Characterization of Delayed Ionization Backgrounds
in the LZ Experiment

Eli Mizrachi (they/them)

On Behalf of the LZ Collaboration

2023-11-07

Introduction

What are Electron Trains?

What do we know about them?

Electron Trains (AKA “e-trains”)

  • “Electron trains” are a form of background noise in dual-phase TPCs
  • Spurious single electrons (SEs) observed for at least a second after S2s
    • 30% livetime loss vetoing electron/photon-trains in LZ Science Run 1 (SR1) (Linehan [1])
  • \(\Delta t_{S2} < t_{drift} \rightarrow\) dominated by electrons from photoionization of TPC liquid & grids
  • \(\Delta t_{S2} > t_{drift} \rightarrow\) dominated by electrons captured & released by impurities in drift path?

Single scatter S2 followed by SEs for at least 40ms (GIF)

Top PMT array, left, showing position-correlated SEs (GIF)

Electron Train Hypotheses

  • Top: “Drift” liquid events generate electron trains, not photoionization in gas or below cathode
  • E-train rates are [2] [3]:
    • Bottom: correlated with electron lifetime
      • \(\color{red}{\bullet} \rightarrow \star \implies\) increasing time since S2
    • Anti-correlated with drift time of progenitor (shown later)
  • \(\implies\) Liquid bulk origin, not liquid surface
    • Unclear physics; electrons captured and released by impurities in drift path?

 

Electron trains in XENON1T [2]

E-train electron lifetime dependence in LUX [3]

Electron Trains in LZ

Position Dependence of SEs after “Progenitor” S2s

  • Prog \(r\) < 55cm, area > 1e4 phd (~200 SE)
    • Skip if < 200ms after any pulse > 5e3 phd
  • [Hz/cm2] because larger \(\Delta r \implies\) larger area in XY \(\implies\) more child pulses
  • Define position-correlated and uncorrelated child pulses
    • p-corr: \(\Delta r\) < 10cm
    • p-uncorr: 20cm < \(\Delta r\) < 30cm
  • Position-correlated region captures power law (next slide), position-uncorrelated avoids power law and walls
  • Prog drift time within fiducial volume

 

SE Rate vs. radial distance \(\Delta r\) between progenitor and child

SE Rates vs. Time Since S2

  • Top: P-corr flux is more intense and appears to follow a power law
  • Bottom: Fit power law \(\alpha t^{-\beta}\) to p-corr rates
    • \(\beta\) consistent with LUX [3], XENON1T [2]
  • Dip in rates prior to 1ms is known artifact of pulse pile-up from photoionization

 

Drift Time and Progenitor Area

  • Top: SE Rate vs. progenitor area
    • \(SE_{R}\): Progenitor size in electrons extracted
  • Bottom: SE Rate vs. progenitor drift time
    • Normalize by \(SE_{I} = SE_{S} \exp(t_{drift} / \tau_{e^-})\)
    • \(SE_{S} \rightarrow SE_{R}\), corrected for extraction efficiency
  • \(\Delta t\) [s]: (0.003, 0.3) avoids photoionization in p-uncorr rates
  • P-uncorr pulses show virtually no correlation for either prog area or drift time
    • Favors explanation of uncorrelated pulses coming from previous e-trains (XENON1T [2])

 

Electron Lifetime

  • Rates in SR1 exhibit dependence on electron lifetime up to 8ms

Electron Loss Normalization

Advancing the Liquid Bulk Hypothesis

Electron Lifetime and Drift Time

  • Electron lifetime and drift time dependence hint at liquid bulk origin
  • Normalizing by \(SE_{I}\) does not cancel out dependence

 

Electron Loss Normalization

  • Normalizing by \(e_{loss} = SE_{I} - SE_{S}\) unifies drift time and electron lifetime
  • Clear indication that power law originates with liquid bulk impurities
  • Also shows (again) lack of correlation for uncorrelated backgrounds

 

