Quantum Computing Modalities: Trapped-Ion QC

(For other quantum computing modalities and architectures, see Taxonomy of Quantum Computing: Modalities & Architectures)

What It Is

Trapped-ion quantum computing uses individual atomic ions, suspended in vacuum by electromagnetic fields, as qubits. Each ion’s internal quantum state (typically two hyperfine energy levels of the atom’s ground-state electron configuration) encodes |0⟩ and |1⟩. Ions are manipulated with precisely tuned laser beams or microwave pulses to perform quantum gates, and read out by detecting whether each ion fluoresces when illuminated.

This was one of the earliest quantum computing proposals. Ignacio Cirac and Peter Zoller’s 1995 paper outlined how to implement a CNOT gate between two trapped ions using a shared vibrational mode as a quantum bus. That same year, a team at NIST led by David Wineland demonstrated the basic elements experimentally, producing the first quantum logic gate in any physical system. Thirty years later, trapped ions hold the record for two-qubit gate fidelity (99.97% two-qubit by Oxford Ionics, with IonQ reporting >99.99% on subsequent hardware), have demonstrated the most efficient error correction on real hardware (2:1 physical-to-logical ratio on Quantinuum Helios), and are one of only two modalities (alongside superconducting qubits) with a credible roadmap to fault-tolerant quantum computing by the end of the decade.

trapped ions hold the record for two-qubit gate fidelity (99.97% two-qubit by Oxford Ionics, with IonQ reporting >99.99% on subsequent hardware)

From a CRQC perspective, trapped ions present an interesting profile. Their gate speeds are slow (microseconds to milliseconds, versus nanoseconds for superconducting systems), which means running Shor’s algorithm on a trapped-ion machine would take substantially longer than on a comparably sized superconducting processor. But their error rates are so low that they need far fewer physical qubits per logical qubit to achieve the same logical error rate. That tradeoff shapes the entire competitive picture between the two modalities in the race toward cryptographically relevant computation.

How It Works

Trapping and Cooling

Ions are confined in electromagnetic traps, most commonly linear Paul traps that use oscillating radio-frequency electric fields to create a pseudo-harmonic potential. The ions sit in a line, spaced 5–15 micrometers apart by mutual Coulomb repulsion, suspended in ultra-high vacuum (<10⁻¹¹ mbar) to prevent collisions with background gas molecules that would knock them out of position.

Before computation begins, the ions must be cooled to near their motional ground state. Doppler cooling brings them to millikelvin temperatures, and sideband cooling (resolved sideband or Raman cooling) reduces the motional quantum number to near zero. This is essential because the two-qubit gate mechanism relies on the ions’ shared vibrational modes being in a well-defined quantum state.

The commonly used ion species each have different advantages. ¹⁷¹Yb⁺ (ytterbium) has a microwave-frequency hyperfine splitting (~12.6 GHz) that provides a stable qubit encoding, and its transitions are compatible with established laser technology at 369 nm. ¹³⁷Ba⁺ (barium) offers transitions at more convenient visible wavelengths and stronger photon scattering for faster readout; Quantinuum switched to barium for Helios. ⁴⁰Ca⁺ (calcium) is used extensively in academic labs (particularly in Innsbruck and Oxford) for its relatively simple level structure. ⁸⁸Sr⁺ (strontium) and ⁹Be⁺ (beryllium) appear in specific research programs.

Single-Qubit Gates

A single-qubit gate rotates the ion’s internal state around an axis on the Bloch sphere. For hyperfine qubits (like ¹⁷¹Yb⁺), this can be done directly with microwave pulses at 12.6 GHz, or via stimulated Raman transitions using two laser beams whose frequency difference matches the qubit splitting. For optical qubits (like the narrow ⁴⁰Ca⁺ quadrupole transition at 729 nm), a single laser drives the qubit directly.

Each approach has its strengths. Microwave-driven gates are inherently uniform across the ion chain (the microwave field is spatially broad), which simplifies calibration but makes individual addressing harder. Laser-driven gates can be tightly focused to address individual ions, enabling parallel operations on different qubits, but require sub-MHz linewidth lasers with tight intensity stabilization.

Single-qubit gate fidelities in trapped ions routinely exceed 99.99%. Oxford researchers demonstrated gate errors at the 10⁻⁷ level in 2025, a record for any qubit technology.

