Project : Matrix Product State

Quantum Computation and Information Course Project Report

P797: QUANTUM INFORMATION AND QUANTUM COMPUTATION

Student Information
Full Name Roll No.
Swarup Kuamr Giri 22134006

Lab Instructor: Prof. Anamitra Mukherjee

Teaching Assistant: Mr. Gopal Mistry

Department of Physics
National Institute of Science Education and Research (NISER)
Bhubaneshwar, India
May 5, 2024
Contents

1 Introduction 1
1.1 The importance of matrix product states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2 Theoretical Consequence Points 2

2.1 Quantum mechanics review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.2 Entanglement Quantification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
2.3 Tensor networks and their corresponding diagrams . . . . . . . . . . . . . . . . . . . . . . 3
2.4 The AKLT matrix product state . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
2.5 Operators of matrix products . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
2.6 An Iterative search for the ground states . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
2.7 Developing the Effective Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.8 Magnetism, correlations, and entanglement entropy . . . . . . . . . . . . . . . . . . . . . . 9

3 The use of MPS in physical models 10

3.1 The one-dimensional ising model in a transverse field . . . . . . . . . . . . . . . . . . . . . 10

4 THEORETICAL CONCEPTS 11
4.1 Solution 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

5 APPLICATION OF MPS TO PHYSICAL MODELS 12

6 References 12

A Appendix A title 13
 Project : Matrix Product State

1 | Introduction
1.1 | The importance of matrix product states
Quantum many-body systems, particularly strongly correlated systems within condensed matter physics,
are extremely important in modern physics. Examples of systems with intriguing properties include
spin glasses [16], frustrated magnets [18], and superconductors [3]. Even simple models of these systems
present a challenge to both analytical and numerical solution approaches due to the large number of
interacting degrees of freedom. The dimension of the underlying Hilbert space increases exponentially
with the number of particles in a system. If it can be described by differential systems on a symplectic
manifold, and there are as many constants of motion as degrees of freedom, it can be solved analytically,
and the system is called ”integrable” [5].
However, the vast majority of such models are not integrable and must be treated numerically. To
determine the eigenvalues and eigenvectors of a Hamiltonian scale disproportionate to the number of
particles or total system length, ”brute force” diagonalization methods can only be applied to small
systems because of limited computing power. Fortunately, most physical systems can be adequately
described by focusing on local interactions, which locate the important ground states in a small portion of
the vast Hilbert space.
For such a system, the matrix product state (MPS) formalism proves to be an effective variational approach
for rebuilding the ground states at zero temperature with minimal processing power. It is analogous
to parametrizing a more compact submanifold of Hilbert space, describing only the sectors with less
entanglement between the system’s subparts by construction. Interesting states include those with low
energy and a few excited thermal states. It has been demonstrated [6] that the ground and thermal states
of Hamiltonians with local interactions follow the area law for entanglement, which immediately connects
low-energy states to those with low entanglement.
The popularity of MPS in recent years stems from the DMRG method, developed by S. White et al. in
the 1990s [20].
The primary goal of this thesis is to provide an introduction to tensor networks and the reformulated
DMRG as part of the MPS formalism from a quantum information perspective. A DMRG program will
be tested for the Ising and BLBQ Heisenberg models, comparing the numerical results to the available
analytical solutions.

Page 1
 Project : Matrix Product State

2 | Theoretical Consequence Points

To introduce the MPS formalism, we will proceed as follows: The fundamentals of quantum mechanics
will be covered in order to understand the definition of entanglement. The next section explains how to
measure entanglement in a quantum mechanical system. After that, the fundamentals of tensor networks
will be covered to introduce the graphical representation for MPS. Understanding a simple MPS requires
only a brief overview of the Affleck, Lieb, Kennedy and Tasaki (AKLT) model. The purpose of the
following section is to lay the groundwork for the implementation of a DMRG algorithm reformulated in
the MPS language.

Some paragraphs on measuring observables with the MPS obtained by the DMRG software are a crossover
to the demonstration section.

