The Signal
Researchers Extend NV-Spin Memory to 565 Microseconds
A team led by researchers at Ulm University reports extending coherence memory in an ensemble of diamond NV centers from 49 to 565 microseconds. Its pulse sequence also improves detection of weak alternating magnetic fields.
Quantum.world Team · Written with AI · fact-checked ·
“We experimentally demonstrate an improved signal contrast and an order-of-magnitude increase in memory time with the optimal sequence compared to the widely used XY8 with standard rectangular pulses.”
Rajan Paul et al., “Optimized Dynamical Decoupling for Spin Ensembles with Large Inhomogeneities”
The authors’ result, posted to arXiv on October 29, comes from changing how microwave control pulses are shaped and sequenced—not from changing the diamond. In an ensemble of negatively charged nitrogen-vacancy (NV) centers, their best protocol preserves spin coherence for 565 microseconds, compared with 49 microseconds using the familiar XY8 sequence with standard rectangular pulses.
The paper is by Rajan Paul, Genko Genov, Oliver Schmid and Fedor Jelezko of Ulm University, alongside Shinobu Onoda, Yutaka Kobayashi and Junichi Isoya, whose affiliations include Japanese research institutions and Sumitomo Electric Industries. The work tests dynamical decoupling: timed spin flips designed to suppress disturbances from the surrounding environment. In an ensemble, differences in microwave-field strength and resonance frequency mean that nominally identical pulses can affect different spins unequally.
Why XY8 lost coherence in the ensemble
The experiment used a diamond ensemble under ambient conditions, with a bias magnetic field of 1.1 millitesla. The researchers measured a 27% standard deviation in microwave-drive amplitude across the ensemble, a source of pulse error that accumulates as a sequence applies more spin flips. With standard rectangular pulses in XY8, the measured memory time was about 49 microseconds.
They compared that baseline with phase-adjusted sequences, smoothly shaped pulses and rapid adiabatic passage (RAP) pulses. RAP changes both the pulse amplitude and frequency, sweeping the drive through resonance so spins can invert even when individual centers experience different drive strengths or frequency offsets. The team combined those pulses with KDD16, a sequence whose pulse phases are arranged to compensate for control errors.
KDD16-RAP reached 565 microseconds
KDD16 paired with RAP pulses produced the longest measured coherence time: 565 microseconds, more than 11 times the 49-microsecond XY8 result. Other steps improved the measurement too. KDD16 with sine-shaped pulses reached 262 microseconds, while a parabolic pulse shape reached 251 microseconds; the RAP combination more than doubled the sine-pulse result.
RAP also raised the contrast in a separate Rabi-oscillation measurement from 17.84% with standard driving to 24.5%. That comparison is distinct from the memory-time measurement, but it points to improved control across the ensemble. The researchers optimized the RAP duration to about one microsecond and selected a 7.5-megahertz frequency sweep for the subsequent experiments; a wider 20-megahertz sweep gave higher Rabi contrast but not the longest coherence.
Weak-field sensing improved, with a trade-off
The group tested the sequence on an applied alternating magnetic field. With KDD16-RAP, it could distinguish a signal set to −35 dBm from the no-signal decay curve; standard XY8 required about −15 dBm for a comparable distinction in their setup. Those values describe the test signal’s power setting, not a universal magnetic-field detection threshold.
The experiment also exposes a limit to what “memory time” means here. The 565-microsecond figure is a coherence-decay measurement under a particular pulse sequence, not a demonstration of a device storing and retrieving arbitrary quantum states. The paper notes that a CPMG measurement reaches a 902-microsecond relaxation limit, but CPMG loses fidelity for arbitrary initial states and is not suited to the team’s amplitude-sensing task.
For this diamond ensemble, the result identifies control-field inhomogeneity and pulse errors—not spin-spin interactions—as the main obstacles tested: sequences designed to cancel spin-spin interactions performed worse than KDD16 and related alternatives. The team’s finding is therefore specific and practical: under its 1.1-millitesla operating conditions, pairing rapid adiabatic passage with KDD16 raised measured coherence time from 49 to 565 microseconds while improving sensitivity to a weak test field.