 2 a

1 d 1 d
ii) δ P⃗ = (− + ) − (− ) = (8)
2 a 2 a

It worked! We now have a constant value that signifies a polarization related property and

it is independent of our choice of unit cell.

Effect of applied electric field on a 3D unit cell

Let’s qualitatively see a 3D example to get an intuitive understanding of polarization induced

by an applied electric field. Considering the example of PbTiO3 we can demonstrate the effect

of polarization on this unit cell and how we can determine the change in polarization when

we flip the electric field. For the 3D case, we will replace a by Ω for the calculations because

now the polarization will be defined by dipole moment per unit volume, not length.

(a) (b) (c)

Pb Ti O

Figure 1: (a) Is the unpolarized case (no applied electric field). (b) and (c) represent the
situation after applying electric field in opposite directions.

6
Considering the electrons on the atoms

So far we have considered the movement of ions alone and assumed that the electron clouds

cling to these ions and move at the exact same position when the electric field is applied,

should obviously not be the case when we consider electrons as clouds enveloping the nuclei.

The charge density in a solid is a continuous function, but for our ease we define the position

of the electrons using Wannier functions. Wannier functions are localized in space and hence

allow us to give a somewhat confined picture of electrons that we can then use to calculate

the polarization. Wannier function, wn (r) in unit cell R associated with band n is defined

similar to a fourier transform

Z Z
Ω Ω
wn (r − R) = d3 ke−ik.R ψnk = d3 ke−ik.(r−R) unk (r) (9)
(2π)3 BZ (2π)3 BZ

where ψnk (r) = e−ik.(r−R) unk (r) are the Bloch functions with the integral over the Brillouin

zone.

The average position of the electron, calculated by taking the average of the Wannier

functions is knows as a Wannier center r⃗n

Z
r⃗n = wn∗ (⃗r)⃗rwn (⃗r)d3⃗r (10)

using momentum representation of the position operator ⃗r = −i(δ/δ⃗k), we can write the

equation for Wannier center as


 δun⃗k
Z  
Ω 3⃗ −i⃗k.R
⃗
r⃗n = i d ke un⃗k 
 (11)
(2π)3 BZ δ⃗k

When we consider the electrons as a separate unit from the ions, our understanding of 1D

ionic chain will need some alteration. Now we consider all the atomic sites as sites of positive

charges and the electron clouds carry the negative charges. Now, lets take into account the

ionic bonding to determine the location of the electronic clouds. We know form the energy

level diagram of NaCl, that the pair of electron responsible for bonding is localised near the

7
bonding orbital that lies near Cl as compared to Na. This means that our electrons must

be somewhere near the Cl ions in the atomic chain. Notice the difference from the previous

arrangement here. The electron clouds are near the Cl ionic site, not on like we previously

considered giving Cl ion the -1 charge.

+ _+_ + _+_ + _+_ + _+_

Now the the negative charge centers have a different location (the Wannier centers) than

what we previously considered, and the dipole moments, and the polarization will have dif-

ferent values than before. To cater for these changes, we modify the equation of polarization

as follows

1
P⃗ = (Σ(qi xi )ions + Σ(qn r⃗n )W Cs ) (12)
a

For the case with no applied electric field, we should not see any difference in polarization

as the net polarization should be zero. Let’s check for that. From the diagram above we can

see that the electron clouds are spread symmetrically around the Cl ion. This will result in

the Wannier center being on the Cl ion’s location itself containing a total charge of -2e.

1 a 3a a 1
P⃗ = (+1 × + 1 × −2× )= (13)
a 4 4 4 2

The result is exactly as we expected so we are good to move ahead to the fun part. We now

apply an electric field and disturb the equilibrium position of charges. The electric field will

distort the electron clouds and the Wannier center will shift its position from the Cl ion by

a distance ∆ as shown in the figure below.

_ _ _ _
+ +_ + +_ + +_ + +_

New charge center

8
 Taking into consideration this new arrangement, our polarization becomes

1 a 3a a 1 d ∆
P⃗ = (+1 × + 1 × ( + d) − 2 × ( − ∆)) = + + 2 (14)
a 4 4 4 2 a a

Notice this additional 2∆/a factor in the polarization equation? This is because the electron

clouds moved in the opposite direction to the motion of the cation, resulting in a grater

effective displacement of positive charge, and a greater value of polarization. The spontaneous

polarization in this cases will be

If we just stick to the ionic picture, can take care of this new, increased value of polarization

by defining an increased value of charge at the old ionic site. This new value of charge is

called Born effective charge and is defined by the following formula

∗ Ω δPi
Zij = (16)
e δdj

For a prototype model of 3D lattice of PbTiO3 that we saw before, gives these effective

charges as Pb +3.9, Ti +7.1, O -5.8 as compared to Pb +2, Ti +4, O -2 using the formal

charges.

Now that we have all the ingredients, let’s write a general equation for the spontaneous

polarization because in the end that is the value that actually matters.

P⃗s = δ P⃗

= P⃗f − P⃗o
1 (17)
⃗
= (Σi (qif xfi )ions + Σn (qnf rnf )W Cs − Σi (qio xoi )ions + Σn (qno r⃗no )W Cs )
Ω
1 ⃗
= (Σi (qif xfi − qio xoi )ions + Σn (qnf rnf − qno r⃗no )W Cs )
Ω

Now we substitute the expression for Wannier centers r⃗n in the above equation

9
 
*  f +
 δu ⃗
Z
1 Ω ⃗ ⃗
P⃗s = (Σi (qif xfi − qio xoi )ions + Σn (qnf (i 3
d3⃗ke−ik.R uf ⃗  nk )
Ω (2π) BZ nk δ⃗
 k

 δuo
Z * +
o Ω 3⃗ −i⃗k.R ⃗ o  ⃗
− qn (i d ke u  nk ))W Cs )
n⃗k  ⃗
(2π)3 BZ δk


*  f + *  o +
1 Ω
Z δu ⃗  δu
3⃗ −i⃗k.R
⃗ ⃗
= (Σi (qif xfi − qio xoi )ions + Σn ((qnf + qno )(i d ke f 
( u ⃗ nk o 
− un⃗k  nk ))
Ω (2π)3 BZ
nk δ⃗
 k  ⃗
δk


*  f + *  o +
1 1
Z
⃗ ⃗ δu ⃗
 δu
⃗
Σi (qif xfi − qio xoi )ions + Σn ((qnf + qno )(i d3⃗ke−ik.R ( uf ⃗  nk − uon⃗k  nk )

=

Ω (2π)3 BZ
nk δ⃗
 k  δ⃗k

An implication that should be mentioned here is that the Bloch functions and hence the

Wannier functions are not uniquely defined due to an inherent phase factor. But this will

not bother us since we are summing over all the Wannier centers in a unit cell and it is single

valued and well defined.

ABINIT

ABINIT is a software designed to simulate molecular properties of materials. It can be used

to calculate all the properties that we discussed here precisely.

Resources

[1] Nicola A. Spaldin,

A beginner’s guide to the modern theory of polarization,

Journal of Solid State Chemistry.

[2] Rabe, Karin M., et al.

”Modern physics of ferroelectrics: Essential background.” Physics of Ferroelectrics.

10
Springer, Berlin, Heidelberg, 2007.

11