Years of academic activity is now transforming itself into a wave of spin-outs and startups. Trapped ion proponents point to the very high fidelity gates that have been achieved and additional opportunities for qubit connectivity. However the best gates have historically required precision Raman laser setups that come with their own significant challenges.
Strip away an electron from an atom and the positively charged ion that remains can be trapped and manipulated by a system of electric fields. Ions with just a single electron remaining in their outer shell are chosen to provide a convenient spectrum of energy levels. Two long-lived energy levels are chosen to form the basis of the qubit system.
Commercial approaches vary widely on how they balance the goals of fidelity and scalability.
For a detailed recent review of trapped ion qubit technology see [ ].
- Sector leading 2Q gate fidelities, 99.9% for laser gates [ ].
- Sector leading 1Q fidelities, 99.9999% for microwave gates [ ].
- Long qubit coherence lifetimes, 1-50s [ ]; lifetime/gate speed implies c 100,000 cycles.
- Longer ‘memory’ qubit lifetimes up to 10mins [ ] and recently 1 hour [ ] have been reported.
- Extreme cryogenic cooling is not required; laser cooling is used instead.
- Ions form fundamentally identical qubits (though trap fabrication naturally introduces imperfections in their control).
- Microfabricated surface-electrode traps allow a variety of possible qubit geometries, including 1D ions string and 2D arrays; QCCD designs allow the creation of specialised zones for loading, gates and memory.
- Ion shuttling and individual ion addressing provide mechanisms to enhanced effective qubit connectivity, avoiding gate-based swap operations [ ]. Further leading edge demonstrations are expected soon [ ].
- High fidelity and enhanced connectivity are attractive for NISQ applications.
- Photonic interconnects promise ease of connection to the Quantum Internet; entanglement between an ion and a photon has already been demonstrated over 50km of standard optical fibre [ ].
- Fast coupling between Rydberg states of ions offers one potential avenue to achieve significantly faster gate speeds. Initial experiments have demonstrated a 700ns 2Q gate with fidelity of 78% (with expectations up to 99.8%) [ ].
- High fidelity and connectivity may allow the use of more efficient error correcting codes, reducing the overhead required to achieve FTQC.
- Gate times are at best moderate and in some variations very slow (though mitigated by long qubit lifetimes).
- Laser driven gates offer very high fidelity but the alignment of such laser systems creates a key engineering challenge for scaling.
- Individual qubit/ion spacing can be c. 5μm; individual trap footprint c. 1mm (with multiple ions). Ultimately the number of ions that can be placed in a single linear trap is limited.
- Ultra-high vacuum required, when used at scale this can lead to long downtime cycles (to ease this effect, moderate cryogenic cooling is often employed).
- Footprint for large scale FTQC installations could be very large, ‘building scale’, machines.
- Photonic interconnects still look like a bottleneck, faster connections have recently been demonstrates but still only with an entanglement rate of 182/s (c.5.5ms) [ ]. There is however no known fundamental obstacle to making this faster.
- Slow underlying gate speed would be a disadvantage if another technology can be built at an otherwise similar level of performance.
Key Players and Approaches
To understand the differences between key approaches currently being pursued, it is necessary to look at exactly how the qubits are being defined and the method used to drive gates.
Hyperfine qubits and Raman Laser Gates
Ions with non-zero nuclear spin induce ‘hyperfine’ splitting in the electron energy level structure. Qubits defined on hyperfine states can have long natural lifetimes. Gate operations are controlled by Raman laser beams.
Notable Commercial Players: IonQ, Honeywell
- Long intrinsic qubit coherence lifetimes in the lab, 50s [ ].
- Sector leading 2Q fidelities in lab at 99.9% [ ].
- Very good 1Q fidelities in lab at 99.99% [ ].
- Moderate 2Q gate time in lab c. 50μs [ ]; still giving sector leading lifetime/gate speed at c.1,000,000 cycles.
- Faster 2Q gates have recently been demonstrated at 1.6μs and 99.8% fidelity [ ]; though these require higher Raman laser power.
- IonQ have successfully demonstrated a 11Q device based on a 1D ion string configuration achieving 2Q gate fidelity of 97.5% (avg) in this first device [ ].
- Honeywell have demonstrated a 4Q QCCD device with 2Q gate fidelity 99.2% [ ], this device is now operating in 6Q mode.
- These platforms must demonstrate a practical solution for scaling-up the laser driven gates on which they depend.
Honeywell Chief Scientist Patty Lee shocked QT Digital Week by announcing that their 6Q system has now achieved QV 64, surpassing the QV 32 achieved by IBM’s latest 27Q devices. This is less surprising than it sounds, as QV justifiably rewards high fidelity operations and enhanced connectivity. With due respect to both teams, this should not hype expectations for what can imminently be achieved by the Honeywell device. It should remind us of just how limited the performance of all early devices actually is: 6 qubits are easy to simulate on a conventional computer.