2.1 | Quantum mechanics review

P
The probability is always conserved ( Pi = 1). Composite systems are described by tensor products of
finite spaces.
Ψ(1) ∈ H1 , ψ (2) ∈ H2 , composite system Ψ(3) ∈ H3 = H1 ⊗ H2 . . In terms of basis states, e.g. qubits

{|0⟩1 , |1⟩1 } ∈ H1 , {|0⟩2 , |1⟩2 } ∈ H2 , {|0⟩1 |0⟩2 , |1⟩1 |1⟩2 , |0⟩1 |1⟩2 , |1⟩1 |0⟩2 } ∈ H3 , (2.1)

dim(H1 ) = dim(H2 ) = 2, dim(H3 ) = dim(H1 ) × dim(H2 ) = 4.

 some possible states for |ψ3 ⟩ are:

Therefore,

 |0⟩1 ⊗ |1⟩2 ,
 √1 (|0⟩1 ⊗ |0⟩2 + |0⟩1 ⊗ |1⟩2 ) ≡ |0⟩1 |+⟩2 , |+⟩ = √1 (|0⟩ + |1⟩),

2 2
|ψ3 ⟩ = √1 (|0⟩1 ⊗ |0⟩2 − |0⟩1 ⊗ |1⟩2 − |1⟩1 ⊗ |0⟩2 + |1⟩1 ⊗ |1⟩2 ) ,

 2
 √1 (|0⟩1 ⊗ |0⟩2 + |1⟩1 ⊗ |1⟩2 ) .

2
Whenever a state is not described by a product of states of the corresponding subsystems, the system is
called entangled. Otherwise, it is called separable.
|ψ⟩ ∈ H1 ⊗ H2 is entangled iff. ∄ |ψi ⟩ ∈ Hi ⇒ |ψ⟩ = |ψ1 ⟩ ⊗ |ψ2 ⟩ Examples of separable and entangled
states are: 
 |0,
 0⟩ → separable
 √1 (|0, 0⟩ + |0, 1⟩ + |1, 0⟩ + |1, 1⟩) = |+, +⟩ → separable ,

2
|ψ⟩ = √1 (|0, 0⟩ + |1, 1⟩) ≡ |Φ+ ⟩ → entangled ,

 2
 √1 (|1, 0⟩ − |0, 1⟩) ≡ |Ψ− ⟩ → entangled .

2

The Einstein-Podolsky-Rosen (EPR) paradox describes entangled states as having a distance-independent

correlation. Consider a |ψ − ⟩ state in which one qubit is sent to Darmstadt and the other remains in
Mainz. A spin-”0” measurement in Mainz will immediately collapse 1 of the qubit in Darmstadt to |1⟩.
Entanglement is not a local attribute of |ψ⟩. Therefore, it does not change under local unitary operators,
for example, E (|ψ1,2 ⟩) = E (U1 ⊗ U2 |ψ1,2 ⟩), where U1 , U2 are unitary operators.

2.2 | Entanglement Quantification

There always exists a so-called Schmidt decomposition. There always exists a so-called Schmidt decompo-
sition.
m
X
|ψA,B ⟩ = λα |α⟩A |α⟩B
α

assuming m ≤ min (dim (HA ) , dim (HB ))and ⟨α | α′ ⟩A = ⟨α | α′ ⟩B = δαα′ . The singular value decomposi-
tion results in an orthonormal basis for both HA and HB .
1. |α⟩A , |α⟩B are known as Schmidt vectors.
2. m = 1 ⇔ |ψA,B ⟩ is separate.
3. Schmidt rank, or m, is a first measure of entanglement. The more m, the more intertwined the state.

The entropy can be determined by decomposing the state using SVD to read out the singular values,
which directly refer to the entanglement of the bond between the two subsystems, or by computing the

Page 2
 Project : Matrix Product State

reduced density matrix. The reduced density matrix of a state with common orthonormal basis |ψ⟩ of two
systems A and B is defined as the partial trace of the common density matrix ϱAB = ϱA ⊗ ϱB ,

ϱA/B ≡ trB/A (ϱAB ) .

