English
Menu

Superconducting vs Trapped-Ion Qubits: Gate Speed and Coherence

Both are leading ways to build the qubits of a quantum computer. The superconducting approach uses a superconducting circuit made of Josephson junctions and capacitors (such as a transmon) as the qubit and controls it with microwave pulses. It can be made on a chip with the same lithography as semiconductors, but it must be cooled to about 10 mK in a dilution refrigerator. The trapped-ion approach uses energy levels of the electrons in ions held in vacuum by electric fields as the qubit and controls them with lasers or microwaves. In general, superconducting qubits have faster gates, while trapped-ion qubits have longer coherence times and connect to each other more freely.

AspectSuperconducting qubitTrapped-ion qubit
What the qubit isThe two lowest energy levels of a superconducting circuit made of a Josephson junction and a capacitor (such as a transmon)Two electronic energy levels of an ion (such as ytterbium or calcium) trapped in vacuum
How it is madeFabricated as a circuit on a chip by lithographyAtoms are ionized, then trapped and lined up by electric fields from electrodes
Operating environmentCooled to about 10 to 20 mK in a dilution refrigeratorUltra-high vacuum, with the ions laser-cooled; no millikelvin refrigerator is needed (some systems cool to a few kelvin to improve the vacuum)
Control methodMicrowave pulses at a few GHzLasers (or microwaves)
Typical gate speedFast: two-qubit gates take about tens to hundreds of nsSlow: two-qubit gates take about a few µs to hundreds of µs, often 2 to 3 orders of magnitude slower than superconducting gates
Typical coherence timeAbout tens to hundreds of µsLong: seconds, and longer under some conditions
ConnectivityCoupled through circuits on the chip, so mainly between neighboring qubitsIons in the same trap share common vibrations, so two-qubit gates can be applied to any pair
Variation between qubitsAs fabricated circuits, small differences in the junctions spread the frequencies, so each qubit needs calibrationAll ions of the same species are identical, so there is no fabrication variation
Scaling challengesThe number of wires into the refrigerator and the heat they bring in, crosstalk between qubits, losses from material defectsThe limited number of ions that fit in one trap, and slow gates; approaches that shuttle ions or link traps with light are being studied

Figures are typical orders of magnitude and vary by process generation and product.

When to use which

Superconducting qubit
When the fast gates should be used to run many circuits in a short time. Because it builds on semiconductor fabrication, it makes it easier to place many qubits on a chip.
Trapped-ion qubit
When long coherence times and free connectivity between qubits matter. Each gate is slow, but circuits with many qubit-to-qubit interactions are easier to run without extra rerouting operations.

FAQ

Why must superconducting qubits be cooled to about 10 mK?

Not only to make the aluminum of the circuit superconducting (below about 1.2 K), but also to make thermal excitation and noise small enough. The energy hf of a 5 GHz qubit corresponds to a temperature of about 0.24 K, so the temperature must be well below that, and the roughly 4 K of liquid helium is not enough.

Why can trapped-ion systems apply gates between distant qubits?

Ions lined up in the same trap push on each other through the Coulomb force, so they share the vibrations of the whole chain. By using lasers to couple each ion's internal state to these vibrations, a two-qubit gate can be applied between any two ions, not just neighbors. As the number of ions grows, however, the vibrational modes crowd together and gates become harder, so the number of ions in one trap is limited.

How should gate speed and coherence time be compared?

What matters is not only the absolute speed but how many gates fit within the coherence time, and the gate fidelity. Superconducting qubits have fast gates but shorter coherence, while trapped ions have slow gates but longer coherence, so the approaches cannot be compared by a single number. Real performance is judged together from fidelity, connectivity, readout errors and more, not just the number of qubits.

Check your understanding

Which correctly describes the relationship between the dephasing time T2 and the energy relaxation time T1 of a qubit?

  1. AT2 is set independently of T1 and is often more than 100 times T1
  2. BT2 is equal to T1 for every qubit
  3. CT2 does not exceed 2T1
Show answer and explanation

Answer: C. T2 does not exceed 2T1

T2 is the time until the information on the relative phase of a superposition α|0⟩ + β|1⟩ is lost. Energy relaxation also destroys phase information, so 1/T2 = 1/(2T1) + 1/Tφ, where Tφ is the pure dephasing time due to frequency fluctuations and similar causes, and therefore T2 ≤ 2T1. For example, with T1 = 100 µs and Tφ = 150 µs, T2 is about 86 µs. In real qubits, frequency fluctuations caused by flux or charge noise often make T2 considerably shorter than 2T1.