Optical Qubits with Laser Driven Gates
Qubits can also be defined based on ‘optical’ transitions in the electron level structure. Gates are implemented directly by lasers at red/IR wavelengths.
Notable Commercial Players/Consortia: AQT, AQTION, NextGenQ
- Very high 2Q fidelities in the lab, 99.6% [ ].
- Anticipated improvement to 99.98% 2Q gate fidelity and 15μs gates [ ].
- Direct optical control requires fewer lasers than Raman based techniques.
- The longer wavelength of red/IR light makes on-chip integrated photonic components significantly easier to fabricate [ ]. Such a system recently demonstrated 2Q gate fidelity of 99.3% [ ].
- AQT provide cloud access to their prototype 1D ion string device, boasting compatibility with a particularly wide range of emerging quantum software platforms: IBM Q, Cirq, Pennylane, Quest, pytket.
- Moderate coherence times e.g. 0.2s (with expectation of up to 2s) [ ] but these are intrinsically shorter than for hyperfine qubits; lifetime/gate speed currently c.4000 cycles is set to improve.
AQT is well placed to benefit from close association with the EU Quantum Flagship through the AQTION consortium. Fact Based Insight is particularly excited by the prospect of seeing integrated photonic laser delivery removing a traditional drawback of this platform.
Hyperfine qubits with Near-Field Microwave Gates
Gates on long-lived hyperfine qubits can also be driven by microwave antenna adjacent to each trap zone.
Notable Commercial Players/Consortia: Oxford Ionics
- Very high 2Q fidelities in lab 99.7% [ ]. In principle microwave control bypasses the photon-scatter limit on gate fidelity, so this is set to improve.
- Sector leading 1Q fidelities in lab 99.9999% [ ].
- Long qubit coherence lifetimes in lab e.g. 50s [ ].
- Demonstrated 2Q gate speed is slow 3.25ms [ ]; lifetime/gate speed c.15,000 cycles. However, this represents an early stage of development for this technique compared to others, speeds comparable to laser driven gates are anticipated.
- No multi-qubit device of this type has yet been demonstrated. Some will worry that crosstalk will be an issue. However wavelengths of 10cm are long enough compared to the traps that qubits experience a very smooth field, but still short enough that current is modelled as crowding tightly around the trap structures. This promises to restrict crosstalk to nearest neighbours at worst [ ].
It’s pertinent to note that Oxford Ionics’ founders Chris Balance and Thomas Harty are no strangers to other flavours of trapped ion technology. Their names stand on experiments that hold or share all of the key records for laser driven trapped ion performance. That they have chosen to champion microwave gate technology is something that will grab investor’s attention.
Dressed states with Global-field Microwave Gates
Each trap gate zone has a local magnetic field gradient. The exact position of an ion within the gradient is determined by a voltage individually applied to that gate zone. The voltage therefore modifies the local magnetic field the specific ion experiences and can place it in resonance with globally applied microwave fields that then drive a particular choice of quantum gate. Different gates are executed by applying different voltages. Because magnetically sensitive ion states are required, ‘dressed states’ are used to supress susceptibility to noise.
Notable Commercial Players/Consortia: Universal Quantum, MicroQC
- The number of radiation fields for quantum gate execution, e.g. lasers or microwave fields, does not scale with the number of qubits. This makes this approach very appealing for scaling to large numbers of qubits.
- Long qubit coherence lifetimes in lab e.g. 1s [ ].
- Optimised ion shuttling across the 2D trap grid promises an improvement in QV compared to gate-based qubit swaps [ ].
- Microwave fields are applied via global emitters, mitigating the need for impedance matching that is typically required if microwaves are applied to chip electrodes.
- Highly modular design. Universal Quantum proposes to extend the QCCD approach to incorporate wafer-to-wafer ion transport, promising connection speeds between modules orders of magnitude faster than typical photonic interconnects. Photonic interconnects also remain an option.
- Promising 2Q fidelities in lab e.g. 98.5% [ ].
- Target 2Q fidelity of 99.9% and gate speed of 361 microseconds [ ].
- Some worry that magnetic field gradient fluctuations across the extended global field may present challenges [ ].
- Current gate speeds are slow, 2700ms [ ].
This technology needs the potential for improved 2Q gate fidelity to be demonstrated. That aside, Fact Based Insight believes it stands out for offering a complete vision of how a large scale FTQC device can be built by solving a series of purely engineering scale challenges [ ].
NextGenQ plan to build an initial device using optical qubits, and then a second device using this approach.