Quantum Hardware and Error Correction: 20 Questions Answered

The leading types are superconducting circuits, trapped ions, neutral atoms, photonics and spins in silicon, with topological qubits still a research bet. Each trades off speed, connectivity, error rates and manufacturability.

They are tiny electrical circuits cooled until they lose all resistance, and their lowest energy levels act as 0 and 1. Microwave pulses control them. They are fast and can be made with chip fabrication, but they need extreme cooling. See superconducting qubits.

Individual charged atoms are held in place by electric fields in a vacuum, and lasers set and read their states. Ions of the same element are identical, which gives very high accuracy. Gates are generally slower than superconducting ones. See trapped ions.

Neutral atoms are held in arrays by focused laser beams called optical tweezers, and can be moved around and entangled with lasers. The approach scales to large arrays of atoms in a flexible layout. See neutral atoms.

They encode information in particles of light and process it with optical components. Photons suffer little from heat noise and can travel through fibre, which suits networking. The challenge is making photons interact and losing as few as possible. See photonic quantum computing.

A topological qubit would store information in a global property of a material, making it naturally resistant to local noise. It is an attractive idea, but demonstrating one convincingly has been difficult, and claims have been debated. See topological qubits.

A physical qubit is one real device, such as an atom or circuit. A logical qubit is a more reliable qubit encoded across many physical ones with error correction. Useful algorithms need logical qubits. See logical vs physical.

Quality matters as much as quantity. A thousand noisy qubits may do less than a hundred excellent ones, because errors pile up as programs grow. Useful comparisons look at error rates, connectivity, speed and how many operations can run before failing.

Fidelity measures how close an operation gets to the ideal result. A two-qubit gate fidelity of 99.9 percent means roughly one error in a thousand operations. Leading platforms have reported two-qubit fidelities at or above that level, though results vary by system and operation.

Decoherence is the loss of a qubit's quantum behaviour as it interacts with its surroundings. The time a qubit stays usable is called its coherence time. Longer coherence and faster gates mean more operations before the information is lost.

It is a method of protecting quantum information by encoding it redundantly and repeatedly checking for errors using extra qubits, without directly reading the data. Unlike copying bits, it must work around the rule that quantum states cannot be cloned. See the guide.

The surface code arranges qubits in a grid where neighbouring checks reveal errors. It tolerates relatively high error rates and needs only nearby connections, which makes it a popular choice, though it uses many physical qubits per logical one. See the surface code.

Error correction only helps if physical error rates are below a threshold. Below it, adding more qubits to the code makes the logical qubit better. Google reported this behaviour on its Willow chip in 2024, a widely noted milestone. See Willow.

It depends on the code and the physical error rate. With surface codes, estimates run from dozens to over a thousand physical qubits per logical qubit. Newer, more efficient codes aim to cut that overhead. See resource estimation.

Yes, at small scale. Several companies and labs have shown a handful to dozens of logical qubits, and some have shown logical error rates better than the underlying physical ones. Large-scale, fault-tolerant computing has not yet been shown. See Helios and logical qubits.

A decoder is the classical software and hardware that reads error-check results and decides which corrections to apply, fast enough to keep up with the quantum machine. Real-time decoding is a big engineering task. See real-time error correction.

Superconducting and some spin qubits need temperatures around a hundredth of a degree above absolute zero. Dilution refrigerators reach that by mixing helium isotopes. Ions and atoms avoid this by using lasers and vacuum. See cooling.

Key hurdles include lowering error rates, wiring thousands or millions of qubits, building fast control electronics, real-time decoding, manufacturing consistency and supply of components. Each is an active engineering effort rather than a fundamental barrier to physics.

In principle, yes, using photons to connect modules, and research groups have demonstrated links over short distances. Networking modules is a leading route to larger machines. See distributed quantum computing.

Nobody knows, and there may not be one winner. Different platforms may suit different jobs, and the leaders have changed over the years. Treat confident predictions with caution, and look at independent benchmarks. See benchmarks explained.