Two-Qubit Gates

This is where trapped ions are most distinctive. Two-qubit entangling gates exploit the Coulomb interaction between ions: because the ions share vibrational modes (they’re coupled through their mutual electric repulsion), applying a state-dependent force on two ions can entangle their internal (qubit) states via the motional degree of freedom.

The dominant gate scheme is the Mølmer-Sørensen (MS) gate. Two off-resonant laser beams create a spin-dependent force that displaces the ions’ collective motion in a way that depends on the product of their spin states. When the gate duration is chosen so that the motional state returns to its initial position (a closed loop in phase space), the net effect is a geometric phase that entangles the qubits while leaving the motion disentangled. The result is an effective σₓσₓ interaction that produces maximally entangled Bell states.

Alternative schemes include the Cirac-Zoller gate (sequential, uses a single motional mode as a bus, slower but conceptually simpler) and the light-shift gate (used by IonQ, based on a geometric phase from off-resonant coupling to motional modes).

The critical advantage: because the shared vibrational modes couple all ions in a chain, any pair of ions can be entangled directly, regardless of their positions. This gives trapped ions effective all-to-all connectivity within a single trap zone, a property that superconducting and silicon spin qubits, with their nearest-neighbor coupling, cannot match without SWAP overhead.

Two-qubit gate fidelities have reached 99.921% all-pairs on Quantinuum Helios (¹³⁷Ba⁺, 98 qubits) and 99.97% on a two-qubit Oxford Ionics system, with IonQ subsequently reporting >99.99% on its own ¹⁷¹Yb⁺ hardware. These are the highest two-qubit fidelities demonstrated in any quantum computing platform.

Readout

Measurement uses state-dependent fluorescence. A cycling transition is driven by a resonant laser: one qubit state (say |1⟩) scatters many photons and appears “bright” on a detector (PMT or EMCCD camera), while the other state (|0⟩) is off-resonant and appears “dark.” The photon count difference provides a clear binary signal. Readout fidelities exceed 99.9% and the measurement is projective (destructive to the superposition but non-destructive to the ion itself, which can be re-initialized by optical pumping).

Scaling: The QCCD Architecture

The fundamental scaling challenge for trapped ions is that adding more ions to a single trap chain slows down two-qubit gates (the vibrational mode spectrum becomes denser, making it harder to isolate individual modes for clean gate operations). Beyond roughly 20–50 ions in a single chain, gate fidelity degrades.

The solution, pioneered by Quantinuum, is the Quantum Charge-Coupled Device (QCCD) architecture. The name draws an analogy to the classical CCD sensor: ions are shuttled between specialized zones on a microfabricated surface-electrode trap chip by smoothly varying the voltages on segmented electrodes. The chip is divided into functional zones: storage zones (where ions wait in long chains), gate zones (where small groups of 2–4 ions are isolated for entangling operations), and readout zones (where fluorescence measurement occurs). After an operation, ions are transported back to storage or to a different gate zone for the next interaction.

Shuttling mechanics. Moving an ion from one zone to another takes 10–100 µs depending on distance and the voltage waveform applied. The key constraint is that fast shuttling heats the ion’s motional state: the acceleration and deceleration add kinetic energy (motional quanta) that must be removed before the next gate operation. Careful waveform optimization (using smooth, jerk-minimized voltage ramps) reduces this motional excitation but cannot eliminate it entirely. After each transport, the ion typically needs a brief recooling step to return to the motional ground state.

Junction routing. Linear shuttling moves ions along a straight track, but a scalable architecture requires branching paths, which means junctions. A junction is the ion-trap equivalent of a highway interchange: the ion must turn a corner where two trap channels meet. This is the hardest shuttling operation because the electric potential at a junction has a saddle point where confinement weakens, exposing the ion to heating and potential loss. Quantinuum’s Helios is the first commercial processor with 2D junction-based routing, enabling arbitrary ion rearrangement across the chip surface. The engineering required to maintain ion integrity through junction transits at high speed is one of Quantinuum’s core proprietary capabilities.

Sympathetic cooling. As ions are shuttled and gated, they accumulate motional energy. Recooling them with the qubit laser would destroy the quantum information stored in their internal state. The solution is sympathetic cooling: a different ion species (or a different isotope) is co-trapped alongside the qubit ions. The coolant ions are laser-cooled continuously, and because they share vibrational modes with the qubit ions through Coulomb coupling, they extract motional energy from the qubit ions without disturbing their quantum states. Quantinuum uses this technique in Helios, and it is essential for sustaining high fidelity over long computational sequences.

