
Inside a Transmon Qubit: Why Its Energy Levels Are Not Evenly Spaced
A transmon pairs a Josephson junction with a large shunt capacitance. Its weakly anharmonic energy ladder provides a controllable qubit transition while limiting sensitivity to charge noise.
A transmon is a superconducting circuit whose energy comes in discrete steps, a little like an atom. A Josephson junction gives the circuit its nonlinearity, while a large shunt capacitance makes it less sensitive to stray charge. Engineers use the lowest two energy levels as a qubit.
From a circuit to an energy ladder
A useful model for the transmon is H = 4E_C(n - n_g)^2 - E_J cos(phi). Here n counts Cooper pairs on the island, phi is the superconducting phase difference across the junction, n_g is an offset charge, E_C = e^2/(2C_Sigma) is the charging energy, and E_J is the Josephson energy. Adding capacitance lowers E_C, so a transmon is operated with E_J much larger than E_C. The original transmon design showed why this helps: charge dispersion falls exponentially as E_J/E_C grows, while anharmonicity shrinks more gradually, leaving enough difference between transitions to control the qubit selectively. Koch and colleagues introduced the design in 2007.
Why does that small difference matter? A perfectly harmonic oscillator has equally spaced levels. A pulse that drives the ground-to-first-excited transition would also match the next transition, making it hard to isolate a two-level system. In a transmon, E_01 = E_1 - E_0 is slightly larger than E_12 = E_2 - E_1. A microwave pulse can target the first transition, though overly strong or poorly shaped pulses can still leak population into the |2> level.
Calculate the first three levels
The cosine term couples neighboring charge states. The code below builds a small matrix representation of the model and diagonalizes it. Energies are in arbitrary units with E_C set to 1, so the output compares transition energies rather than reporting a real device frequency.
import numpy as np
EJ, EC = 30.0, 1.0
ng = 0.0
nmax = 20
charges = np.arange(-nmax, nmax + 1)
H = np.diag(4 * EC * (charges - ng) ** 2)
for k in range(len(charges) - 1):
H[k, k + 1] = H[k + 1, k] = -EJ / 2
levels = np.linalg.eigvalsh(H) # sorted from lowest to highest
E01 = levels[1] - levels[0]
E12 = levels[2] - levels[1]
print(f"E01 / EC = {E01 / EC:.2f}")
print(f"E12 / EC = {E12 / EC:.2f}")
print(f"anharmonicity / EC = {(E12 - E01) / EC:.2f}")
For these parameters, the first two transition energies differ. That is the feature a control pulse uses. The finite charge basis is only a small numerical approximation; increasing nmax is a useful convergence check, and this model does not include noise, fabrication details, or pulse errors.
How the circuit is controlled and read
On a chip, microwave pulses drive the transmon between levels. A coupled microwave resonator provides a common readout route: in the dispersive regime, the resonator response shifts depending on the qubit state, so a probe pulse carries information about whether the transmon was in |0> or |1>. An introductory circuit-QED tutorial walks through the drive and readout picture.
At the circuit-programming level, an instruction such as an X gate is an abstract operation. The hardware control stack must realize it with calibrated pulses while respecting the transmon's nonuniform spectrum. That is why an ideal gate simulator can look simple even when reliable control of the physical device takes careful calibration.