Calcium-43 ions as long-lived qubits

Short answer: Calcium-43 ions store a qubit in hyperfine levels with very long coherence. A 2008 preprint reported a 1.2 second coherence time, with an inferred spin-echo time of at least 45 seconds, and a 2014 Physical Review Letters paper from Oxford reported 99.93% preparation and readout fidelity, a 50 second memory coherence time and 99.9999% single-qubit gate fidelity. All were obtained on single ions in a laboratory.

2007 to 2008: first long coherence in calcium-43

The arXiv abstract 0710.4421 (submitted October 2007, "A long-lived memory qubit on a low-decoherence quantum bus", Lucas and coauthors) demonstrates long-lived coherence in the internal hyperfine states of a single calcium-43 ion, with T2 of 1.2 seconds (uncertainty 0.2). It also reports coherence of 0.18 seconds in the external motional states of a calcium-40 ion in the same apparatus. The motional decoherence rate is consistent with the measured heating rate of 3 quanta per second.

The authors say the internal-state T2 in calcium-43, which had not been used as a trapped-ion qubit before, is about one thousand times longer than for physical qubits based on calcium-40. A single spin-echo pulse re-phases the internal state, and they could detect no decoherence after 1 second, implying an effective spin-echo coherence time of at least 45 seconds. They compare this with about 1 microsecond for single-qubit operations and about 10 microseconds for two-qubit operations in this trap. This is a preprint abstract, and the 2008 listing for arXiv 0810.1036 could not be read, so no claims are made from it.

2014: preparation, gates, memory and readout

The Physical Review Letters paper (Harty and coauthors, Oxford, published November 2014) reports implementing all single-qubit operations with fidelities significantly above the minimum threshold for fault-tolerant quantum computing, using a qubit in hyperfine atomic clock states of calcium-43. The authors measure a combined state preparation and single-shot readout fidelity of 99.93%, a memory coherence time T2 of 50 seconds, and an average single-qubit gate fidelity of 99.9999%. The results were achieved in a room-temperature microfabricated surface trap, without magnetic field shielding or dynamic decoupling to overcome technical noise.

How the two compare

The 2014 memory coherence of 50 seconds is reported without a spin echo or decoupling, while the 2008 paper reached 1.2 seconds without a spin echo and a lower bound of 45 seconds with one. The two experiments used different traps. The abstracts do not compare the two directly.

Why hyperfine clock states

Both abstracts point to hyperfine states as the source of the long memory. The 2014 authors describe their qubit as stored in hyperfine atomic clock states, and the 2008 authors contrast the calcium-43 internal-state coherence with that of calcium-40, which they say is about a thousand times shorter.

Limits

Single-qubit performance is a necessary ingredient and not a complete one. Neither abstract reports a two-qubit gate, and the fidelity numbers are the authors' own measurements.

FAQ

Why calcium-43?

Its hyperfine clock states give long coherence, as both papers report.

How long did the qubit stay coherent?

1.2 seconds in 2008, with at least 45 seconds inferred by spin echo, and 50 seconds in 2014.

How good were the single-qubit gates?

99.9999% average fidelity in 2014, per the authors.

Did it need shielding?

The 2014 authors say no magnetic field shielding or dynamic decoupling.

Does this make a computer?

No. Two-qubit gates are not covered in these abstracts.