Some of the most accurate quantum computers ever built store their information in single atoms suspended in a vacuum. This note builds from a high-school-physics picture up to the industry view — the ions IonQ, Quantinuum and AQT use, how the laser light is delivered, and how quantum gates are actually performed.
Start with a normal atom. Knock one electron off it and the atom becomes an ion — an atom carrying a net electric charge. That charge is the whole trick, because charged objects can be pushed and pulled with electric fields, the same way a balloon rubbed on your hair sticks to a wall.
A trapped ion is a single charged atom held almost perfectly still inside an ultra-high vacuum chamber — emptier than outer space — by carefully shaped electric fields. You can't hold a charged particle still with steady fields alone (it always squirts out somewhere), so traps use a rapidly oscillating field that, on average, pushes the ion back toward the centre from every direction — like spinning a marble around the inside of a bowl so fast it never falls out. Line up several ions and they space themselves into a neat row, because like charges repel and each ion pushes on its neighbours.
Now, where is the "0" and "1"? Each ion's electrons can sit in different energy levels. We pick two of those levels and call one "0" and the other "1". That pair of levels is the qubit. Because every atom of a given type is exactly identical, every qubit starts out perfectly the same — there is no manufacturing variation, unlike chips.
Everything else is done with laser light:
A trapped-ion quantum computer is a row of identical atoms held in place by electric fields and controlled with finely tuned light — atoms make perfect, identical qubits, which is why they hold the records for accuracy.
Only a handful of elements make good ion qubits — they need a convenient electronic structure and a single outer electron that lasers can address cleanly. The favourites are ytterbium, barium, calcium and strontium. But the more important distinction is which two energy levels you choose for 0 and 1:
The 0 and 1 are two very closely spaced "ground" states, split by the interaction between the electron and the atom's nucleus (this needs an isotope with nuclear spin, e.g. ¹⁷¹Yb⁺ or ¹³³Ba⁺). The energy gap is tiny — a microwave frequency — so these qubits are driven with microwaves or with a pair of laser beams (a "Raman" transition). Their great virtue is coherence: they can hold a superposition for seconds to minutes.
The 0 and 1 are a ground state and a long-lived "metastable" excited state, separated by a much larger, optical-frequency gap (e.g. ⁴⁰Ca⁺). A single ultra-stable laser flips between them. Simpler control, but the qubit lifetime is set by how long the excited state survives, and it demands an exceptionally quiet laser.
A practical wrinkle drives a lot of recent engineering: the wavelength of light each ion needs. Ytterbium's key transition sits in the deep ultraviolet (369 nm), where optics are finicky and lasers are expensive. Barium and calcium work in the visible range (green/red), which is far easier and cheaper to generate, stabilise and route. That single fact is reshaping the industry.
The trapped-ion field is unusually concentrated, with two commercial leaders and a strong European cohort. Their choices of ion and control method are the clearest fingerprints of their strategy.
| Company | Qubit ion | Control | Architecture | Note |
|---|---|---|---|---|
| IonQ | Barium ¹³³Ba⁺ (from ytterbium) | Laser Electronic | Linear chain, all-to-all | Moved Yb→Ba (Tempo); acquired Oxford Ionics (~$1.07B) for chip-based electronic control |
| Quantinuum | Barium ¹³⁷Ba⁺ (Yb-171 coolant) | Laser | QCCD (ions shuttle between zones) | Helios: 98 qubits, all-to-all; switched Yb→Ba for visible-light control |
| AQT | Calcium ⁴⁰Ca⁺ (optical) | Laser | Linear chain, room-temp rack | Innsbruck spinout; rack-mounted IBEX / LYNX, available on AWS Braket |
| Oxford Ionics (now IonQ) | Calcium | Electronic | Integrated trap chip | "Electronic Qubit Control" — gates from on-chip electronics, ~99.99% two-qubit fidelity |
| eleQtron | Ytterbium | Microwave | Linear chain | Germany; "MAGIC" magnetic-gradient microwave gates; €57M Series A (2026) |
| Universal Quantum | Ytterbium | Microwave | QCCD, modular | UK; microwave control aimed at large-scale modular machines |
Two big bets are playing out: a move from ytterbium to barium (visible light instead of fragile UV), and a move from laser-based to electronic/microwave control (putting the control onto fabricated chips). Both are really about the same goal — making the machine manufacturable and scalable rather than a physics-lab optical table.
