Quantum Memory

Quantum memory is to the quantum internet what RAM is to a classical computer: a place to hold information temporarily while other parts of a system catch up. Without it, quantum networks are severely limited. With it, they become practical.

The central challenge is preserving quantum coherence, the property that allows a qubit to exist in superposition, for long enough to be useful. Coherence is fragile: any interaction with the environment tends to destroy it. Building a quantum memory means engineering a system where controlled, deliberate interactions with photons are possible while uncontrolled, environmental interactions are minimized.

Why quantum memory is hard

Storing a classical bit is easy: a capacitor holds a charge, a magnetic domain holds an orientation, and the value can be refreshed indefinitely by reading it and writing it back. None of those options exist for a qubit.

A qubit in a superposition cannot be read without collapsing the superposition. It cannot be copied to a backup location, because the no-cloning theorem forbids exact copying of unknown quantum states. And it interacts with its environment, losing coherence over time through a process called decoherence. A quantum memory must somehow isolate the qubit from the environment well enough to preserve its state, hold it for as long as needed, and then release it on demand with high fidelity.

The fundamental tension is that the same interactions that allow a qubit to be written into memory and read back out also allow the environment to disturb it. Engineering a quantum memory means finding physical systems where controlled interactions are possible but unwanted environmental coupling is suppressed.

Physical implementations

Atomic ensembles store quantum information in the collective spin state of a large cloud of atoms. Individual atoms contribute to a collective excitation, providing some natural robustness against loss of a single atom. The DLCZ protocol (Duan, Lukin, Cirac, Zoller, 2001) showed how atomic ensembles could be used to build practical quantum repeaters: a weak laser pulse creates a single collective excitation in the ensemble, heralded by detection of a scattered photon, and that excitation can later be retrieved as a photon on demand.

Rare-earth-doped crystals use ions such as europium or praseodymium embedded in a crystal lattice. The crystal provides a rigid, predictable environment, and the rare-earth ions have optical transitions that couple well to photons. The atomic frequency comb (AFC) protocol tailors the absorption spectrum of the ensemble to store multiple temporal modes simultaneously, making these memories well-suited for high-bandwidth quantum networks. Storage times of seconds have been demonstrated in some rare-earth systems.

Individual atoms in optical cavities confine a single atom inside a high-finesse optical resonator. The cavity enhances the coupling between the atom and photons, making it possible to efficiently absorb a single photon into the atom’s internal state and emit it later. These systems offer excellent control and long coherence times but are technically demanding.

Nitrogen-vacancy (NV) centers in diamond are point defects in a diamond crystal where a nitrogen atom replaces a carbon atom adjacent to a lattice vacancy. The electron spin associated with the NV center can store a qubit for milliseconds at room temperature, and nuclear spins nearby can extend this to seconds. NV centers have the significant advantage of operating at room temperature and being addressable with standard laser and microwave equipment.

Trapped ions store quantum information in the electronic or hyperfine states of individual ions held in electromagnetic traps. Coherence times can exceed minutes, making trapped ions among the longest-lived quantum memories available. They are also used as qubits in quantum computers, so a trapped-ion quantum memory could in principle directly interface with a trapped-ion quantum processor.

Key metrics

When comparing quantum memories, four metrics matter most:

Storage time is how long the memory can hold a quantum state before fidelity degrades below a useful threshold. For quantum repeater applications, the required storage time is set by the time needed to establish entanglement on neighboring links, which depends on distance and link efficiency. Current demonstrations range from microseconds to several seconds.

Efficiency is the probability that a photon written into the memory is successfully retrieved when requested. Low efficiency means many attempts are wasted. Practical repeater designs require efficiencies well above 90%; current systems achieve 50-80% in lab conditions.

Fidelity measures how close the retrieved state is to the stored state. Errors from decoherence, imperfect coupling, and retrieval operations all reduce fidelity. High-fidelity storage is essential because errors accumulate across multiple repeater nodes.

Multimode capacity refers to how many independent quantum states the memory can hold simultaneously. A memory that stores many temporal modes can accept a burst of photons and release them in sequence, greatly improving the throughput of a quantum network. AFC memories can store dozens to hundreds of temporal modes.

Relevance to quantum repeaters

Quantum memory is the enabling technology for second-generation quantum repeaters. Without memory, a repeater can only attempt link establishment sequentially: wait for one link to succeed, then attempt the next. The probability of all links succeeding simultaneously drops exponentially with the number of links.

With memory, a node stores one end of an entangled pair after a link succeeds, and waits for neighboring links to complete before performing entanglement swapping. Links can be attempted in parallel. The memory requirement is that the stored state must remain coherent long enough for all neighboring links to succeed, which imposes a minimum storage time that grows with network distance.

Relevance to distributed quantum computing

Beyond repeaters, quantum memory enables distributed quantum computing by buffering qubits at the interface between a quantum processor and a quantum network. A processor may finish computing a result that needs to be teleported to a remote node, but the entangled pair required for teleportation may not yet be available. A quantum memory holds the result while the network establishes the link, then releases it for teleportation when the resource arrives.

Common misconceptions

Misconception: quantum memory and qubit storage in a quantum computer are the same thing. Quantum processors store qubits in their physical qubits (trapped ions, superconducting circuits, etc.) during computation, but these are not quantum memories in the networking sense. A quantum network memory is specifically engineered to accept, hold, and re-emit photons that arrive from an external optical channel. This requires an efficient interface between flying qubits (photons) and stationary qubits (the memory medium), which is a distinct engineering challenge from building a good quantum processor.

Misconception: longer storage time always means better memory. A memory that stores a qubit for hours but releases it with 50% efficiency and 90% fidelity may be far less useful for a quantum repeater than one that stores for milliseconds with 95% efficiency and 99% fidelity. All four metrics matter, and the right balance depends on the network architecture and the distance being spanned.

Misconception: any quantum system that stores a qubit is a quantum memory. The term is sometimes used loosely, but in the context of quantum networking it specifically refers to a device with an optical interface that can accept photons from a quantum channel, store the encoded quantum state, and re-emit it on demand. A qubit in a quantum computer that sits idle between gates is not a quantum memory in this sense, even if it preserves its state for a long time.

See also

• Quantum Repeater
• Quantum Internet
• Decoherence
• Entanglement
• Qubit