Nuclear-Spin Quantum Memory Poised to Take the Lead

See allHide authors and affiliations

Science  08 Jun 2012:
Vol. 336, Issue 6086, pp. 1239-1240
DOI: 10.1126/science.1223439

Exploiting quantum mechanics for processing information requires balancing two opposing criteria. Quantum systems must be isolated to prevent decoherence—destruction of the quantum state—but must interact with other quantum systems if the stored information is to be accessed and processed. One way to overcome this challenge is to transfer quantum states between two different systems—one for efficient processing and readout, and the other for long-term storage. Two papers in this issue—by Steger et al. on page 1280 (1) and Maurer et al. on page 1283 (2)—show that solid-state quantum memories based on nuclear spin states can have extremely long storage times that approach those of ion traps in a vacuum. This capability is enabled in part by using highly isotopically pure semiconductor materials to make “semiconductor vacuums” that isolate nuclear spins in near solitude.

There are many approaches for quantum information processing (3). Those based on isolated spins in semiconductors exploit the wealth of expertise from more than half a century of materials development and processing. They generally follow a similar path. An electron or nuclear spin is localized in a semiconductor, either by incorporating electrostatic gates (4) or by using naturally occurring “defects,” such as phosphorus donors in silicon (Si:P) (5) or nitrogen-vacancy (NV) centers in diamond (6). Applying a magnetic field then allows the eigenstates of the spin to be used as a logical basis for storing quantum information—a quantum bit (qubit). Serendipitously, during the early development of spin-resonance techniques, ensembles of such isolated spins in semiconductors were studied, revealing extremely long phase-coherence times (orders of milliseconds were already observed half a century ago), which now motivate the use of these systems for quantum technologies.

Quantum information processing requires qubit initialization, control, and readout. For NV centers, luminescence provides for optical single-spin readout (6). In silicon, optical spectroscopy and electric currents have been used to read ensembles of isolated spins (1, 7), and recently such approaches have been scaled to the single-spin limit (8). Spin resonance techniques are generally used for control.

Nuclear-spin-based quantum memory.

The quantum bits created in diamond by Maurer et al. and in silicon by Steger et al. can both be read out optically, and both couple to nearby nuclear spins, which can be used as a long-lived quantum memory. Both types of qubits are contained in crystals that naturally include isotopes of silicon (29Si) and carbon (13C), which contribute unwanted spins, leading to decoherence of the quantum memories. Isotopic engineering can substantially reduce this background, which, when combined with dynamic decoupling sequences and control of the hyperfine coupling between qubit and quantum memory, can result in quantum storage for more than 3 min for 29Si.


Stored quantum information must also retain its “quantumness” on time scales that allow information processing. Spins couple easily to other spins in their local environment, which can extend for hundreds of nanometers. Both carbon and silicon have isotopes with nuclear spin, and interactions between spin qubits and these unwanted spins limit coherence. To overcome these environmental perturbations, both studies reported here use isotopically engineered materials—the pure silicon 28Si comes from the Avogadro project (9). The aim is to remove all nuclear spins that can cause decoherence—except one.

It is this sole remaining nuclear spin that lies at the heart of these approaches. Because of their small size (femtometer scale) and their small magnetic moments, nuclear spins are much more isolated than electron spins and have correspondingly longer coherence times, which has led to them being recognized as ideal systems in which to store quantum information (7, 1012). In the two experiments described in this issue, a nuclear spin is coupled to an electron qubit via the hyperfine interaction (see the figure). Using careful control sequences, the quantum state of the electron qubit can be transferred to the nuclear spin, stored there for some time, and then transferred back again (12) or read out directly. Maurer et al. used a 13C nucleus ∼1.7 nm from the NV center, whereas Steger et al. used a 31P nucleus and its very own donor electron spin. In both cases, dynamic decoupling sequences, which manipulate the nuclear spins in an acrobatic fashion to compensate environmental fluctuations, were used to remove the impact of the trace amounts of remaining impurities.

Quantum memory must be isolated from the outside world as much as possible. Thus, the hyperfine interaction used for the quantum states transfer seems a glaring problem. Indeed, both the quantum memories reported here are limited by spin-flip relaxation time (T1) of the electron qubit (which is much longer in silicon than in diamond), mediated by the hyperfine interaction. However, this interaction is needed only for the transfer of quantum information in and out of the memory, and Maurer et al. present an elegant method for controlling this interaction by optically exciting the electron into a state that has no spin, and no hyperfine coupling, to the nuclear memory. This method allowed them to extend the lifetime of their quantum memory by nearly two orders of magnitude, to ∼1 s. Recently, Dreher et al. (13) demonstrated that this approach can also be effective in extending the coherence of Si:P-based qubits by optically ionizing the 31P donor.

So, in the end, how good are these memories? With isotopic purification and dynamic decoupling, quantum states were stored in a 31P ensemble for more than 3 min [compared with ion trap lifetimes in the second-to-minute range (3)]. This remarkable achievement advances nuclear spins in silicon as one of the most coherent systems in nature. It will be interesting to see if this time scale can be maintained when the infrastructure necessary for single-donor readout is incorporated near a single nuclear spin.

The 13C-based nuclear memory, with a storage time of 1.4 s, might seem less impressive, but such an opinion would vastly undervalue this result, which is observed at room temperature and on a single quantum system. Indeed, with the ability to entangle NV centers and photons (14), and progress toward entangling distant NV centers (15), the availability of an isolated, long-lived nuclear-spin quantum memory represents an important advance for this technology. Maurer et al. conclude on an extremely optimistic note, estimating (quite rigorously) that improvements to the isotopic purity of diamond, the hyperfine decoupling, and the dynamic decoupling may result in quantum storage times exceeding 1 day. Such long-lived quantum systems could disruptively change the way we think of information security. Quantum-secured tokens could prevent credit-card skimming or be incorporated into provably unforgeable identity cards.

Solid-state nuclear-spin quantum memory is progressing rapidly, and the two studies described here show that it may well be the most effective way to store quantum information. The demonstration of quantum coherence lasting for time scales relevant to real-world conditions holds promise for as yet unimagined quantum-enabled technology becoming as ubiquitous as classical electronics are today.


View Abstract

Stay Connected to Science

Navigate This Article