Background Determination for the LUX-ZEPLIN (LZ) Dark Matter Experiment

J. Aalbers,1, 2 D.S. Akerib,1, 2 A.K. Al Musalhi,3 F. Alder,4 S.K. Alsum,5 C.S. Amarasinghe,6 A. Ames,1, 2
T.J. Anderson,1, 2 N. Angelides,4, 7 H.M. Araújo,7 J.E. Armstrong,8 M. Arthurs,6 A. Baker,7 J. Bang,9
J.W. Bargemann,10 A. Baxter,11 K. Beattie,12 P. Beltrame,4 E.P. Bernard,12, 13 A. Bhatti,8 A. Biekert,12, 13
T.P. Biesiadzinski,1, 2 H.J. Birch,6 G.M. Blockinger,14 B. Boxer,15 C.A.J. Brew,16 P. Brás,17 S. Burdin,11
M. Buuck,1, 2 R. Cabrita,17 M.C. Carmona-Benitez,18 C. Chan,9 A. Chawla,19 H. Chen,12 A.P.S. Chiang,7
N.I. Chott,20 M.V. Converse,21 A. Cottle,3, ∗ G. Cox,18 O. Creaner,12 C.E. Dahl,22, 23 A. David,4 S. Dey,3
L. de Viveiros,18 C. Ding,9 J.E.Y. Dobson,4 E. Druszkiewicz,21 S.R. Eriksen,24 A. Fan,1, 2 N.M. Fearon,3
S. Fiorucci,12 H. Flaecher,24 E.D. Fraser,11 T. Fruth,4 R.J. Gaitskell,9 J. Genovesi,20 C. Ghag,4 R. Gibbons,12, 13
M.G.D. Gilchriese,12 S. Gokhale,25 J. Green,3 M.G.D.van der Grinten,16 C.B. Gwilliam,11 C.R. Hall,8 S. Han,1, 2
E. Hartigan-O’Connor,9 S.J. Haselschwardt,12 S.A. Hertel,26 G. Heuermann,6 M. Horn,27 D.Q. Huang,6 D. Hunt,3
C.M. Ignarra,1, 2 R.G. Jacobsen,12, 13 O. Jahangir,4 R.S. James,4 J. Johnson,15 A.C. Kaboth,16, 19 A.C. Kamaha,28
arXiv:2211.17120v1 [hep-ex] 30 Nov 2022

D. Khaitan,21 I. Khurana,4 R. Kirk,9 D. Kodroff,18, † L. Korley,6 E.V. Korolkova,29 H. Kraus,3 S. Kravitz,12

L. Kreczko,24 B. Krikler,24 V.A. Kudryavtsev,29 E.A. Leason,30 J. Lee,31 D.S. Leonard,31 K.T. Lesko,12
C. Levy,14 J. Lin,12, 13 A. Lindote,17 R. Linehan,1, 2 W.H. Lippincott,10, 22 X. Liu,30 M.I. Lopes,17 E. Lopez
Asamar,17 B. López Paredes,7 W. Lorenzon,6 C. Lu,9 S. Luitz,1 P.A. Majewski,16 A. Manalaysay,12
R.L. Mannino,32 N. Marangou,7 M.E. McCarthy,21 D.N. McKinsey,12, 13 J. McLaughlin,23 E.H. Miller,1, 2
E. Mizrachi,8, 32 A. Monte,10, 22 M.E. Monzani,1, 2, 33 J.D. Morales Mendoza,1, 2 E. Morrison,20 B.J. Mount,34
M. Murdy,26 A.St.J. Murphy,30 D. Naim,15 A. Naylor,29 C. Nedlik,26 H.N. Nelson,10 F. Neves,17 A. Nguyen,30
J.A. Nikoleyczik,5 I. Olcina,12, 13 K.C. Oliver-Mallory,7 J. Orpwood,29 K.J. Palladino,3, 5 J. Palmer,19 N. Parveen,14
S.J. Patton,12 B. Penning,6 G. Pereira,17 E. Perry,4 T. Pershing,32 A. Piepke,35 D. Porzio,17, ‡ S. Poudel,35
Y. Qie,21 J. Reichenbacher,20 C.A. Rhyne,9 Q. Riffard,12 G.R.C. Rischbieter,6, 14 H.S. Riyat,30 R. Rosero,25
P. Rossiter,29 T. Rushton,29 D. Santone,19 A.B.M.R. Sazzad,35 R.W. Schnee,20 S. Shaw,10, 30 T. Shutt,1, 2
J.J. Silk,8 C. Silva,17 G. Sinev,20 R. Smith,12, 13 M. Solmaz,10 V.N. Solovov,17 P. Sorensen,12 J. Soria,12, 13
I. Stancu,35 A. Stevens,3, 4, 7 K. Stifter,22 B. Suerfu,12, 13 T.J. Sumner,7 N. Swanson,9 M. Szydagis,14 R. Taylor,7
W.C. Taylor,9 D.J. Temples,23 P.A. Terman,36 D.R. Tiedt,27 M. Timalsina,20 Z. Tong,7 D.R. Tovey,29
J. Tranter,29 M. Trask,10 M. Tripathi,15 D.R. Tronstad,20 W. Turner,11 U. Utku,4 A.C. Vaitkus,9 A. Wang,1, 2
J.J. Wang,35 W. Wang,5, 26 Y. Wang,12, 13 J.R. Watson,12, 13 R.C. Webb,36 T.J. Whitis,10 M. Williams,6
F.L.H. Wolfs,21 S. Woodford,11 D. Woodward,18 C.J. Wright,24 Q. Xia,12 X. Xiang,9, 25 J. Xu,32 and M. Yeh25
(The LUX-ZEPLIN (LZ) Collaboration)
1
SLAC National Accelerator Laboratory, Menlo Park, CA 94025-7015, USA
2
Kavli Institute for Particle Astrophysics and Cosmology,
Stanford University, Stanford, CA 94305-4085 USA
3
University of Oxford, Department of Physics, Oxford OX1 3RH, UK
4
University College London (UCL), Department of Physics and Astronomy, London WC1E 6BT, UK
5
University of Wisconsin-Madison, Department of Physics, Madison, WI 53706-1390, USA
6
University of Michigan, Randall Laboratory of Physics, Ann Arbor, MI 48109-1040, USA
7
Imperial College London, Physics Department, Blackett Laboratory, London SW7 2AZ, UK
8
University of Maryland, Department of Physics, College Park, MD 20742-4111, USA
9
Brown University, Department of Physics, Providence, RI 02912-9037, USA
10
University of California, Santa Barbara, Department of Physics, Santa Barbara, CA 93106-9530, USA
11
University of Liverpool, Department of Physics, Liverpool L69 7ZE, UK
12
Lawrence Berkeley National Laboratory (LBNL), Berkeley, CA 94720-8099, USA
13
University of California, Berkeley, Department of Physics, Berkeley, CA 94720-7300, USA
14
University at Albany (SUNY), Department of Physics, Albany, NY 12222-0100, USA
15
University of California, Davis, Department of Physics, Davis, CA 95616-5270, USA
16
STFC Rutherford Appleton Laboratory (RAL), Didcot, OX11 0QX, UK
17
Laboratório de Instrumentação e Fı́sica Experimental de Partı́culas (LIP),
University of Coimbra, P-3004 516 Coimbra, Portugal
18
Pennsylvania State University, Department of Physics, University Park, PA 16802-6300, USA
19
Royal Holloway, University of London, Department of Physics, Egham, TW20 0EX, UK
20
South Dakota School of Mines and Technology, Rapid City, SD 57701-3901, USA
21
University of Rochester, Department of Physics and Astronomy, Rochester, NY 14627-0171, USA
22
Fermi National Accelerator Laboratory (FNAL), Batavia, IL 60510-5011, USA
23
Northwestern University, Department of Physics & Astronomy, Evanston, IL 60208-3112, USA
24
University of Bristol, H.H. Wills Physics Laboratory, Bristol, BS8 1TL, UK
 2

25
Brookhaven National Laboratory (BNL), Upton, NY 11973-5000, USA
26
University of Massachusetts, Department of Physics, Amherst, MA 01003-9337, USA
27
South Dakota Science and Technology Authority (SDSTA),
Sanford Underground Research Facility, Lead, SD 57754-1700, USA
28
University of Califonia, Los Angeles, Department of Physics & Astronomy, Los Angeles, CA 90095-1547
29
University of Sheffield, Department of Physics and Astronomy, Sheffield S3 7RH, UK
30
University of Edinburgh, SUPA, School of Physics and Astronomy, Edinburgh EH9 3FD, UK
31
IBS Center for Underground Physics (CUP), Yuseong-gu, Daejeon, Korea
32
Lawrence Livermore National Laboratory (LLNL), Livermore, CA 94550-9698, USA
33
Vatican Observatory, Castel Gandolfo, V-00120, Vatican City State
34
Black Hills State University, School of Natural Sciences, Spearfish, SD 57799-0002, USA
35
University of Alabama, Department of Physics & Astronomy, Tuscaloosa, AL 34587-0324, USA
36
Texas A&M University, Department of Physics and Astronomy, College Station, TX 77843-4242, USA
(Dated: December 1, 2022)
The LUX-ZEPLIN experiment recently reported limits on WIMP-nucleus interactions from its
initial science run, down to 6.5 × 10−48 cm2 for the spin-independent interaction of a 30 GeV/c2
WIMP at 90% confidence level. In this paper, we present a comprehensive analysis of the back-
grounds important for this result and for other upcoming physics analyses, including neutrinoless
double-beta decay searches and effective field theory interpretations of LUX-ZEPLIN data. We con-
firm that the in-situ determinations of bulk and fixed radioactive backgrounds are consistent with
expectations from the ex-situ assays. The observed background rate after WIMP search criteria
were applied was (6.3 ± 0.5) × 10−5 events/keVee /kg/day in the low energy region, approximately
60 times lower than the equivalent rate reported by the LUX experiment.