I analyzed the Helios architecture in detail, including its rotatable memory ring and dual processing regions.

Photonic Interconnects: The Distributed Path

An alternative (and complementary) scaling path connects separate trap modules via photonic links. Each module contains a small number of high-fidelity ions. To entangle ions in different modules, each module emits a photon entangled with one of its ions. The two photons are interfered on a beam splitter, and a coincidence detection heralds successful entanglement between the remote ions.

The entanglement rate is low (the process is probabilistic, succeeding perhaps once in every 10,000–100,000 attempts), but because the success is heralded, failed attempts can simply be discarded and retried without corrupting the computation. Over time, enough inter-module entanglement is generated to support distributed quantum computation.

IonQ is pursuing this approach through its Lightsynq acquisition (completed 2025), which provides advanced photonic interconnect technology. The vision is a quantum computer built from many small, perfect trap modules connected by optical fiber, rather than one enormous trap chip with thousands of zones. This architecture has the theoretical advantage of keeping each local module small enough to maintain peak gate fidelity, while scaling the total system to millions of qubits.

I cover the integration details in my deep dive on building a trapped-ion quantum computer.

Key Academic Papers

Foundational

Cirac & Zoller (1995). The proposal that launched the field. Showed how to implement a universal CNOT gate between two trapped ions using shared vibrational modes as a quantum bus. Published in Physical Review Letters.

Monroe, Meekhof, King, Itano & Wineland / NIST (1995). First experimental demonstration of a quantum logic gate, implementing a controlled-NOT between the internal and motional states of a single trapped ⁹Be⁺ ion. Published in Physical Review Letters.

Sørensen & Mølmer (1999). Introduced the Mølmer-Sørensen gate, which entangles ions through a geometric phase without requiring ground-state cooling. This gate scheme is used in nearly every modern trapped-ion processor. Published in Physical Review Letters.

Scaling and Performance (2016–2025)

Ballance, Harty et al. / Oxford (2016). Demonstrated 99.9% two-qubit gate fidelity using ⁴³Ca⁺ ions, the first to cross the surface-code error threshold with a two-qubit gate. Published in Physical Review Letters.

IonQ / 99.99% breakthrough (2025). Oxford Ionics demonstrated 99.97% two-qubit gate fidelity on a ⁴³Ca⁺ system, and IonQ subsequently reported >99.99% on its own ¹⁷¹Yb⁺ hardware. The highest two-qubit fidelities in any platform. My analysis covers the implications for CRQC timelines.

Oxford / 10⁻⁷ single-qubit error (2025). Oxford researchers demonstrated single-qubit gate errors at the 10⁻⁷ level on ⁴³Ca⁺, three orders of magnitude below the surface-code threshold. Published in a peer-reviewed journal.

Error Correction and Logical Qubits (2025–2026)

Quantinuum Helios / Skinny Logic (November 2025). 98 physical ¹³⁷Ba⁺ trapped-ion qubits yielding 48 error-corrected logical qubits via “Skinny Logic” codes at a 2:1 physical-to-logical ratio. All-pairs two-qubit gate fidelity 99.921%; single-qubit 99.9975%. The most efficient quantum error correction demonstrated on any hardware platform. My architecture analysis.

Quantinuum / 94 logical qubits (March 2026). Quantinuum squeezed 94 logical qubits from the same 98-physical-qubit Helios system. My analysis examined what this result actually means for practical computation.

Quantinuum / magic state distillation (June 2025). First high-fidelity magic state distillation on a commercial trapped-ion system. Magic states are the resource needed for non-Clifford operations (T gates), which are required for universal fault-tolerant computation. A critical CRQC capability.

IonQ / Walking Cat fault-tolerant blueprint (April 2026). IonQ published a complete architectural blueprint for fault-tolerant trapped-ion quantum computing using “walking cat” codes. My analysis covers the resource estimates and how they compare to superconducting approaches.

The Vendor Landscape (May 2026)

The trapped-ion market is smaller than superconducting but features two of the strongest quantum computing companies in the world, plus a growing European ecosystem.

