# Using Qiskit VQE for solving an eigenvalue problem

I am a new user of Qiskit and I believe there is a simple answer to my question but I have had a very hard time finding a straightforward answer. I am trying to transform a given $$3 \times 3$$ (Hermitian) Hamiltonian matrix into a qubit operator and then use the built-in VQE solver for evaluating the minimum eigenvalue for said matrix.

I have been successful in doing something similar using the qiskit.chemistry.FermionicOperator built-in class (see example below) but for this example the Hamiltonian is mapped to a Fermionic one-body qubit Hamiltonian which has $$2^n$$ eigenvalues instead of $$n$$ eigenvalues (where $$n=3$$ for this case). I am hoping there is a simple way to directly map a matrix to a qubit operator. My vague understanding is I need to write the matrix as a sum of weighted Pauli operators but I would expect there would exist some predefined function for doing this? (I have looked at the source code for the FermionicOperator class but it did not help me much).

import numpy as np
import scipy
import h5py

from qiskit.aqua.algorithms import VQE, NumPyEigensolver
from qiskit.chemistry import FermionicOperator
from qiskit import Aer

backend = Aer.get_backend("qasm_simulator")

c1=1
c2=2
c3=3

ff=np.zeros((3,3))
ff[1,1]=2
ff[2,2]=1

n=3

Hamil=np.zeros((n,n))
Hamil[1,1]=c2
Hamil[2,2]=-c3/2+c2
Hamil[0,2]=np.sqrt(2)*c1
Hamil[1,2]=np.sqrt(2)*c1
Hamil[2,0]=np.sqrt(2)*c1
Hamil[2,1]=np.sqrt(2)*c1

vals,vecs=np.linalg.eig(Hamil)

print("Standard Eigenvalues: ")
print(vals)

Hamil_op = FermionicOperator(h1=Hamil)
Hamil_ops = Hamil_op.mapping(map_type='parity', threshold=1e-12)
result = NumPyEigensolver(Hamil_ops,k=int(2**n)).run()

print("Quibit Eigenvalues=")
print(result['eigenvalues'])

vqe = VQE(operator=Hamil_ops)
vqe_result = np.real(vqe.run(backend)['eigenvalue'])
print("VQE Eigenvalue")
print(vqe_result)


With result:

Standard Eigenvalues:
[-1.45426242  0.90078898  3.05347344]
Quibit Eigenvalues=
[-1.45426242+0.j -0.55347344+0.j  0.        +0.j  0.90078898+0.j
1.59921102+0.j  2.5       +0.j  3.05347344+0.j  3.95426242+0.j]
VQE Eigenvalue
0.18608074335373637

• Welcome to the community! Dec 2, 2020 at 20:03
• If I understand correctly, you are trying to decompose your $3 \times 3$ Hamiltonian into Pauli strings? If this is the case then it is not possible as product of Pauli matrices are in dimension of $2^n$. Dec 2, 2020 at 21:43
• Thank you for making this point clear. Still, the question stands: is it possible to represent a 3x3 matrix as a qubit operator in qiskit for use in the VQE? Dec 2, 2020 at 22:08
• Hello, I understand you want to create an Operator from a matrix, did you try the Operator class in qiskit.quantum_info? Here is the tutorial from Qiskit, does this help? qiskit.org/documentation/tutorials/circuits_advanced/…
– Lena
Dec 3, 2020 at 9:52
• Thanks for the suggestion Lena, I tried looking through that documentation but it does not seem to be compatible with the NumPyEigensolver class (at least not in a straighforward way that I can figure out). Dec 4, 2020 at 19:37

I believe to have answered my own question through further attempts. Per the comment from KAJ226, it is true that a $$(3 \times 3)$$ matrix cannot be represented as a string of Paulis so I have nested the $$(3 \times 3)$$ matrix into a $$(4 \times 4)$$ matrix where the final row and column are all zeros. From this I can omit the useless eigenvalue/eigenvector and just retain the relevant results for the $$(3 \times 3)$$ matrix.

My solution mostly came from using the WeightedPauliOperator, MatrixOperator, and op_converter classes. Below is a sample of my code with corresponding outputs. Thanks to those who gave suggestions to help me out!

import numpy as np
import scipy
import h5py

from qiskit.aqua.algorithms import VQE, NumPyEigensolver
from qiskit.aqua.operators import WeightedPauliOperator, MatrixOperator, op_converter
from qiskit import Aer

backend = Aer.get_backend("qasm_simulator")

c1=-0.5
c2=.75
c3=2

n1=3
Hamil=np.zeros((n1,n1))
Hamil[1,1]=c2
Hamil[2,2]=-c3/2+c2
Hamil[0,2]=c1
Hamil[1,2]=c1
Hamil[2,0]=c1
Hamil[2,1]=c1

print("(3x3) Hamiltonian")
print(Hamil)

vals,vecs=np.linalg.eig(Hamil)

print("Standard Eigenvalues: ")
print(vals)

n2=4
Hamil=np.zeros((n2,n2))
Hamil[1,1]=c2
Hamil[2,2]=-c3/2+c2
Hamil[0,2]=c1
Hamil[1,2]=c1
Hamil[2,0]=c1
Hamil[2,1]=c1

print("(4x4) Hamiltonian")
print(Hamil)

Hamil_Mat=MatrixOperator(Hamil)
Hamil_Qop = op_converter.to_weighted_pauli_operator(Hamil_Mat)

q_vals = NumPyEigensolver(Hamil_Qop,k=4).run()

print("Qubit Op Eigenvalues: ")
print(q_vals['eigenvalues'])

vqe=VQE(Hamil_Qop)
vqe_result=vqe.run(backend)

print("VQE Eigenvalue: ")
print(vqe_result['eigenvalue'])


With result:

(3x3) Hamiltonian
[[ 0.    0.   -0.5 ]
[ 0.    0.75 -0.5 ]
[-0.5  -0.5  -0.25]]
Standard Eigenvalues:
[-0.75  0.25  1.  ]
(4x4) Hamiltonian
[[ 0.    0.   -0.5   0.  ]
[ 0.    0.75 -0.5   0.  ]
[-0.5  -0.5  -0.25  0.  ]
[ 0.    0.    0.    0.  ]]
Qubit Op Eigenvalues:
[-0.75+0.j  0.  +0.j  0.25+0.j  1.  +0.j]
VQE Eigenvalue:
(-0.705078125+0j)