← Quantum Store · Guides · Lessons · Glossary · Packages · About · Türkçe
Quantum machine learning and hardware course · Lesson 4 / 5 · 12 min
Kinds of quantum hardware: from superconductors to photons
Qubits can be made from superconducting circuits, trapped ions, neutral atoms, photons and electrons in silicon. Compare the strengths and weaknesses of each approach.
Superconducting qubits
The superconducting qubits used by IBM and Google are small electrical circuits cooled to about 10–20 millikelvin, only about a hundredth of a degree above absolute zero. Their gates are very fast (tens of nanoseconds) and the chips are made with semiconductor manufacturing techniques.
Their weaknesses are short coherence times, each qubit being connected to only a few neighbours, and the need for large refrigerators. The surface code you saw in this course was tried on superconducting chips precisely because it fits this neighbour structure.
Ions and neutral atoms
The trapped ions used by Quantinuum and IonQ are atoms held in mid-air by electric fields and controlled with lasers. Because every ion is identical the qubits are very consistent; two-qubit gate fidelities have passed 99.9% and every qubit can interact with every other. The price is speed: gates take hundreds of times longer than on superconductors.
The neutral atoms used by QuEra, Pasqal and Atom Computing are held in arrays by laser beams called “optical tweezers”. Arrays of thousands of atoms have been built, and atoms can be moved during a computation. In 2023 a Harvard and QuEra team used this approach for an error-correction experiment with 48 logical qubits.
Photons, silicon and topological qubits
Photonic qubits (PsiQuantum, Xanadu, Quandela) are particles of light; they can travel at room temperature and be carried far through fibre, but photons interact with each other only with difficulty and can be lost on the way. Spin qubits in silicon (for example Intel’s) can be made in today’s chip factories and are very small, but have so far been shown with only a few qubits.
Microsoft is working on topological qubits, which are expected to be naturally resistant to noise, and announced its first chip in 2025; this claim is still debated among scientists. It is not clear which approach will win; different hardware may also come out ahead for different applications.
Try it yourself
On every kind of hardware the hardest operation is the two-qubit gate: the CNOT in this Bell circuit. When judging a device, the first numbers to look at are the two-qubit gate fidelity and which qubits can do a CNOT with each other directly. Remove the CNOT: the entanglement disappears.
Open the circuit in your browserThe circuit playground is in Turkish. Button guide: “1000 kez ölç” = measure 1000 times, “Durum vektörünü göster” = show the state vector, “Bu devrenin Qiskit kodu” = this circuit’s Qiskit code, “Geri al” = undo, “Temizle” = clear. Click a gate in the circuit to remove it.
Key points
- Superconducting qubits are fast but need extreme cold and are connected only to their neighbours.
- Trapped ions are very accurate and fully connected, but their gates are slow.
- Neutral atoms scale to arrays of thousands of qubits, and the atoms can be moved.
- Photons, silicon spins and topological qubits are other routes promising different advantages.
Test yourself
What is the most distinctive advantage of trapped-ion qubits?
- Very high gate fidelity and every qubit being able to interact with every other
- Having the fastest gates
- Needing no cooling or vacuum at all
Show the answer
Very high gate fidelity and every qubit being able to interact with every other. Being identical atoms they are very consistent and fully connected; the price is that their gates are slow.
← Quantum kernels: comparing data with quantum states · The roadmap: fault-tolerant quantum computers and your next step →