I. INTRODUCTION potential signal and background events. In the case of

WIMPs, their few to tens of keV single NRs can be well
LUX-ZEPLIN (LZ) is an experiment optimized for distinguished from the majority background ERs of sim-
the observation of signals from Weakly Interacting Mas- ilar energy, which typically occur towards the edges of
sive Particles (WIMPs) having masses in excess of about the detector and often happen as part of MS events.
5 GeV/c2 . In addition to WIMPs, LZ is sensitive to a Two additional detectors form an anti-coincidence, ac-
range of other hypothetical processes as well as physics tive veto system: a 2-tonne liquid xenon “Skin” directly
beyond the Standard Model including, but not limited to, surrounding the TPC and a 17-tonne gadolinium-loaded
neutrinoless double-beta decay (0νββ) [1, 2], axions and liquid scintillator (GdLS) outer detector (OD). The Skin
axion-like particles [3]. To maximize discovery potential provides an extra, instrumented buffer layer between the
for WIMPs or any other search candidates, or to set reli- TPC and outside radiation, leveraging the high proba-
able upper limits on their interactions, sources that could bility that gamma rays which traverse this detector will
produce similar signatures must be well-understood. scatter within it. The design of the OD is optimized for
The LZ experiment is described in detail in Refs. [4–6]. the tagging of neutrons, with gadolinium having an ex-
The central detector is a liquid-gas xenon time projection tremely high thermal neutron capture cross-section. The
chamber (TPC) of 7 tonne active mass, wherein particle Skin and OD are therefore effective at tagging gamma
interactions or “events” are observed via the collection rays and neutrons that enter or exit the TPC, which
of light. Two separate signals arise from the detection would signify events from conventional sources.
of prompt scintillation and delayed electroluminescence LZ operates in the Davis Cavern of the Sanford Under-
light from a given interaction, the latter of which is cre- ground Research Facility (SURF) in Lead, South Dakota,
ated by charge extracted into the gas phase. These are where it is well-shielded from cosmic rays by 4300 meters
known as S1 and S2, respectively. The combination of S1 water equivalent (m.w.e.) rock overburden [8]. Further-
and S2 can be used to reconstruct the energy and position more, the entire detector configuration resides within a
of an event, as well as indicate the type of interaction, tank filled with 238 tonnes of ultrapure water, providing
whether it was an electronic recoil (ER) or nuclear re- > 1.2 m.w.e of passive shielding in every direction. Back-
coil (NR) on xenon, the target medium. The excellent, ground events in the TPC thus result predominantly from
O(mm) position resolution facilitates the identification of radioactivity internal to the LZ assembly.
single scatter (SS) and multiple scatter (MS) event classi- An extensive radioassay campaign was undertaken to
fications, and offers accurate detector fiducialization [7]. inform the material selection for the construction of the
All of these TPC features help to discriminate between experiment, to ensure and confirm low WIMP-search
background burden arising from the detector compo-
nents [9]. Construction of the TPC was undertaken in a
∗ [email protected]
class 1000, radon-reduced cleanroom under strict clean-
† [email protected] liness protocols to limit the plate-out of radon progeny
‡ [Deceased] to < 0.5 mBq/m2 and the deposition of airborne dust to
 3

< 500 ng/cm2 on detector surfaces. The xenon is purified particular focus on response matching for low energy
in situ via a hot zirconium getter and an inline radon re- events pertinent to the WIMP search. Among the im-
duction system [10]. Before underground deployment, portant parameters tuned for SR1 were the photon gain,
the xenon also underwent charcoal chromatography at g1 = 0.1136 ± 0.0020 phd/photon, and the charge gain,
SLAC to reduce the Kr and Ar contamination in order g2 = 47.07±1.13 phd/electron. The full set of SR1 detec-
to control the beta backgrounds from 85 Kr and 39 Ar. The tor parameters that were provided to NEST can be found
expectations under these background mitigation strate- in Ref. [16]. A uniform electric field strength of 193 V/cm
gies and associated requirements, as they pertain to a was used for calculating light and charge yields, as well
WIMP search, were formerly set out in our sensitivity as the electron-ion recombination probability.
projections in Ref. [11]. Non-uniformity in the field near the TPC walls leads
This paper reports the current understanding of our to curvature of the drift paths of the ionization electrons.
backgrounds from inferences made with LZ’s initial sci- This was modeled in LZLAMA using a drift map as de-
ence run (SR1): 89 live days of data taken between scribed in Ref. [12], tuned so that the mean reconstructed
December 2021 and May 2022. It details the general position of the wall as a function of drift time matched
background observations and how their determinations that of an injected 83m Kr calibration source. Finite res-
compare to previous ex-situ estimates, before motivat- olution in position reconstruction was incorporated via a
ing how they affected the physics conclusions of the ex- model that was similarly constrained using the position
periment’s first WIMP search results in Ref. [7]. Sec- variation from events at the wall.
tion II discusses the simulations and modeling that un- Pulses that appear close together in time may be re-
derpin the latest calculations of our expected background constructed as a single pulse. This pulse merging was ap-
events; Section III covers measurements of our back- proximated in simulation using two independent models
grounds from examining distributions in energy space for the merging of S1 and S2 pulses, both of which were
(external gamma-ray, beta radiations and electron cap- dependent on relative pulse time, pulse size, and interac-
tures from noble radioisotopes within the xenon, alpha tion position. 83m Kr calibration events were used to tune
particles from radon and its daughters); Section IV doc- the S1 model, given the high likelihood of producing S1
uments backgrounds which can be counted within our pulses close together in time, due to the two-step struc-
data (muons and neutrons); Section V explains how ture of its decay. AmLi calibration events were chosen
these backgrounds impacted our first WIMP analysis, as for the S2 model tuning, given that they often consist of
well as describes our physics backgrounds informed by multiple interactions in close proximity where S2s could
other experiments and calculations (136 Xe two-neutrino merge. The tunings of both models achieved a better
double-beta decay (2νββ), 124 Xe two-neutrino double- than 5% agreement when compared to these respective
electron capture (2νDEC), solar and atmospheric neutri- calibrations. Simulated background events were found to
nos), sources specific to such low energy searches (37 Ar, be far less likely than 83m Kr or AmLi events to contain
accidental coincidences and wall backgrounds) and the multiple interactions close enough in space and/or time
choice of the backgrounds-driven fiducial volume (FV) to be edge cases in the pulse merging implementation.
for SR1; Section VI provides a conclusion. This suggested that the fraction of events producing SSs,
or the SS efficiency, should be faithfully reproduced in
simulations. Indeed, the consistency of fits to background
II. BACKGROUND SIMULATIONS data conducted with simulation outputs, as discussed in
Section III, implied good agreement in the SS efficiency
across all energies.
Detailed Monte Carlo simulations employing two in-
The background simulations include 1144 detector vol-
house software packages were used to estimate the back-
ume and radioisotope pairings. Results from a compre-
ground contributions from detector materials, xenon con- hensive radioassay and screening campaign of all detec-
taminants, and the laboratory environment. The first tor materials in Ref. [9] informed these fixed contami-
package, BACCARAT [12], contains the LZ detector ge- nants. The 238 U chain was broken into “early” and “late”
ometry, which has seen updates since the sensitivity stud- parts, separated by 226 Ra, since its long half-life of 1600
ies to closely match the as-built detectors. It tracks par- years would delay the re-establishment of secular equi-
ticles using Geant4 [13] and identifies their interaction librium within the chain following chemical processing of
points in the detectors. Energy depositions were recorded the materials. 232 Th was similarly split into early and
and passed to the second package, LZLAMA (LZ Light late chains at the level of 224 Ra.
Analysis Montecarlo Application). LZLAMA models the The simulations were designed to be as extensive and
detector response based on NEST (Noble Element Simu- as granular as possible. However, more materials were
lation Technique) [14, 15] and returns observables such as assayed than were represented in the BACCARAT ge-
S1 and S2 pulse size, in photons detected (phd), timing, ometry; the smaller components by mass and size, for
and reconstructed event location. which there was no direct corresponding volume in the
The TPC detector response model in LZLAMA was simulations, were proxied by nearby larger components.
tuned with tritium calibration data and verified with The simulation statistics of these larger components and
deuterium-deuterium (DD) neutron generator data, with
 4

their initial normalizations, as assumed for background events, and for the model fitting of higher energy detector
fitting, were accordingly scaled to account for the esti- gamma-ray radiation, where there was a need to ensure
mated activities of these smaller volumes. sufficient energy resolution up to the 2.6 MeV 208 Tl line.
Gamma rays from the cavern rock can penetrate the The xenon activation peak analysis in Section III B
water tank and irradiate the TPC to constitute an ex- was pursued with both SS and MS events. Since energy
ternal background. The simulation here relied on the resolution was not a concern for this particular study, the
multi-staged process detailed in Ref. [12], and the initial inclusion of MS events was justified as increased statistics
normalizations on the in-situ measurements described in were required for accurate event counting during later
Ref. [17]. Additionally, a range of xenon contaminants, time periods of SR1, when xenon activation isotopes had
activation lines, and neutrino fluxes were considered. largely decayed. In the case of MS events, in which an S1
is paired with two or more S2s, Equation 1 instead used
the summed total of the S2c pulse sizes, and the S1 was
III. GLOBAL TPC BACKGROUND corrected by calculating the position correction weighted
DETERMINATION by the pulse sizes of each S2.
Exceptions to Equation 1 were made for a selection of
To provide a global overview of the backgrounds the radon-chain alpha analyses and the cavern gamma-
present in SR1, the full dataset was analyzed in recon- ray studies. In these cases, energy was calculated using
structed energy space to identify and constrain features S1 information only. This was necessary for the latter
particular to certain backgrounds. These include mono- analysis since it was performed in a xenon gas environ-
energetic peaks from total gamma-ray or alpha energy ment prior to the application of electric fields, when no S2
depositions, as well as more complicated spectra with su- signals were available. For the radon-chain alpha stud-
perposed beta and gamma-ray contributions. Fits were ies, S2 information was often difficult to reconstruct due
either conducted with template shapes to these spectral to the interaction location or topology, and many events
features, or with spectra simulated using the framework of interest were not classified as SS or MS. In this case,
outlined in Section II. an S1-derived energy scale was found to provide adequate
The majority of these fits were conducted with SS resolution for alpha events, given the large S1s associated
events, as this was the classification focused on for the with the O(MeV) alpha energy depositions.
SR1 WIMP search and consequently the one for which Different energy ranges and volumes were explored
the event reconstruction and analysis cut efficiencies had to best constrain specific backgrounds. The radon al-
been well vetted [7]. SS event energies were reconstructed pha peaks were examined in an otherwise background-
using poor region above 3 MeV using both S1-only and S1+S2-
based analyses (Section III A). Separately, the identifi-
  able xenon activation peaks were isolated in an interme-
S1c S2cbot diate energy range of 200-450 keV (Section III B). These
E=W + , (1)
g1 g2bot and internal backgrounds, including the beta spectra
from 214 Pb in the 222 Rn chain and the 212 Pb from the
where a W-value of 13.5 eV [18] was assumed, the c 220
Rn chain, whose rates were informed by the radon
denotes the S1 and S2 pulse sizes have been corrected
alpha fits, were then constrained in an inner one-tonne
for position, via the processes described in Ref. [7], and
volume in the 80-700 keV range (Section III C). Gamma-
the subscript “bot” signifies the quantity was derived us-
ray backgrounds were examined in commissioning data,
ing only light collected in the bottom photomultiplier
when cavern gamma rays were expected to be domi-
tube (PMT) array. g1 was as reported in Section II and nant (Section III D 1). The gamma-ray contributions
g2bot = 14.89±0.48 phd/electron, the bottom array-only were then fitted with SR1 data in the 1-2.7 MeV region
equivalent of the charge gain. Bottom array-only S2s (Section III D 2), before a final global fit in the SR1 FV
were used since, when reconstructing events with energy was attempted, incorporating the internal backgrounds,
of O(100 keV) and higher, the large amount of light seen spanning 80-2700 keV (Section III E).
in the top array can cause amplifier saturation, which in
turn leads to an under-reporting of the S2 pulse size.
For fits that were particularly position-sensitive, extra
corrections were employed for the S1s and S2s to provide A. The Radon Alpha Region (>3 MeV)
an adequately homogenized response across the whole de-
222
tector for improved energy resolution. Correction factors Rn and 220 Rn emanate from primordial 238 U and
232
were derived by fitting the position dependence of the Th decay chains present in detector materials and
S1 and S2 signals produced by 5.49 MeV alpha particles dust, dispersing within the active liquid xenon volume.
from 222 Rn decays in the active TPC volume. The re- Radon and many of its progeny provide the only sources
sultant correction functions, smooth in all three position of alpha particles in the TPC. These alpha particles are
coordinates, are henceforth referred to as 222 Rn alpha- highly energetic, producing O(MeV) energy depositions;
based corrections. These correction factors were utilized they are also densely ionizing in liquid xenon, leading to
for the subset of radon-chain alpha studies that used SS high recombination yields for their interactions. These
 5

222
Rn
S1 pulse sizes were corrected by fitting a 3rd degree poly-
1 218
210
Po 214 nomial to each alpha S1 area-TBA band and normal-
Po Po
izing it to the vertical center of the detector. A TBA
Rate [mHz]