The Schmidt decomposition simplifies this to:

r
* r
+
X X
2
ϱA = λα |α⟩A α|A , ϱB = λ2α |α ⟨ α|B .
α=1 α=1 B

The von Neumann entropy is calculated using the reduced density matrix ϱA/B .



SA/B ≡ − tr ϱA/B log ϱA/B
Xr
λ2α log λ2α .


=−
α

2.3 | Tensor networks and their corresponding diagrams

A tensor is a system of complex numbers organized in multiple arrays. Its rank is determined by the
number of indexes. Thus, a single integer x is a tensor of rank zero, a vector xα of rank one, the Kronecker
delta δil of rank two, and the Levi-Civita tensor ϵijk of rank three. To introduce tensor networks and
tensor network contractions, a graphical depiction of tensors is typically used (see Fig.2. 1).

Figure 2.1: Tensor graph examples. Scalar, vector, matrix, and rank-3 tensor are shown in order from
left to right. Each index is represented by a tensor leg.

Tensor contractions are sums of identical indices from distinct tensors.

D
X
Cαλ = Aαβ Bβλ (2.2)
β=1

Equation 2.2 illustrates a contraction of index β. The non-contracted indices α and λ are termed open
indices. These standards produce a graphical depiction of contractions. As a result, Figure 2 shows a
matrix multiplication that is equivalent to Equation 2.2.

Figure 2.2: Contraction 2.2 is shown graphically.

This notation visually represents a complex contraction and highlights several aspects that are hidden in
a large number of indices, such as the cyclic property of a trace of a matrix product (see Fig. 3).

Figure 2.3: Trace of a matrix product in graphical and analytical notation, using sum convention.

Assuming A and B are (m × m) matrices, matrix multiplication in Eq. 2.2 requires O(m3 ) operations for
contraction.This number of operations cannot be lowered using simple matrix multiplication methods,

Page 3
 Project : Matrix Product State

however, when dealing with more complicated tensor contractions, the sequential arrangement must be
considered. For example, both contractions in Fig. 4
provide the same outcome but require a different number of operations. This is necessary for programming
the contraction and determining the optimized computation time.

Figure 2.4: Trace of a matrix product in graphical and analytical notation, using sum convention.

2.4 | The AKLT matrix product state

The AKLT model’s valence-bond-states provide an excellent example of MPS concepts and physical
understanding.
Consider the Hamiltonian
B 2
X 
⃗l S
 
⃗l+1 + β S⃗l S
⃗l+1 1
H= α S and α = 1, β = (2.3)
3
l=1

For a 1D spin-1 chain of length L, B = L − 1 for an open boundary condition (OBC) and B = L
for a periodic boundary condition (PBC), with L + 1 ≡ 1. Each particle can inhabit one of three
orthonormal states. |σϵ{|+⟩, |0⟩, |−⟩}. The spin operators can be expressed component by component:
⃗l = S x , S y , S z .T


S l l l
Slj = I ⊗ I ⊗ . . . ⊗S j ⊗ I · · · ⊗ I jϵ{x, y, z}
| {z } | {z }
l−1 L−l

Using their explicit forms

     
0 1 0 0 −i 0 1 0 0
1 1
Sx = √  1 0 1  , Sy = √  i 0 −i  , S z =  0 0 0 .
2 0 1 0 2 0 i 0 0 0 −1

In Eq. 2.3, each Hamiltonian term takes the form

2
⃗B + 1 S
  
Hobc,l = S⃗A · S ⃗A · S
⃗B ,
3
which may be reduced to
1 2
2 2

 1 2

SA SB = (SA + SB ) −(SA + SB = Stot − 4
2 2 
0 − 4 Stot = 0
1 
= · 2 − 4 Stot = 1
2 
6 − 4 Stot = 2

2 4 2
⇒ Hobc,l = − (P0 + P1 ) + P2 = − ⊮ + 2P2 ,
3 3 3
using the spin projection operators Ps . In OBC ( open boundary condition), the ground state of Hl is
four-fold degenerate with the states.

