The Signal
Nanjing Researchers Design a Bosonic Memory That Can Sense Force
A theoretical proposal uses one bosonic oscillator to sense a force while protecting a stored logical qubit from photon loss. Its cavity-noise simulations favor different code sizes depending on whether Kerr evolution is reversed, while end
Quantum.world Team · Written with AI · fact-checked ·
A sensor must let a weak force leave a measurable trace; a quantum memory must correct unwanted changes before they disturb its stored state. A paper posted October 29 by researchers including Tairong Bai and Wen Zheng proposes a way for one bosonic oscillator to do both, storing a logical qubit in separated wave-packet sites while registering the force in a shared internal coordinate. The authors analyze the scheme under photon loss and model its performance with superconducting-cavity noise; they do not report an experimental demonstration.
One oscillator, two jobs
Bosonic codes store quantum information in the states of an oscillator, such as a microwave mode in a superconducting cavity. A force can shift that mode’s position, but codes designed to correct displacement-like errors can also suppress the very response a sensor needs to measure. The paper’s proposal separates those functions within the oscillator: discrete site amplitudes carry the logical information, while a shared coordinate inside each wave packet responds to the force.
For the four-packet example, the logical basis states distribute their amplitudes across four separated sites with binomial weights. The force shifts the shared internal wave-packet profile, and an endpoint interaction transfers information about that shift to a probe qubit. The probe’s readout is designed to be independent of the stored logical state, including when the logical qubit is entangled with an external reference.
“By storing logical information in binomial site amplitudes while accumulating the force signal in a shared internal coordinate, our protocol achieves exact state-independent readout under photon loss and suppresses memory-probe correlations one order beyond the leading logical dephasing at every correction order.”
The authors write in “Force Sensing in a Recoverable Bosonic Quantum Memory.”
Recovery keeps the readout separate
Photon loss complicates the separation because it can distort a wave packet and couple the site index to the force-sensitive coordinate. The researchers pair the binomial encoding with a recovery procedure that removes packet-dependent distortions from the probe signal while correcting logical errors. For a code with 2r+2 packets, their analysis predicts memory infidelity scaling as O(γr+1) and residual memory–probe correlations scaling as O(γr+2), where γ represents accumulated photon loss.
Numerical simulations illustrate the scaling from four through ten packets. In the four-packet case, at γ = 10−3 and a dimensionless force impulse λT = 0.2, the simulated memory infidelity is 1.16 × 10−5. Adding packets improves the predicted loss protection, but it also raises photon number and number variance—properties that make the encoding more sensitive to other cavity imperfections.
Higher sensitivity brings a narrower range
The authors also study reading out the force at higher integer harmonics. For the four-packet code at γ = 10−3, their calculation raises normalized quantum Fisher information from 3.73 at the first harmonic to 9.13 at the second. The paper compares those values with a coherent-state displacement reference of about 2 in the same normalization.
That gain comes with a constraint: the probe’s phase repeats as the force changes, so the measurement can become ambiguous outside a defined range. For the parameters used in the example—site spacing L = 3 and packet width w = 1.8—the second harmonic still gives an unambiguous mapping across the protected interval. Higher harmonics would require a narrower prior range for the force and stronger endpoint kicks.
Cavity noise changes which code wins
To compare encodings under hardware-like conditions, the team simulates a superconducting cavity using a reported cavity lifetime of 25.6 milliseconds, a residual Kerr coefficient of K/(2π) = 3.6 hertz, and a conservative photon-number dephasing time of 0.5 seconds. Under that modeled combination, the six-packet code has the lowest memory infidelity at a 20-microsecond sensing time without Kerr compensation. With an ideal endpoint operation that reverses Kerr evolution, the ten-packet code performs best at that duration, with simulated infidelity of 2.86 × 10−5.
The ranking is not simply “more packets is better.” Larger codes reduce some loss and dephasing errors, but their higher photon numbers increase Kerr sensitivity; even with Kerr reversal, the six-packet code overtakes the ten-packet code at longer simulated durations. The calculations also find that Kerr and number dephasing can make the probe readout weakly dependent on the logical input, complicating force estimation when that input is unknown.
Control overhead remains outside the ideal code
The paper’s practical caveat is that its main performance simulations assume ideal endpoint controls, while estimating some control-related risks separately. Its estimate for an echoed conditional-displacement readout kick is 1.1–2.2 microseconds, excluding pulse-shaping overhead and transient photon excursions. For a 20-microsecond sensing interval, the authors estimate a probability of 2.17 × 10−4 for a thermal excitation of the parked ancillary transmon—larger than the ideal simulated error of the ten-packet code after Kerr reversal.
Compiling the spatial decoder into finite-bandwidth control pulses, accounting for preparation and recovery faults, and testing the protocol in hardware remain next steps. The paper’s results establish a theoretical route to sensing while preserving a bosonic logical state; they do not yet show that a device can perform the full cycle. Its numerical comparisons make the engineering trade-off explicit: at 20 microseconds, six packets lead without Kerr compensation, while ten lead only when the model grants ideal Kerr reversal.