Harvard Extends Diamond Qubit Coherence Threefold With Sound Instead of Microwaves

August 25, 2026 – Researchers at the Harvard John A. Paulson School of Engineering and Applied Sciences demonstrated a method for extending the coherence time of a diamond-based qubit using mechanical vibrations rather than conventional microwave pulses, according to a paper published online July 15 in Nature Physics.

The team, directed by Marko Lončar, the Tiantsai Lin Professor of Electrical Engineering, applied a continuous acoustic field to a silicon-vacancy (SiV) center in diamond, creating what the researchers call a “dressed” qubit – a state that is less sensitive to the low-frequency environmental noise that degrades stored quantum information. The approach extended the SiV spin dephasing time (T₂*) from 680 ± 30 nanoseconds to 2.2 ± 0.1 microseconds – a 3.2× improvement.

Eliza Cornell and Zhujing Xu contributed equally to the experimental work. Cornell, a recent Ph.D. graduate from the Lončar lab, is now a postdoctoral researcher at Boston University. Xu is a former postdoctoral scholar in the group.

The result addresses an engineering conflict specific to phononic quantum networks: conventional coherence-protection methods rely on microwave pulses that do not work well on qubits housed in the phononic cavities needed to trap and route acoustic signals between qubit nodes. The Harvard team demonstrated its all-mechanical coherence protection on a surface-acoustic-wave (SAW) device – an AlN-on-diamond platform with interdigital transducers that focus acoustic energy onto a single SiV spin – in a protocol designed to be compatible with the phononic cavities that would connect qubit nodes in a chip-scale network. That cavity integration has not yet been shown.

“We are solving two problems,” Cornell said. “We want the spin to have strong interaction with phonons, and we want the spin to have a long coherence time. Our paper demonstrates a method of extending the coherence time that is compatible with the silicon-vacancy center being in a cavity.”

The team also achieved a mechanically driven Rabi frequency reaching 800 MHz at cryogenic temperatures – reported as the fastest ever for direct resonant driving of an SiV spin – demonstrating that ultrafast acoustic control is possible on this platform, though at a different operating point from the dressed-state coherence protection.

The research was a collaboration across Harvard, the University of Chicago, the Walther-Meißner-Institut at the Bavarian Academy of Sciences, the Technical University of Munich, and Argonne National Laboratory. Co-authors include Zhaoyou Wang, Hana K. Warner, Eliana Mann, Michael Haas, Smarak Maity, Graham Joe, Liang Jiang, Peter Rabl, and Benjamin Pingault.

The Harvard Office of Technology Development is pursuing patent protection and commercialization related to the work.

My Analysis

A Modest Gain on a Hard Engineering Problem

A 3.2× extension of dephasing time – from 680 nanoseconds to 2.2 microseconds – is incremental by any absolute measure. Resource estimates for a cryptographically relevant quantum computer are unchanged by it, and diamond color centers remain far below the qubit counts of superconducting, trapped-ion, and neutral-atom platforms. Anyone writing “breakthrough” in their headline is reaching.

The value of the result comes from the engineering problem it addresses.

The Lončar lab has spent years building the infrastructure for phonon-based quantum networking in diamond. Their previous work demonstrated phononic cavities – structures that trap mechanical vibrations and force them to interact with the electron spin of an SiV center. Phonons, the quantum particles of sound, have real advantages over photons at the chip scale: at a given frequency, phonon wavelengths in diamond are roughly four orders of magnitude shorter than light wavelengths, enabling far smaller devices and tighter integration. They couple readily to both solid-state spins and electromagnetic fields, which makes them useful connectors in hybrid quantum systems that combine different qubit types.

The problem was that the very coupling that makes SiV centers useful for phononic networks also makes them vulnerable. A qubit coupled strongly to engineered phonons also interacts with the uncontrolled thermal vibrations in its environment, which cause decoherence. The standard fix in quantum information science is dynamical decoupling: applying rapid sequences of microwave pulses that average out the environmental noise. But microwave pulses conflict with phononic cavities. You cannot protect the qubit and network it at the same time using the conventional toolkit.

Harvard’s answer is to replace the microwave pulses with a continuous mechanical driving field, demonstrated on a surface-acoustic-wave device. The acoustic field shifts the qubit into a dressed basis – a set of states defined by the qubit-plus-field system rather than by the bare qubit alone. These dressed states are inherently less sensitive to the low-frequency noise that dominates decoherence in solid-state spin systems. And because the protective field is itself mechanical, the method is designed to operate inside the same phononic cavity structures that would connect the qubit to its neighbors in a network – though that cavity integration has not yet been demonstrated.

In this architecture, phonons would serve a dual function: carrying quantum information between nodes and protecting that information while it is stored. That dual-use capability is what a practical on-chip quantum phononic network requires, and this paper demonstrates that the coherence-protection half works on a real SiV spin – a necessary precursor to dual-use operation inside a phononic cavity.

Three Diamond Results From 2026

Two other groups reported results on diamond qubits in 2026, and each addressed a different engineering problem for quantum networking.

At the University of Pennsylvania, Lee Bassett’s group published in Nature Nanotechnology a single-gate method for generating four-qubit entangled states in a nitrogen-vacancy (NV) center register at room temperature. Their parallel gate completed the operation in 14.8 microseconds – roughly ten times faster than a conventional sequence of two-qubit gates – with a fidelity of 0.92(4). The technique exploits crosstalk that conventional gate sequences treat as a source of error, turning it into a resource for simultaneous multi-qubit control.

