Before I start, I would like to say sorry for this possibly stupid question of mine. I just recently got into the topic of quantum computing and try me there.

I'm currently trying to display a SU(2) rotation using Qiskit.

from ibm_quantum_widgets import CircuitComposer

from qiskit.quantum_info import Statevector
from qiskit.visualization import plot_bloch_multivector

from qiskit import QuantumRegister, ClassicalRegister, QuantumCircuit
from numpy import pi
import numpy as np
from qiskit.quantum_info.operators import Operator, Pauli

qreg_q = QuantumRegister(1, 'q')
circuit = QuantumCircuit(qreg_q)

clreg = ClassicalRegister(1)


After the implementations, I tried to achieve such a rotation with some matrices. This is one of them:

enter image description here

def rn_su2_5(theta, n1,n2,n3):
    #This represents a matrix operator that will evolve() a Statevector by matrix-vector multiplication and will evolve() a DensityMatrix by left and right multiplication
    return Operator([
    [np.cos(theta/2)- 1j*n3*np.sin(theta/2), -1j*(n1-1j*n2)*np.sin(theta/2)],
    [-1j*(n1+1j*n2)*np.sin(theta/2), np.cos(theta/2)+1j*n3*np.sin(theta/2)]
    ],input_dims=(2, 1), output_dims=(2, 1))



Operator([[ 6.123234e-17-1.j, -1.000000e+00-1.j],
          [ 1.000000e+00-1.j,  6.123234e-17+1.j]],
         input_dims=(2, 1), output_dims=(2, 1))

**ExtensionError: 'Input matrix is not unitary.'**

I would be very happy if you could help me with this problem.

Later the rotation should also be visible on a Blochsphere.

If the code is not sufficient, I will submit it later if necessary.

  • $\begingroup$ For single qubits (representable in a Bloch sphere), you can use the usual rotation gates rx,ry,rz to perform special unitary rotations. $\endgroup$
    – Mauricio
    May 24, 2022 at 20:01
  • $\begingroup$ @Fation: It seems that your question has an academic background investigating physical properties under the hood. You should mention that you are aware of the possibilities to use out-of-the-box gates, but you want investigate $SU(2)$ matrices. $\endgroup$ May 25, 2022 at 5:42
  • $\begingroup$ In fact, my question is specifically related to the use of SU(2) matrices. I know there are out-of-the-box options. But I can't use this for my further development. $\endgroup$
    – Fation
    May 25, 2022 at 7:05
  • $\begingroup$ It might be helpful to check what the is_unitary_matrix function does. It computes A^dagger.A and checks if it is identity matrix: See Code here. $\endgroup$ May 26, 2022 at 6:38

1 Answer 1


The problem is that the vector n=[n1,n2,n3] must have an magnitude (length) of $1$, see for example this document (page 8, equation 50).

I implemented a debug line that shows you wether the matrix is unitary or not. When choosing n=[0,0,1] or n=[1/np.sqrt(3), 1/np.sqrt(3), 1/np.sqrt(3)] the matrix becomes unitary.

# Matrix taken from
# https://docplayer.org/117986458-Die-symmetriegruppen-so-3-und-su-2.html (p. 11, equation 46)
# https://www.uni-muenster.de/Physik.TP/archive/fileadmin/lehre/teilchen/ws1011/SO3SU2.pdf (p. 8, equation 50)

def rn_su2(theta, n):
    n1 = n[0]
    n2 = n[1]
    n3 = n[2]
    return Operator([
        [np.cos(theta/2) - 1j*n3*np.sin(theta/2), -1j*(n1 - 1j*n2)*np.sin(theta/2)],
        [-1j*(n1 + 1j*n2)*np.sin(theta/2), np.cos(theta/2) + 1j*n3*np.sin(theta/2)]
    ], input_dims=(2, 1), output_dims=(2, 1))

# Magnitude of the vector n must be 1
n = [0,0,1]
n = [1/np.sqrt(3), 1/np.sqrt(3), 1/np.sqrt(3)]

# Debug: check if the matrix is unitary
mat = np.array(rn_su2(5, n))
# Compute A^dagger.A and see if it is identity matrix
mat = np.conj(mat.T).dot(mat)

# construct the operator
rotated = circuit.unitary(rn_su2(pi, n), 0)

I checked in the notebook here at Github.

The rotation by $\frac{\pi}{4}$ around the $z$-axis can be performed as follows (for this I oriented myself to this lecture note):

rot_angle = pi/4
n = [0, 0, 1]
rot_operator = rn_su2(rot_angle, n)
rot_matrix = np.array(rot_operator)
start_vec = [1, 0, 0]

_bloch = Bloch()
_bloch.vector_color = ['blue', 'red']

sv = []
vec = start_vec

spherical_vec = to_spherical(vec)
ϕ = spherical_vec[1]
θ = spherical_vec[2]

sx = msigma(1)
sy = msigma(2)
sz = msigma(3)
M_q = (np.sin(θ)*np.cos(ϕ)*sx + np.sin(θ)*np.sin(ϕ)*sy + np.cos(θ)*sz)
U_n = np.eye(2)*np.cos(rot_angle/2) -1j*(n[0]*sx+n[1]*sy+n[2]*sz)*np.sin(rot_angle/2)
M_q_rotated = U_n*M_q*np.matrix(U_n).H
cos_θ_rotated = float(N(re(M_q_rotated[0,0])))
θ_rotated = np.arccos(cos_θ_rotated)

#e^(ix) = cos(x) + i*sin(x)
#see https://en.wikipedia.org/wiki/Euler%27s_identity
temp = float(N(re(M_q_rotated[1,0])))
temp = temp/np.sin(θ_rotated)
ϕ_rotated = np.arccos(temp)

vec = np.array(to_cartesian([1, θ_rotated, ϕ_rotated]))


which looks as follows:

enter image description here

Note that in the code above some polishing still need to be done such as exception handling of corner cases (e.g. in trigonometric functions).


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.