Picture a single atom — one lone charged atom, an ion — hanging motionless in a vacuum, pinned in place by nothing but invisible electric fields, and glowing on command when a laser touches it. Now line up a dozen of them in a perfectly even row, each one an identical twin of the last because they are literally the same kind of atom, straight from the periodic table. That row of floating atoms is a working quantum computer. This is the trapped-ion platform, built by companies like IonQ and Quantinuum — and it holds the records for the cleanest, longest-lived, most-connected qubits anyone has ever made.

The qubit is two energy levels of one atom

In a trapped-ion machine the qubit is not a manufactured circuit — it is a single ion, an atom that has had one electron stripped away so it carries a net charge (which is what lets electric fields grab hold of it). Common choices are ytterbium (\mathrm{Yb}^+) or calcium (\mathrm{Ca}^+).

The two states |0\rangle and |1\rangle are two internal energy levels of that ion — two of the discrete states its outermost electron is allowed to sit in. Choose a pair that are extraordinarily stable (a "hyperfine" or "optical clock" pair, the very same physics behind atomic clocks) and you have a qubit whose |0\rangle and |1\rangle barely leak into each other for seconds at a time.

Holding the ion still is the job of a Paul trap: a set of electrodes driven by a rapidly oscillating electric field. You cannot trap a charged particle with static fields alone, but a field that flips back and forth fast enough produces an effective bowl that the ion settles into. Many ions dropped into the same trap repel one another (like charges) and space themselves out into a neat, evenly spaced line — a "Coulomb crystal."

Identical by nature — no calibration spread

Here is the quiet superpower of the platform. Every \mathrm{Yb}^+ ion in the universe is exactly the same as every other: same mass, same energy levels, same transition frequencies, to a precision no factory could ever match. Nature does the fabrication.

Contrast this with superconducting qubits, which are patterned onto a chip: no two come out quite alike, and each needs its own calibration to pin down its slightly-different frequency. A trapped-ion chain sidesteps that whole problem. This directly answers one of the DiVincenzo criteria — a well-characterised qubit — almost for free.

Initialise and read out with light

Both "reset to |0\rangle" and "measure" are done with lasers, by a trick called fluorescence:

• Initialisation. Shine a carefully tuned laser and the ion is optically pumped — it cycles through excited states and reliably ends up parked in a single chosen level, giving you a clean |0\rangle every time.
• Readout. Shine a different "detection" laser that only the |0\rangle state can absorb. If the qubit is in |0\rangle it scatters thousands of photons and lights up bright; if it is in |1\rangle it cannot absorb that colour and stays dark. A camera or photodetector simply asks "bright or dark?" — a direct |0\rangle / |1\rangle measurement, and one of the most reliable readouts in all of quantum computing.

Gates: single-ion light, two-ion motion

Single-qubit gates are the easy part: aim a laser (or microwave) pulse at one ion and its precise frequency and duration rotate that qubit's state to wherever you like on the Bloch sphere. One ion, one beam.

Two-qubit gates are the clever part, and the heart of the whole platform. Two ions sitting in the same trap are not wired together — so how do they talk? Through their shared motion. Because the ions repel each other by the Coulomb force, the whole chain behaves like beads on a spring: push one and they all sway together in collective motional modes (quantised vibrations of the line). A two-qubit gate — the famous Mølmer–Sørensen gate — uses laser pulses to momentarily couple each ion's internal qubit state to this shared vibration, let the two qubits interact through that common "bus," and then decouple the motion again, leaving the two ions entangled.

The motional bus is what makes trapped ions special: any ion in the trap can be coupled to the shared vibration, so any ion can be entangled with any other — a property called all-to-all connectivity.

The picture

Step through the schematic: first the trap and its oscillating electrodes, then the line of identical ions, then the laser beams that drive them, and finally the "springs" — the shared vibrations that let any two ions talk to each other.

Strengths: coherence, fidelity, connectivity

Trapped ions are the accuracy champions of quantum hardware, on three fronts at once:

• The longest coherence times of any platform. Because the qubit lives in a well-shielded internal state of a floating atom, it can hold its quantum information for seconds — an eternity next to the microseconds of most solid-state qubits.
• The highest gate fidelities. Both single- and two-qubit gates routinely exceed 99.9\% accuracy, among the best ever measured — helped by those perfectly identical, well-understood atoms.
• All-to-all connectivity. Any ion can be entangled directly with any other through the motional bus, with no need to route information through intermediaries.

Worked example 1: how many operations fit in one coherence time?

A qubit is only useful for as many gates as you can run before it forgets — so the number that matters is the ratio of the coherence time to the gate time. Take generous but realistic trapped-ion figures: a coherence time of T \approx 1\ \text{second} and a two-qubit gate that takes about \tau \approx 100\ \mu\text{s} = 100\times 10^{-6}\ \text{s}. Then

Ten thousand gates before the qubit decoheres — and with longer coherence and faster gates that budget can climb into the hundreds of thousands. The generous ratio is exactly why trapped ions are prized for the long, deep circuits that error correction and serious algorithms demand.

Worked example 2: what all-to-all connectivity saves you

Why does all-to-all connectivity matter? Because on hardware with only nearest-neighbour links, two qubits that need to interact but sit far apart must first be shuffled next to each other with a chain of SWAP gates — and each SWAP costs about three two-qubit gates.

Say you want to entangle qubit 1 and qubit 10 in a line of ten. On nearest-neighbour hardware you must SWAP one of them past the 8 qubits in between to bring the pair together — roughly

just for the routing (and as many again to move it back). On a trapped-ion chain the answer is one gate: the Mølmer–Sørensen gate couples ions 1 and 10 directly through the shared motional bus. Every one of those saved gates is a chance for error avoided — a big reason ion traps can run deeper circuits than their raw gate count alone would suggest.

Weaknesses: slow gates, hard-to-scale chains

No platform wins everything, and trapped ions pay for their precision in two coins:

• Gates are slow. A trapped-ion gate takes microseconds to milliseconds, because it works through the ions' physical motion, which is sluggish. That is often a thousand times slower than a superconducting gate (tens of nanoseconds). The coherence-to-gate ratio is still excellent — but wall-clock throughput lags.
• One long chain does not scale. Pack more ions into a single trap and the collective motional modes grow crowded and closely spaced, making the shared bus harder to address cleanly. You cannot simply keep lengthening one line forever.

The industry's answer is not a longer chain but more traps: the QCCD architecture (Quantum Charge-Coupled Device) splits the machine into many small trapping zones and physically shuttles ions between them — moving qubits to wherever a gate needs to happen, then moving them on. Quantinuum's machines are built this way. Scaling ion traps is an exercise in traffic management, not just in adding atoms to a row.

Summary