Scaling the Quantum Frontier: Meet Helios

Author: Denis Avetisyan

A 98-qubit trapped-ion processor pushes the limits of quantum computation with a novel architecture and robust performance.

Within the Helios 2D surface trap, ninety-eight atomic ions—arranged in a quincunx—become visible only when coaxed into resonance by laser light, a demonstration of how even the most fundamental particles yield their secrets only through carefully calibrated interaction, and betray the hand of the physicist who designed the trap itself.

Researchers detail the design and benchmarking of Helios, a transport-based trapped-ion quantum processing unit demonstrating advances in qubit control, connectivity, and error mitigation.

Despite ongoing challenges in scaling quantum systems while maintaining high fidelity, significant progress is being made in trapped-ion quantum computing—as demonstrated in ‘Helios: A 98-qubit trapped-ion quantum computer’. This work introduces Helios, a 98-qubit processor utilizing a quantum charge-coupled device architecture and achieving component infidelities predictive of surpassing the capabilities of classical simulation through random circuit sampling. With all-to-all connectivity and improved operational speeds, Helios represents a new benchmark in scalability and fidelity. Will this advancement pave the way for more complex quantum algorithms and fault-tolerant quantum computation?

Beyond Connectivity: Architecting for Quantum Scale

Existing trapped-ion quantum computers struggle with qubit connectivity and control complexity, hindering their ability to execute complex algorithms and scale. Current architectures often require extensive routing, increasing error rates and overhead. Helios, a next-generation trapped-ion system boasting 98 qubits, overcomes these obstacles through architectural innovation. Its segmented trap—a ring-shaped storage region and central processing zone—efficiently loads and manipulates qubits. This staged approach minimizes ion transport and improves connectivity. Helios decouples storage from computation, reducing crosstalk and simplifying control – a shift from overcoming physical limitations to managing information flow.

The Helios design facilitates qubit manipulation through a staged loading process, where ions are rotated within a ring storage system and transferred to a cache region for quantum operations, as illustrated by the sequential stages of loading, trap configurations, and crystal alignment.

Compartmentalization: A Blueprint for Qubit Control

Helios employs a novel architecture with dedicated regions for qubit storage—Ring Storage, Leg Storage, and Cache—and quantum logic. This compartmentalization optimizes performance for distinct computational stages. Efficient qubit routing is enabled by the X-Junction, a programmable crossroad that dynamically reconfigures pathways, avoiding the bottlenecks of fixed-connectivity designs. This approach optimizes qubit allocation, reduces control complexity, and allows for parallel operations, increasing computational throughput.

A breakdown of transport operations within a 98-qubit program reveals that the time is distributed among global shifts, quantum operations, storage leg movements, four-ion shifts, and static operations, providing insight into the program’s execution profile.

Precision and Stability: Mitigating the Noise of Reality

Maintaining qubit coherence demands precise control over their states. Helios employs State Preparation and Frequency Selective State Preparation to initialize and manipulate qubits, minimizing unwanted transitions. Spatial Phase Tracking actively compensates for magnetic field fluctuations, enhancing qubit frequency stability. Protected Measurement, Ternary Measurement, Mid-Circuit Measurement, and Reset functionalities minimize errors and improve readout fidelity, yielding a measured Single-Photon-Addition/Multi-Photon-Subtraction (SPAM) error of 5.3(51) x 10^-4.

Analysis of 98-qubit mirrored RCS circuits demonstrates an exponential decay in fidelity with increasing circuit depth, consistent with gate-counting predictions, while cost estimations reveal that classical sampling via tensor-network contraction requires significantly more time and power compared to sampling with Helios, particularly when limited by memory constraints.

Validation and Benchmarking: Measuring Against the Impossible

Helios’s performance is rigorously evaluated using standard benchmarks, including Random Circuit Sampling and Random Clifford Circuits. Tensor Network Contraction approximates the complexity of these circuits, simulating behavior beyond classical capabilities. The transition from Ytterbium-171 to Barium-137 Ions demonstrably improves performance and scalability, yielding an effective 2-qubit gate error of 2.00(6) x 10^-3—a significant advancement in fidelity.

Time budget analysis for a depth-10 random program executing on 98 qubits indicates that ion transport, ground-state cooling, and quantum operations each contribute significantly to the total time per layer, with the total time varying based on the number of active qubits and the density of two-qubit gates.

Dynamic Control and the Illusion of Order

The Helios Runtime software dynamically allocates qubits and schedules gates, optimizing computational efficiency. This adaptive system manages resources based on real-time needs, minimizing idle time. The Molmer-Sørensen gate, driven by Raman beams, enables high-fidelity two-qubit entanglement. A 3 ms ground state cooling time minimizes decoherence and maintains qubit stability. By addressing limitations in connectivity, control, and calibration, Helios surpasses the capabilities of current supercomputers on specific quantum tasks, ultimately offering not solutions, but a temporary reprieve from the anxiety of the unknown.

An example four-qubit program demonstrates the arbitrary permutation of qubits and the application of single- and two-qubit gates at each layer, showcasing the program's operational structure. — An example four-qubit program demonstrates the arbitrary permutation of qubits and the application of single- and two-qubit gates at each layer, showcasing the program’s operational structure.

The pursuit of Helios, a 98-qubit system, exemplifies a predictable human tendency: the relentless scaling of ambition despite inherent uncertainty. Everyone calls these endeavors ‘rational progress’ until the costs—in resources, time, and perhaps, eventual disillusionment—become undeniably apparent. Werner Heisenberg observed, “The very act of observing changes an object.” This resonates deeply with quantum computing; the act of measurement, integral to verifying Helios’s fidelity through RCS benchmarking and mid-circuit measurement, fundamentally alters the quantum state. It’s not merely about building a more powerful machine, but acknowledging that every attempt to quantify reality introduces a layer of subjective influence, transforming the objective into the observed.

What’s Next?

The construction of Helios, a 98-qubit device, represents a predictable escalation. Each additional qubit isn’t simply a technical achievement; it’s an exercise in managing escalating complexity – and, more subtly, an assertion of control. Every chart is a psychological portrait of its era, revealing a deep-seated human need to map chaos onto order. The benchmarks detailed within speak not just to fidelity, but to the enduring faith in the possibility of perfect measurement, a faith consistently undermined by the very quantum mechanics being harnessed.

The emphasis on transport-based qubits, while offering scalability, merely shifts the problem. Error correction, the holy grail, remains elusive not because of a lack of ingenuity, but because errors aren’t random glitches—they’re symptoms of interaction, of the universe refusing to be neatly contained. Mid-circuit measurement, a clever workaround, acknowledges this fundamental friction.

Future iterations will undoubtedly pursue higher qubit counts and improved fidelity. However, the truly interesting questions lie not within the hardware itself, but in the evolving assumptions baked into the software and algorithms. The limitations aren’t physical; they’re cognitive. Humans consistently overestimate their ability to predict and control complex systems, and quantum computing is, perhaps, the ultimate test of that hubris.

Original article: https://arxiv.org/pdf/2511.05465.pdf

Contact the author: https://www.linkedin.com/in/avetisyan/