Background-only fit to the observed numbers of events (black points) as a function of the diphoton invariant mass $m_{\gamma\gamma}$.

Expected and observed 95% CL upper limits on the signal fiducial cross section $\sigma_{\textrm{fid}}$, assuming 100% branching ratio for ALP decay into two photons, as a function of the hypothetical ALP mass $m_{\textrm{X}}$. The 1$\sigma$ and 2$\sigma$ confidence intervals are shown by the coloured bands.

Expected and observed 95% CL upper limits on the the ALP coupling constant, assuming 100% branching ratio for ALP decay into two photons, as a function of the hypothetical ALP mass $m_{\textrm{X}}$. The 1$\sigma$ and 2$\sigma$ confidence intervals are shown by the coloured bands. Contours of the ALP natural width $\Gamma$ are illustrated by the smooth blue solid lines.

Fiducial region acceptance of signal events for coupling constant f$^{-1}=0.05$ TeV$^{-1}$. EL, SD, and DD stand for the exclusive, single-dissociative, and double-dissociative processes. The acceptances are parameterised as $p_0 + p^{i}_{1} e^{p_{2}^{i} m_{\textrm{X}}} + p_{3}^{i} e^{p_{4}^{i} m_{\textrm{X}}} $ , where $i$ specifies the process type, exclusive, single-dissociative, or double-dissociative.

Correction factors $A_{\textrm{X}}/A_{0}$ and $C_{\textrm{X}}$ for the exclusive signal events with ALP coupling constant f$^{-1}$=0.05 TeV$^{-1}$, where acceptance $A_{\textrm{X}}$ and efficiency $C_{\textrm{X}}$ are defined in arXiv:2102.13405. $A_{\textrm{X}}$/$A_{0}$ and $C_{\textrm{X}}$ are parameterised as $p_0 + p^{i}_{1} e^{p_{2}^{i} m_{\textrm{X}}} + p_{3}^{i} e^{p_{4}^{i} m_{\textrm{X}}}$, where $i$ specifies the process type.

Simulated ALP signal cutflow for $m_{\rm{X}}$=300 GeV and $m_{\rm{X}}$=1200 GeV. The yields are normalized to a luminosity of 14.6 fb$^{-1}$. Selections are applied for each side after the selection on acoplanarity. The union of the sets of selected events is taken finally.

Expected and observed 95% CL for the 1SFH mu Majorana model.

Expected and observed 95% CL for the 2QDH NH Dirac model.

Expected and observed 95% CL for the 2QDH NH Majorana model.

Expected and observed 95% CL for the 2QDH IH Dirac model.

Expected and observed 95% CL for the 2QDH IH Majorana model.

Cutflow for six simulated signal channels showing the weighted number of expected events based on the single-flavour mixing model in the Majorana limit. Each column uses the generated signal sample with the mass hypothesis $m_N = 10$ GeV and proper decay length $c\tau_N = 10$ mm.

Cutflow for the six channels in data showing the number of events passing each successive signal selection for Majorana HNLs.

The event selection efficiency for each mass-lifetime point in all six studied channels. Shown is the fraction of the produced MC simulation events that pass all signal region selections. An entry of 0 indicates no events were selected.

The dominant signal uncertainty is due to differences in reconstruction of displaced vertices and tracks between data and MC. This is evaluated by comparing $K^{0}_{S} \rightarrow \pi^+\pi^-$ event yields in the VR and in MC produced with Pythia8.186 in bins of $p_\mathrm{T}$ and $r_\mathrm{DV}$. The data/MC ratio is normalized to the bin nearest the IP where the tracking and vertexing reconstruction algorithms are expected to be most robust. The symmetrized difference from 1.0 is applied to each signal vertex as a per-event systematic variation.

Expected and observed yields in the different analysis regions (prefit) for the 1SFH e Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 1SFH e Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (prefit) for the 1SFH e Majorana model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 1SFH e Majorana model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (prefit) for the 1SFH u Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 1SFH u Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (prefit) for the 1SFH u Majorana model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 1SFH u Majorana model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (prefit) for the 2QDH (NH) Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 2QDH (NH) Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (prefit) for the 2QDH (NH) Majorana model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 2QDH (NH) Majorana model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (prefit) for the 2QDH (IH) Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 2QDH (IH) Dirac model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (prefit) for the 2QDH (IH) Majorana model (10 GeV, 10mm).

Expected and observed yields in the different analysis regions (postfit) for the 2QDH (IH) Majorana model (10 GeV, 10mm).

The total displaced vertexing efficiency as a function of $r_{DV}$ for the custom configuration used in this analysis. The definition of the secondary vertex efficiency can be found in defined in \cite{ATL-PHYS-PUB-2019-013}. The efficiency is shown for $\mu-\mu\mu$, $\mu-\mu e$ and $\mu-ee$ signals with $m_N=10$~GeV and $c\tau_N=10$~mm.