What is quantum memory?

Quantum memory is the equivalent of classical computing’s memory, except it’s far more difficult to implement in real-world settings because of the volatile nature of quantum states. Subtle environmental changes, including the act of observing quantum information, can alter its properties.

Classical computing components such as RAM, solid-state drives and hard disk drives store information as bits, represented through binary code with 0s and 1s. But quantum computers use qubits, which take advantage of the properties of quantum mechanics. On a quantum level, particles can exist in multiple states at the same time, containing elements of both 0 and 1 until measured. This phenomenon is known as superposition. Quantum memory is responsible for storing and preserving the information encoded on qubits.

A quantum system resolves to either a 0 or 1 when external forces act upon it, such as when an observer views it. Qubits are made up of a combination of two basis states: One corresponding to 0 and another corresponding to 1. Each basis state has an associated probability amplitude—a complex number that can be used to assess the likelihood that the quantum system will become a 0 or 1 upon viewing.

Each probability amplitude contains both magnitude and phase information. The square of the magnitude describes the probability of measuring the outcome that each basis state represents (either 0 or 1). The probabilities of each basis state must always add up to 100%. For example, a valid quantum system might have a 90% probability of resolving to 0 and a 10% probability of resolving to 1, or a 50% probability of each outcome.

The phase information describes how the probability amplitudes of each basis state interact with each other. Amplitudes can cancel each other out (destructive interference) or reinforce each other (constructive interference) to either decrease or increase the likelihood of different outcomes—like how a pair of waves can interact to either strengthen or dampen their respective signals. (While phases can alter the probability of different outcomes, the sum of these outcomes must still add up to 100%.)

To run different quantum computations, scientists can use quantum logic gates to manipulate probability amplitudes, and thus make some outcomes more likely than others. They can also place qubits into entangled states, where multiple qubits contribute to a shared quantum state, even when separated by long distances. This phenomenon enables quantum computers to treat multiple qubits as a unified entity so that those qubits can be deployed to collectively solve problems more efficiently.

Researchers have developed various quantum memory techniques to reliably store qubits without destroying them and to enable sharing between multiple quantum modules and chips. While classical components can generally operate at room temperature, many quantum memory devices store quantum information in temperatures that approach absolute zero. This method reduces the risk of thermal interference and can help eliminate electrical resistance in cases where superconductive materials are used to store quantum information.

“Quantum phenomena in general is something that’s inherently quite unstable,” said Matt Hollister, Head of Cryogenic Systems Engineering for IBM Quantum, on a recent episode of The Coherence Times podcast. “So to maintain the quantum phenomena that we’re using in these computing devices, it helps to operate at very low temperatures in order to reduce the phenomena of noise coming in from the environment.”

Quantum memory is a necessary part of functional, synchronized and scalable quantum networks, where multiple quantum computers can reliably maintain and exchange quantum information. Although largely used in experimental settings today, this capability might soon help researchers perform operations faster and with a higher degree of complexity than would be possible with classical computing. Practical applications range from cybersecurity to materials science, drug discovery, advanced mathematics and machine learning.

While classical memory and quantum memory are both concerned with storing, preserving and retrieving data, quantum memory introduces additional complexity due to the vulnerable, error-prone nature of quantum states. Major differences include:

Quantum memory aims to securely capture a quantum system, hold it in a stabilizing medium and release it without interfering with the quantum information itself.

First, scientists must choose which phenomenon they want to encode with quantum information. Subatomic particles exhibit quantum behaviors, as do atoms and molecules. Researchers might record the polarization of a photon, the spin of an electron or the energy level of an atom (whether the atom is in a ground state or an excited state).

Different particles can be ideal for different use cases. For example, photons can travel over long distances without being disturbed, making them a good fit for networking and communication applications. Meanwhile, trapped ions are highly stable and controllable, which make them better for long-term storage.

Quantum computers can use different types of particles to preserve the same quantum state across the collection, storage and retrieval stages. For example, researchers might use photons to send quantum information through a high-speed fiber optic cable before transforming this data and preserving it in a more stable, ion-based medium.

Researchers generally classify quantum memory devices based on their storage medium—regardless of which type of quantum phenomenon they initially record. Common quantum memory approaches include:

Solid-state quantum memories embed quantum information in solid materials, such as crystals, superconducting circuits or spin defects (intentionally placed irregularities in a substance’s atomic structure, which can be used to trap quantum information). One of the most promising approaches uses rare-earth ion-doped crystals, where crystals are infused with rare-earth elements, which are well suited to preserve quantum states by protecting them from environmental conditions. An alternative technique inserts a nitrogen-vacancy center (a nitrogen atom paired with an empty space) in the crystal lattice of a diamond to capture and preserve the spin of an electron.

Because solid-state memories use solid objects as their storage medium, they can be a good option for integrating quantum capabilities onto physical chips and hardware. They can also support scalable quantum computers, potentially operating on a smaller scale than gas-based quantum memory solutions, with infrastructure packed densely onto a single chip.

