Quantum Transport:

Atom to Transistor

Prof. Supriyo Datta

ECE 659
Purdue University

02.17.2003

Lecture 15: Basis Functions, Density Matrix II

Ref. Chapter 4.3 & 4.4

Network for Computational Nanotechnology

 00:00
Electron Density

• Next topic shall be “Band Structure,” but
before going on we will spend another lecture
going over the important points of the
equilibrium density matrix
• Recall, electron density is given by
• The expression for electron density
may be viewed as a special case
(diagonal elements) of a much more
general concept, i.e. the equilibrium
density matrix:
[?] = f 0 ( [H] - µ[ I ] )

n (x ) = ∑ (x)
2
φα fα
• Note: [?] can be written in any basis
α or “space,” including momentum,

where fα = f 0 (ε α −µ ) frequency, and 3d Cartesian

coordinates. For any space the
diagonal elements of [?] will always
1
f0 ( E ) =
give the electron density.
and
e E kBT
+1
02.17.2003
 05:10
Function of a Matrix

• Recall, to evaluate the function of a 4. Return back to the original basis to get
matrix we must do the following … the final answer:

 1 2   1 2    †
f 0    =V 
1. Given:
f 0       V
  2 3   2 3   
(Note: Here, a Hermitian matrix is given, • So, to evaluate
(H*)T = H. Hermitian matrices in [?] = f 0 ( [H] - µ[ I ] )
diagonalized form contain only real
numbers)
We simply take
2. Diagonalize:  0. 2361 0 
 − 4 .2361
 0  f0 (ε1 − µ ) 
3: Evaluate the Diagonal Matrix: V  f 0 (ε 2 − µ ) V †
 
 f0( ) 0   O

 0 f0 ( )
02.17.2003
 Real Space Density of a 10:20
Matrix

• Summarizing, electron density in real • Another common expression for the

space is given by Vf0 (Hdiag – µI) V† density matrix in real space is:

• Recall, V is structured as follows: ρ(x, x′) = ∑φα (x) f0 (εα − µ )φα ( x′)
α
eigenstates • Finally, for any basis remember that the

[L ||| L ]
x diagonal elements of ? always give the
V = electron density “n” in that basis

where the eigenvectors are columns

• As well, the ordering of V and V† can be • What are the off diagonal elements
remembered through the mnemonic: for? Firstly, they are needed to
transform to and from basis sets.
Secondly, they are important for
 new   new   old  † calculating observables (more on this
 old   new   new  =VHdiag V later)
   
02.17.2003
 Physical Meaning of
Electron Density
• Electron density can be used to analyze
the characteristics of many systems
• For example, consider the wave
functions of the periodic and infinite wall
Boundary Conditions (B.C.)
? with periodic and box B.C
? periodic
Periodic
Box
0 x
02.17.2003
16:16
• The electron densities of the periodic and
infinite wall B.C ’s are identical at the center
but differ at the ends
Electron Density in Periodic and infinite
wall boundary Conditions
Box
Electron
Density
x
0
• From this we see intuitively that boundary
conditions can be ignored in large systems,
since it is often what occurs far from the
boundaries that we are interested in (i.e.
solids)
 Electron Density and 18:28
“Wave Function”

• Recall, given a wave function φα the • The only way this could work is to write

Ψ = ∑ φα (x )Cα
probability distribution of this wave
2
function is φ
α
α
• Hypothetically, suppose we could write And view C’s as stochastic numbers
the wave function of a system as Such that
= 0, α ≠ β
*
such that Cα Cβ
Ψ = ∑φα ( x) fα And
α
= fα
*
Cα Cβ
n = Ψ Ψ = ∑ φα (x) fα
* 2

• In a way, Ca Cß* is the density matrix in the

But this is not possible, since ? *?
eigenstate basis (?aß = fa daß)
produces cross terms

