PHYS 90507 Topology and Dirac fermions in condensed matter Assaf

Lecture 4: Edge states between two media of opposite mass:

1 0 0 0 0 0 0 1 0 0 0 −𝑖
𝐻2𝐷 = 𝑚(𝑥)𝑣 (0
2 1 0 0 ) + ℏ𝑣𝑘 (0
𝑥 0
0 1 0) + ℏ𝑣𝑘 ( 0 0
𝑦 0 −𝑖
𝑖 0)
0 0 −1 0 1 0 0 0 0
0 0 0 −1 1 0 0 0 𝑖 0 0 0
The mass changes sign at x=0, so that:
𝑚(𝑥 > 0) = +𝑚2
𝑚(𝑥 < 0) = −𝑚2

𝑚(𝑥)𝑣 2 0 0 ℏ𝑣(𝑘𝑥 − 𝑖𝑘𝑦 )

2 ℏ𝑣(𝑘 + 𝑖𝑘 ) 0
0 𝑚(𝑥)𝑣 𝑥 𝑦
𝐻2𝐷 𝜓 = 𝜓 = 𝐸𝜓
0 ℏ𝑣(𝑘𝑥 − 𝑖𝑘𝑦 ) −𝑚(𝑥)𝑣 2 0
(
ℏ𝑣(𝑘𝑥 + 𝑖𝑘𝑦 ) 0 0 −𝑚(𝑥)𝑣 2 )
The Hamiltonian can be decoupled into two equations:
𝑚(𝑥)𝑣 2 ℏ𝑣(𝑘𝑥 − 𝑖𝑘𝑦 )
𝐻𝜓 = ( ) 𝜓 = (𝐻1𝐷 + ℏ𝑣𝑘𝑦 𝜎𝑦 )𝜓 = 𝐸𝜓
ℏ𝑣(𝑘𝑥 + 𝑖𝑘𝑦 ) −𝑚(𝑥)𝑣 2

And,
𝑚(𝑥)𝑣 2 ℏ𝑣(𝑘𝑥 + 𝑖𝑘𝑦 )
𝐻†𝜓 = ( ) 𝜓 = 𝐸𝜓 = (𝐻1𝐷 − ℏ𝑣𝑘𝑦 𝜎𝑦 )𝜓 = 𝐸𝜓
ℏ𝑣(𝑘𝑥 − 𝑖𝑘𝑦 ) −𝑚(𝑥)𝑣 2

It seems obvious that if we can separate the variable x and y, the two equations describe identical states
propagating with opposite velocities ±𝑣.
It is more convincing to get to this solution in a more rigorous way. We first look at the +𝑣 Hamiltonian
and look for solutions of the form:
𝜑+ +
𝜓+ (𝑥 > 0, 𝑦) = ( 1+ ) 𝑒 𝑖𝜅𝑥 𝑥+𝑖𝑘𝑦 𝑦
𝜑2
𝜑− −
𝜓− (𝑥 < 0, 𝑦) = (𝜑1− ) 𝑒 𝑖𝜅𝑥 𝑥+𝑖𝑘𝑦 𝑦
2

As before, we will solve this problem with boundary condition constraints at x=0 and x=±∞. The system
is a continuum in the y-direction.
Solution for x>0
𝐻2𝐷 𝜓+ = 𝐸𝜓+
Leads to the following:
𝑚2 𝑣 2 − 𝐸 ℏ𝑣(𝑘̂𝑥 − 𝑖𝑘̂𝑦 ) 𝜑1+ 𝑖𝜅+𝑥+𝑖𝑘 𝑦
( ) ( +) 𝑒 𝑥 𝑦 = 0
ℏ𝑣(𝑘̂𝑥 + 𝑖𝑘̂𝑦 ) −𝑚2 𝑣 2 − 𝐸 𝜑2

Replacing the wavevector operators 𝑘̂𝑥,𝑦 = −𝑖𝜕𝑥,𝑦 and having them act on the wavefunction leads to:
PHYS 90507 Topology and Dirac fermions in condensed matter Assaf

