Radion--QCD Interference in $t\bar t$ Production at the HL-LHC: Finite-Top-Mass Effects and Projected Sensitivity
We investigate the interference of a heavy radion with the QCD continuum in top-quark-pair production at the High-Luminosity Large Hadron Collider, restricting the numerical study to the pure-radion limit $ξ=0$. The gluon-fusion amplitude combines the QCD trace anomaly with the exact finite-top-mass loop form factor, whose coherent sum fixes both the magnitude and the phase of the production coefficient. Above the top threshold the loop develops an absorptive part, rendering the production coefficient complex and allowing the interference to remain non-zero on the resonance pole, where a purely real point-like coefficient would give none. The complex coefficient is validated independently using a native loop-induced implementation and a common-event phase-basis construction. The resulting signature is a peak--dip deformation rather than a positive bump, while detector smearing turns the narrow truth-level structure into a broad sub-percent distortion. For $3\,\mathrm{ab}^{-1}$ at $\sqrt{s}=14$ TeV, we construct an ATLAS-anchored phenomenological response and perform an exact binned Poisson Asimov analysis with a free background normalization and a correlated shape nuisance. For a $0.1\%$ shape benchmark with a $50$ GeV correlation length, the profiled median $95\%$ $CL_s$ reach in $Λ_r$ is $2.49$, $2.82$, $3.00$, $2.56$, and $2.19$ TeV for radion masses of $600$, $800$, $1000$, $1200$, and $1500$ GeV, respectively. A diagnostic decomposition at $800$ GeV gives $2.18$ TeV for the resonance-squared term alone and $2.83$ TeV for the interference term alone. A selected-background deformation stress test largely removes the apparent maximum near $1$ TeV, showing that its location and prominence are normalization-prescription dependent, while leaving the central conclusion unchanged: the projected sensitivity is predominantly interference driven.