However, solid objects are often subject to a higher degree of noise because particles held within the solid can interfere with the stored quantum state. As a result, it can take more resources to protect the quantum information from decoherence and environmental disturbances. Rare-earth ion-doped crystals are a notable exception because their special shielding characteristics can help preserve quantum information for longer periods, improving coherence times. This capability makes them a promising candidate for practical, real-world quantum solutions outside of a lab setting.

Cold atom quantum memory is a gas-based solution that uses clouds of atoms as its storage mechanism. These atomic ensembles must be cooled to exceptionally low temperatures to limit thermal motion and noise. Researchers can use a combination of mechanisms, including laser-cooling systems, vacuum chambers and electromagnetic fields, to trap and preserve quantum information within the atom cloud and to reduce noise.

This approach delivers high reliability because scientists can maintain a greater degree of control over the environment compared to solid-state techniques. As a result, quantum information can remain coherent for longer periods of time. However, it can be more difficult to implement and scale cold atom solutions, as they require specialized equipment and infrastructure that’s more difficult to maintain in non-experimental environments.

One common cold atom approach called electromagnetically induced transparency (EIT) uses a laser field to temporarily transform an opaque cloud of atoms into a transparent one. When photons enter the atomic cloud, researchers turn off the laser to temporarily map the photons’ quantum state onto the atomic array. To retrieve the quantum information, researchers turn the laser on again, which rewrites the quantum data into photon form.

EIT can be a strong fit for optical quantum memory applications, which pair the efficiency and speed of photon-based transportation with the reliability and predictability of atom-based storage.

Instead of embedding ions or atoms within a solid or gas-based host material, trapped ion quantum computers (TIQCs) remove the host material entirely and instead isolate ions inside a vacuum chamber, typically by using an electromagnetic field. Quantum information is then encoded onto these trapped ions.

This approach provides a high degree of stability and a low risk of error because there are few potential sources of noise within the storage environment. As a result, trapped ion methods often have very long coherence times and the highest fidelity of any quantum memory approach. However, like cold atom approaches, they are difficult to maintain and scale, relying on resource-intensive, expensive components.

TIQCs can also be difficult to pair with photon-based delivery mechanisms. Photons must interact with a single ion or a small number of ions compared to the larger atomic ensembles present in cold atom and solid-state memories. As a result, efficiently transforming quantum states between ions and photons often requires additional configuration and specialized hardware. This limitation means that, although trapped ion approaches are highly reliable, they can be difficult to deploy inside larger quantum networks.

Some quantum memories might blend elements from different storage approaches to take advantage of each technology’s distinctive benefits. For example, a quantum network might send and receive quantum information by using photonic qubits, perform calculations with highly reliable trapped ion processors and store quantum states with relatively efficient and accessible solid-state mechanisms.

Quantum memory plays a key role in quantum communication by acting as an interface to safely store quantum states before and after they are shared across disparate quantum nodes. Quantum memories are often embedded in quantum repeaters, devices that send quantum information from one quantum computer to another while maintaining the integrity of the original quantum state.

Networks of quantum computers must be precisely synchronized to preserve entangled states and to enable entanglement swapping, where pairs of particles are brought into entanglement without needing to physically interact, expanding the quantum network’s potential reach. Quantum memories can serve as holding stations for quantum states so that they can remain stable even as multiple quantum computers are brought into alignment.

With quantum networks, researchers can coordinate quantum information processing across disparate nodes to improve performance or to enable more complex computations. Use cases include running physics simulations, encrypting data and collecting measurements with more precision and on a smaller scale than is possible with classical hardware. For example, quantum networks can improve the accuracy and synchronization of atomic clocks, which are used in geographic positioning systems, telecommunication infrastructure and the detection and measuring of quantum-scale phenomena.

Deploying quantum memory takes significant resources and development time because quantum states are delicate, difficult to observe and prone to interference. Major challenges include:

One complicating factor is the no-cloning theorem, which states that an arbitrary unknown quantum state cannot be perfectly copied—unlike a classical bit, which can be freely copied. This limitation is a consequence of the linearity of quantum mechanics, which prevents any single operation from accurately reproducing an arbitrary superposition. Therefore, researchers must preserve the integrity of the already-existing quantum information as it passes from one repeater to another.

Monitoring and measuring quantum activity can also be difficult because the measurement process itself can disturb the quantum state and collapse its superposition, rendering a classical outcome (either 0 or 1, but not both, as is possible with superposition).

To work around this measurement-disturbance problem, scientists use a technique called quantum state tomography to estimate the characteristics of a quantum state. With this approach, researchers can re-create the same quantum state thousands or millions of times and conduct experiments across each replica. By compiling their results, they can obtain a better understanding of the qubit’s attributes, including its probability amplitudes and phase relationships.