−1
10
212 cut was applied to remove events from the grids and be-
10− 2
Po
low the cathode, and all S1s above 30,000 phd were an-
10
−3 alyzed. The alpha peaks were simultaneously fit, with
222
Rn, 218 Po, and 214 Po modeled as the sum of double
10− 4 Gaussians, and 212 Po as a single Gaussian. Double Gaus-
10 5 6 7 8 9 10
222 Energy [MeV] sians were used to account for the observed skewness of
Rn 218
Po
1 some of the peaks as a result of the radially-dependent
Rate [mHz]

10 −1
212
light collection efficiency. The 212 Po alphas were often
Po
10− 2
216
Po 214
Po merged with 212 Bi betas, thus slightly skewing the peak
−3
to higher energies. However, given the low rate of 212 Po,
10
it was found a Gaussian fit worked sufficiently well to
10− 4
5 6 7 8 9 10
capture these 212 Po events. 210 Po was fit with a modi-
Energy [MeV] fied Crystal Ball function — a Gaussian core with a low
energy exponential tail as defined in Ref. [19] — to ac-
FIG. 1. Fitted radon alpha spectra in S1-linear calibrated count for alphas that lose energy before interacting in
energy. Top: All alpha events in the TPC are shown with only the xenon, having come from decays embedded in the
TBA cuts applied to remove excess grid and below-cathode TPC PTFE walls, and to account for events on the wall
alpha events. The 222 Rn, 218 Po, and 214 Po were modeled as surfaces that suffer from poor S1 light collection.
double Gaussians, 212 Po as a single Gaussian, and 210 Po as The classification-agnostic fit is illustrated in the top
a modified Crystal Ball function. Bottom: Alpha spectra for
panel of Figure 1. The results were used to populate the
SS events in the SR1 fiducial volume (see Section V A), where
position information from the S2 pulse was used to correct
“TPC Rate” column of Table I, having been normalized
the spectra, resulting in improved resolution. The sum of five by the mass of the fit volume, assuming it to be repre-
Gaussians was used to model the 222 Rn, 218 Po, 216 Po, 214 Po, sentative of the whole TPC. 216 Po and 220 Rn were not
and 212 Po peaks. included in this analysis as 216 Po was not well-resolved
in this space, and the 220 Rn contribution, which overlaps
218
Po, was too subdominant to be identified.
In the second, SS-only fit, S2 information could be ex-
two facts mean that radon alphas produce extremely
ploited to more accurately inform the position of the in-
large and readily identifiable S1 signals, far removed from
teraction. Whilst significant alpha populations that were
those of other backgrounds.
not classified as SS were lost from the analysis, a bet-
The analyses that follow rely on these S1s for investi-
ter correction of the pulse size as a function of position
gations of the Rn progeny. Two separate fits were per-
could be implemented for those still present. The 222 Rn
formed on S1 pulse sizes, each with different motivations
alpha-based corrections discussed in the introduction to
and merits: one which was agnostic to the classification
Section III was employed for the alpha bands to correct
of the alpha event, and another which used solely events the S1 size to the vertical and radial center of the TPC.
classified as SS. This improved the resolution of the alpha peaks and re-
The classification-agnostic fit had the advantage of be- vealed the presence of 216 Po, expected to manifest as a
ing sensitive to more populations than the SS-only fit SS the majority of time, that was previously obstructed
since, for many alpha decays, the event may not be SS- by alphas from other classifications. The alpha decay SS
classified. For example, alpha decays that occur on or spectra in the SR1 FV (illustrated in Figure 6 and de-
near the wall may lose much of their S2 signal due to elec- fined in Section V A) are shown in the bottom panel of
tron attachment on the polytetrafluoroethylene (PTFE), Figure 1, where a simultaneous fit with five single Gaus-
leading to misclassification. Bi-Po events, in which both sians was performed.
the 214 Bi (212 Bi) and 214 Po (212 Po) decays occur within To be able to use the results of the SS-only fit to quote
the event window, are complicated due to the possibilities activities for each radon-chain radioisotope, the chance of
of overlapping signals from the beta and alpha decays, its alpha events being classified as SS had to be taken into
and multiple scatters from the gamma-ray daughters of account. The SS efficiency was determined by identifying
the Bi decay. In these events, the top-bottom asymme- and fitting the given alpha band in S1-TBA space sep-
try (TBA) of the S1 signal, i.e. the ratio of the difference arately for each classification type, and then calculating
between S1 light collected by the top and bottom PMT the fraction of total events that were SS. The obtained
arrays to the total S1 light collected, can be leveraged. values are listed in the “Single Scatter Efficiency” column
Regardless of event classification, alphas produce distinct of Table I. Less than half of 212 Po, and a small fraction
bands in S1 area-TBA space, where the size of the signal
of 214 Po, was observed as SS due to the nearly instanta-
is primarily determined by the light collection efficiency
neous Bi-Po decay that produces overlapping signals.
as a function of depth in the detector.
Investigation of the SS events revealed that the posi-
In order to measure alpha rates in the full TPC, the
 6

8 8
140 140
7 7
120 120
6 6
100 100
5 5

µBq/kg

µBq/kg
Z [cm]

Z [cm]
80 80
4 4
60 60
3 3

40 40
2 2

20 1 20 1

0 0 0 0
202 302 402 502 602 702 202 302 402 502 602 702
R2 [cm2] R2 [cm2]
(a) Observed 218 Po Distribution (b) Observed 214 Po Distribution

8 8
140 140
7 7
120 120
6 6
100 100
5 5
µBq/kg

µBq/kg
Z [cm]

Z [cm]
80 80
4 4
60 60
3 3
40 40
2 2

20 1 20 1

0 0 0 0
202 302 402 502 602 702 202 302 402 502 602 702
2 2 2 2
R [cm ] R [cm ]
(c) Simulated 214 Pb Distribution (d) Simulated 214 Po Distribution

FIG. 2. The observed and simulated distributions of selected 222 Rn daughters. The 218 Po distribution in data was found from
SS events and is shown in panel (a); the 222 Rn distribution can be considered to be nearly identical to that of 218 Po. The
measured 214 Po distribution is shown in panel (b), for which the S1-S2 pair of 214 Po was extracted from each Bi-Po event to
reconstruct its position analogously to a SS event. Simulated 214 Pb and 214 Po distributions in panels (c) and (d), respectively.
The robustness of the toy model was validated from the agreement between the simulated and observed 214 Po distributions. The
FV boundary is shown with a dashed gray line on all panels, whereas for the observed distributions, two additional dotted lines
further separate the FV into upper, lower, and outer regions. These volumes are used in Table I to quantify the non-uniformity
of the observed alphas.

tion distributions of 222 Rn and its progeny in the TPC be contrasted to the distributions of later radioisotopes
were highly non-uniform. The degree of inhomogeneity in the 222 Rn chain, such as that of 214 Po (Figure 2b).
for 218 Po and 214 Po in the data can be seen in Figure 2, The tendency towards the bottom of the TPC was due
panels (a) and (b), respectively. To quantify the level to charged ion movement: radon-chain decays often pro-
of anisotropy, the SS-based fits were performed in each duce positively charged progeny that drift toward the
of the lower, upper and outer sub-volumes marked out cathode under the influence of the drift electric field [20].
in these sub-figures, with the results detailed in the last This charged progeny motion, coupled with the extant
three columns of Table I. inhomogeneity of the parent nuclei, defines the observed
The non-uniformity was the result of thermodynamic position distribution.
conditions and xenon flow in the TPC. A “slow-mixing” The incidence of charged progeny and their movements
inner region, which experiences a lower event rate from were characterized through the studies of 222 Rn-218 Po
radon-chain isotopes than the rest of the detector, was decay pairs. The observed decay pairs form vectors de-
attributed to the fact that the xenon mixing timescale scribing the motion of 218 Po, from its production to its
was longer than the half-life of 222 Rn. The observed dis- decay. Neutral 218 Po progeny move slowly with the liq-
tribution of 222 Rn alpha events was very similar to the uid, whereas positively charged 218 Po progeny move more
218
Po illustrated in Figure 2a, with the two essentially quickly, primarily along the drift field and towards the
identical with this bin resolution, since there is only a cathode.
3.1 minute half-life between the two decays. This can The charged progeny fraction of 0.46 ± 0.04 was cal-
 7

TABLE I. Table of measured alpha activities. Lower, upper, and outer sub-volumes are defined as shown in Figure 2. Fits in
the FV and its sub-volumes were performed with SS events and the rates are scaled to be representative of the full population
using the listed SS efficiency factor.

Radon TPC Rate Single Scatter FV Rate Lower Rate Upper Rate Outer Rate
Isotope [µBq/kg] Efficiency [µBq/kg] [µBq/kg] [µBq/kg] [µBq/kg]
222
Rn 4.78 ± 0.33 0.96 ± 0.03 4.62 ± 0.87 2.64 ± 0.60 3.38 ± 0.76 6.32 ± 1.33
218
Po 4.82 ± 0.34 0.98 ± 0.03 4.53 ± 0.84 2.64 ± 0.60 3.69 ± 0.83 6.26 ± 1.31
−3 −3 −4 −3
216
Po (8.2 ± 0.6)·10 0.56 ± 0.25 (4.69 ± 3.15)·10 (6.43 ± 4.39)·10 (2.48 ± 1.69)·10 (7.63 ± 5.18)·10−3
214
Po 2.65 ± 0.19 (1.14 ± 0.38)·10−3 2.07 ± 0.95 0.76 ± 0.45 1.09 ± 0.65 3.37 ± 1.99
−2 −2 −3 −2
212
Po (3.7 ± 0.3)·10 0.34 ± 0.08 (1.49 ± 0.73)·10 (5.72 ± 2.89)·10 (1.43 ± 0.72)·10 (2.89 ± 1.44)·10−2