φα φβ fα f β
*

02.17.2003
 25:06
P in the Eigenstate Basis

• i.e. ih ∂ Ψ1 = ε Ψ
∂Ψ 2
• Why is it that ? is diagonal only in ih = ε 2 Ψ2
the eigenstate basis? ∂t
1 1
∂t
∂ Ψn
• To answer this question, in part, L ih = ε n Ψn and
begin by looking at the Schrödinger ∂t
Ψ j (t ) = Ψ j (0 )e
− iε j t h
Equation
∂Ψ
ih = H Ψ or
∂t
op
• Hence, there exists no overlap between
Ψ1  ε1  Ψ1  members of the eigenstate basis; so the
     Hamiltonian must be diagonal
∂Ψ Ψ2   ε2 Ψ2 
ih  = • From this one might intuitively infer that the
∂t  M   O  M  density matrix must also be diagonal with the
Ψn   εn Ψn  diagonal elements given by the appropriate Fermi
function for the state that each eigenvalue
represents. What isn’t obvious, and will be
which by separation of variables explained later, is exactly why we use fa and
(time and space) gives n linearly where it comes from.
independent equations

02.17.2003
Summary, and on to an
Example
• Main Points:
– Fermi distribution used to write
? in the eigenstate basis
– The general formulation for the
EQUILIBRIUM density matrix, in
any basis, is ?=f 0 (H – µI)
– Most properties of a material
can be calculated from the
density matrix
02.17.2003
32:00
• Example: suppose you want to
calculate current flow
I = (-q)nv,
where n is the electron density and
v is the velocity. Given that n is
known from the density matrix ?,
we must find v. To calculate v we
must develop a velocity operator
 32:47
Example: Current Flow

• The expectation value of a • Let us expand upon <v>…

v v
v = ∫ dr ∑ φα* (r )Cα* vop
quantity, such as v, is
v
v = ∫ dr Ψ *vop Ψ α
v
⋅ ∑ φβ (r )C β = ∑∑ vαβ ραβ
v
Back to devising op … we know α β
that the operator for momentum in where, in general, the matrix elements
one dimension is p op = − i h ∂ for any operator are
∂x
v v
or in general p op = − i h ∇ Aαβ = ∫ dr φα* Aopφβ

∴ v = ∑ [vρ ]αα = Trace(vρ )

And, velocity is related to
momentum by p=mv, thus
v α
v = − ih ∇
m

02.17.2003
 Box with Periodic 40:52
Boundary Conditions

• Consider a box with periodic boundary • Likewise, the state carries current in the
conditions –x direction and has negative momentum
• So, the velocity matrix is
• Recall, such a box has the appearance
hk 
 0 
V =  m
hk 
 0 − 
 m 

àx ∂
where p op = − i h
∂x
and the eigenvectors e +ikx and e -ikx Given two electrons, at equilibrium each
state is equally occupied …
• Intuitively, the state e +ikx can be said to 1 0
ρ= 
posses positive momentum and carry
0 1 
current in the +x direction

02.17.2003
 Box with Periodic 45:30
Boundary Conditions

• When voltage is applied, • In this case the velocity matrix is

-positive terminal at the +x end-
states move such that there is Cos(kx) Sin(kx)
more current to the right and less
 0 i hk 
to the left. i.e.
ν = 2 mL  where k = (2np )/L
− ihk 0 
1 .5 0   2 mL 
ρ =  • Here it is not at all clear what the eigenvalues are,
 0 0.5  due to the off-diagonal nature of V. To solve
situations such as this, particularly on a much larger
• However, interpreting ? scale, the full machinery of QM must be employed.
becomes more difficult when
working in another basis. For • For this reason, we often take a semi-classical
example, suppose instead of approach in large scale simulations. Meaning, we
(e+ikx, e -ikx) we use (cos(kx), use orthogonal basis sets providing near diagonal
sin(kx)) operator matrices. This often makes results
conceptually simpler to understand.

02.17.2003