quantum computing

Nuclear spins entangled via electron exchange in silicon

If you've followed the race to build practical quantum computers, you’ve probably heard a common refrain: nuclear spins in silicon make phenomenal quantum memories but are hard to wire up at scale. A new Science paper flips that script by entangling two donor nuclear spins not by squeezing them around a single shared electron, but by letting two separate electrons talk to each other and quietly pass the message along. It’s an elegant, semiconductor‑friendly way to make nuclear spins do real computational work.

What just happened

A team working with phosphorus donors in silicon demonstrated genuine, scalable entanglement between two 31P nuclear spins separated by about 20 nm. Each donor has its own bound electron. Those two electrons are weakly exchange‑coupled—think of it as a tunable magnetic handshake around 10–12 MHz. Because each electron also feels its local nucleus via the hyperfine interaction, the combined electron resonance becomes conditional on the joint nuclear state. By driving the electrons through a carefully chosen loop in parameter space, the device accumulates a controlled phase only when both nuclei are in a particular configuration—a geometric controlled‑Z (CZ) gate. Result: nuclear Bell states with a reported fidelity of roughly 76 ± 5% and concurrence of 0.67 ± 0.05, achieved on microsecond timescales. Source: Science (2025), doi: 10.1126/science.ady3799; arXiv: 2503.06872.

Why this is different

Earlier demonstrations of nuclear–nuclear entanglement in silicon typically required both nuclei to share a single electron. That works for one pair but quickly runs into a wiring nightmare as you try to scale beyond two spins. Here, each nucleus keeps its own electron, and the electrons exchange information. That opens a path to larger arrays: as long as you can resolve the electron spin resonance conditioned on both nuclei, you can mediate a nuclear CZ without demanding nanometer‑perfect donor placement or a single shared electron. In other words, the architecture tolerates realistic variability and remains compatible with silicon foundry processes.

The simple idea behind the trick

  • Hyperfine interaction: the local electron’s energy depends on whether its nucleus points up or down, like two gears with teeth that must mesh.
  • Exchange coupling: two nearby electrons weakly interact; this lets their energies—and thus their resonant frequencies—shift when you flip one or the other.
  • Conditional phase: by driving the electrons along a specific loop (a geometric operation), you accumulate a phase only when the nuclei are in a particular joint state. That phase is the heart of a controlled‑Z gate, now acting on the nuclei.

Gate times are on the order of microseconds. That’s leisurely compared to electron gates, but perfectly fine for nuclear spins whose coherence can stretch from seconds to minutes in isotopically purified silicon—long enough to be world‑class quantum memories and now, increasingly, active qubits. See, for example, long‑coherence measurements by Muhonen et al. (Nat. Nanotechnol. 2014) and single‑nuclear‑spin control by Pla et al. (Nature 2013).

How we got here: a short lineage

  • The blueprint: Bruce E. Kane proposed a silicon nuclear‑spin quantum computer using phosphorus donors in 1998—a landmark that put silicon on the quantum map (Nature 393, 133; doi: 10.1038/30156).
  • Single‑atom control: Andrea Morello’s team demonstrated single‑shot readout of an electron spin bound to a phosphorus donor (Nature 2010; doi: 10.1038/nature09392) and later high‑fidelity control and readout of the accompanying 31P nuclear spin (Pla et al., Nature 2013; doi: 10.1038/nature12011).
  • Record memory: Nuclear spins in silicon devices stored quantum states for tens of seconds (Muhonen et al., Nat. Nanotechnol. 2014; doi: 10.1038/nnano.2014.211), cementing their role as robust memories.

This new Science work closes a persistent gap: it shows a concrete, scalable route to entangle nuclear spins using exchange‑coupled electrons as mediators, squarely within a CMOS‑friendly platform.

Why this matters

  • Scalable building blocks: Arrays of donor nuclei can be linked by engineered chains of exchange‑coupled electrons—potentially extended via quantum dots—without resorting to fragile shared‑electron tricks.
  • Memory–compute synergy: Use electrons for fast logic and nuclei for long‑lived storage, shuttling quantum information back and forth when needed. That hybrid mode is attractive for error‑corrected architectures.
  • Manufacturing advantage: Donor qubits live in the world’s most mature material system—silicon. Even modest tolerance to donor placement and exchange variability is a big deal for practical fabrication.

If silicon quantum chips can combine long‑coherence nuclear memories with exchange‑mediated entangling gates, the platform becomes a realistic contender for error‑corrected machines that could accelerate materials discovery, optimize energy grids, and model catalysts for greener chemistry.

What to watch next

  • Calibration at scale: How robust is the geometric CZ as devices age, temperatures shift, or exchange drifts?
  • Crosstalk and crowding: Can we keep electron resonances clean when many donor pairs live nearby?
  • Modular links: Can exchange buses be extended through quantum dots to stitch together larger nuclear‑spin registers?
  • Error correction: How do these nuclear CZ gates integrate with surface codes or LDPC codes, and what’s the overhead?

A closing thought

Silicon has always been good at scaling. By teaching two quiet nuclear spins to entangle through a whispering pair of electrons, this result suggests a future where the most reliable qubits we know—nuclear spins—can also compute at scale. What new algorithms would you design if fast electron logic and ultra‑stable nuclear memory lived side by side on the same chip?

Sources

  • Scalable entanglement of nuclear spins mediated by electron exchange. Science (2025). doi: 10.1126/science.ady3799; arXiv: https://arxiv.org/abs/2503.06872
  • Kane, B. E. A silicon‑based nuclear spin quantum computer. Nature (1998). doi: 10.1038/30156
  • Morello, A. et al. Single‑shot readout of an electron spin in silicon. Nature (2010). doi: 10.1038/nature09392
  • Pla, J. J. et al. High‑fidelity readout and control of a nuclear spin qubit in silicon. Nature (2013). doi: 10.1038/nature12011
  • Muhonen, J. T. et al. Storing quantum information for 30 seconds in a nanoelectronic device. Nat. Nanotechnol. (2014). doi: 10.1038/nnano.2014.211

— Source: https://doi.org/10.1126/science.ady3799

Read more