culated as the ratio of pairs in which the 218 Po was ob- timate for 222 Rn activity and the values reported here,
served to drift downward to all 222 Rn-218 Po pairs. The given the observed position distribution of the 222 Rn, is
ion mobility of 0.242 ± 0.031 cm2 /(kV·s) was determined that the measured excess originates from the titanium
from the average velocity, having fitted to the distribu- cryostat. During the integration of the TPC, radon em-
tion of observed velocities along the field, considering the anation studies of the partially assembled detector were
strength of the drift field and the liquid xenon density. conducted, as described in Ref. [21], from which it was de-
These measured values for the charged progeny fraction duced that the cryostat contributes 17.2 ± 4.4 mBq. Pos-
and ion mobility agree to within one sigma of those re- sible explanations for this high emanation rate include
ported by EXO-200 [20]. contamination near or on the surface of the titanium,
Toy Monte Carlo mobility simulations using these val- introduced during intensive assembly and integration ac-
ues were created to understand the position distribution tivities; that radon has a much larger diffusion length in
of sequential radon-chain decays. Starting 218 Po posi- titanium than is typically seen in metals; or that, rather
tions were sampled from the distribution in Figure 2a, than being uniformly distributed throughout the tita-
and each subsequent ion was allowed to drift given the nium, radium has been concentrated near surfaces, which
assumed mobility and charged fraction probability. For would elevate the emanation rate into the liquid xenon-
these simulations, convective flow was ignored as it was filled cryostat from radium-decay radon recoils [22]. The
subdominant to charged progeny mobility. The simu- relative contributions of these effects to the in-situ mea-
lated 214 Po distribution is shown in Figure 2d. The drift surement reported here, and emanation from titanium in
model was validated against the observed 214 Po distribu- general, requires further investigation.
tion in Figure 2b, with the activity of simulated 214 Po The measured 210 Po rate can also be compared to that
within the fiducial volume found to be within 2% of the from LZ’s design goals for radon daughter plate-out and
measured value summarized in Table I. Given the good surface contamination. During the fabrication and as-
agreement between simulations and data, the toy mobil- sembly of detector components, 222 Rn daughters could
ity model was used to generate an a priori estimate for plate-out on detector surfaces, leading to long-lived 210 Pb
the position distribution of 214 Pb (Figure 2c). This was on (and embedded in) the TPC PTFE walls. 210 Po de-
used to evaluate the rate of 214 Pb decays in the FV, a cays can pose a low energy NR background for the WIMP
major background for the WIMP search (Section V B 1). search when the alpha is emitted into the PTFE, leav-
The 222 Rn activities found in these studies can be com- ing a recoiling 206 Pb nucleus to deposit O(100 keVnr )
pared to ex-situ estimates compiled during the construc- energy in the active xenon volume. These events can
tion phase of the experiment. In sensitivity projections, largely be removed via fiducialization (Section V A). Nev-
the 222 Rn decay rate was estimated to be 1.8 µBq/kg, ertheless, LZ instituted a target plate-out rate for 210 Po
or equivalently a total of 12.6 mBq in the active xenon of ≤ 0.5 mBq/m2 on the TPC walls and ≤ 10 mBq/m2
volume. To determine this value, a combination of room on other surfaces [5]. Using the previously discussed fit
temperature emanation measurements for several compo- of a modified Crystal Ball function, as shown in the
nents and literature values for components that had not top panel of Figure 1, the measured 210 Po rate was
been screened at the time of publication were used [11]. 2.32 ± 0.15 mBq in the TPC. This corresponds to an up-
Assumptions were made for this calculation about the fi- per limit of 0.35 ± 0.02 mBq/m2 on the TPC walls, as-
nal detector surface cleanliness, the performance of the suming this to be the sole origin of the 210 Po, well below
inline gaseous radon reduction system, and the expected the design requirement.
reduction of radon emanation from some materials and
dust particles at cryogenic versus room temperature.
One hypothesis for the tension between the original es-
 8

01-12 02-04 02-27 03-23 04-15 236 keV peak which heavily overlaps 127 Xe and thus was
127
Xe
127
Xe: fixed t 1/2 = 36.4 d
inferred from the difference of the 127 Xe peaks compared
107 131m 131m
129m
Xe
129m
Xe: fixed t 1/2 = 11.8 d with what would have been expected, having normal-
Xe Xe: fixed t 1/2 = 8.88 d
Rate [counts ⋅ tonne-1 ⋅ yr-1]

133
Xe
133
Xe: fixed t 1/2 = 5.2 d ized for their branching ratios. The 133 Xe spectra is de-
fined by the emission of a 346 keV endpoint beta decay
106
to ground state 133 Cs, which immediately relaxes via a
81 keV gamma ray. The 133 Xe rate in the SR1 FV was
Circulation measured by counting events in a “flat” section of the
105 Event
energy spectrum from 90–120 keV and and normalizing
to the full spectrum.
The observed rates in time are shown in Figure 3 along-
104 DD Calibrations
side fixed exponential functions with the measured half-
lives of 36.4, 11.8, 8.88, and 5.2 days for 127 Xe, 131m Xe,
129m
0 20 40 60 80 100 120
Xe, and 133 Xe, respectively [23–25]. The estimation
129m
Time Since Start of SR1 [days] for Xe broke down towards the end of SR1 when
the number of expected 129m Xe events was comparable
to the statistical uncertainties in the 127 Xe peak integral.
FIG. 3. Rates of activated 127 Xe, 131m Xe, 129m Xe, and 133 Xe Similarly, the counting of 133 Xe events became unreliable
as a function of time in 1.5 day bins during the SR1 exposure
when its rate was comparable to that of 214 Pb and 136 Xe
beginning 23 December 2021. The two hiatuses are due to the
mid-SR1 DD calibration, which resulted in additional neutron (see Figure 4 for reference), and therefore 133 Xe data
activation of the active xenon, and a circulation interruption below a rate of 2 × 104 counts/tonne/yr were excluded.
that led to a decrease in purity, which affected energy recon- As 133 Xe was almost entirely repopulated by neutron ac-
struction. Exponential trends for the measured half-lives are tivation from the mid-SR1 DD calibration, data points
superposed. immediately following this calibration were also removed
as a few days were needed for 133 Xe to uniformly mix
within the fiducial volume.
B. The Xenon Activation Region (200–450 keV) The steps up in rate for 131m Xe, 129m Xe, and 133 Xe af-
ter the DD calibration were expected, given the produc-
tion cross-section and abundances of the isotopes. The
Cosmic ray-induced activation of the xenon occurs dur- production rate of 127 Xe from neutron activation by the
ing its storage on the surface and transportation to the DD calibration was estimated to be an order of mag-
experimental site. The suppression of atmospheric par- nitude smaller than that of the metastable states, and
ticle fluxes by the rock overburden renders further cos- thus any rate increase is imperceptible given its high ex-
mogenic activation underground insignificant. However, tant concentration [26, 27]. This implies that the 127 Xe
given the time between the last delivery of xenon to rate can be safely extrapolated backwards in time and as-
SURF (31 August 2021) and the start of science data-
sumed to be of entirely cosmogenic origin: the activity at
taking (23 December 2021) in relation to the half-lives
the start of SR1 was calculated as 80.4 ± 9.0 µBq/kg. A
of the activation products, these radioisotopes were ex- 127
Xe activity of 76.9 µBq/kg at the start of SR1 was in-
pected to be prevalent in SR1. Moreover, neutron cal-
ferred from a measurement of the 127 Xe rate in LUX by
ibrations that were carried out before and during the
scaling the equilibrium value of 127 Xe according to the
science run resulted in further expected activation of the
duration and change in cosmic ray flux during surface
xenon.
transportation and underground storage. The details of
To understand the rates of the decays of 127 Xe, 129m Xe
this process are explained in Ref. [28]. A separate calcula-
and 131m Xe, and their time evolution over the SR1 ex-
tion of cosmogenic activation with the ACTIVIA simula-
posure, their mono-energetic peaks were analyzed in the
tion package [29] yields a starting activity of 52.1 µBq/kg.
SR1 FV with fits to the combined energy spectra con- Disagreements between ACTIVIA and experimental re-
structed from events labelled as either SS or MS. This sults for 127 Xe have been previously reported in Ref. [30].
dual classification selection was needed since in the case
of 127 Xe, which decays via electron capture, the de-
excitation gamma ray of the daughter 127 I can scatter
sufficiently far from the X-ray or Auger electron cascade C. The Beta Background Region (80–700 keV)
as to be identified as a distinct interaction site. The
127
Xe 375 keV gamma-ray interaction with 5.2 keV L- The SR1 data were examined in a central one-tonne
shell and 33.2 keV K-shell cascades was modeled with region of the TPC in order to directly inform the rates
the sum of two Gaussians plus a linear background. Sep- of sources internal to the xenon. This study aimed to
arately, the 164 keV peak of 131m Xe, along with the 127 Xe constrain beta radiation sources that could contribute
203 keV gamma-ray interaction with L-shell and K-shell to the WIMP search backgrounds, as well as reaffirm the
cascades, were modeled with the sum of three Gaus- xenon activation product observations from Section III B.
sians plus a linear background. 129m Xe manifests as a The central one-tonne sub-volume was defined as a right
 9

5 5
10 10
]

]
127 127
ee

ee
Rate [tonne-1 ⋅ year-1 ⋅ keV -1

Rate [tonne-1 ⋅ year-1 ⋅ keV -1

104 Xe Data 104 Xe Data
129m 129m
131m Xe 131m Xe
Xe
127 Model Xe
127 Model
10
3 Xe 10
3 Xe
125
Xe

102 133
102 133
Xe 136
Xe 214
Pb Xe 136
Xe 214
Pb

10 212
Pb 10 212
Pb

85 85
1 Kr 1 Kr
ν -ER ν -ER

0.6 0.6
Model-Data

Model-Data
0.4 0.4
0.2 0.2
Data

Data
0.0 0.0
− 0.2 − 0.2
− 0.4 − 0.4
− 0.6 − 0.6
100 200 300 400 500 600 700 100 200 300 400 500 600 700
Reconstructed Energy [keVee] Reconstructed Energy [keVee]

FIG. 4. Fit results for the SR1 exposure for the inner one-tonne region of the TPC. Data are shown in black and the summed
background model in purple. Left: Pre-DD calibrations Right: Post-DD calibrations.

cylinder of R ≤ 45 cm and 35 ≤ Z ≤ 95 cm, where the butions were minimal, whereas the lower energy bound
assumption was made that rates measured in this region was chosen to avoid the 124 Xe 2νDEC and WIMP effec-
were extendable to the entire TPC. The exceptions to tive field theory (EFT) search regions, where analyses are
this were 214 Pb and 212 Pb, which are known to be inho- in preparation [32, 33].
mogeneously distributed (Section III A), and the short- Two separate fits were performed to the SR1 data
lived DD activation product 125 Xe, which did not persist split by the mid-SR1 DD calibrations, using simulated
long enough to mix into the bulk xenon. Energy spec- energy spectra for all components other than for 127 Xe
tra of SS events in this volume were analyzed within the and 131m Xe, where Gaussian functions were adopted. A
80 to 700 keV region. The upper energy bound, along background-only version of the likelihood in Equation 2
with the volume definition, ensured gamma-ray contri- (Section V H) was used, and 125 Xe, which appears as a
new contribution following the DD calibration, was un-
constrained in the post-DD version of the fit. The re-
sults of the fits, and their statistical errors, in the two
TABLE II. Inner one-tonne volume fit results for SR1, deter-
mined separately for two periods, one before and one after
time periods are shown in Figure 4 and Table II. The
the DD-calibration. The errors shown in the table are the goodness-of-fit was assessed using the χ2λ,p prescription
statistical errors reported from the minimization. Reported in Ref. [34]. The number of degrees of freedom (NDF) is
rates are those extrapolated to the start of SR1 (23 Decem- defined as the number of data points minus the number
ber 2021). A 10% systematic in addition to these values can of fit nuisance parameters. The χ2 /NDF values for the
be assumed to account for uncertainties associated with the pre- and post-DD fits were found to be 216.72/136 and
exposure estimation and event reconstruction. 221.54/135, respectively. The fit results were found to be
Component Half-life Pre-DD Fit Post-DD Fit consistent in their reported activity of 214 Pb, despite the
[days] [µBq/kg] [µBq/kg] enhancement of shorter-lived activation products follow-
127 ing the DD calibration.
Xe 36.4 92.88 ± 0.38 89.65 ± 0.48
131m The rates of 212 Pb, 85 Kr, and solar neutrino scatters
Xe 11.8 18.87 ± 0.13 108.11 ± 0.74
129m
were not further constrained from these fits, given they
Xe 8.9 4.91 ± 0.23 193.04 ± 6.93 were sub-dominant components in this energy space. The
133
Xe 5.2 2.01 ± 0.11 1467.15 ± 22.21 214
Pb and 136 Xe two-neutrino double beta (2νββ) decay
125
Xea 0.7 - 26.70 ± 1.74 contributions were highly anti-correlated. However, the
214
Pb - 3.05 ± 0.12 3.10 ± 0.10 tight constraints placed on the 136 Xe rate in turn limited
212
Pb - 0.13 ± 0.01 0.11 ± 0.01 the 214 Pb result. The determined 214 Pb rate was further
136
Xe - 3.89 ± 0.18 3.96 ± 0.17 used for analysis of its contribution to the WIMP search,
85
Kr - (4.21 ± 0.42)·10−2 (4.18 ± 0.42)·10−2 with consideration of its position distribution and thus
how this one-tonne estimation should be scaled for the
a Note that 125 Xe has a 16.9 hour half-life and is only full TPC (Section V B 1).
measurable in the post-DD period [31]. With little time to
homogenize, its distribution was seen to be highly non-uniform
following DD calibrations, constrained to the upper third of the
TPC, therefore its one-tonne estimate is not representative.
 10