𝑚2 𝑣 2 − 𝐸 ℏ𝑣(𝜅𝑥+ − 𝑖𝑘𝑦 ) 𝜑1+

( ) ( +) = 0
ℏ𝑣(𝜅𝑥+ + 𝑖𝑘𝑦 ) −𝑚2 𝑣 2 − 𝐸 𝜑2

The root of determinant of the matrix Hamiltonian can now be extracted to determine 𝜅𝑥+ :

𝐸 2 − 𝑚22 𝑣 4 − (ℏ𝑣)2 [(𝜅𝑥+ )2 + 𝑘𝑦2 ] = 0

Which leads to,

2
√𝐸 2 − 𝑚22 𝑣 4 − (ℏ𝑣𝑘𝑦 )
𝜅𝑥+ =± (1)
ℏ𝑣
The eigenvector can now be found using the first row of the matrix Hamiltonian:

(𝑚2 𝑣 2 − 𝐸)𝜑1+ + ℏ𝑣(𝜅𝑥+ − 𝑖𝑘𝑦 )𝜑2+ = 0

Or,

ℏ𝑣(𝜅𝑥+ − 𝑖𝑘𝑦 )𝜑2+

𝜑1+ =
𝐸 − 𝑚2 𝑣 2
So the wavefunction in the x>0 region takes the form:

ℏ𝑣(𝜅𝑥+ − 𝑖𝑘𝑦 )
+
𝜓+ (𝑥, 𝑦) = ( 𝐸 − 𝑚2 𝑣 2 ) 𝑒 𝑖𝜅𝑥 𝑥+𝑖𝑘𝑦 𝑦
1

Solution for x<0

𝐻2𝐷 𝜓− = 𝐸𝜓−
Leads to the following:
−𝑚1 𝑣 2 − 𝐸 ℏ𝑣(𝑘̂𝑥 − 𝑖𝑘̂𝑦 ) 𝜑1− 𝑖𝜅−𝑥+𝑖𝑘 𝑦
( ) (𝜑− ) 𝑒 𝑥 𝑦 = 0
ℏ𝑣(𝑘̂𝑥 + 𝑖𝑘̂𝑦 ) +𝑚1 𝑣 2 − 𝐸 2

Replacing the wavevector operators 𝑘̂𝑥,𝑦 = −𝑖𝜕𝑥,𝑦 and having them act on the wavefunction leads to:

−𝑚1 𝑣 2 − 𝐸 ℏ𝑣(𝜅𝑥− − 𝑖𝑘𝑦 ) 𝜑1−

( ) (𝜑− ) = 0
ℏ𝑣(𝜅𝑥− + 𝑖𝑘𝑦 ) 𝑚1 𝑣 2 − 𝐸 2

The root of determinant of the matrix Hamiltonian can now be extracted to determine 𝜅𝑥− :

𝐸 2 − 𝑚12 𝑣 4 − (ℏ𝑣)2 [(𝜅𝑥− )2 + 𝑘𝑦2 ] = 0

Which leads to,

2
√𝐸 2 − 𝑚12 𝑣 4 − (ℏ𝑣𝑘𝑦 )
𝜅𝑥− = ± (2)
ℏ𝑣
PHYS 90507 Topology and Dirac fermions in condensed matter Assaf

The eigenvector can now be found using the first row of the matrix Hamiltonian:

(−𝑚1 𝑣 2 − 𝐸)𝜑1− + ℏ𝑣(𝜅𝑥− − 𝑖𝑘𝑦 )𝜑2− = 0

Or,

ℏ𝑣(𝜅𝑥− − 𝑖𝑘𝑦 )𝜑2−

𝜑1− =
𝐸 + 𝑚1 𝑣 2
So, the wavefunction in the x<0 region takes the form:

ℏ𝑣(𝜅𝑥− − 𝑖𝑘𝑦 )
−
𝜓− (𝑥, 𝑦) = ( 𝐸 + 𝑚 𝑣 2 ) 𝑒 𝑖𝜅𝑥 𝑥+𝑖𝑘𝑦 𝑦
1
1
Boundary conditions at x=0
At x=0, the wavefunction needs to be continuous. In a conventional Schrödinger problem, the first
derivative of the wavefunction should also be continuous, however, with the mass changing as we move
across x=0, this condition is no longer satisfied. The derivative weighted by the mass has to be continuous
instead. This results from the current probability continuity across x=0. This is detailed in Appendix A at
the end of the notes for lecture 4 (this file). In the end, the wavefunction needs to satisfy:

𝜓+ (𝑥 < 0) = 𝜓− (𝑥 < 0) (3𝑎)

And,
1 𝑑𝜓+ 1 𝑑𝜓−
| = | (3𝑏)
𝑚(𝑥 > 0) 𝑑𝑥 𝑥=0 𝑚(𝑥 < 0) 𝑑𝑥 𝑥=0
The condition 3(a) gives:

ℏ𝑣(𝜅𝑥+ − 𝑖𝑘𝑦 ) ℏ𝑣(𝜅𝑥− − 𝑖𝑘𝑦 )

=
𝐸 − 𝑚2 𝑣 2 𝐸 + 𝑚1 𝑣 2
And the condition 3(b) gives:
1 𝑑𝜓+ 1 𝑑𝜓−
| =− |
𝑚2 𝑑𝑥 𝑥=0 𝑚1 𝑑𝑥 𝑥=0

𝑖𝜅𝑥+ ℏ𝑣(𝜅𝑥+ − 𝑖𝑘𝑦 ) 𝑖𝜅𝑥− ℏ𝑣(𝜅𝑥− − 𝑖𝑘𝑦 )

= −
𝑚2 𝐸 − 𝑚2 𝑣 2 𝑚1 𝐸 + 𝑚1 𝑣 2
Combining the two boundary conditions leads to a simple equation that we can solve to find the energy:
𝜅𝑥+ 𝜅𝑥−
=−
𝑚2 𝑚1
Given the square root that is present the expressions for 𝜅𝑥− and 𝜅𝑥+ it is convenient to square the last
equation,
PHYS 90507 Topology and Dirac fermions in condensed matter Assaf

2
𝜅𝑥+ 𝜅𝑥− 2
( ) =( )
𝑚2 𝑚1

Using equations (1) and (2), we find:

2 2
𝐸 2 − 𝑚22 𝑣 4 − (ℏ𝑣𝑘𝑦 ) 𝐸 2 − 𝑚12 𝑣 4 − (ℏ𝑣𝑘𝑦 )
=
𝑚22 𝑚12

Or,
2 2
[𝐸 2 − 𝑚22 𝑣 4 − (ℏ𝑣𝑘𝑦 ) ] 𝑚12 = [𝐸 2 − 𝑚12 𝑣 4 − (ℏ𝑣𝑘𝑦 ) ] 𝑚22
2 2
[𝐸 2 𝑚12 − 𝑚12 𝑚22 𝑣 4 − (ℏ𝑣𝑘𝑦 ) 𝑚12 ] = [𝐸 2 𝑚22 − 𝑚12 𝑚22 𝑣 4 − (ℏ𝑣𝑘𝑦 ) 𝑚22 ]

We get a simple equation to solve for E:

2
𝐸 2 (𝑚12 − 𝑚22 ) − (ℏ𝑣𝑘𝑦 ) (𝑚12 − 𝑚22 ) = 0

As long as, 𝑚12 ≠ 𝑚22 , we have,

𝐸 = ±ℏ𝑣𝑘𝑦

Approach 1
Dirac edge states
Let us now find the shape of the wavefunction at both sides of the interface:
With the expression found for the energy, we get for x>0:

√−𝑚22 𝑣 4 ±𝑖𝑚2 𝑣
𝜅𝑥+ = ± = = ±𝑖𝜆+
ℏ𝑣 ℏ
And,
±𝑖𝑚2 𝑣
ℏ𝑣 ( − 𝑖𝑘𝑦 )
ℏ
𝜓+ (𝑥, 𝑦) = ( ) 𝑒 ±𝜆+𝑥+𝑖𝑘𝑦 𝑦
𝐸 − 𝑚2 𝑣 2
1
Since 𝜓+ (𝑥 → ∞) → 0, we have to reject the −𝑖𝜆+solution to get:
𝜅𝑥+ = 𝑖𝜆+

𝑖(𝑚2 𝑣 2 − ℏ𝑣𝑘𝑦 )
𝜓+ (𝑥, 𝑦) = ( 𝐸 − 𝑚2 𝑣 2 ) 𝑒 −𝜆+𝑥+𝑖𝑘𝑦 𝑦
1
It is easy to check that 𝜓+ is an eigenvector that satisfies ℏ𝑣𝑘𝑦 𝜎𝑦 𝜓+ = 𝐸𝜓+ if and only if,

𝐸 = +ℏ𝑣𝑘𝑦

For which,
PHYS 90507 Topology and Dirac fermions in condensed matter Assaf

−𝑖
𝜓+ (𝑥, 𝑦) = ( ) 𝑒 −𝜆+𝑥+𝑖𝑘𝑦 𝑦
1

Similarly, for x<0, we get:

−𝑖𝑚1 𝑣
𝜅𝑥− = = −𝑖𝜆−
ℏ
−𝑖(𝑚1 𝑣 2 + ℏ𝑣𝑘𝑦 )
𝜓− (𝑥) = ( 𝐸 + 𝑚1 𝑣 2 ) 𝑒 𝜆−𝑥+𝑖𝑘𝑦 𝑦
1
For
𝐸 = +ℏ𝑣𝑘𝑦
−𝑖
𝜓− (𝑥) = ( ) 𝑒 𝜆−𝑥+𝑖𝑘𝑦 𝑦
1
Negative velocity branch
We get the negative velocity branch simply by changing the sign of v:

𝑖(𝑚2 𝑣 2 + ℏ𝑣𝑘𝑦 )
𝜓+ (𝑥, 𝑦) = ( 𝐸 − 𝑚 𝑣 2 ) 𝑒 −𝜆+𝑥+𝑖𝑘𝑦 𝑦
2
1

−𝑖(𝑚1 𝑣 2 − ℏ𝑣𝑘𝑦 )
𝜓− (𝑥) = ( 𝐸 + 𝑚1 𝑣 2 ) 𝑒 𝜆−𝑥+𝑖𝑘𝑦 𝑦
1
𝐸 = −ℏ𝑣𝑘𝑦

Yields,
−𝑖
𝜓+ (𝑥, 𝑦) = ( ) 𝑒 −𝜆+𝑥+𝑖𝑘𝑦 𝑦
1
−𝑖
𝜓− (𝑥, 𝑦) = ( ) 𝑒 𝜆−𝑥+𝑖𝑘𝑦 𝑦
1
Approach 2
Using the 4x4 matrix and the following expression for 𝜆± :
𝑚(𝑥)𝑣
𝜆± =
ℏ
𝑚(𝑥)𝑣 2 − 𝐸 0 0 𝑖(𝑚(𝑥)𝑣 2 − 𝑖ℏ𝑣𝑘𝑦 )
0 𝑚(𝑥)𝑣 2 − 𝐸 𝑖(𝑚(𝑥)𝑣 2 + 𝑖ℏ𝑣𝑘𝑦 ) 0
2 𝜓=0
0 𝑖(𝑚(𝑥)𝑣 − 𝑖ℏ𝑣𝑘𝑦 ) −𝑚(𝑥)𝑣 − 𝐸
2
0
2
(𝑖(𝑚(𝑥)𝑣 + 𝑖ℏ𝑣𝑘𝑦 ) 0 0 −𝑚(𝑥)𝑣 2 − 𝐸 )

We can check that indeed the inner block gives the -v branch and the out block gives the +v branch:
PHYS 90507 Topology and Dirac fermions in condensed matter Assaf

For the positive branch:

𝑚(𝑥)𝑣 2 − 𝐸 0 0 𝑖(𝑚(𝑥)𝑣 2 − ℏ𝑣𝑘𝑦 )

−𝑖
0 𝑚(𝑥)𝑣 2 − 𝐸 𝑖(𝑚(𝑥)𝑣 2 + ℏ𝑣𝑘𝑦 ) 0
(0)=0
0 𝑖(𝑚(𝑥)𝑣 2 − ℏ𝑣𝑘𝑦 ) −𝑚(𝑥)𝑣 2 − 𝐸 0 0
2
(𝑖(𝑚(𝑥)𝑣 + ℏ𝑣𝑘𝑦 ) 0 0 −𝑚(𝑥)𝑣 2 − 𝐸 ) 1

The first line gives

−𝑖(𝑚(𝑥)𝑣 2 − ℏ𝑣𝑘𝑦 ) + 𝑖(𝑚(𝑥)𝑣 2 − ℏ𝑣𝑘𝑦 ) = 0

It is indeed zero.
For the negative branch:

𝑚(𝑥)𝑣 2 − 𝐸 0 0 𝑖(𝑚(𝑥)𝑣 2 − ℏ𝑣𝑘𝑦 )

0
0 𝑚(𝑥)𝑣 2 − 𝐸 𝑖(𝑚(𝑥)𝑣 2 + ℏ𝑣𝑘𝑦 ) 0 −𝑖
( )=0
0 𝑖(𝑚(𝑥)𝑣 2 − ℏ𝑣𝑘𝑦 ) −𝑚(𝑥)𝑣 2 − 𝐸 0 1
2
(𝑖(𝑚(𝑥)𝑣 + ℏ𝑣𝑘𝑦 ) 0 0 −𝑚(𝑥)𝑣 − 𝐸 ) 0
2

From the second line, we get,

−𝑖(𝑚(𝑥)𝑣 2 + ℏ𝑣𝑘𝑦 ) + 𝑖(𝑚(𝑥)𝑣 2 + ℏ𝑣𝑘𝑦 ) = 0

Indeed zero.
Therefore, the two solution can be thought of as 1D states that propagate with opposite velocity along ky
and penetrate into each side by an amount defined by the characteristic depth 𝜆± .

Edge states
It is impressive how edge states systematically emerge from the Jackiw-Rebbi treatment at the interface
between two ‘Dirac fermion’ systems having opposite masses. The physics community never appreciated
this fact, until the realization that ‘Dirac fermions can be made using materials’ came to be.
The simplest definition of topological matter emerges from this treatment and can be stated as follows:
A change in the sign of the mass (or gap) at an interface, yields states that are pinned to this interface
and confined to a direction of motion parallel to the interface.
Definition of topological non-triviality
Note however, that the changing sign of the Dirac mass is only relative in this treatment, in the sense that
there is no ‘stand-alone’ topologically non-trivial state. In this form, one may wonder whether m1 or m2 is
have the unconventional sign.
Nevertheless, in semiconductor systems that host this sign inversion of the mass (or gap), this issue is
quickly solved by the fact that the band structure is almost never purely Dirac-like. The matrix Hamiltonian
for most semiconductors is given by a combination of Dirac and Schrödinger terms. It is of the following
form:
PHYS 90507 Topology and Dirac fermions in condensed matter Assaf

0 𝑘𝑥 − 𝑖𝑘𝑦
1 0 0
ℏ 𝑘 2 2 0 𝑘𝑥 + 𝑖𝑘𝑦 0
𝐻2𝐷 2
= (𝑚(𝑥)𝑣 + ) (0 1 ) + ℏ𝑣
2𝑚
̃ −1 0 0 𝑘𝑥 − 𝑖𝑘𝑦
0 0
0 −1 𝑘𝑥 + 𝑖𝑘𝑦 0
( )
𝑚
The sign of 𝑚
̃
then allows us to determine the topological character of the system. The exact nature of the
𝑚
̃ term will be revealed when we discuss real topological materials.