Why prompt-photon predictions still need fragmentation functions
Prompt photons — photons produced directly in a hard partonic scattering — are one of the cleanest handles we have on the gluon and quark content of colliding protons and nuclei. In practice, experiments never measure a purely “direct” photon: a quark or gluon can radiate a photon collinearly as it fragments, and an isolation cut (removing hadronic activity near the photon) only suppresses this fragmentation contribution, it does not eliminate it.
That residual piece is controlled by parton-to-photon fragmentation functions, and the standard sets used to describe them (BFG, GRV, GDRG) are decades old fits with only loose experimental constraints. In arXiv:2608.15944 I took a more pragmatic route: rather than fitting data directly, extract an effective fragmentation function straight from PYTHIA’s own parton shower, flavour by flavour, and package it as an LHAPDF grid so it can be dropped into existing fixed-order codes exactly like any other fragmentation set. Along the way, converting the legacy BFG/GRV/GDRG sets into the same LHAPDF grid format turned up a more mundane but useful finding: LHAPDF’s default log-cubic interpolation reproduces the tabulated functions noticeably better than the linear interpolation used in older, hand-rolled implementations — a reminder that infrastructure choices (interpolation, grid spacing) can matter as much as the physics input itself.
See the photon fragmentation research page for where this is headed next.