D. The High Energy Gamma Region polynomial was evaluated to normalize the pulse sizes to
(700–3000 keV) those at 0 TBA, which approximately corresponds to the
midway point between the top and bottom TPC PMT
Gamma rays from the decays of naturally occurring ra- arrays. A linear energy calibration was performed by fit-
dioisotopes in the cavern rock and detector components, ting the mean position in S1 space of each of the visible
as well as those from anthropogenic radionuclide decays peaks of the cavern gamma-ray spectra.
within the LZ assembly, can reach the TPC and interact To provide spectra to fit to the data, BACCARAT was
in the xenon. 238 U and 232 Th, and their progeny, 40 K and modified for the simulation of interactions under commis-
60
Co are the most prevalent: screening results for these sioning detector conditions. These changes can be sum-
decay chains were used as the basis of starting values for marized in four parts: 1) the target material of the TPC
the fits of these sources, as outlined in this section. The was changed to gaseous xenon (GXe) and the tempera-
contribution of gamma rays from the cavern walls was ture, pressure and hence density were set to the values
separately informed through a dedicated analysis under- measured during data acquisition; 2) the reflectivity of
taken during early detector commissioning, when it was the PTFE-lined surfaces inside the TPC were changed
expected to be the dominant background. These exam- to be representative of a GXe-PTFE interface [35]; 3)
inations did not include consideration of (α,γ) or (α,n) the water tank and OD were emptied; 4) a GXe NEST
interactions that could result in high energy gamma rays, model was used. The activity of the ∼149 tonne rock
more energetic than the 2.6 MeV 208 Tl gamma ray that is shell volume was set using previously measured activi-
encompassed by the upper energy limit of these studies. ties of 232 Th and 238 U decay chains and 40 K decay [17].
Geant4 particle and optical simulations of the cavern
Data Model rock gamma-ray background were then completed, fol-
1 40
K
40
K peak
238
U
lowing the prescription in Ref. [12]. Only gamma rays
232
Th
208
Tl peak
208
Tl double escape peak
with energy greater than 1 MeV were simulated to limit
the computational burden, given the prominent gamma-
10− 1
ray lines suitable to constrain to are above this energy.
Rate [Hz/keV]

To determine the contribution of each decay chain, the

10− 2
component energy spectra were fitted to the data. The
simulated spectra for 232 Th and 40 K were split into the
−3
10 Compton continuums and the gamma-ray photopeaks
and escape peak features, to be treated as separate com-
10− 4 ponents in the fit, as per the prescription followed in
0.6
Ref. [17]. Similar to that analysis, the photopeak to total
Model-Data

0.4 event ratios were required to remain within 20% of what

Data

0.2
0.0
− 0.2
they were before the simulated spectra were decomposed.
− 0.4
− 0.6
The final fit is illustrated in Figure 5, and a breakdown
1200 1400 1600 1800 2000 2200 2400 2600 of how fitted and predicted rates for each of the decay
Reconstructed Energy [keVee]
chains compare in Table III. The fitted rates listed con-
sist of the sum of the post-fit continuum and peak contri-
FIG. 5. Fit of the cavern gamma spectra to data taken dur- butions, e.g. the 232 Th total rate was constructed from
ing technical commissioning when the TPC was filled with
summing the Compton continuum, the 208 Tl photopeak
gaseous xenon and the water tanks and OD were empty. The
reconstructed energy scale is based on the observed S1 pulses, and the 208 Tl double escape peak results. The ratio of
using a linear energy calibration derived with the gamma-ray fitted and predicted rates in the case of all three spectra
photopeak signals. considered are approximately the same. Taken together,
a normalization of 0.65±0.14 on the total cavern gamma-
ray rate was required to scale simulation to match data.

1. Cavern Gamma Ray Measurement

TABLE III. Fitted and predicted rates of cavern wall radioac-

Cavern rock gamma ray normalizations were con- tivities, with the former derived from Figure 5.
strained via TPC measurements taken when the xenon
target was in gaseous phase and the OD and water tank Predicted Fitted Ratio
Isotope/ Rate Rate (Fitted/
were both empty, therefore providing no external shield-
Chain (Hz/keV) (Hz/keV) Predicted)
ing. The event rate in the TPC was thus dominated by 40
K 4.2 ± 1.1 2.79 ± 0.40 0.67 ± 0.20
the O(10 kHz) rate of the cavern gamma-ray energy depo- 238
U 3.9 ± 2.0 1.95 ± 0.53 0.49 ± 0.29
sitions: with no drift or extraction field applied, only S1 232
Th 6.1 ± 1.4 4.51 ± 0.43 0.74 ± 0.18
signals were observed. To correct for light collection ef-
ficiency differences in the detector, the variation in TBA
of the S1s was fitted with a 3rd order polynomial. The
 11

TABLE IV. Cumulative source activities for different group-

ings of sources used in the FV fit. The screening estimates
were found combining the material assay information given
in Ref. [9] for the components in each grouping, whereas best
fit numbers were derived from the FV fit in Figure 7. Only
components that have >1% of their decays create events in
the FV and contribute >1% of all estimated counts in the FV
are considered in the totals.

Isotope/ Region Screening Best fit [Bq]

Chain estimate [Bq]
Top 1.13 ± 0.11 1.05 ± 0.11
60
Co Side 1.18 ± 0.12 1.12 ± 1.02
FIG. 6. The simulated position distribution of all single scat- Bottom 0.81 ± 0.08 1.53 ± 0.19
ter ER events from detector components and cavern gamma Total 3.11 ± 0.18 3.71 ± 1.04
rays. Overlaid are the three sub-volumes (upper, lateral,
Top 7.63 ± 0.76 2.94 ± 1.66
lower) and the SR1 FV in which fits were performed. The in- 40
ner one-tonne volume, used for the fitting of internal sources K Side 2.56 ± 0.26 6.32 ± 0.61
in Section III C, is also shown. Bottom 6.54 ± 0.65 5.58 ± 2.19
Total 16.73 ± 1.04 14.85 ± 2.81
Top 0.28 ± 0.03 0.33 ± 0.29
232
2. SR1 Gamma Ray Background Fitting Th-early Side 0.66 ± 0.07 0.66 ± 0.49
Bottom 0.22 ± 0.02 0.23 ± 0.17
Assessing the gamma-ray background in SR1 was com- Total 1.16 ± 0.07 1.22 ± 0.59
plicated by the large number of possible contributions to Top 0.25 ± 0.02 0.11 ± 0.16
232
consider, their various geometric locations, and the po- Th-late Side 1.05 ± 0.10 2.57 ± 1.75
sition dependence of the signals across the TPC. Within Bottom 0.30 ± 0.03 0.32 ± 0.27
the contributions involving the same decay chain, vari- Total 1.59 ± 0.11 3.00 ± 1.78
ation was seen in the simulation outputs in the relative Top 2.37 ± 0.24 3.70 ± 1.80
peak heights of gamma rays within their energy spec- 238
U-early Side 1.99 ± 0.20 3.92 ± 1.53
tra due to the differing trajectories involved to reach the Bottom 1.86 ± 0.19 2.72 ± 1.40
xenon target. To simplify the problem, sources were
Total 6.21 ± 0.36 10.34 ± 2.75
grouped together as single components to be consid-
ered for the fitting if they were of the same radiogenic Top 0.84 ± 0.08 0.63 ± 0.30
238
origin and if their spectra were judged to be similar U-late Side 0.54 ± 0.05 3.01 ± 0.61
enough in shape, as would often be the case for isotope- Bottom 0.95 ± 0.09 1.28 ± 0.73
location pairs originating in close physical proximity to Total 2.32 ± 0.14 4.92 ± 1.00
each other. The metric for similarity was defined by a
Kolmogorov-Smirnov (KS) test statistic [36, 37], with any
sources below a threshold of 0.1 grouped together. As an
example, 40 K sources from the top TPC array PMTs, reducing and compartmentalizing the 1141 simulated de-
nearby thermometers and position sensors, and the tita- tector volume-decay chain pairs such that no more than
nium plate that they were all affixed to were combined a dozen parameters or groupings were fitted at any one
together in a single grouping. time in a given sub-volume fit.
The fitting was also broken down into separate sub- Fits were conducted in sequence beginning with the
volumes, with the idea of limiting the number of promi- lateral and ending in the upper volume, using the 222 Rn
nent components to be considered in one go, before a alpha-based corrections for the S1s and S2s used in re-
final fit using their combined inferences was attempted constructing energy, and a background-only version of
in the SR1 FV. Within each sub-volume, components the likelihood in Equation 2 (Section V H). The consid-
that did not satisfy the spectral similarity requirement ered energy range spanned from 1 MeV to 2.7 MeV in
and contributed fewer than 10% of events for a given order to eliminate contributions from sources internal to
isotope were deemed subdominant, and hence were fixed the xenon aside from 136 Xe 2νββ decay, and to capture
in their respective fits. Three sub-volumes were defined the highest common gamma-ray photopeak, the 2.6 MeV
as illustrated in Figure 6: a lateral volume to capture peak from 208 Tl. Resulting fit parameters from a given
contributions from the TPC and cryostat walls; a lower sub-volume were used as constraints for those compo-
and an upper volume to constrain the various compo- nents in successive fits. In the vast majority of cases, the
nents present in the vicinity of the respective TPC PMT exclusivity of the sub-volumes implies that the compo-
arrays. In all, these considerations were successful in nent was best fitted in a single region. This was typically
 12

Data Cavern gammas 60 Co 40 K

Rate [tonne −1 · year −1 · keV −ee1 ]
104 Fit Internals
136 Xe 2vbb
232 Th-early
232 Th-late
238 U-early
238 U-late

3
10
102
101
100
10-1
10-2
1
Model − Data
Data

1
0 500 1000 1500 2000 2500
Reconstructed Energy [keV ee ]

FIG. 7. Fitted detector component spectra in the SR1 FV, following the prescribed sequence of sub-volume fits. The fit results
are consistent with an average residual of approximately zero, with fluctuations at certain peaks arising from imperfections
in energy resolution matching between data and simulations. The contribution from internals is fixed here according to the
outputs of the central sub-volume fit. Gray shading is used to obscure data < 80 keVee to avoid inferences in the regions of
interest for future EFT and 124 Xe 2νDEC searches.

