|q⟩ Bad Qubits

intermediate · Physics · Time-Independent Perturbation Theory

Two-Level Perturbation

A qubit is a two-level system

Every two-level quantum system — a spin-1/2, a two-state molecule, a superconducting transmon in its lowest two levels — is mathematically a qubit. Its most general Hermitian Hamiltonian is a combination of the Pauli matrices,

H^=aσz+bσx+cσy+d1.\hat{H} = a\,\sigma_z + b\,\sigma_x + c\,\sigma_y + d\,\mathbb{1}.

The identity term d1d\,\mathbb{1} only shifts all levels equally, so we drop it. Take the simplest nontrivial case, a "longitudinal" splitting plus a "transverse" coupling:

H^=aσz+bσx=(abba).\hat{H} = a\,\sigma_z + b\,\sigma_x = \begin{pmatrix} a & b \\ b & -a \end{pmatrix}.

The aσza\,\sigma_z part is our solvable H^(0)\hat{H}^{(0)} (eigenstates 0,1|0\rangle, |1\rangle with energies ±a\pm a); the bσxb\,\sigma_x part is the perturbation H^\hat{H}' that mixes them.

Perturbative vs exact

Because σx\sigma_x is purely off-diagonal, 0H^0=0\langle 0|\hat{H}'|0\rangle = 0, so there is no first-order energy shift. The leading correction is second order:

E0(2)=1bσx02E0(0)E1(0)=b2a(a)=b22a.E_0^{(2)} = \frac{|\langle 1|b\sigma_x|0\rangle|^2}{E_0^{(0)} - E_1^{(0)}} = \frac{b^2}{a - (-a)} = \frac{b^2}{2a}.

Wait — for the ground state we want the lower level. With a>0a > 0 the lower unperturbed level is 1|1\rangle at a-a, and its second-order shift is b2/(aa)=b2/(2a)b^2/(-a - a) = -b^2/(2a), pushing it down. This two-level model is special because we can also solve it exactly: the eigenvalues are

E±=±a2+b2.E_\pm = \pm\sqrt{a^2 + b^2}.

Expanding the lower one, a2+b2=a1+b2/a2ab22a-\sqrt{a^2 + b^2} = -a\sqrt{1 + b^2/a^2} \approx -a - \tfrac{b^2}{2a}, exactly reproduces the perturbative result. The two-level system is the cleanest place to see perturbation theory agree with the truth, term by term.

The exact eigenstate as a rotation

The exact ground state is a rotation of 0|0\rangle about the yy-axis of the Bloch sphere. Writing it as cos(θ/2)0+sin(θ/2)1\cos(\theta/2)|0\rangle + \sin(\theta/2)|1\rangle, the angle is fixed by the eigenvector of the lower eigenvalue. For a=1a = 1, b=0.5b = 0.5 one finds θ2.678\theta \approx -2.678 rad. In the small-perturbation limit bab \ll a, θπ\theta \to \pi (the ground state 1\approx |1\rangle), and the small admixture of 0|0\rangle is exactly the first-order state correction.

Try it

Build a 1-qubit circuit that prepares the exact ground state of H^=σz+0.5σx\hat{H} = \sigma_z + 0.5\,\sigma_x. Use a single RY(θ)RY(\theta) rotation with the angle determined by the lower-eigenvalue eigenvector (θ2.678\theta \approx -2.678 rad). The simulator will compare your prepared state to the reference.

Run your code to see the quantum state.

Sign in on the full site to ask questions and join the discussion.