Page 4
 Project : Matrix Product State

1
|1, +1⟩ = √ (|+⟩l |0⟩l+1 − |0⟩l |+⟩l+1 ) ,
2
1
|1, 0⟩ = √ (|+⟩l |−⟩l+1 − |−⟩l |+⟩l+1 ) ,
2
1
|1, −1⟩ = √ (|−⟩l |0⟩l+1 − |0⟩l |−⟩l+1 ) ,
2
1
|0, 0⟩ = √ (|+⟩l |−⟩l+1 + |−⟩l |+⟩l+1 − |0⟩l |0⟩l+1 ) .
3
The three triplet states are defined as
1
|+⟩ = | ↑↑⟩, |0⟩ = √ (| ↑↓⟩ + | ↓↑⟩), |−⟩ = | ↓↓⟩. (2.4)
2
To understand the formalism and structure of MPS, consider the introduction of a mapping, which results
in a simple non-trivial tensor network as a solution for the ground state.
According to Eq. 2.4, the Hamiltonian’s ground state in Eq. 2.3 should be made up of linkages that never
result in a total spin-2 state. A spin-1 chain can achieve this state by projecting two symmetric spin-1/2
subcomponents onto a triplet state at each lattice point. All spin-1/2 pairings between adjacent spin-1
systems contract to a singlet state with spin zero.

Figure 2.5: AKLT is the state for Open boundary conditions. The grey tensors with a leg at the bottom
represent the contraction of two (blue) virtual spin-1/2 to one physical spin. The zigzag line represents
singlet boundaries between adjacent spin-1/2. By construction, the boundary spin-1/2 is free.

Let |a⟩ = |a1 . . . aL ⟩ , |b⟩ = |b1 . . . bL ⟩ denote the left and right spin1/2 on each point. The state |ψ⟩ can
be expressed as its subcomponents. XX
|ψ⟩ = cab |ab⟩.
a b

One can consider the bonding condition.

E X
Σ[i] = Σbi ai+1 |bi ai+1 ⟩
bi ai+1

and rewrite the four constants Σbi ai+1 as a 2 × 2 matrix

!
0 √1
Σ= 2 .
− √12 0

It is now possible to write the complete state using singlet-bonded spin-1/2 particles.
XX
|ψΣ ⟩ = Σb1 a2 Σb2 a3 . . . ΣbL−1 aL |ab⟩
a b

The state is represented as a matrix product, with local states made up of two constituents. To incorporate
σ
the projection to physical spin-1, design a mapping operator Mab ′ that acts on |σ⟩⟨ab|. In matrix form,

this yields !
0 √1
   
+ 1 0 0 2 − 0 0
M = , M = √1
, M = .
0 0 2
0 0 1

The total state can be written as

XX
|ψ⟩ = Maσ11b1 Σb1 a2 Maσ22b2 Σb2 a3 . . . ΣbL−1 aL MaσLLbL |σ⟩
σ a,b

Page 5
 Project : Matrix Product State

or equivalently by the AKLT matrices Aσ ∝ M σ Σ, normalised in the thermodynamic limit (system size
L→∞)  q 
q ! 1
!
0 2 − 3 0 0 0
+ 0 −
A = 3 , A = q , A = q .
0 0 0 1 − 23 0
3

An explicit representation of a wave function

[1] [2] [L−1] [L]
X
|ψ⟩ = .Aσ1 Aσ2 . . . AσL−1 .AσL |σ1 σ2 . . . σL−1 σL ⟩ , (2.5)
σ1 ...σL

with rank 3 tensors. A[i] represents site-dependent dimensions. The 1D matrix product state is represented
[i]
by (mi × mi+1 × di ). Except for the borders, the matrices Aσi in the AKLT model are all equivalent
and have the dimension ((mi = 2) × (mi+1 = 2) × (di = 3)). As a result, the whole phrase simplifies. To
summarize briefly: The ground state can be rewritten as an MPS by introducing maximally entangled
subcomponents.