In a laser-controlled ion computer, light does almost everything — and a single machine juggles many wavelengths at once, each with a job:
The light typically begins in tunable diode or Ti:Sapphire lasers, often frequency-doubled to reach the needed colour and locked to an ultra-stable reference (an optical cavity or frequency comb). It is then routed to the trap through free-space optics and optical fibre. Along the way, acousto-optic and electro-optic modulators (AOMs / EOMs) act as the high-speed shutters and tuners — switching beams on and off in nanoseconds and setting their exact frequency, intensity and phase, which is how a pulse becomes a precise gate.
Hitting one ion in a row without disturbing its neighbours is the hard part — individual addressing. IonQ, for example, uses acousto-optic deflectors to steer and split a beam into many independently controlled spots, one per ion.
All of this lives on a vibration-isolated optical table of mirrors, lenses and fibres. It works beautifully for tens of ions and becomes unwieldy for thousands. The fix the whole field is chasing: photonic integrated circuits that route the light through waveguides built into the trap chip itself, plus electronic control (Oxford Ionics, eleQtron) that replaces gate lasers with on-chip fields. This is where the real supply chain lives — lasers, optics, PIC foundries, vacuum and trap fabrication — and the subject of our next report.
To rotate one qubit, you hit it with a pulse tuned to the 0↔1 transition — microwaves for a hyperfine qubit, a laser (or Raman beam pair) for an optical one, or an engineered field in the electronic-control approach. The duration and phase of the pulse set exactly how far around the Bloch sphere the state rotates. These gates are now routinely better than 99.99% accurate.
This is the elegant heart of trapped-ion computing. Because the ions are charged and packed in a line, they can't move independently — push one and the whole chain ripples, like beads on a taut string. These shared vibrations (motional modes) are a communication channel every ion can access.
A two-qubit gate uses laser light to apply a force on two chosen ions that depends on their qubit states, briefly exciting a shared vibration and then returning it to rest. The motion does the talking: when it comes back to where it started, the two qubits are left entangled, having picked up a quantum phase that depends on both their states. The standard recipe is the Mølmer–Sørensen gate, which drives this with a two-tone laser field and is robust to the ions' residual heat.
Because the motional bus connects every ion in the chain, trapped ions get all-to-all connectivity for free — any qubit can interact directly with any other, with no need to shuffle data through neighbours. That, plus record fidelities and identical qubits, is the platform's core advantage. The cost is speed: gates take microseconds to milliseconds, far slower than superconducting qubits, so ion machines win on quality-per-operation rather than raw clock rate.
A single ion chain can't grow forever — pack in too many ions and the shared vibrations become a crowded, noisy mess. The industry's answers to that define the competitive map:
For investors and analysts, trapped ions occupy a clear position: the accuracy and connectivity leaders, which is why they post the field's best logical-qubit results today, traded against slower gates and a genuinely hard scaling problem. The bet on any given company is really a bet on how it plans to scale — shuttling, photonic networking, or moving control onto silicon — and on the unglamorous supply chain of lasers, optics, vacuum and trap chips that makes any of it manufacturable.
How does a trapped-ion quantum computer work?
It holds charged atoms (ions) almost still in an ultra-high vacuum using electric fields, then uses precisely tuned laser or microwave pulses to cool the ions, run quantum gates, and read them out by state-dependent fluorescence.
Which companies build trapped-ion quantum computers?
IonQ and Quantinuum are the commercial leaders, with AQT, eleQtron and Universal Quantum in Europe. The ions used include ytterbium, barium and calcium.
What is a Mølmer–Sørensen gate?
It is the standard trapped-ion two-qubit gate: a two-tone laser field applies a state-dependent force through the ions' shared vibration (motional mode), entangling them via a geometric phase.
Company dossiers: IonQ · Quantinuum. Related explainers: Quantum Error Correction · QEC Code Architectures.
A GroundState market-insight report on the trapped-ion ecosystem — the key suppliers of lasers, optics, photonic integrated circuits, foundries, cryogenics and vacuum hardware that quietly determine who can scale. See all explainers →