the case as the components normally included sources fit was not extended below 80 keV so as not to impinge
that originated from similar locations, leading to strong on the energy range and expose data being considered
position dependence of their events in the active xenon. for EFT dark matter and 124 Xe 2νDEC analyses.
The exceptions were the cavern gamma rays and sources All simulated components were collected into cate-
from the cryostat, which irradiate a wide region of the gories of side, top and bottom, based on their relative
TPC, and thus whose normalizations were seen to up- event rates in each of the three sub-volumes. This led
date from fit to fit. to 18 groupings: side, top and bottom for each of 60 Co,
40
The results of these fits were used in a final SR1 FV K, 232 Th-early, 232 Th-late, 238 U-early, 238 U-late. The
fit (Section III E), in which the most substantial conclu- results from the sub-volume fits were worked in as follows:
sions could be drawn about the gamma-ray background if the component had been previously constrained by a
contributions. sub-volume fit, its starting normalization was reweighted
according to that outcome, before it was combined with
other components. The uncertainties on all components
E. The SR1 FV Fit (80–2700 keV) in a group, from the sub-volume fit if appropriate, else
from assay measurements [9], were combined to be used
Parameters produced by the series of sub-volume fits as a single Gaussian constraint for that group in the fit.
in Section III D 2 were consolidated in a final fit within The results of the final SR1 FV fit are illustrated in
the SR1 FV. The SR1 FV was selected, as opposed to the Figure 7, showing the total fitted outcomes for each iso-
full TPC, for primarily two reasons: consistency with the tope or decay chain. Table IV reports the total assay-
SR1 WIMP search, which set a target mass in excess of estimated material activities for each of the aforemen-
what might be used in subsequent higher energy physics tioned 18 groupings, under the “Screening Estimate” col-
searches, and in an effort to minimize unavoidable posi- umn, and their fitted equivalents. To ensure the compar-
tion effects at the edges of the detector, such as charge isons are meaningful, only values for components that
loss due to field non-uniformities. To achieve a more were thought to be reasonably constrained by the fit were
comprehensive picture of our background contributions, accumulated. The two criteria to assess this were that
the lower energy bound of the fit was lowered to 80 keV > 1% of the decays of that component should result in
to encompass the rising edge of the detector components events in the FV, and that those events should comprise
and the internal sources, whose rates were fixed to the > 1% of all FV events
values previously obtained by the inner one-tonne fit in In examining position distributions of events prior to
Section III C. For the same reason as for that analysis, the fitting, evidence was seen of a potential mismodeling is-
 13

sue, with proportionally more events recorded towards 102

the side of the TPC than the top of it in simulations BG Data
10
compared to data, and vice versa. This potential ex-
AmLi Data
change of activity between side and top components may 1
be responsible for some of the differences observed in Ta-

Rate [Hz/phe]
ble IV, for example in the case of 40 K. The most direct 10−1
proof of this in the fit results was in the incompatibility 10−2
between reported scaling factors of the cavern gamma-
ray contributions, single components that irradiate the 10−3
full TPC, in the lateral (3.88 ± 0.80) versus the upper
10−4
sub-volumes (0.46+0.73
−0.46 ). The value reported in the up-
per volume is consistent with the value of 0.65 ± 0.14 10−5
for the GXe study discussed in the previous subsection;
10−6
for reference, the SR1 FV fit yields a scaling factor of 0 200 400 600 800 1000 1200 1400 1600 1800 2000
+1.55
0.72−0.72 . OD Pulse Size [phe]
The fit procedure was seen to preferentially raise the
activities of certain components with respect to the ini- FIG. 8. Total pulse size spectra in the OD for both SR1
tial estimates from radioassay measurements. The origin background data and an AmLi calibration of duration 7.4
of some of these rises is understood: the elevated rates hours.
of 60 Co in the bottom TPC PMTs are associated with
unintentional activation when a number of them were
stored nearby neutron sources. The fitted activities of tectors, where the uncertainty quoted is solely that from
238
U-early are amplified across all positional groupings. the measurement. Analysis is ongoing to optimize the
However, given their small contributions, it is likely that selection criteria and quantify their efficiencies in order
the fit is simply not as sensitive to these components as to determine the muon flux at the LZ location [39].
the constraints would suggest. The total 238 U-late contri- Muons do not themselves pose a problem for a physics
bution is also elevated compared to the prediction, which search. However, atmospheric muons that traverse the
may be partially explained by the quality of fit around rock surrounding the Davis Campus can produce ener-
the main gamma-ray photopeaks, where energy resolu- getic neutrons which impact the LZ detector. The at-
tion was more mismatched between simulations and data. tenuation of the cosmic ray muon flux by the rock over-
The χ2 /NDF for the fit is 13914.20/847, which can be burden implies at least a three order of magnitude lower
compared to the pre-fit version of 187741.23/847. The production rate of muon-induced neutrons than that of
quality of the fit implies that it can be used to attribute radiogenic neutrons in the rock [40]. On the other hand,
the observed counts to each decay chain, a feature useful muon-induced neutrons have harder energy spectra up
for future high energy searches. to GeV energies, and thus can penetrate the shielding
to reach the TPC. An earlier simulation study found
that LZ would see 1.4 ± 0.2 events in 1000 days from this
IV. BACKGROUNDS COUNTED IN THE SR1 source before analysis cuts were considered [12]. There-
EXPOSURE fore, muon-induced neutrons are not explored further
here.
A. Muons

Muons are readily identified in the data due to the high B. Neutrons
energy nature of their interactions. A preliminary study
selected candidate muon events in the TPC by mandating Radiogenic neutrons can arise from 238 U spontaneous
that there should be coincident signals in both the OD fission and (α, n) reactions on light nuclei in detector ma-
and Skin. Additionally, the signals in all three detectors terials. To look for these neutrons in the SR1 data, two
were required to meet a threshold on either the maximum facts of their interactions were leveraged: first, neutrons
or total integrated pulse size in the event. Applying these with MeV energies have a mean free path of O(10 cm)
cuts, 1061 candidate muon events were found. This cor- in liquid xenon, and thus will typically scatter multiple
responds to ∼12 muon events per day in the TPC, seen times, with distinctly separated interaction sites; second,
in all three detectors, assuming high muon selection ef- the majority of these neutrons are expected to scatter out
ficiency and that close to all events identified would be of the TPC and be captured and detected in the OD. An
muons. For comparison, current simulations, using the effective search strategy was therefore to limit the pre-
framework detailed in Ref. [12], combined with the most selection of events to those that were classified as MS and
recent measurement of the total muon flux at SURF by with correlated signals in the OD.
the Majorana Demonstrator [38], would indicate a rate of The signal spectrum in the OD was dominated by the
13.4 ± 0.4 muons per day which interact in all three de- Davis Cavern gamma-ray flux. A total pulse rate of 43 Hz
 14

4.75
140
4.50
120
log 10 (S2c [phd]) 4.25
100
4.00

Z [cm]
80
3.75
60
3.50
40
3.25
5 keV nr
25 keV nr

50 keV nr

75 keV nr

100 keV nr

125 keV nr

150 keV nr
20
3.00
2.75 0
0 50 100 150 200 250 300 0 10 20 30 40 50 60 70
S1c [phd] R [cm]
150
150
100 100

50
50
Y [cm]

Z [cm]
0
0
50
100 50
150
100
100 0 100 0 20 40 60 80 100 120
X [cm] R [cm]

FIG. 9. Locations of MS neutron events identified in the SR1 dataset, correlated across all three detectors. Top: Distribution
of the 10 identified neutron events in log10 (S2c)-S1c space overlaid with the MS NR band, as well as their averaged positions in
the TPC. White crosses denote three example events displayed in detail on the second row. Bottom: Chains of reconstructed
scatters demonstrating inter-detector coincidences in tagging neutron events. Working outwards: the red outline indicates
the SR1 FV; the gray curve highlights the TPC wall boundary in reconstructed space; the black box indicates the physical
edges of the active xenon volume; the teal profile denotes the liquid xenon Skin; the outermost green region represents the OD
acrylic tanks containing the GdLS. As the exact chronology of the event could not be determined, interactions were ordered
by drift time. Black circles denote the locations of the scatters with shortest drift time in the given neutron MS chain, with
empty circles showing the positions of other interactions in the TPC. Scatters in the Skin and OD are shaded in blue and green
respectively. Neutron captures in the OD are marked as a ∗, and resulting gamma-ray splashes observed in the Skin are labelled
with a pink cross. OD points are randomly assigned radial positions as XY reconstruction there is often biased towards the
centre, a correction for which is under development.

above 37.6 phe (∼200 keV) was observed throughout the not all the energy from the 7.9 MeV cascade of gamma
SR1 exposure, consistent within uncertainties with the rays released following capture on 155 Gd or the 8.5 MeV
sum of simulations based on cavern flux measurements from capture on 157 Gd is fully contained. Often, a few
performed with a NaI detector [17], and radioassay- of these gamma rays will travel back into the Skin and
normalized simulations of GdLS internal backgrounds TPC, causing responses in these detectors simultaneous
[41]. Figure 8 illustrates this spectrum, contrasted with with the OD capture signal.
an example spectrum from an AmLi neutron calibration. A further criterion was applied on the time delay be-
To enhance the probability of seeing neutrons over other tween the TPC interaction (the S1 observation) and the
backgrounds, a cut was made to select only events with OD capture. When entering the GdLS, neutrons are usu-
OD pulse sizes greater than 400 photoelectrons [phe]. ally well above thermal energies. Before capture becomes
This made use of two features particular to the neutron likely, they must thermalize through collisions with hy-
spectrum resulting from neutron captures in the GdLS: drogen in the GdLS. The time constant to capture in a
the 450 phe hydrogen capture peak, corresponding to the pure GdLS volume, with 0.1% Gd concentration, is 30 µs.
emission of a single 2.2 MeV gamma ray, and the con- However, as the OD is finite in dimension, a small frac-
tinuum followed by an end point around 2000 phe that tion of neutrons can find themselves in the acrylic of the
is attributable to captures on Gd. The continuum is OD tanks and the water surrounding the vessels and take
observed as, due to the thickness of the OD (∼60 cm), much longer to capture. Simulations have shown that the
 15

time-to-capture distribution contains a tail of up to a mil- up to O(TeV/c2 ). The SS classification efficiency, as it

