A new arXiv preprint reports a global analysis that extracts deeply virtual Compton scattering (DVCS) amplitudes—quantities that summarize the real and imaginary response—from published measurements of protons and neutrons. Its two modeling strategies generally agree, but the real parts of some amplitudes remain poorly constrained.
The fit included 12,822 proton-target and 117 neutron-target data points. It kept only measurements with Q² above 1 GeV² and −t/Q² below 0.4, and left existing timelike Compton-scattering (TCS) data out of the main fit because using them would require fixed-order relations between the two reactions.
Two routes to the same signal
Within the PARTONS framework, one strategy was theory-agnostic and the other theory-augmented, adding theory-based constraints. The analysis used them to extract real and imaginary Compton form factors—the fitted quantities that represent the amplitudes—including helicity-conserving and helicity-flip components.
Uncertainty was estimated with 101 fits: one using the original data and 100 using artificially fluctuated pseudodata. Reported total unweighted chi-squared values—a measure of model-data mismatch—were 1.29 for the theory-agnostic proton fit and 1.36 for the theory-augmented fit. For neutron data, the corresponding values were 0.76 and 0.70.
The two models generally agreed for both targets. Adding theory usually narrowed uncertainties in poorly constrained regions, except for the real part of the E-tilde form factor, and the GK model agreed better with the extraction at next-to-leading order than at leading order.
The helicity-flip form factors H0+ and H+− were generally consistent with zero, with a possible negative H0+ signal at high ξ. The authors interpret the possible feature as a sign that higher-twist effects may become more relevant, but the analysis does not establish that those effects caused it.
The clearest limits are in the real parts
The theory-augmented model estimated the asymptotic term of the subtraction constant at −0.69 for the proton, with printed upper and lower terms of +6.92 and −3.13. The neutron estimate was −0.86, with +7.25 and −5.59 terms.
The extracted subtraction constant was compatible with lattice-QCD predictions, but its uncertainty was much larger because the real parts of the Compton form factors were poorly constrained. Lattice artefacts and truncation effects were not included in the lattice uncertainty band.
A cautious test beyond the main fit
Although TCS observations were excluded from the primary fit, the researchers projected the extracted DVCS results onto TCS observables. The projections generally agreed with exploratory TCS data. For the theory-augmented model, chi-squared per point was 0.91 at leading order and 0.47 at next-to-leading order.
The lower next-to-leading-order value is consistent with the authors’ view that these corrections are important and that the comparison supports the shared description of DVCS and TCS amplitudes. The authors caution, however, that the TCS data were exploratory and that the next-to-leading-order projections had larger uncertainties.
In the region where proton and neutron results overlapped, the real part of the proton-minus-neutron H++ difference was weakly favoured negative. The imaginary parts of the H++ and E++ differences were preferentially positive.
A baseline, not a full map
The study extracts Compton form factors rather than generalized parton distributions, so it does not yet provide a completed three-dimensional extraction of the nucleon. The authors present the results instead as a baseline for future efforts to deconvolve the underlying distributions.
The paper is a version 1 arXiv preprint posted on 20 August 2026. Funding came in part from the National Science Centre in Poland and the French National Research Agency.
Paper data and sources
Original title: Extraction of DVCS amplitudes off the nucleon
Authors: C. Mezrag, P. Sznajder, J. Wagner
Journal/Repository: arXiv
Status: Preprint, not yet peer-reviewed
First online: 2026-08-20
DOI: Not available
Original paper · Full text