BPT and VO87 diagrams showing the comparison between JADES z ∼ 6 galaxies and samples across various redshifts. Individual measurements are shown as solid purple points, while values derived from stacked JADES spectra are shown as the open magenta diamonds. The background grey 2D PDF shows z ∼ 0 galaxies from SDSS. Extremely metal-poor z ∼ 0 galaxies from Izotov et al. (2018, 2019) are shown as yellow pentagons. MOSDEF z ∼ 2 galaxies are shown as green circles (Kriek et al. 2015). The z ∼ 5.6 composite CEERS spectra presented by Sanders et al. (2023) are shown as the blue open circles. The solid navy lines in each panel show the theoretical maximum starbust lines from Kewley et al. (2001), while the dotted line in the left panel shows the empirical demarcation derived by Kauffmann et al. (2003). Left: classical BPT or ‘N2–BPT’ – Purple triangles provide 3-σ upper limits on the locations of our JADES galaxies since we are unable to detect [N II] λ6583 in any of the 18 galaxies for which we have G395M/F290LP spectral coverage of the Hα complex, although the R3 ratio is very well constrained in all cases from PRISM data. Even after stacking the grating spectra of all galaxies, we do not recover a robust detection of [N II] λ6583. Centre: in the S2–VO87 diagram (often referred to as ‘S2–BPT’), three out of 20 of our JADES galaxies show detections of [S II] in individual spectra. The tight 3-σ upper limits we place on our non-detections highlight that there must be more than an order of magnitude in scatter in the S2 ratio within the sample. Right: we find one tentative detection of [O I] λ6300 in our O1–VO87 diagram.