 Solution

A plausible explanation is partial dissociation leaving a residual surface species, such as an NH${}_x$ fragment, adsorbed nitrogen or hydrogen, after molecular ammonia has finished desorbing. A small residual population can maintain a surface electric dipole moment and a $0.2$ eV work function offset. Heating farther can remove, recombine or rearrange those species.

Test this hypothesis by recording temperature-programmed desorption signals for hydrogen and nitrogen as well as ammonia, and by measuring residual nitrogen with X-ray photoelectron spectroscopy. A surface-sensitive vibrational spectrum could distinguish NH${}_x$ groups from atomic nitrogen. Isotopically labelled ammonia would help assign desorption products. Correlating disappearance of the residual spectral signal with recovery of the clean-surface work function would be more diagnostic than the absence of an ammonia signal alone. This is a proposed explanation, not a uniquely established identification.