Major Vendors

Quantinuum. The performance leader. Formed from the merger of Honeywell Quantum Solutions and Cambridge Quantum in 2021, Quantinuum builds QCCD trapped-ion processors. The company switched from ¹⁷¹Yb⁺ (used in H1 and H2) to ¹³⁷Ba⁺ for Helios, gaining faster readout (barium’s stronger photon scattering produces a brighter fluorescence signal) and the ability to use visible-wavelength lasers instead of UV.

The current flagship is Helios: 98 physical qubits, 48 error-corrected logical qubits via Skinny Logic codes, all-pairs two-qubit gate fidelity of 99.921%, single-qubit fidelity of 99.9975%. The Helios architecture separates qubit storage from logic operations on the chip, using a rotatable memory ring that can feed ions into two parallel processing regions. A new classical control engine compiles quantum programs dynamically at runtime, enabling real-time conditional logic (via the Guppy programming language) rather than pre-compiled static circuits. This is essential for QEC, where the next operation depends on the outcome of the previous measurement.

Helios is available via cloud and on-premise deployment, with a system being installed in Singapore by 2026 under a partnership with A*STAR. Quantinuum’s roadmap: Sol (~192 physical qubits, first 2D-grid commercial QCCD, 2027), Apollo (thousands of physical qubits, universal fault-tolerant quantum computing, 2029–2030). In May 2026, Quantinuum announced an accelerated roadmap confirming the 2030 target for fully fault-tolerant quantum computing. Funding: an $800 million round (2025–2026) with participation from Fidelity and backing from parent company Honeywell.

The jump from Sol to Apollo is the biggest risk in the roadmap. Sol roughly doubles the physical qubit count from Helios; Apollo represents an order-of-magnitude leap. Quantinuum has not yet detailed how that jump will be achieved, whether through a single larger trap chip, photonic interconnects between modules, or a hybrid approach. This is worth watching closely.

IonQ. The most aggressively scaling trapped-ion company, and the one pursuing the most acquisitive growth strategy. IonQ went public via SPAC in 2021 and has since assembled a quantum stack through four major acquisitions:

1. Oxford Ionics ($1.075 billion, June 2025, completed September 2025). The most consequential deal. Oxford Ionics’ “eQC” (electronic qubit control) chip replaces lasers with electronic microwave/RF fields for gate operations. The chip is manufactured on standard Infineon semiconductor lines, making it compatible with industrial-scale production. Oxford Ionics demonstrated 99.97% two-qubit fidelity with eQC. If this approach scales to hundreds of qubits while maintaining that fidelity, it eliminates the laser infrastructure bottleneck that has constrained trapped ions since the field began. IonQ’s entire scaling roadmap from 2026 onward depends on eQC working at scale.
2. Lightsynq (completed 2025). Photonic interconnect technology for entangling ions across separate trap modules via optical fiber. This is IonQ’s path to millions of qubits: instead of building one enormous trap, connect many small modules.
3. Capella Space (completed 2025). Satellite infrastructure for space-based quantum key distribution (QKD) networks. This extends IonQ from quantum computing into quantum communications.
4. Qubitekk (earlier acquisition). Quantum networking hardware, including entangled photon sources.

Current systems: Forte Enterprise (36 physical qubits, ¹⁷¹Yb⁺, rack-mountable for data centers, #AQ 36) and Tempo (64+ physical qubits on 100-qubit hardware, #AQ 64, IonQ’s first barium system, achieved three months ahead of schedule in September 2025). IonQ achieved 99.997% two-qubit gate fidelity in 2025, a world record. The Walking Cat fault-tolerant blueprint (April 2026) provides IonQ’s complete architectural vision for FTQC.

Combined roadmap: 256 physical qubits at 99.99% accuracy by 2026 (incorporating eQC), 10,000 physical qubits with logical fidelity 99.99999% by 2027, 800 logical qubits by 2027, 2 million physical qubits and 80,000 logical qubits by 2030. These targets are extremely aggressive; the 2027 targets in particular represent a 40× jump in physical qubit count from Tempo in roughly 18 months. Available on AWS Braket, Azure Quantum, and Google Cloud.