The UPenn work addresses a different piece of the diamond-networking puzzle. The quantum-sensing and quantum-memory nodes at the endpoints of a distributed quantum network would need to entangle qubits quickly and with high fidelity, and UPenn’s gate does this at room temperature. The 0.92 fidelity is strong for a room-temperature solid-state register, and the parallel gate approaches the fundamental speed limit set by the hyperfine coupling strength. The room-temperature operation eliminates the cryogenic overhead that currently limits diamond SiV systems to laboratory settings, though the NV centers used in the UPenn work have their own trade-offs in optical properties.

Meanwhile, QuTech at TU Delft demonstrated earlier this year a cryo-CMOS system-on-chip that controls both the electron and nuclear spins of diamond NV centers with gate fidelities of 99.3% (electron) and 99.8% (nuclear) and dynamical-decoupling-extended coherence times exceeding 50 ms. That result was presented at ISSCC in February 2026 as part of a Fujitsu-funded project. It attacks the classical-control bottleneck: the electronics that drive the qubits need to work at cryogenic temperatures without degrading qubit performance, and QuTech showed that a single CMOS chip can do it.

All three groups work with color centers in diamond: an SiV center at Harvard, NV centers at UPenn and QuTech. Harvard demonstrates coherence protection compatible with phononic cavities. UPenn demonstrates fast entanglement at room temperature. QuTech demonstrates classical control at cryogenic temperatures. A quantum computer capable of running Shor’s algorithm is no closer after any one of these results. Collectively, they are assembling the engineering stack for diamond-based quantum networking.

Where Diamond Fits

NV and SiV color centers in diamond trail the leading platforms in qubit count and are not the likeliest path to a standalone CRQC. The qubit counts are small (Harvard’s result involves a single SiV center; UPenn’s involves four physical qubits), and the scaling challenges are fundamentally different from those facing superconducting or neutral-atom systems. QuTech and Fujitsu are pursuing error-corrected diamond-spin machines. Fault-tolerant operation of one logical qubit in diamond has been demonstrated. Cryptanalysis requires thousands of logical qubits.

What diamond offers is something those larger-qubit-count platforms need but do not yet have: a credible interface between stationary qubits and traveling information carriers. Every architecture for a quantum internet – and every architecture for distributed quantum computing that links multiple processors – requires nodes that can store quantum information and entangle it with a traveling particle. Each node then sends that particle to another node. Diamond color centers, which have optically addressable spins and nuclear-spin memories with long coherence times, are among the strongest candidates for those nodes.

The Harvard result fits precisely into this picture. A phononic quantum network needs nodes that maintain coherence while coupled to the network’s own communication channel. Demonstrating all-mechanical coherence protection that is compatible with the structures that would transmit quantum information is a necessary engineering milestone, regardless of the modest absolute coherence improvement.

Limits of the Result

The 3.2× dephasing-time improvement, while a proof of principle, needs to grow substantially before it enables high-fidelity multi-qubit operations within phononic networks. The absolute dressed-state T₂* of 2.2 microseconds is useful but far from the millisecond-scale coherence achieved in NV centers with conventional decoupling. The authors describe the result as a first step toward phonon-mediated quantum gates, which the paper does not demonstrate.

The experiment operates at millikelvin temperatures on a SAW device, not inside a phononic cavity. Moving from a propagating acoustic wave on a bulk diamond surface to a confined mode inside a high-Q phononic resonator is a separate engineering challenge. The UPenn room-temperature result uses NV centers, a different defect type from the SiV centers in the Harvard work. The two defect types have complementary strengths: NV for room-temperature operation and long nuclear-spin memory, SiV for superior optical properties and stronger strain coupling. The diamond-networking community has not converged on which color center, or which combination, a practical network will use.

The 800 MHz Rabi frequency and the dressed-state coherence protection, while demonstrated on the same device, operate at different drive conditions – the Rabi record uses strong bare-spin driving, while the coherence protection operates in the dressed basis at lower Rabi rates (~9 MHz with a ~76 MHz dressing drive). Integrating ultrafast control and coherence protection into a single operating regime remains an open problem.

Harvard’s OTD pursuing patent protection and commercialization is standard for a university result and signals nothing about commercial timelines. The gap between a laboratory demonstration on a single qubit and a deployed quantum networking product is measured in years and in engineering problems that this paper does not address: fabrication yield, inter-node entanglement rates, multiplexing, and classical-control integration at scale.

Diamond as a Network Node

The CRQC that eventually threatens RSA and ECC will most likely run on a superconducting, trapped-ion, or neutral-atom processor – platforms where fault-tolerant architectures are under active development and qubit counts are growing toward the thousands required for cryptanalysis.

Diamond color centers are building toward a different role: the network nodes that connect quantum processors over fiber and distribute entangled resources between them. Trapped-ion and neutral-atom groups have demonstrated remote entanglement through their own spin-photon interfaces, so diamond is one candidate among several. It combines an optically addressable spin, nuclear-spin memories with minute-scale coherence, and – after this Harvard result – phonon-compatible coherence protection in a single node. As far as I know, it is the only platform that offers all three in one node.

Today’s metropolitan-fiber QKD networks rely on trusted nodes and use no quantum memory. The Harvard, UPenn, and QuTech results in 2026 are steps toward the next generation: quantum repeater nodes that extend range and enable distributed computing through entanglement. That generation is decades away.