Quantum states are sensitive to environmental conditions, including heat, light, electric and magnetic fields, mechanical vibrations, sound waves and even cosmic rays. Quantum information can also leak outward and interact with nearby particles, disturbing phase relationships, a process known as decoherence.

These challenges are compounded when multiple qubits are in a state of entanglement because noise affecting one qubit can change the entire system’s phase relationships and thus destroy the quantum state. Phase relationships can also degrade or fall out of alignment over time, interrupting the precise interference patterns that enable quantum computers to perform calculations.

Quantum error correction (QEC) entails identifying errors and disturbances as they occur and correcting them, ideally in real time, so that the quantum system can remain functional and reliable. Fault tolerance systems can help address various errors, including phase-flip and bit-flip errors (when a phase or bit becomes inverted), unintended energy loss, gate errors (when control signals send too much or too little energy) and other problems.

According to IBM Quantum, fault tolerance systems can already be effective at smaller scales. However, scaling these systems remains a challenge. With larger systems, it can be more difficult to keep errors below a particular fault-tolerance threshold, where noise can be sufficiently controlled to preserve and maintain the quantum state. More qubits necessarily require more error correction resources and can add additional operational complexity. As a result, fault tolerance remains a barrier to deployable quantum systems that can complete real-world tasks accurately and consistently at scale.

However, it’s important to distinguish between physical qubits and logical qubits, which are abstracted qubit composites that researchers use to perform error-corrected calculations. If scientists can sufficiently suppress errors on a per-qubit basis (below the fault-tolerance threshold), introducing more physical qubits can help improve the reliability of the logical qubit. This is because additional physical qubits can act as redundancies, presenting more opportunities to find and eliminate errors before they interrupt the preserved quantum state.  

Scientists are developing more robust QEC and fault tolerance techniques, building from foundational approaches. These include syndrome extraction, where researchers spot clues about how an error might have occurred without disturbing the underlying quantum state, and decoding, where researchers apply corrections to the qubit based on the symptoms they uncovered.

Current quantum deployments are resource-intensive and largely confined to experimental environments, but researchers are developing ways to reduce the operational strain needed to run quantum memories at scale.

Today, many configurations rely on monolithic networks of cryostats, which are designed to support specific quantum computing deployments, limiting scalability. “If we decided that we want to make the system 50% larger, there’s a lot of inefficiency in then having to reengineer everything from square one in order to accommodate that,” said Hollister on the The Coherence Times.

For its part, IBM is aiming to improve scalability by enabling organizations to add more capacity as needed, instead of requiring an entirely new cooling system to work alongside larger-scale quantum networks. “At all levels, everything in our scaling plans points toward making the system modular,” Hollister added. “In the cryogenic space, at least from the point of view of systems operating at millikelvin (mK) temperatures, that’s not something that’s really been done before.”

Other researchers are experimenting with quantum memories that can maintain a high level of fidelity (meaning the retrieved quantum state closely matches the original) even at room temperature.

Quantum memories might also help pave the way for the quantum internet. While quantum networks connect quantum computers spread across a limited area or region, the quantum internet can theoretically enable quantum communication across distributed global systems. This capability supports advanced cryptography (such as quantum key distribution), and quantum cloud computing, where users can access offsite quantum resources through the cloud.

What’s the difference between quantum memory and qRAM?

Quantum memory’s primary goal is quantum information storage. qRAM, meanwhile, typically incorporates quantum memory, but adds more routing and cataloging components that enable quantum computers to query multiple data records simultaneously.

Specifically, qRAM can place multiple data addresses into a unified superposition, enabling multiple simultaneous data retrievals through a single request. However, qRAM largely remains a theoretical concept today because it requires complex error handling and longer coherence times that are difficult to achieve outside of highly controlled, experimental environments.

Do quantum memories store gigabytes of data like a hard disk?

No, quantum memories preserve fragile quantum states but do not provide sheer data storage in the same way that classical hard disk drives do. Researchers tend to compare different quantum memories based on the number of qubits they can store, how accurately they can retrieve quantum information and how long they can preserve qubits.

Can a quantum memory permanently store data?

No, quantum information will inevitably decohere over time due to interactions between the quantum state and other particles in the environment. Different quantum systems can preserve quantum states for different lengths of time, ranging from milliseconds to hours. Longer coherence times give researchers more time to perform calculations before the qubit itself becomes corrupted or destroyed.

Will quantum memory eventually replace classical memory?

It’s unlikely that quantum memory will completely replace classical memory, as they are useful for different tasks. Classical memory can securely store data volumes for long periods of time. Quantum memories, meanwhile, are designed to help solve a narrower set of use cases through quantum mechanics. For example, IBM and University of Chicago recently demonstrated a quantum system that took just 15 minutes to solve a problem that would otherwise take “infeasible amounts of time” for a classical computer to crack.