By Andre Saraiva, UNSW

Superconducting qubits (Google, IBM and Rigetti): This is perhaps the easiest comparison. Amazon’s qubits are pretty much the same size, use similar control and readout techniques and are also setup in a 2D array with fixed addresses for each qubit. The difference is mostly the mode of operation, which Amazon claims will give them added qubit-level protection against errors. This claim hasn’t yet been demonstrated experimentally, despite initial efforts.

Advantage: Amazon’s qubit is coupled to a long-lived nanomechanical resonator, providing it with bit flip error rates that are far better than the traditional transmon approach. This is core to their plans for magic state generation and, potentially, could even lead to better alternative methods of error correction (see Part 2 of our article series). In other words, this qubit would show its potential advantage when quantum processors reach the 1k-100k qubit count.

Disadvantage: This qubit intrinsically relies on the integration between technologies in a way that is not yet proven (see Part 1 of our article series). This disadvantage might turn out to be irrelevant if in practice transmon qubits also require integration with other quantum technologies in order to scale up. Perhaps the most immediate disadvantage is from the scientific standpoint – transmons are well understood by now, while cat states have unknown limitations.


Spin qubits (Intel, QuTech, SQC): Even though these two systems are similar in that they are based in solid state devices, a direct comparison between the two technologies is hard because an in-depth analysis of the implementation of error correction in a realistic spin-based quantum computer hasn’t been performed yet. One similarity between the two technologies is the error bias – spin qubits take much longer to suffer from bit flip than from phase flips.

 Advantage: Amazon’s plan for a quantum processor has a well-defined strategy for averting qubit crosstalk. Spins, on the other hand, are minute entities that need to be crammed together at no more than tens of nanometres apart. Since there hasn’t yet been a demonstration of a large-scale spin qubit processor, it is unknown whether the resulting crosstalk will be a major limitation.

Disadvantage: The first clear difference between the two technologies is the size of the qubits. Spin qubits require 100 to 10 000 times smaller area, which means more die area can be made available for quantum error correction. Moreover, spin qubits can be operated at higher temperatures, which makes a large difference in cooling power and cost. When it comes to simultaneously controlling thousands of qubits, the cooling requirements quickly grow.


Ion trap qubits (IonQ, Honeywell): This is a much more difficult comparison to make. Even though both approaches are based on a “kosher” form of gate-based quantum computation, the geometries are wildly different.

Advantages: The main advantage is a clear pathway for scalability. Amazon’s qubits can be scaled up by simply juxtaposing more and more qubits in the same chip. Also, because of the nature of the resonators, wiring between two chips is somewhat direct. There have even been demonstrations of superconducting resonators in different fridges being wired together and interacting with good performance. Ion traps, on the other hand, can only faithfully hold and operate a few tens of qubits before they start losing ions. That means that the future of ion traps is pending the demonstration of methods to couple different traps or to reconfigure traps electrostatically.

Disadvantages: The main engineering disadvantage of Amazon’s qubits in comparison to ion traps is the need for a dilution refrigerator. Ions are cooled down in a vacuum chamber entirely by lasers, making the whole setup more compact and imposing less restrictions in control, such as limits in wiring. From the quantum information point of view, ion traps have two major advantages over most other qubit technologies: extremely high-fidelity control of their qubits because they work with atomic clock transitions; and all-to-all connectivity within a trap, which means that two-qubit entangling gates can be performed in a single shot. Amazon’s qubits only interact with its nearest neighbors, and its fidelity is limited by the fact that their mode of operation relies on engineered interactions with the environment, which can lead to instabilities.

If you missed Part 1 of this series that describes the qubit architecture, you can view it here. If you missed Part 2 of this series that describes the error correction architecture, you can view it here.


Dr. Saraiva has worked for over a decade providing theoretical solutions to problems in silicon spin quantum computation, as well as other quantum technologies. He works on MOS Silicon Quantum Dot research and commercially-oriented projects at the University of New South Wales (UNSW).

January 11, 2021