lisecond [42]. For the WIMP analysis, where it was im- varies with energy in the WIMP ROI, was evaluated us-
portant to ensure a high neutron veto efficiency, a cut of ing neutron calibration data as discussed in Ref. [7].
< 1200 µs between the TPC S1 and OD signal was used The SR1 data were subjected to further analysis cuts
to reject events. For the neutron search detailed here, to improve the quality of the persistent data and to re-
a < 400 µs time separation selection was adopted to en- move events with identifiable background features within
compass the vast majority of possible timings and avoid this space. Veto anti-coincidence cuts preserved events
an influx of events with accidental coincidences between that did not see a response in the OD greater than
TPC and OD. 37.6 phe (200 keV) within 1200 µs after the TPC S1, mit-
The TPC signal region of interest for this search was igating neutrons, and those events that did not see a
restricted to 2.5 < log10 (S2c) < 5.5 and S1c < 500 phd. signal within ± 0.3 µs (0.5 µs) of the TPC S1 pulse in the
Neutrons interact with O(keV) energy depositions, which OD (Skin), that could be potentially created by gamma
will populate the lower end of this range, but the ex- rays associated with the TPC event.
tension to 500 phd covers the possibility of decays of A fiducial volume was implemented to combat wall
the short-lived states of 129 Xe (39.6 keV) and 131 Xe backgrounds that could be misreconstructed towards the
(80.2 keV) stimulated by inelastic scattering. In addi- centre of the TPC, and to remove external backgrounds,
tion, a subset of the S1 and S2 shape and parameter cuts whose position distribution was skewed towards the edges
developed in Section V F and relevant to MS events was of the detector. The FV definition was 86 < drift time <
applied to eliminate obvious accidental contributions. 936.5 µs and radius > 4.0 cm from the TPC wall in ob-
After all selection criteria and handscanning to confirm served space, with extra cut outs of radius > 5.2 cm for
event classification, 10 events were found and are illus- drift time < 200 µs and > 5.0 cm for drift time > 800 µs.
trated in Figure 9, demonstrating the excellent position Being strongly background-motivated, the FV choice is
reconstruction in the TPC and the correlations between discussed in Section V A.
all three detectors. Figure 9 also shows where events A suite of S1 and S2 pulse-based cuts were developed
are in log10 (S2c)-S1c space with respect to the MS NR to tackle pulse pathologies that typically manifest in ac-
band, which was determined with our calibration-tuned cidental coincidence events, in which unrelated S1s and
simulations, using the simulation-derived average multi- S2s would be falsely paired and classified as a SS event:
plicity of scatters for neutron events. One event is seen these are discussed in further detail in Section V F. In
clearly above the MS NR band at S1c of 279 phd, and addition, a series of live time vetoes was applied that dis-
is ascribed as a 129 Xe inelastic scatter event. Another carded high pulse rate periods, wherein the probability
is seen well below at S1c of 210 phd and is a candidate of such fake SS events was elevated due to the increased
“neutron-X” event, wherein the neutron, and likely the incidence of pulses that can mimic S1s and S2s formed
coincident gamma rays associated with the original re- by the pile-up of other pulses. Combined, these vetoes
action, has scattered in the S2-insensitive region below result in a reduction of the live time to an effective 60 ± 1
the cathode, enhancing the S1 signal relative to the S2s. days.
The OD neutron veto efficiency at a threshold of 400 phe After all cuts were applied, 335 events remained in the
and a TPC-OD delay time of < 400 µs was established SR1 WS dataset on which a statistical inference was per-
with AmLi neutron calibration analysis as 48.2 ± 0.2%. formed in log10 (S2c)-S1c space (Section V H). The anal-
Therefore, the total number of MS neutron interactions ysis therefore required information on the expected dis-
this observation implies were present in SR1, agnostic to tribution of each background in this observable space, as
veto response, is ∼20 events. The observed weak bias well as their rates. The former was obtained through sim-
of the events towards the bottom of the TPC is being ulations of each component (Section II), on which cuts
investigated. were applied in accordance with the data treatment. As
only S1 and S2 sizes and not pulse shape information were
simulated, acceptance functions were defined for pulse-
V. BACKGROUND IMPACT ON THE SR1
based cuts through studies on calibration data and were
WIMP SEARCH applied to the simulation outputs. Normalizations for
these probability distribution functions (PDFs) were in-
formed through analysis of the SR1 data: many of these
To perform the WIMP search (WS), a number of se- rely on more global observations of the backgrounds as
lections were made on the SR1 data to exclude non- outlined in Section III. These rates were characterized
candidate events, as described in detail in Ref. [7]. The outside of the WIMP ROI to avoid experimenter bias.
analysis was conducted using SS-classified events, in a Aside from the wall background rate, which was fac-
region of interest (ROI) of 3 < S1c < 80 phd, log10 (S2c) tored into the FV choice in Section V A, specific deriva-
< 5, and S2raw > 600 phd, where ‘raw’ specifically de- tions of the normalizations are described in Sections V B
notes the pulse has not been position-corrected. An ad- to V G, with the various categories of background cov-
ditional three-fold PMT coincidence requirement was im- ered in descending order of expected dominance in the
posed on the S1s. This ROI encompasses the expected WS background budget.
signals of WIMP interactions for masses of a few GeV/c2
 16

A. Fiducial Volume Definition and the Wall stant before gradually decreasing on approaching the de-
Background tector edge, where the reconstruction uncertainty smears
the reported positions. The wall position was defined
The development of the fiducial volume cut for SR1 where the reconstructed R2 distribution reaches half of
was highly driven by background considerations. Exter- the uniform density of the interior of the detector.
nal radiation is attenuated as it enters the TPC due to This procedure was repeated for drift time slices of
the self-shielding power of the xenon. Event rates in SR1 width 32 µs, and the full reconstructed wall position ob-
were therefore elevated towards the edges of the target tained by fitting these drift time bin values with an
and could be effectively mitigated through fiducializa- empirical function. The wall appears curved in recon-
tion. In the vertical direction, cuts were defined at drift structed coordinates because of the curvature of the un-
times of 936.5 µs and 86 µs, above and below which events derlying drift field that governs the XY positions re-
were excluded. These drift times correspond to heights ported for the S2s. No azimuthal dependence was seen
2.2 cm above the cathode at the bottom of the forward in the 83m Kr events, and therefore the observable wall
field region of the TPC, and 12.8 cm below the gate grid, position was defined as a function of drift time only.
respectively [6]. The lower drift time cut ensured the re-
moval of cathodic events in the active region of the detec-
tor. The upper drift time cut was motivated by the need 1.6
Empirical Fit
to remove SS events originating close to the liquid-gas in-
SR1 S2 threshold (600 phd)
terface, as well as those from the gaseous phase between 1.4
the anode grid and the top PMT array that had been
misreconstructed into the active volume. These cut def-

σ∆ r [cm]
1.2
initions were adopted to be equivalent to those assumed
for projected sensitivity studies in Ref. [11] and therefore 1.0
not further optimized for SR1.
Updated simulation-based studies using this lower drift 0.8
time bound confirmed an expected zero counts in the SR1
exposure of so-called “gamma-X” events. These events
0.6
are a potential WIMP background in which the gamma
ray also scatters in the S2-insensitive region below the 0 500 1000 1500 2000 2500 3000
cathode, adding to the S1 signals and producing SS- S2raw [phd]
classified events with more NR-like ratios between the S1
and S2. Given the centimeters of liquid xenon between
FIG. 10. Dependence of the radial position resolution of the
the cathode and the bottom edge of the fiducial volume, wall events on S2 pulse size. Events below the NR band and in
only gamma rays with energies O(100 keV) and greater the WS sideband region (80 < S1c < 500 phd) were selected
contribute to the gamma-X rate, causing this background to determine the radial position resolution of wall events, and
in the S1c < 80 phd region to be virtually non-existent. were observed to depend on the S2 pulse size based on the
The more critical dimension to determine was the ra- equation: σ∆r (S2) = √aS2 + b, where the fitting parameters
dial boundary because of the possibility of backgrounds a and b were 21.04 ± 0.65 and 0.22 ± 0.02 respectively.
physically outside the boundary being misconstructed
within it, due to the resolution of the S2-determined XY In establishing the S2 position resolution model, charge
position. As smaller S2s have worse position resolution, loss events below the ER band in the WIMP ROI adja-
this problem was exacerbated by charge loss at the walls. cent region of 80 < S1c < 500 phd were considered. It
Small, local field non-uniformities existing between the was verified that their distribution was centered on the
field cage rings that establish the TPC drift field prevent calculated wall position before their observed positions
the full extraction of the charge signals near the walls, were used to evaluate the position resolution for bins of
resulting in smaller amplitude S2s. different S2 pulse sizes. The results are illustrated in Fig-
Simulations of the drift field indicated that events up √
ure 10 and follow an expected 1/ S2 dependence. As the
to 3 mm away from the TPC wall could exhibit charge position resolution is independent of S1 pulse size, this
loss. Hence, not only events originating from the surface model was adopted to investigate the wall background
contamination on the PTFE could be impacted, but also rate as a function of potential fiducial radius within the
those of external gamma rays stopped within this outer WIMP ROI of S1c < 80 phd.
layer of xenon. The concern is if affected events have The radial fiducial volume cut and the S2 threshold
increased S1-S2 ratios such that they manifest in the NR were simultaneously optimized, with the end goal of pre-
band. serving as large a fiducial mass as possible whilst ensuring
The position of the TPC active volume wall in observ- that any potential wall NR background could be safely
able space for SR1 was defined examining 83m Kr calibra- ignored in a WIMP analysis. The estimated wall leakage
tion events with a reconstructed radius > 65 cm. Moving was therefore desired to be < 0.01 events.
radially outwards, the event distribution remains con- The study to set the radial boundaries used events in
 17

the sideband region of 100 < S1c < 500 phd that passed on the prescription in Ref. [44].
all data quality cuts used for the WIMP analysis, apart
from the S2 threshold. To account for the observed over-
densities of events at the top and bottom of the wall, 1. 222
Rn Chain (214 Pb)
three drift time bins were considered separately in the ra-
dial optimization: < 200 µs, 200–800 µs, > 800 µs. Gaus-
The 222 Rn chain contains beta-emitting radioisotopes,
sian functions were fitted to the populations in each of
which could generate low energy ERs in the WIMP ROI.
these bins in order to determine suitable options for these 214
Bi is not a concern as it is followed in quick succes-
radial cuts to obtain a 0.01 wall event count. The radii se-
sion by 214 Po, which has a half-life of 164 µs. Given the
lected were 5.2 cm, 4.0 cm and 5.0 cm away from the wall,
4.5 ms event acquisition window used in SR1, the prob-
respectively in each of the drift time bins, with an S2raw
ability of the 214 Bi beta signal being isolated in its own
threshold of 600 phd. These boundaries were adopted for
event window and being in the WIMP ROI is negligible.
SR1, along with extra cuts to remove events clustered 210
Pb has a 22-year half-life and thus will be effectively
within a 6.0 cm locus of the high-activity TPC field cage
removed from the bulk before decaying. However, it can
resistors. These latter cuts had minimal impact on the
plate out on surfaces. The possibility of 210 Pb and its
resultant fiducial mass, which was evaluated as 5.5 ± 0.2
tonnes. progeny leaching off of surfaces has not been investigated
An additional analysis, using the S2 position resolu- here. Our measurement of the 210 Po rate on the TPC
tion model, was conducted to corroborate these radial wall (2.32 ± 0.15 mBq, Section III A) is lower than that
cut choices, directly examining the WIMP ROI events to reported for LUX Run 3 (> 14.3 mBq) in Ref. [43] for a
evaluate the expected wall event leakage. Within each much greater surface area and target mass, indicative of a
of the three drift time bins, events below the NR band lower surface contamination rate. A first-principles mod-
(S2raw < 3000 phd) were further sub-divided into bins of eling effort aimed at quantifying low-energy grid back-
100 phd. The wall events within rwall − σ∆r were counted grounds revealed that the gate and cathode data were
for each S2 bin, where rwall was the value obtained for consistent with an origin primarily from the 210 Pb chain,
the empirical function defining the TPC active volume with an activity per grid area of 7.3±0.4 µBq/cm2 and
wall and the σ∆r was obtained from the S2 position res- 4.3±0.4 µBq/cm2 , respectively [45]. Again, these num-
olution model defined in Figure 10. For the SR1 radial bers are lower than reported for LUX, where no evidence
FV boundaries, this method calculated a wall leakage of of 210 Pb mobility was observed in Ref. [46]. Therefore,
0.015 events. the only contribution of 222 Rn chain decays considered
The radial fiducial cut will impact other backgrounds relevant to the SR1 WS ROI is that of 214 Pb.
214
that were not used to motivate its choice: decays near Pb undergoes “naked” beta decay with a branching
the FV edge where only partial energy is captured in the ratio of 11.0% [47]. This is a decay mode where no accom-
panying gamma rays are observed, and which therefore
xenon as associated gamma rays escape the TPC (“semi-
directly contributes to the WIMP search background.
naked” decays); the nuclear recoils of 206 Pb ions expelled
The beta decay topologies of 214 Pb are difficult to iden-
into the TPC from decays of 210 Po embedded on the
tify in situ, due to the lack of sharp spectral features
walls. Semi-naked decays of concern to the WS are those
in the beta decay continuum, thus constraints are de-
of 127 Xe and 214 Pb: the rates of these were constrained
rived indirectly through the alpha-emitting isotopes that
in Sections V E and V B 1, respectively. 206 Pb recoils fall
precede and follow the 214 Pb, and through energy spec-
under the selection of events discussed for charge loss
trum fitting. The 218 Po and 214 Po rate in the TPC were
and were not separately considered. Though once an
measured at 4.82 ± 0.34 µBq/kg and 2.65 ± 0.19 µBq/kg
important background in noble TPC-based dark matter
(Section III A), respectively, which bound the 214 Pb rate.
experiments [43], given the confirmation that the 210 Po
These were used to inform the fitting, which yielded a
PTFE surface activity met our construction requirement
rate of 3.10 ± 0.10 µBq/kg in a central, one-tonne volume
(Section III A), the expected 206 Pb NR background in
SR1 was < 0.003 counts. (Section III C). The distribution of 222 Rn was found to
be inhomogeneous in the TPC due to xenon circulation
patterns, and successive radioisotopes could be positively
ionized, drifting downwards towards the cathode. Sim-
B. Beta and Gamma Ray Backgrounds ulations of this effect were verified for 214 Po and used
to infer the position distribution of 214 Pb, illustrated in
Background contributions from beta and gamma-ray Figure 2c. Accounting for this result, the rate in the FV
sources produce flat spectra that are essentially indistin- was evaluated at 3.26 ± 0.09 µBq/kg.
guishable in the low energy WIMP ROI. For the statisti- “Semi-naked” events of gamma ray-accompanied decay
cal analysis detailed in Section V H, these contributions modes, in which the gamma ray(s) is not observed in the
were therefore combined into a single component. For TPC, were also considered. This estimate used the infor-
more accurate evaluations of the fraction of each beta mation gained from the study of this topology for 127 Xe
spectrum that is captured in the WIMP ROI, the spec- (Section V E), given the similar principal gamma-ray en-
tral shapes of 214 Pb, 212 Pb and 85 Kr were updated based ergies of the two decays. Conservatively integrating the
 18

