An experiment has reported direct images of spatial fluctuations in a quantum field whose strength changed with scale in the way expected for a vacuum state. The field was the spin degree of freedom in a two-component atomic Bose–Einstein condensate, treated in the experiment as a massive relativistic sine-Gordon analog in the interaction-dominated regime. The authors interpret the images as a visual demonstration of vacuum fluctuations associated with Heisenberg uncertainty.
The result is presented in arXiv preprint version 1 dated 20 Aug 2026. It is a laboratory observation in an atomic condensate analog, so its evidence concerns that engineered quantum field rather than a direct measurement of a naturally occurring field.
The question was whether fluctuations in the field’s ground state—the lowest-energy state—would have a scale-dependent amplitude and an inverse-frequency spectrum, instead of the pattern expected from thermal occupation. That distinction matters because the experiment is not simply asking whether the images contain noise; it is asking whether the noise follows a particular quantum prediction.
The field inside the box
The laboratory sample was a two-component condensate made from 39 K atoms. It was confined to a two-dimensional plane inside a square box trap. The measured quantum field was the condensate’s spin degree of freedom, which gave the experiment a way to study field fluctuations through atomic images.
State-selective density imaging measured population imbalance—the difference in the populations of the two components. To access relative phase, the experiment used radio-frequency pulses to rotate the atoms’ Bloch vectors before taking the state-selective image. In this way, the same imaging approach could probe two complementary parts of the spin field.
Data were collected as stacks of images from independent repetitions. The images were mean-subtracted and Fourier-transformed before the noise spectra were analyzed. Fourier analysis let the researchers sort fluctuations by wave vector, a measure related to spatial scale, rather than treating the entire image as one undifferentiated signal.
To set the scale of the measured noise, the spectrum was normalized against reference images of weakly interacting, single-component gases. Those reference samples had the same mean optical density and uncorrelated, shot-noise-dominated fluctuations. The comparison provided a common reference for judging how much fluctuation appeared at each mode.
Two ways to test the spectrum
In one protocol, the field began in the Rabi regime. The coupling between the components, written Ω′, was then quenched to different final values. According to the reported behavior, the quench amplified phase fluctuations while suppressing population-imbalance fluctuations.
For the Rabi-quench example, the time-dependent structure factor S_Y(k,t) was extracted from approximately 70 images for each time point. A structure factor is a compact measure of fluctuation strength at a chosen wave vector, so its time dependence showed how different spatial modes evolved after the quench.
Those measurements followed the comparison trend, but their amplitudes were systematically lower by a factor of approximately 0.7. The authors suggest that imperfect atom detection is the likely reason for the attenuation. The result therefore supplied a calibration issue as well as a test of the predicted pattern: the trend agreed with the comparison, while the measured size was reduced.
A second protocol was designed to read out the ground state more directly. The field was initialized in the Rabi regime, then Ω′ was ramped down linearly over 9 ms to different final couplings. Approximately 120 images were used for each time point in the adiabatic-ramp example.
Across a wide range of wave vectors and final couplings, the extracted amplitudes showed inverse-frequency scaling. Their magnitudes were consistent with the zero-temperature theoretical prediction after the same detection factor was applied. In plain terms, the experiment found both the expected change with frequency and a size that could be reconciled with the vacuum calculation once the reduced signal was included.
The researchers also fitted the data while allowing for a nonzero temperature. Outside the region where the adiabaticity condition failed, the fit gave 6 ± 6 nK, or 0.2 ± 0.2 µ′s in the paper’s spin-energy units, with mode occupation Nₖ ≲ 0.1 for the included points. Because the uncertainty is as large as the central temperature estimate, the fit overlaps the zero-temperature prediction; it does not establish that the thermal contribution is exactly zero.
What the measurement can and cannot show
The correction is central to how the result should be read. The Rabi measurements were lower by approximately 0.7, and the adiabatic comparison used that same attenuation factor. The paper identifies imperfect atom detection as the likely explanation for the shortfall, but the supplied analysis does not describe an independent randomized control that would establish that explanation.
The shaded adiabaticity-failure region is another boundary on the interpretation. Data in that region cannot be treated in the same way as points satisfying the ramp condition, and the temperature fit excluded them. That makes the inverse-frequency comparison strongest over the range where the protocol remained adiabatic, rather than across every measured point.
This was not a comparison between two randomized groups of atoms. The experiment compared the measured spectra with vacuum-state and thermal theoretical predictions, while the single-component reference images were used to normalize the noise. That design can test whether the observed scale dependence matches a prediction, but it does not turn the result into evidence that one experimental treatment caused a change in another.
The broader claim is therefore narrower than the phrase “imaging the vacuum” might suggest. The measured object was a two-dimensional atomic analog: a condensate spin field selected because its fluctuations could be imaged and compared with a field-theory calculation. The observation does not demonstrate vacuum fluctuations in a natural electromagnetic, gravitational or cosmological field.
Nor does it show that any proposed future analog phenomenon has occurred. The authors describe the observation as a step toward using the platform to simulate outstanding quantum-field problems, including nonperturbative and nonequilibrium phenomena. That is a direction for follow-up work, not a result reported by this experiment.
A result that still needs testing
The immediate technical test is whether the amplitude can be reproduced with detection efficiency fully calibrated. Independent replication would also help determine whether the approximately 0.7 attenuation is a property of the apparatus rather than a feature of the underlying fluctuation signal.
The preprint’s central contribution is a measurement strategy: image the condensate’s spin field, separate its modes by spatial scale, and compare the resulting amplitude with vacuum and thermal expectations. Within the stated limits, the reported spectrum is consistent with vacuum-state theory; the remaining uncertainty lies in the detector correction, the adiabaticity boundary and the fact that the system is an atomic analog.
Paper data and sources
Original title: Imaging the vacuum fluctuations of a quantum field
Authors: Yansheng Zhang, Feiyang Wang, Yi Jiang et al.
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text