European Ecosystem

AQT (Alpine Quantum Technologies, Innsbruck). Spun out of the University of Innsbruck (Rainer Blatt, Peter Zoller, Thomas Monz). Builds compact, rack-mountable ⁴⁰Ca⁺ trapped-ion systems. The PINE system fits in two 19-inch racks, supports up to 50 qubits, and operates at room temperature with <2 kW power. Demonstrated Quantum Volume 128 in September 2025. AQT was selected as a EuroHPC quantum computer vendor, delivering the PIAST-Q system to PSNC in Poland and a system to the Munich Quantum Valley/LRZ. AQT does not yet have a PostQuantum.com company profile but is an increasingly important European player.

eleQtron (Germany). Pursuing a laser-free approach: microwave-controlled trapped ions using individually addressed magnetic field gradients, eliminating the entire laser subsystem. Pre-commercial stage. The EGALE system is under development. This approach, if it scales, would simplify the trapped-ion hardware stack.

Infineon + Oxford Ionics + Cyberagentur. The Mini-Q program: a €35 million contract for a portable quantum computer for German defense applications. Uses Yb⁺ ions with Oxford Ionics’ semiconductor-manufactured eQC chip. This is the first trapped-ion procurement explicitly targeting defense mobility.

Universal Quantum (UK). Modular trapped-ion architecture using electric-field shuttling between connected trap modules. Pre-commercial stage, but backed by significant UK government funding.

What You Cannot Buy (Yet)

Unlike the superconducting QOA ecosystem, there is no established market for standalone trapped-ion QPUs that an independent integrator can purchase and assemble into a custom system. Quantinuum and IonQ sell integrated systems or cloud access. The closest thing to a component-level offering is the socketed ion-trap-on-carrier from Infineon/Oxford Ionics (now integrated into IonQ’s pipeline), but it is not available for general purchase. AQT’s PINE SET-UP is the most accessible trapped-ion hardware for independent researchers.

This means that the trapped-ion supply chain is vertically integrated, not horizontally modular. If you want a trapped-ion quantum computer, you buy (or rent cloud time from) Quantinuum or IonQ. The implications for procurement and vendor lock-in are different from the QOA model available in superconducting. I discuss this contrast in my deep dive on building a trapped-ion quantum computer.

Quantum Error Correction: The 2:1 Ratio

Trapped ions hold the most dramatic QEC result demonstrated on real hardware as of May 2026. Quantinuum’s Helios converted 98 physical qubits into 48 error-corrected logical qubits using “Skinny Logic” codes, a 2:1 physical-to-logical ratio. In March 2026, the same hardware produced 94 logical qubits from 98 physical.

Compare this to the superconducting state of the art: Google’s Willow demonstrated below-threshold surface code error suppression, but surface code at distance 7 requires 97 physical qubits per logical qubit. IBM’s qLDPC Gross code targets a 10:1 ratio. Quantinuum achieved 2:1 on shipping hardware. The gap is structural: trapped ions’ all-to-all connectivity enables codes (like the Skinny Logic and iceberg codes) that require long-range stabilizer measurements, which planar nearest-neighbor architectures cannot implement without routing overhead.

The implication for CRQC timelines is direct. If you can achieve a 2:1 physical-to-logical ratio, the Gidney 2025 estimate of ~1,400 logical qubits for RSA-2048 translates to roughly 2,800 physical trapped-ion qubits. With surface code on superconducting hardware at current overhead ratios, the same computation requires 1–2 million physical qubits. The trapped-ion path to a CRQC is narrower in physical qubit count but longer in wall-clock time, because each gate operation takes microseconds instead of nanoseconds.

Quantinuum also demonstrated high-fidelity magic state distillation on Helios in June 2025, a critical capability for universal fault tolerance. IonQ’s Walking Cat blueprint provides an alternative architectural approach to fault tolerance using a different code construction. I assess these developments through my CRQC Quantum Capability Framework, particularly the capabilities for magic state production and below-threshold operation.

Comparison to Other Modalities

Trapped Ion vs. Superconducting

The central comparison in quantum computing. Both modalities are commercially shipping, both have demonstrated QEC milestones, and both have credible fault-tolerance roadmaps.

Trapped ions win on fidelity (99.921% all-pairs vs. 99.67% best pair for Google Willow), coherence (minutes to hours vs. hundreds of microseconds), connectivity (all-to-all vs. nearest-neighbor), and QEC efficiency (2:1 ratio vs. ~100:1 for surface code). Superconducting qubits win on gate speed (nanoseconds vs. microseconds, a 1,000× advantage), fabrication scalability (lithographic, chip-based, amenable to semiconductor mass production), qubit count (1,121 on IBM Condor vs. 98 on Helios), and the HPC integration ecosystem (NVQLink, QRMI, and the full QOA supply chain).