2.4.1 | Sequential Decomposition of Singular Values

The MPS structure of Eq. 2.5 can be obtained for any arbitrary spin-1 state using sequential singular
value decomposition (SSVD) with 3L coefficients cσ1 ...σL . Figure 6 graphically depicts the main notion,
which comprises of the phases listed below:

Figure 2.6: Starting from Ψσ1 ,...,σL , the MPS can be obtained using sequential reshaping, SVD, and
permutation, as needed.

1. Create a tensor Ψσ1 ...σL with dimensions (3 × 3 · · · × 3 × 3) to represent the 3L coefficients cσ1 ,...,σL .
| {z }
L
2. Transform the tensor into a matrix Ψσ1 ...σL → Ψ̃σ1 (σ2 ...σL ) .

Page 6
 Project : Matrix Product State

3. A SVD Ψ̃σ1 (σ2 ...σL ) . produces

r1
Uσ1 α1 Λα1 α1 Vα†1 (σ2 ...σL ) .
X
Ψ̃σ1 (σ2 ...σL ) =
α1 =1

4. Reshape Uσ1 α1 into a tensor Uασ11 ≡ Γσα11 , and reshape the matrix Λα1 α1 Vα+1 (σ2 ...σL ) → Ṽ(α1 σ2 )(σ3 ...σL )
and create an additional SVD to gain

V(α1 σ2 )(σ3 ...σL ) . = U(α1 σ2 )α2 λα2 α2 Vα+2 (σ3 ...σL )

σ
5. Reshape U(αl σl+1 )αl+1 → Γαl+1
l αl+1
. then do the SVD with Λαl+1 αl+1 Vα+l+1 (σl+2 ...σL ) for l = 1, . . . (L − 2).
6.ReshapethematrixΛαL−1 αL−1 Vα+L−1 σL → ΓσαLL−1 .
Note: At each step, the matrices U, Λ, V , and Γ vary. Thus, the coefficients cσ1 ...σL are obtained by the
contraction of L separate Γ matrices.
[1] [2] [L−1] [L]
Cσ1 ...σL = Γσ1 Γσ2 . . . ΓσL−1 ΓσL

At each step, the matrices U, Λ, V , and Γ vary. Thus, the coefficients cσ1 ...σL are obtained by the
contraction of L separate Γ matrices.
[1] [2] [L−1] [L]
Cσ1 ...σL = Γσ1 Γσ2 . . . ΓσL−1 ΓσL (2.6)

A naı̈ve decomposition would result in the splitting of matrices with exponential rising dimensions.
However, the SVD shows that these dimensions may be greatly reduced since information between two
subsystems is constrained. In the AKLT condition, the Schmidt rank is even constant at m = 2.

2.5 | Operators of matrix products

Given a physical state in MPS form (Eq. 2.6),
[1] [2] [L−1]
⟨σ | ψ⟩ = Γσ1 Γσ2 . . . ΓσL−1 ΓσL .[L] (2.7)

To compute expectation values, it is more advantageous to use an analogue representation for operators O
[1] [2] [L−1] [L]
⟨σ|O|σ ′ ⟩ = Ωσ1 σ′ Ωσ2 σ′ . . . ΩσL−1 σ′ Ωσ L σ ′
1 2 L−1 L

Any operator can be written as:

X X
O= c(σ1 ...σL )(σ′ ...σ′ ) |σ⟩ ⟨σ ′ |
1 L
σ1 ...σL σ ′ ...σ ′
1 L

may be decomposed similarly to Fig. 6 using SSVD, but with two physical indices σi , σi′ instead of only
one. This means that in graphical notation, each tensor of a matrix product operator (MPO) at site j has
rank-4, and the pictorial representation consists of two vertical lines, one for an ingoing quantum number
and one for an outgoing one. The finished MPO includes two physical legs vertically and two auxiliary
legs horizontally (see Fig. 7). The boundary tensors can also be written using dummy indices.

σ ,σ ′
Figure 2.7: The MPO chain has L sites. At site j, each tensor Ωαjj ,βlj contains two physical indices
σj , σj′ as well as two auxiliary indices αj , βj .