“Drift” Field Dependence

Two TPCs for the Price of One

  • Distinguish between extraction liquid (EXL) and drift liquid (DRL) single scatters with drift time
  • Use top-bottom asymmetry to exclude gas events
  • EXL: a second TPC where the gate is a “cathode” i.e. \(E_{drift} = E_{extract}\)
  • Vary extraction field, compare EXL and DRL e-trains
  • Isolate “drift” field dependence with otherwise identical detector conditions!
    • Right: Drift time does not affect exponent
    • Correct for extraction efficiency and increased charge yield in EXL with \(SE_S\) normalization

 

Fits to power law at different drift time bands show exponent does not depend on drift time

\(\Delta V_{Extract}\) 7, 8kV - EXL Events

  • E-trains from extraction liquid have much weaker delayed correlated pulses
  • Try subtracting flux from uncorrelated pulses for slightly cleaner power law
  • Liquid field \(\approx\) 3400, 3900 V/cm for 7, 8kV respectively; radial field variation ~few %

\(\Delta V_{Extract}\) 7, 8kV - Uncorrelated Subtraction

  • Exponent is steeper than typical ~1.0-1.1 from drift liquid events (~0.5-1 sigma difference)
  • Gate grid at 2.5us drift time; no change in exponents with drift time cut at 2us

\(\Delta V_{Extract}\) 7, 8kV - DRL Events

  • Rate curves shown here from same datasets, different drift time cut
  • Appears compatible with “weak” extraction field dependence reported by XENON1T [2]
  • \(\implies\) Field in drift path could influence a time constant in power law exponent

Conclusions

Summary of General Characterization

  • Power law observed for rate of single electron pulses following S2s in drift region
  • P-uncorr pulses also uncorrelated w/other progenitor characteristics e.g. area and drift time
  • Bottom: \(e_{loss}\) normalization appears to unify electron lifetime and drift time dependence
    • Strong evidence for liquid bulk impurities as dominant factor in power law

Summary of Possible “Drift” Field Dependence

  • Simultaneous analysis of e-trains in extraction liquid (EXL) and drift liquid (DRL)
  • Steeper exponent at field > 3 kV/cm?
    • Apparent agreement with result from Akimov et al. [4]
  • Exponent in DRL at field \(\approx\) 180 V/cm matches literature
  • Studies of this effect are worth pursuing to gain a better understanding of e-train physics and modeling!

Note: value for study conducted by Kopec et al. [5] was reported for 500 V/cm and “unchanged” for other fields

Acknowledgements 👏

 

 

Thank you to our sponsors and 37 participating institutions!

References

[1]
R. E. Linehan, T. Shutt, D. S. Akerib, P. Burchat, and A. Friedland, High Voltage Electrode Development and the LZ Experiment’s WIMP Search, PhD thesis, 2022.
[2]
[3]
D. S. Akerib et al., Investigation of Background Electron Emission in the LUX Detector, Physical Review D 102, 092004 (2020).
[4]
D. Yu. Akimov et al., Observation of Delayed Electron Emission in a Two-Phase Liquid Xenon Detector, Journal of Instrumentation 11, C03007 (2016).
[5]
A. Kopec, A. L. Baxter, M. Clark, R. F. Lang, S. Li, J. Qin, and R. Singh, Correlated Single- and Few-Electron Backgrounds Milliseconds After Interactions in Dual-Phase Liquid Xenon Time Projection Chambers, Journal of Instrumentation 16, P07014 (2021).

Backup

Normalization Reference

Factor Description
\(SE_{R} = S2_{phd} / n_{phd / SE}\) “Raw” (extracted) S2 area in units of single electrons
\(SE_{S} = SE_{R} / e_{eee}\) “Surface” S2 area, i.e. \(SE_{R}\) corrected for extraction efficiency
\(SE_{I} = SE_{S} \exp(t_{drift} / \tau_{e^-})\) “Initial” S2 area, i.e. \(SE_{S}\) corrected for drift losses
\(e_{loss} = SE_{I} - SE_{S}\) Number of electrons lost while drifting
\(\text{cm}^2\) Area of liquid surface subtended by radial selection of pulses

Electron Trains in TPC Regions

\(E_{drift}\) Sweep - DRL Events

\(E_{drift}\) Sweep - “Background” Subtraction in DRL Events

BigDEB Main Algorithm

  • Livetime between windows is not counted unless trigger efficiency of pulse is ~100%