Which element is essential for a superconducting qubit (transmon) to work as a qubit?

  1. AA pn junction diode
  2. BA MOSFET with a floating gate
  3. CA Josephson junction
Show answer and explanation

Answer: C. A Josephson junction

A Josephson junction consists of two superconductors separated by a thin insulating film about 1–2 nm thick; a superconducting current flows through it without producing a voltage, and it acts as a lossless nonlinear inductor. A transmon is a resonant circuit that combines this junction with a capacitor. In an ordinary LC resonator made of a coil and a capacitor the energy levels are equally spaced, so the transition between |0⟩ and |1⟩ cannot be addressed on its own; the nonlinearity of the junction makes the spacing unequal, so the two lowest levels can be used as a qubit. The junction is usually made of aluminum and aluminum oxide formed by oxidizing its surface.

What is the main reason superconducting qubits are cooled to about 10–20 mK in a dilution refrigerator?

  1. ATo make the wiring and junctions superconducting and to make thermal excitation of the qubits and thermal noise sufficiently small
  2. BBecause cooling widens the band gap of silicon and raises the qubit frequency
  3. CBecause electron mobility rises at low temperature, making gate operations 1,000 times faster than at room temperature
Show answer and explanation

Answer: A. To make the wiring and junctions superconducting and to make thermal excitation of the qubits and thermal noise sufficiently small

The aluminum used in transmons becomes superconducting below about 1.2 K, eliminating resistive losses. In addition, the energy hf of a qubit at a frequency of about 5 GHz corresponds to a temperature of about 0.24 K, so unless the thermal energy kT is made much smaller than this, the qubit is thermally excited to |1⟩ and thermal noise enters through the wiring. Liquid helium temperature, about 4 K, is therefore not enough, and the qubits are cooled to about 10 mK with a dilution refrigerator that uses a mixture of helium-3 and helium-4. Qubit speed is not set by mobility, and superconducting qubits do not use the band gap of silicon.

  1. Which correctly describes dispersive readout, which is widely used to read out superconducting qubits?
  2. Which correctly describes a trapped-ion quantum computer?
  3. Which is the most appropriate comparison of superconducting and trapped-ion platforms?

Practice Quantum Computing Hardware questions

Key terms explained

Qubit
The unit that represents information in a quantum computer. It can be in a superposition α|0⟩ + β|1⟩ of |0⟩ and |1⟩, and a measurement gives 0 with probability |α|² and 1 with probability |β|². Qubits are realized with superconducting circuits, ions, neutral atoms, electron spins in silicon, photons and more.
Transmon
A superconducting qubit that combines a Josephson junction with a capacitor. Its frequency is generally a few GHz, and the nonlinearity of the junction makes the energy levels unequally spaced, so the two lowest levels are used as a qubit. Its design makes it insensitive to charge noise, and it is widely used in superconducting platforms.
Josephson junction
A structure in which two superconductors are separated by a thin insulating film or similar layer about 1–2 nm thick. A superconducting current flows through it without producing a voltage, and it acts as a lossless nonlinear inductor. It is used in superconducting qubits, ultra-sensitive magnetic sensors (SQUIDs), voltage standards and more. For qubits it is usually made of aluminum and its oxide.
Dilution refrigerator
A refrigerator that uses a mixture of helium-3 and helium-4 to produce cryogenic temperatures of about 10 mK continuously. It is used to operate superconducting qubits and silicon spin qubits. The cooling power of its coldest stage is small, roughly ten to several tens of µW, so heat flowing in through the wiring is a challenge for scaling up.
Ion trap
A device that holds ions in a vacuum with electric fields or similar means, or the quantum computing approach that uses one. Electronic energy levels of the ions serve as qubits and are controlled with lasers or microwaves. All the qubits are identical, with long coherence times and high fidelities, but gates are slower than in superconducting platforms.
Decoherence
The process by which a qubit, through interactions with its environment, loses the phase and energy information of its superposition and its quantum properties break down. Its speed is described by the energy relaxation time T1 and the dephasing time T2.
Quantum gate
A basic operation that changes the state of qubits. There are single-qubit gates such as X (bit flip), Z (phase flip) and H (Hadamard) and two-qubit gates such as CNOT, and they can be represented as rotations on the Bloch sphere. They are implemented with microwave pulses for superconducting qubits and with lasers and similar means for ions and neutral atoms.

Related comparisons