Rare-Earth Quantum Memories: What Is Real

First piece of a series on the hardware behind quantum networks. Each paragraph carries its sources, with a tag showing how firm the evidence is: Journal (peer-reviewed) or Preprint (on arXiv, not yet formally reviewed).

What researchers are trying to do

Rare-earth ions such as erbium, ytterbium and europium are studied as optical quantum memories and as connection points for quantum networks. A few ions are placed inside a crystal, where they behave like nearly isolated atoms that can be controlled with light. This series asks a plain question: what have experiments already shown, what is still missing, and what conclusions do the data support?

Journal (literature review). Sources: Review "Quantum networks using rare-earth ions" (Quantum Sci. Technol. 2025, arXiv version)

Why erbium stands out

Among the ions most used in quantum memories, erbium is the only one whose optical transition falls in the 1530 nm window that telecom optical fiber uses. Europium, praseodymium, thulium and ytterbium work at other wavelengths and need frequency conversion to talk to fiber. In 2025, a Princeton group entangled the spin of a single erbium ion with a photon emitted at 1532.6 nm and estimated an entanglement fidelity of 73% after 15.6 km of fiber.

Journal. Sources: Review (Quantum Sci. Technol. 2025) · PRX 15, 011071 (2025), single Er ion, Princeton

What storage has been shown in the telecom band

In an erbium-doped lithium niobate chip, an Illinois group stored single-photon-level light pulses for more than 1 microsecond, across up to 20 temporal modes and with an acceptance bandwidth of up to 2.2 GHz. A June 2026 preprint transferred telecom optical excitations to the hyperfine spins of 167Er in Y2SiO5 and retrieved them after up to 25 microseconds, with a transfer efficiency above 12%, at 0.8 kelvin and 1.1 tesla. The authors also measure a hyperfine population lifetime of 25 seconds, but that is not the demonstrated storage time. They identify inhomogeneous broadening as the current limit.

Journal + preprint. Sources: Nano Letters 26, 12056 (2026), Er in lithium niobate · arXiv 2606.15009 (Jun 2026), 167Er hyperfine spins

On duration: the 2025 review reports spin coherence of up to hours in europium (Eu:Y2SiO5). That is a different ion, outside the telecom band, and spin coherence is not the same as an efficient memory for photons. Among the telecom erbium results I could verify, storage times are in the microsecond range.

Journal (review). Source: Review (Quantum Sci. Technol. 2025)

And neodymium

It is the same element found in the neodymium-iron-boron magnets of old hard disk drives, used in the motor that moves the read head. In a quantum memory, though, neodymium is not used as a magnet. It is used as a few Nd3+ ions inside a transparent crystal. There is serious experimental work here, for example a nanophotonic memory built from a neodymium ensemble coupled to a cavity (Science, 2017). Its transition in Nd:YVO4 is at 879 nm, outside the telecom band, so connecting it to fiber would need frequency conversion. The 2025 review does not list it among the ions of greatest current interest for memories, and I found no recent neodymium results in this first pass. So there are fewer reasons to highlight it for now. It stays on the watch list.

Journal + technical source. Sources: Science 2017, nanophotonic neodymium memory · PRB 77, 125111 (2008), Nd:YVO4 at 879 nm · Stanford Magnets: magnets in HDD voice coil motors · Review (Quantum Sci. Technol. 2025)

Other elements, different strengths

In Geneva, an ytterbium memory in Y2SiO5 stored 8235 modes for 63 microseconds over 5.66 km of urban fiber (preprint, August 2026). Another preprint entangled a trapped ytterbium ion with a europium memory 75 m away, with a fidelity of 89%, using frequency conversion. There is no overall ranking: each ion brings a different wavelength, bandwidth and coherence time, and the best choice depends on which part of the network you want to solve.

Preprints. Sources: arXiv 2608.13177 (Aug 2026), Yb in Geneva · arXiv 2603.05836 (2026), trapped ion + Eu memory · Review (Quantum Sci. Technol. 2025)

Quick comparison by ion

• Erbium (Er): telecom-band wavelength (~1530 nm). Results I read: spin-photon entanglement of a single ion (73% after 15.6 km), on-chip storage beyond 1 µs, and 167Er spin retrieval at 25 µs at 0.8 K.
• Ytterbium (Yb): around 980 nm, not telecom. Geneva preprint: 8235 modes stored for 63 µs over 5.66 km of urban fiber.
• Europium (Eu): around 580 nm, not telecom. The 2025 review cites spin coherence of up to hours in Eu:Y2SiO5. Preprint: entanglement with a trapped ytterbium ion at 75 m, 89% fidelity.
• Neodymium (Nd): near 879 nm in Nd:YVO4, not telecom. Nanophotonic cavity memory shown in 2017. I found no recent results in this first pass.
• Praseodymium, thulium: listed in the 2025 review (~606 nm and ~794 nm). Not covered in depth yet.

What can and cannot be concluded

Erbium is the ion that combines a telecom wavelength with experimental results on a chip, and the other ions show different strengths, such as very long spin coherence in europium. From what I read, one ion cannot be called better than another in general: the results use different setups, temperatures and metrics, and are not comparable number to number.

What to watch next

• Telecom storage in erbium: whether it moves from microseconds to longer times, with measured efficiency and fidelity.
• Entanglement between memories at a distance, with reported rates and operating conditions (temperature, magnetic field).
• Ions less present in this first piece, such as neodymium, praseodymium and thulium, and any new materials.