The metric that matters: error-corrected operations per second. Neither raw gate speed nor raw fidelity alone determines which modality produces more useful computation. What counts is how many logical (error-corrected) gate operations the processor executes per unit of time. This depends on three factors: the QEC cycle time (how fast the processor can detect and correct errors), the physical-to-logical overhead (how many physical qubits and gates are consumed per logical operation), and the logical error rate (how often a corrected operation still fails).

Consider a concrete comparison. A superconducting processor running surface code at distance 15 on 450 physical qubits per logical qubit, with a 1 µs cycle time, executes ~10⁶ QEC cycles per second. Each logical gate requires multiple QEC cycles (lattice surgery for a CNOT takes ~d cycles for code distance d), so the logical gate rate is roughly 10⁶/15 ≈ 67,000 logical gates per second. A trapped-ion processor running Skinny Logic codes at a 2:1 ratio, with a ~100 µs cycle time, executes ~10⁴ QEC cycles per second. But because the code is far more efficient, each logical gate requires fewer physical operations. If a Skinny Logic gate requires only 2–4 QEC cycles, the logical gate rate is 10⁴/3 ≈ 3,300 per second.

In this rough estimate, the superconducting processor runs ~20× more logical gates per second. But it needs ~225× more physical qubits per logical qubit. The total physical resources (qubit-seconds) per logical gate may actually favor trapped ions, depending on the specific codes and architectures deployed. The honest answer is that neither modality has a clear overall advantage at this stage. The winner will be determined by engineering progress over the next five years, not by any fundamental physical law.

Trapped Ion vs. Neutral Atom

Neutral-atom systems (particularly QuEra and Pasqal) have closed the gap rapidly. QuEra’s 96 verified logical qubits on 448 physical atoms (November 2025) matched Quantinuum’s logical qubit count, and their 2:1 qLDPC encoding (April 2026) matched the efficiency. Neutral atoms offer larger qubit arrays (1,180 atoms demonstrated by Atom Computing), room-temperature operation, and reconfigurable connectivity via optical tweezers. Trapped ions counter with higher two-qubit fidelity (99.921% vs. ~99.5% for neutral atoms), more mature mid-circuit measurement, and longer coherence. The two modalities are closer competitors than most people recognize.

Trapped Ion vs. Photonic

Photonic systems encode qubits in light and operate at room temperature (except for single-photon detectors). Their entangling gates are probabilistic rather than deterministic, requiring measurement-based or fusion-based architectures. Trapped ions offer deterministic, high-fidelity gates, making them superior for algorithmic quantum computing. Photonic systems may excel at quantum networking and communication. Interestingly, photonic interconnects are a key part of IonQ’s scaling strategy, creating a natural bridge between the two modalities.

Advantages

Highest fidelity of any platform. The 99.921% all-pairs two-qubit gate fidelity on Quantinuum Helios and 99.997% on a two-qubit IonQ system are the best results in any quantum computing modality. Single-qubit fidelities exceed 99.999%. These numbers mean trapped ions need the fewest physical qubits per logical qubit for error correction, which is the metric that actually matters for building a useful machine.

All-to-all connectivity. Any ion in a trap chain can interact with any other via shared vibrational modes. In QCCD architectures, ion shuttling extends this to any-to-any connectivity across the entire chip. This eliminates the SWAP overhead that plagues nearest-neighbor architectures and enables the use of high-rate error-correcting codes (Skinny Logic, iceberg codes) that require non-local stabilizer measurements.

Identical qubits. Every ¹³⁷Ba⁺ ion is physically identical to every other ¹³⁷Ba⁺ ion. There are no fabrication variations, no frequency collisions, no chip-to-chip variability. The calibration burden scales with the number of control channels (lasers, electrodes), not with qubit quality variations.

Long coherence. Hyperfine qubits in trapped ions have coherence times measured in seconds to minutes (hours in specialized configurations). This is four to six orders of magnitude longer than superconducting transmons (~100–500 µs). Even with slow gates, the ratio of coherence time to gate time allows tens of thousands of operations before decoherence becomes significant.

