The Signal
Researchers Use Compression to Map Quantum Matter
A team led by researchers at the University of California, San Diego, used lossless image compression to measure spatial complexity in quantum materials. The method detected peaks near integer and fractional quantum Hall transitions in grap
Quantum.world Team · Written with AI · fact-checked ·
Spatial averages can reveal when graphene crosses a quantum Hall transition, but they do not show how conductive regions connect and reorganize along the way. In a paper submitted to arXiv on October 26, researchers led by the University of California, San Diego, report a way to quantify that missing geometry: compress microscopy images without losing data, then track how much of each image the algorithm can remove.
The resulting computable information density, or CID, rises to a pronounced maximum near quantum Hall transitions in graphene. The team says the measure also distinguishes those transitions from an ordinary conducting-to-insulating crossover, although it does not identify a phase or its mechanism on its own.
Turning microscopy frames into a complexity measure
The method starts with microwave impedance microscopy, which probes local electronic response. The researchers convert each two-dimensional image into a one-dimensional sequence using a Hilbert space-filling curve, designed to keep nearby pixels near each other in the sequence, then apply the lossless LZ77 compression algorithm. CID is the length of the compressed sequence divided by the original length: repeated or orderly patterns compress more, while intricate patterns retain more information.
The team applied the analysis to image series of gated graphene taken at millikelvin temperatures. For the integer quantum Hall transition from filling factor 1 to 2, each image covered a 3-by-3-micrometer area with 75-nanometer pixels; measurements at 5 tesla captured the changing electronic pattern. As conductive filaments emerged near the transition, CID developed a peak.
“The same framework tracks fractional quantum Hall transitions and resolves isolated fractional puddles that coexist with a disordered background.”
Chen Wu et al., “Information-theoretic entropy imaging of quantum matter”
The peak marks a particular kind of reorganization
The peak does not simply restate what a spatial average or correlation length reports. The paper compares the CID signal with a version calculated after randomly shuffling the image pixels, which preserves the distribution of pixel values but destroys their spatial arrangement. The contrast remains pronounced near the quantum Hall transition, supporting the authors’ interpretation that the signal responds to geometry, not just to the mix of bright and dim pixels.
A disordered Hofstadter model reproduces the transition-associated peak. In that model, CID identifies the quantum critical point from images across an energy sequence without averaging over multiple disorder realizations. The authors limit the claim: locating a transition does not yield critical exponents or establish its universality class, which would require systematic finite-size analysis.
A control rules out a generic crossover signal
To test whether any uneven mixture of conducting and insulating regions would produce the same effect, the researchers examined a trivial band-edge crossover in a MoSe₂/WS₂ moiré heterobilayer. The crossover produces coexisting conductive and insulating puddles, and CID rises as the pattern becomes mixed. But the ratio between the measured CID and the shuffled-image CID lacks the pronounced enhancement seen at the quantum Hall transitions.
That difference makes the control important: the compression metric responds to spatial change, but the paper’s results do not support treating every rise in CID as evidence of quantum criticality. The authors say interpretation still depends on conventional measurements and knowledge of the system being imaged.
Local maps reveal fractional puddles
In graphene, the team also applied a sliding-window version of the calculation to images near filling factor 1/3. The resulting spatial CID map shows areas where the fractional state appears locally coherent alongside regions with a more varied mixture of compressible and incompressible responses. That information would be hidden in a single global CID value or a spatial average.
The paper also applies the same approach to existing datasets from electron and X-ray microscopy, magnetic skyrmion imaging, and optical images of bacterial biofilms. The authors present these as demonstrations across different imaging methods, not as proof that CID has the same physical meaning in every system. Its absolute value depends on image size, resolution, encoding choices and noise; comparisons are most reliable within a consistently processed image series.