Qubit architectures, IBM transmons and the quantum.ibm.com platform
Second installment of the series, covering physical qubit architectures, detailed IBM transmon operation, and the quantum.ibm.com platform. Originally published in French.
The core of IBM’s quantum computer relies on the transmon: a Josephson junction (two superconducting layers separated by an insulating barrier) coupled with a capacitor. Cooper pairs tunnel through the barrier, introducing a non-linearity that produces discrete, non-uniformly spaced energy levels. Only the first two levels (|0⟩ ground state, |1⟩ excited state) are selected to form the qubit. Control is achieved through calibrated microwave pulses in a heavy-hex lattice topology.
Brisbane’s Eagle R3 processors show a median T1 of 233 µs, T2 of 152 µs, best 2Q error rate of 0.325%, and 180K CLOPS throughput. The median readout error reaches 1.855%. The T1 > T2 condition is common as coherence is more fragile than simple relaxation.
The article systematically compares superconducting qubits (IBM, Google — mature and fast but noise-sensitive), trapped ions (IonQ — exceptional coherence of hundreds of seconds, but slow), topological qubits (Microsoft — theoretical fault tolerance via Majorana fermions, experimentally undemonstrated), neutral atoms (Pasqal, QuEra — scalability via optical tweezers and Rydberg states), photonic qubits (Xanadu — no cryogenics, transmission speed, but photon loss), and silicon spin qubits (Intel — CMOS compatibility, miniaturization, but limited coherence).
The article highlights the fundamental distinction: current systems are specialized quantum calculators, not universal computers. Quantum computing stands at the 1950s stage of classical computing.
Through the cloud, quantum computing power is now accessible from a personal PC: three Eagle R3 processors (127 physical qubits each) are available online with 10 minutes of computation per month, enabling quantum circuit creation, qubit entanglement, and job execution on real hardware.