Room-temperature operation (mostly). The ion trap operates in a room-temperature vacuum chamber. No dilution refrigerator, no helium-3 supply chain concerns, no cryogenic wiring bottleneck. Some detector subsystems require modest cooling (4 K for superconducting photon detectors in photonic interconnect schemes), but the core QPU is at ambient temperature. This eliminates the entire cryogenic infrastructure stack that dominates superconducting system cost and complexity.

Mid-circuit measurement and feed-forward. Trapped ions excel at measuring individual qubits in the middle of a computation and using the classical result to determine subsequent quantum operations. This capability is essential for quantum error correction, teleportation-based protocols, and adaptive algorithms. Quantinuum’s Helios real-time control engine supports full dynamic quantum-classical programming via the Guppy language.

Disadvantages

Slow gates. Two-qubit gate times range from tens of microseconds (Mølmer-Sørensen, optimized) to hundreds of microseconds, with some schemes in the millisecond range. This is 1,000–10,000× slower than superconducting CZ gates (~20–100 ns). For a computation requiring 6.5 billion Toffoli gates (the Gidney 2025 estimate for RSA-2048), even at 10 µs per gate, the wall-clock time would be ~18 hours assuming perfect parallelism, versus seconds on a comparably sized superconducting processor running at nanosecond gate speeds. In practice, QCCD architectures add shuttling overhead on top of gate time: moving ions to and from gate zones, recooling after transport, and rearranging chains between operations. A realistic trapped-ion QEC cycle takes ~1 ms when shuttling and recooling are included, versus ~1 µs for superconducting surface code. Research into ultrafast laser pulses (~1 ps stimulated Raman transitions) aims to close the raw gate-speed gap, but the shuttling overhead is architectural and will persist in QCCD designs.

Scaling complexity. Beyond ~20–50 ions in a single chain, the vibrational mode spectrum becomes dense and gates slow down further. QCCD architectures solve this by shuttling ions between zones, but each element of the QCCD pipeline adds complexity. Junction transits require precise voltage control and introduce motional heating. Sympathetic cooling ions consume trap capacity without contributing to computation. Readout zones must be optically isolated from gate zones to prevent stray photons from disturbing qubits mid-computation. Scaling to thousands of qubits requires hundreds of trap zones with precise voltage control on thousands of electrodes. The trap chip fabrication itself is not a bottleneck (surface-electrode traps are made with standard microfabrication), but the control electronics driving thousands of independent electrode voltages at microsecond timescales represent a substantial classical engineering challenge.

Laser and optical addressing infrastructure. A trapped-ion system with laser-driven gates requires 5–10 lasers per ion species (cooling, repumping, photoionization, qubit manipulation, readout). Each laser needs sub-MHz linewidth and tight intensity stabilization (cavity-locked ECDLs, frequency combs for Ba⁺). The resulting optical table, with its mirrors, acousto-optic modulators (AOMs), and beam-routing optics, is large and sensitive to alignment drift.

Addressing individual ions within a chain is a particular challenge. Ions are spaced only 5–15 µm apart, and a laser beam focused tightly enough to address one ion will partially illuminate its neighbors, causing cross-talk errors. Addressing techniques include tightly focused beams steered by AOMs or MEMS mirrors, individual addressing with multi-channel AOMs (one channel per ion), and integrated waveguides on the trap chip that route light to specific zones. Each approach has tradeoffs: AOMs are fast but bulky, MEMS mirrors are compact but slow, and integrated photonics are elegant but add fabrication complexity and optical loss.

Oxford Ionics’ eQC approach is the most aggressive mitigation: replacing lasers entirely with electronic microwave/RF fields for gate operations. Because microwave fields can be generated by on-chip electrodes with arbitrary spatial patterns, individual addressing becomes an electronics problem rather than an optics problem. This is why IonQ paid $1.075 billion for Oxford Ionics. If eQC scales to hundreds of qubits while maintaining 99.97% two-qubit fidelity, it transforms the trapped-ion modality from a precision-optics discipline into a semiconductor-electronics discipline. That transformation would change the cost structure, manufacturing scalability, and deployment footprint of trapped-ion systems fundamentally.

Photonic interconnect overhead. For the distributed scaling approach (connecting multiple trap modules via photonic links), the entanglement generation rate is a binding constraint. Current remote ion-ion entanglement rates are on the order of 1–10 Hz: generating one entangled pair per 0.1–1 seconds. A useful distributed computation might require thousands of remote entanglement operations. At 10 Hz, generating 10,000 remote links takes ~17 minutes. Improving this rate requires better photon collection efficiency (currently <1% in most experiments), faster photon detection, and more efficient optical coupling between ions and fibers. Lightsynq’s technology addresses these engineering bottlenecks, but production-rate photonic interconnects are still years away.