Page 7
 Project : Matrix Product State

Fortunately, given an operator described as a sum of local operators (for example, the link model
Hamiltonian or magnetization), the MPO form may be immediately P written′ down. Consider the local
′
block Ωσσ and its projector |σ⟩ ⟨σ ′ | to obtain a local operator Ω̂i = σσ′ Ω.σσ |σ⟩ ⟨σ ′ | applies to the local
quantum numbers σ and σ ′ . The total operator is a composite of operators acting on many local Hilbert
spaces.
O = Ω̂1 Ω̂2 . . . Ω̂L
As a result, contractions of these operator-valued tensors produce a total of tensor products of their
respective operators. A simple example should demonstrate the fundamental notion underlying the
compact Hamiltonian form. The 1D Ising Hamiltonian in a transverse field with OBC is given by
L
X L
X
HIsing = − Jij Six Sjx − Γ Siz . (2.8)
i=1,j=1 i=1

The Hamiltonian is formally represented as S x/z = 12 σx/z . In this thesis, ℏ = c0 = 1 for simplicity. In
the following, we assume that this coupling is constant for each particle pair: Jij = 0 for j ̸= i + 1 and
Ji,i+1 = const. = J.
Specifically, the Hamiltonian has the form

H = − JSx ⊗ Sx ⊗ I ⊗ I ⊗ . . . ⊗ I − hSz ⊗ I ⊗ I ⊗ . . . ⊗ I
− I ⊗ JSx ⊗ Sx ⊗ I ⊗ . . . ⊗ I − I ⊗ hSz ⊗ I ⊗ . . . ⊗ I
...
− I ⊗ I ⊗ . . . ⊗ I ⊗ JSx ⊗ Sx − I ⊗ I ⊗ . . . ⊗ hSz ⊗ I
− I ⊗ I ⊗ . . . ⊗ I ⊗ hSz .

The tensors in Fig. 8 provide an efficient way to express this Hamiltonian. The assignment of matrices to
vertical legs only allows for combinations that exist in the Hamiltonian.

Figure 2.8: The Ising model’s local tensors are represented in MPO. The ”else = 0” condition ensures
that only terms from the sum can appear in the final contraction.

σi σ ′ σj σ ′ P σ σ′ σj σ′
After a contraction of two local Hamiltonian tensors Ωαβ i j = γ Ωαγi i Ωγβ j , multiplications that do
not occur in the sum are forbidden by the last condition in Fig. 8. Furthermore, except for an increase in
matrix size, they maintain the MPS form invariant:
X σ σ′ σ σ′ σL−1 σ ′ σ σ′ σ′ σ′ σ′ σ′
O|ψ⟩ = 1 1
Ω1,b 1
Ωb12,b22 . . . ΩbL−1 bL−1
L−1
ΩbLL,1L Γ1,a
1 L−1
Γ 2 . . . ΓaL−1
1 a1 ,a2 ,aL ΓaL,1 |σ⟩
L

a,b,σ,σ ′
X  σ1 σ ′ σ1′

σ σ′ σ′
  σ σ′
L−1
′
σL−1

σ σ ′ σL ′ 
= Ω1,b1σ1 Γ1,a 1
Ωb12,b22 Γa12,a2 . . . ΩbL−1 bLL−1 ΓaL−1 ,aL ΩbLL,1L ΓaL,1 |σ⟩
a,b,σ,σ ′
 
[1] [2] [L−1] [L] [j] [j]σ
X
= Nσ1 Nσ2 . . . NσL−1 NσL |σ⟩, with Nσj = N(bj ajj ),(bj+1 ,aj+1 ) .
σ

2.6 | An Iterative search for the ground states

In nanotechnology, where materials and devices operate at nanometer scales, understanding and manipu-
lating local kinetic energy is fundamental. This knowledge aids in the design of nanoscale materials and
devices for various applications.

Page 8
 Project : Matrix Product State

2.7 | Developing the Effective Hamiltonian

Future studies on this topic will very certainly incorporate sophisticated computational approaches, such
as density functional theory (DFT), to more precisely determine local kinetic energy. Interdisciplinary
collaborations involving physicists, chemists, and engineers will also be formed to further investigate the
applications of this basic notion.

