Flow-Guided Tags Identify 63% of LZ Lead-214 Background

Radon-chain decays are a persistent internal background for liquid-xenon dark-matter searches because $^{214}$Pb beta decays can resemble low-energy signal candidates. Their locations are not static: neutral atoms follow xenon circulation, while charged daughters also drift in the detector electric field. LZ uses that transport rather than treating it as a nuisance. The paper shows that a stable, low-mixing flow state can turn prior alpha decays into a time-dependent veto for later $^{214}$Pb decays.

Core Contribution

The central contribution is a data-driven transport model for the LZ liquid-xenon time projection chamber (LXe-TPC), built from sequential $^{222}$Rn–$^{218}$Po alpha-decay pairs. These pairs reveal both horizontal liquid flow and the separation between neutral and positively charged progeny. The authors then propagate the inferred daughter trajectories forward from each observed production point, defining a temporally evolving volume in which a subsequent $^{214}$Pb decay is expected.

That is more specific than a conventional spatial fiducial cut or a fixed delayed-coincidence window. The selection follows the atom through the detector for as long as 81 minutes after production. In the analyzed low-mixing state, the paper reports identification of $(63 \pm 6_{\mathrm{stat}} \pm 7_{\mathrm{sys}})\%$ of ground-state $^{214}$Pb decays at a 9.0% exposure cost. For a rare-event search, that trade is useful if the tagged sample is a leading background and the removed exposure is smaller than the background-rejection value it delivers.

Technical Approach

The reconstruction begins with alpha pairs from the $^{222}$Rn decay chain. Their measured displacement over a time separation $\Delta T$ supplies a local velocity measurement. Horizontal displacement is used to isolate correctly paired decays from accidental pairings; Figure 5 shows this population at small separation after subtraction of an unrelated-pair background estimated using a 40-minute time-shifted search. The same figure separates neutral and charged $^{218}$Po populations in vertical velocity. Charged ions move downward under the electric field, while neutral atoms trace the liquid flow.

The analysis is explicitly tied to a low-mixing circulation regime. Figure 2 contrasts high-, low-, and minimal-mixing states in two detector slices. The active region runs from the cathode at $Z=0$ cm to the gate at $Z=145.6$ cm; the maps use $^{222}$Rn–$^{218}$Po pairs with $\Delta T>5$ s to visualize horizontal flow and $^{218}$Po activity inside the 5.5-tonne fiducial volume. Low mixing supplies identifiable neutral and charged pair populations, which is necessary for fitting the transport model.

The fitted model combines liquid velocity and ion drift to predict daughter paths. The paper also uses these flow-informed selections to make a calibration sideband concurrent with science data, rather than reserving separate calibration running. That is operationally relevant: the same chain responsible for the background supplies a tagged control population under the detector conditions of interest.

Results and Analysis

The headline result is a measured tagging efficiency, not merely a simulation projection: 63% of ground-state $^{214}$Pb decays are identified, with 6 percentage points statistical and 7 percentage points systematic uncertainty, while the exposure penalty is 9.0%. The uncertainty is material, but even the lower side of that estimate suggests that the method rejects a substantial fraction of this background for a modest loss of search volume-time.

The transport fit also yields charge-branching estimates: $^{218}$Po$^+$ is reported at $0.49 \pm 0.01$, $^{214}$Pb$^+$ at $0.48 \pm 0.12$, and $^{214}$Bi$^+$ at $0.74 \pm 0.05$. The mean charged-ion lifetime in the LZ TPC is $49 \pm 4$ minutes. These measurements explain why a static geometric approximation would be inadequate: whether a daughter remains in the moving neutral population or becomes an ion changes where it will be when it decays.

The evidence supports the paper's claim within the selected detector state. It is an in-detector demonstration with directly reconstructed alpha-pair populations and quantified uncertainty, rather than an extrapolation from fluid simulation alone. Its practical scope is narrower: the tagging model depends on maintaining and characterizing the flow regime that made the neutral and charged bands separable.