Vertically integrated supply chain. As noted above: you cannot buy a trapped-ion QPU from one vendor and assemble it with independently sourced control electronics and vacuum systems the way you can in the superconducting QOA market. This creates vendor lock-in and limits the buyer’s ability to specify best-of-breed components independently. AQT’s PINE SET-UP is the closest to a component-level offering.

Vacuum and environmental sensitivity. Ultra-high vacuum systems (<10⁻¹¹ mbar) require ion pumps, getter pumps, and days-to-weeks of bake-out at ~200°C when the chamber is opened. Background gas collisions cause ion loss and anomalous motional heating (a phenomenon where ions gain energy from the trap electrodes at rates higher than thermal physics predicts, attributed to surface contaminants on electrode surfaces). The anomalous heating rate scales inversely with the fourth power of the ion-electrode distance, which means smaller trap geometries suffer more. Surface cleaning techniques (ion bombardment, laser cleaning) mitigate this but do not eliminate it. While operational vacuum systems can run for years without intervention, any hardware change that breaks the vacuum seal is a multi-day recovery event.

Impact on Cybersecurity

Trapped ions are one of the two modalities (alongside superconducting) most likely to underpin a CRQC. Their unique profile, extreme fidelity offset by slow gate speed, produces a distinctive threat model.

The Efficiency Argument

The 2:1 physical-to-logical ratio changes the resource calculus for cryptographic attacks. The Gidney 2025 estimate requires ~1,400 logical qubits for RSA-2048. At Quantinuum’s demonstrated 2:1 ratio, that translates to ~2,800 physical ions. At a surface-code ratio typical of superconducting hardware (~1,000:1), the same computation requires ~1.4 million physical qubits. This suggests trapped-ion systems could reach CRQC capability at a much smaller physical scale.

But scale is only half the story. The Gidney estimate also involves ~6.5 billion Toffoli gates. At trapped-ion gate speeds, executing that many operations would take hours to days, depending on parallelism. A superconducting processor with 1,000× faster gates could, in principle, finish the same computation in minutes, albeit requiring a vastly larger machine.

Which path arrives first depends on how quickly each modality can scale physical qubits while maintaining quality. Quantinuum’s roadmap to Apollo (thousands of qubits, 2029–2030) and IonQ’s target of 2 million physical qubits by 2030 suggest that trapped-ion vendors are betting on scale. My CRQC Quantum Capability Framework tracks progress across both modalities against the specific capabilities required.

What Should You Do?

The same answer as for every modality: begin PQC migration now. The relevant standards are NIST FIPS 203 (ML-KEM), FIPS 204 (ML-DSA), and FIPS 205 (SLH-DSA). The Harvest Now, Decrypt Later threat is modality-agnostic. And the real deadline isn’t Q-Day itself: regulators, insurers, investors, and clients are setting their own quantum deadlines.

Future Outlook

2026–2027. IonQ delivers a 256-qubit system incorporating Oxford Ionics’ eQC chip, the first large-scale trapped-ion processor using electronic (laser-free) qubit control. If this works at the demonstrated 99.97% two-qubit fidelity, it changes the scalability picture fundamentally. Quantinuum ships Sol (~192 physical qubits), the first commercial 2D-grid QCCD. AQT delivers additional EuroHPC systems. The race between Quantinuum’s code-efficiency approach and IonQ’s qubit-count approach intensifies.

2028–2030. Quantinuum targets Apollo: thousands of physical qubits, universal fault-tolerant quantum computing. IonQ targets 10,000 physical qubits (2027) and 2 million (2030). These are aggressive roadmap claims. If either vendor achieves fault tolerance with hundreds of logical qubits in this window, the first computations provably beyond classical reach become possible. Whether the trapped-ion or superconducting path reaches that milestone first is the open question that defines the industry’s next five years.

The long game. Trapped ions may not win on qubit count, but they have a structural advantage in qubit quality that translates directly to lower error-correction overhead. If QEC codes continue to improve (reducing the physical-to-logical ratio further), trapped ions could reach CRQC capability with a smaller machine than any superconducting alternative.