2.8 | Magnetism, correlations, and entanglement entropy

Page 9
 Project : Matrix Product State

3 | The use of MPS in physical models

3.1 | The one-dimensional ising model in a transverse field
3.1.1 | Magnetisation
3.1.2 | Correlations
3.1.3 | Entropy
3.1.4 | onclusions and Perspectives

Page 10
 Project : Matrix Product State

4 | THEORETICAL CONCEPTS
4.1 | Solution 4
Analytical Calculation:-
We know, k = 2 and κ = 4
Let’s choose A=1,ℏ = 1,m=1 for our whole calculation.

C 2 B 1 − i κk
= , = (4.1)
A 1 + i κk A 1 + i κk

Again, √
k = 2mE/ℏ (4.2)
p
κ = 2m(Vo − E)/ℏ (4.3)
From Equation (4.2) and Equation (4.3), putting the value of ℏ,m ; we get E=2,Vo =10.
From the total energy,

E = T + Vo (4.4)

Where T=kinetic energy of the particle.

When we apply this only to the left side, we get T=2 as potential Vo = 0.
The wave function continuously decays exponentially in the right site x > 0, for the right site it will be,
T=E-Vo =-8

Figure 4.1: Potential step and propagation direction of particle

Analytical Kinetic Energy (Left side; where potential energy

(4.5)
= 0, x < 0) = 2.0
From Equation (4.4), Analytical Kinetic Energy (Right side,
(4.6)
x > 0) = −8.0
Calculation using Python:-
1

2 import cmath
3 from math import *

Page 11
 Project : Matrix Product State

5 def psistep (x , nametoval ) :

6 # Given values form question
7 kappa = nametoval [ " kappa " ]
8 k = nametoval [ " k " ]
9 C = 2 / (1 + 2 j )
10 A = nametoval [ " A " ]
11 B = (1 - 2 j ) / (1 + 2 j )
12

13

14 if x > 0:
15 return C * cmath . exp ( - kappa * x )
16

17 else :
18 return A * cmath . exp (1 j * k * x ) + B * cmath . exp ( -1 j * k * x )
19

20 def kinetic_step ( psi , x , nametoval , h =0.0005) :

21 psiold = psi (x , nametoval )
22 psip = psi ( x + h , nametoval )
23 psim = psi ( x - h , nametoval )
24 lapl = ( psip + psim - 2. * psiold ) / h ** 2
25 kin = -0.5 * lapl . real / psiold . real # Ensure real values as kinetic energy
is real
26 return kin
27

28 def main () :
29 nametoval = { " k " : 2 , " kappa " : 4 , " A " : 1 , " C " : 2 / (1 + 2 j ) ," B " : (1 - 2 j ) /
(1 + 2 j ) ,}
30 x = -1.0 # Particle located at x < 0
31

32 # Calculate kinetic energy for x < 0

33 k i n e t i c _ene rgy_x _lt_ 0 = kinetic_step ( psistep , x , nametoval )
34 print ( " Kinetic ␣ energy ␣ of ␣ ␣ the ␣ particle , when ␣ it ␣ is ␣ located ␣ at ␣ x ␣ <␣ 0: " ,
k i n etic _ener gy_x _lt_ 0 )
35

36 x = 1.0 # Particle located at x > 0

37 # Calculate kinetic energy for x > 0
38 k i n e t i c _ene rgy_x _gt_ 0 = kinetic_step ( psistep , x , nametoval )
39 print ( " Kinetic ␣ energy ␣ of ␣ ␣ the ␣ particle , when ␣ it ␣ is ␣ located ␣ ␣ at ␣ x ␣ >␣ 0: " ,
k i n etic _ener gy_x _gt_ 0 )
40

41 if __name__ == ’ __main__ ’:
42 main ()

Listing 1: Python Code

output:-

Kinetic energy when x < 0 :1.9999998346321277

Kinetic energy when x > 0 : − 8.00000266694459

5 | APPLICATION OF MPS TO PHYSICAL MODELS

6 | References

Page 12
 Project : Matrix Product State

A | Appendix A title

Page 13