predicted non-naked 214 Pb rate within 5 cm of the radial the experiment. The sampling results were consistent:
FV definition, the radius from which 127 Xe semi-naked accounting for both bottle measurements and residual
events were observed, and taking into account the re- xenon in the circulation system, the mean concentrations
ported 22% veto inefficiency for 127 Xe (Section V E) and were determined to be 144 ppq g/g nat Kr/Xe and 890 ppt
the mean free paths of the gamma rays in xenon, the g/g nat Ar/Xe, with a systematic uncertainty on the sam-
number of counts for SR1 is expected to be < 1. Thus pling rate of 15%. Additionally, periodic sampling during
the semi-naked topology was discounted in the estima- SR1 allowed for assessing the ingress rate of nat Kr and
nat
tion of final SR1 counts. The predicted WS counts for Ar, which was found to negligible for the short SR1
the naked decay mode-only in SR1, regarded as the total exposure. Given a 10.75 year (269 year) half-life and
contribution of 214 Pb, were found to be 164 ± 35, enfold- 2 × 10−11 (8 × 10−16 ) natural abundance [49], the 85 Kr
ing all sources of systematic uncertainty. (39 Ar) rate in the TPC was determined to be 42.3 nBq/kg
(0.876 nBq/kg). 39 Ar was excluded from the final back-
ground model since its contribution was extremely sub-
220
2. Rn Chain (212 Pb) dominant (<1 event in SR1).
The 85 Kr rate was further validated in situ by count-
The 220 Rn chain is analogous to the 222 Rn chain. 212 Bi ing coincident beta-gamma ray decays. 85 Kr undergoes
is again not problematic to the WIMP search as, given naked beta decay to 85 Rb with a 99.57% branching frac-
the 300 ns half-life of its daughter, 212 Po, it cannot be tion. In the other 0.43% of decays, 85 Kr beta decays to
present alone in an event. Aside from 212 Pb, the only metastable 85m Rb, with a 1.015 µs half-life, which subse-
other radioisotope that beta decays is 208 Tl, which does quently relaxes to 85 Rb via emission of a mono-energetic
not have naked decay modes. Thus only 212 Pb was con- 514 keV gamma ray. This delayed coincidence signal is
sidered. Concentrations of 220 Rn were expected to be unique to 85 Kr and provides a distinct event topology of a
much lower than 222 Rn as its shorter half-life impacts small beta S1 followed by a larger gamma-ray S1. Events
the amount that can emanate before it decays in the ma- with at least two S1 signals and at least one S2 signal were
terial [9], implying lower activities of 212 Pb compared therefore considered for this analysis. Strict analysis cuts
to 214 Pb. The suppression of the 220 Rn chain was con- were applied on the coincident S1 signal time separation,
firmed with the measured activities of 216 Po and 212 Po known beta and gamma-ray S1 sizes, and correlation of
versus their 218 Po and 214 Po counterparts in the 222 Rn the TBA of the S1 pulses, based on the expected signal
chain (Table I). Constraining the 212 Pb rate was compli- from the tuned detector simulation model. Considering
cated by the fact that 216 Po can decay upstream in the the combined efficiency of these cuts, the SR1 counts were
circulation system such that the sufficiently long-lived found to be 8.2 ± 4.1 events. This result is equivalent to
212
Pb progeny (10.6 hour half-life) can flow and mix in an average concentration of 136±69 ppq g/g nat Kr/Xe, in
the TPC prior to decaying. The 212 Po position distribu- strong agreement with the sampling results. For the SR1
tion was seen to be similar to 222 Rn, confirming similar WS, the 85 Kr rate derived from sampling measurements
mixing in the TPC. Assuming the EXO-200 β positive was used to determine the contribution to the WIMP
ion fraction of 76.4 ± 5.7% in Ref. [20] derived for the ROI as 32 ± 5 counts.
214
Pb to 214 Bi decay could be applied to 212 Bi, given our
previous confirmation of the α positive ion fraction in
Section III A, analogous flow calculations were performed 4. Detector and Cavern Gammas
to infer the 212 Pb rate given the observed 212 Po rate and
distributions. The 212 Pb rate was thus calculated to be Gamma rays originating from trace amounts of 60 Co,
0.137 ± 0.019 µBq/kg. Considering the naked beta decay 40
K, 238 U, and 232 Th present in the detector materials,
branching ratio of 11.9% [48], and assuming semi-naked in addition to those of 238 U, 232 Th and 40 K from the
decays to be negligible, the number of estimated SR1 cavern walls, contribute to the ER rate in the WIMP
counts was 18 ± 5. ROI and are collectively referred to as the Detector ER
background. With effective fiducialization, the Comp-
85 39
ton plateau contribution is expected to be small. The
3. Kr and Ar extensive simulations undertaken for each component of
the Detector ER background were scaled with the re-
Natural xenon contains trace amounts of the beta- sults from fits at higher energies, in which these contri-
emitting radioisotopes 85 Kr and 39 Ar, which are uni- butions are dominant (Section III D). For the estimation
formly dispersed throughout the liquid xenon and con- of counts, all cuts were applied to the simulation out-
tribute to the total ER background. Both isotopes have puts, including veto rejection. This is the only source for
concentrations that were greatly reduced via a charcoal which veto cuts were applied directly from simulations:
chromatography campaign at SLAC. The concentrations this may introduce a small systematic that has not been
of krypton and argon were measured at both SLAC and quantified or accounted for in the final uncertainty. The
SURF using a liquid nitrogen cold trap, sampling each number of counts in the SR1 WS ROI was estimated as
storage pack that was added to the xenon inventory of 1.4 ± 0.4.
 19

37 5.0
C. Ar

Ar Contour [counts/day]
4.5
37
Ar decays via electron capture to 37 Cl which atom- 4.0 Data
ically relaxes via K-shell (2.82 keV, 90.2%), L-shell 3.5
Flat
(0.270 keV, 8.9%), and M-shell (0.018 keV, 0.9%) cas- Flat+Exp
cades. 37 Ar is produced via cosmogenic spallation dur- 3.0
ing the storage and transit of the xenon from SLAC to 2.5
SURF. The rate of 37 Ar was estimated by calculating the 2.0

37
exposure of the xenon to cosmic rays before it was trans-

Event Rate in
1.5
ported underground, then correcting for the decay time
before the search began [50]. The final prediction for 1.0
37
Ar was 11 nBq/kg, on 23 December 2021 (the start of 0.5
SR1), producing an expected 96 events in the WS. This
0.0
was allowed to float between 0 and 288 events in the sta- 0 20 40 60 80 100
tistical analysis as the uncertainty on the spallation yield Days since the start of SR1
is about a factor of three.
Looking at the data in both the log10 (S2c)-S1c observ- FIG. 11. Time dependence of events in the 2σ contour for
able space and in reconstructed energy, there is a clear 37
Ar in log10 (S2c)-S1c space. (see Figure 15 Black points
peak of events in the expected region for the 37 Ar K- show the data. The orange band shows the best fit to a rate
shell (see Figures 14 and 15). In the fits to the WS ROI, constant in time, including systematic uncertainty from the
the best-fit number of 37 Ar events was 52.5+9.6 −8.9 (Sec-
fit, and the blue band shows the best fit to a rate constant in
tion V H). A post-fit analysis was undertaken to under- time plus an exponential decay with a 35 day half-life. The
stand whether the events in this region were consistent data and predictions are corrected for the live time in each
with the hypothesis of 37 Ar decay. bin.
To perform this analysis, the 85 WS events that
were within the 2σ contour of the 37 Ar location in the
log10 (S2c)-S1c observable space were selected (see orange D. Physics Backgrounds
contour in Figure 15). These events will include both
37
Ar events with a decaying rate and other ER back- Though interesting physics in their own right, 124 Xe
grounds, primarily from 214 Pb, that were constant in DEC, 136 Xe 2νββ decay, solar and atmospheric neutrinos
time. The WS time period was divided into 13 bins: four pose backgrounds in the WS ROI. Given the long half-
before the calibration period in January 2022, seven be- life of both isotopes, these backgrounds were modeled as
tween the calibration and the circulation event in March uniform in the TPC where each decay was assumed to
2022, and two after the circulation event. Two models be a single scatter ER without any interaction in the ve-
were fit to the data: one in which the rate of events in toes. The following subsections describe how each of the
this region is constant in time, and one in which there is component rates were calculated, using literature values,
one component with a decay half-life of 35 days, consis- and which energy spectra were used.
tent with that of 37 Ar, and a flat component to account
for the constant beta background. The model that is
constant in time has a best fit of 1.32 ± 0.14 events per 1. 124
Xe 2νDEC
day. The model with exponential decay has a constant
in time rate of 0.36 ± 0.22 events per day and a starting 124
Xe 2νDEC can occur with a combination of cap-
rate for the exponential model of 2.47 ± 0.61 events per tures from K-, L-, M-, and N- shells: the LL and LM
day. decay modes are background contributors in the WIMP
Figure 11 shows the best fit with uncertainty for each ROI, with total energies of 10.00 keV and 5.98 keV, re-
of these two models. The constant in time model has a spectively [51]. The first observation of 124 Xe 2νDEC was
χ2 /NDF = 32.58/12, which corresponds to a p-value of reported by the XENON1T experiment in Ref. [52]: the
0.0033, which is not consistent with the data. The model XENON collaboration has since made further measure-
with exponential decay has a χ2 /NDF = 13.25/11, which ments to place its half-life at (1.1±0.2stat ±0.1sys )×1022
corresponds to a p-value of 0.43, therefore consistent with years [53], and also estimated the branching ratios of the
the data. LL mode as 1.4% and the LM mode as 0.8%. With a nat-
This analysis indicates that in SR1, 37 Ar formed a sig- ural abundance of 9.52×10−4 [54], the activity of 124 Xe in
nificant part of the overall background, but is decaying the TPC was therefore predicted to be 8.72±2.44 nBq/kg.
away and will be substantially less prominent in future The log10 (S2c)-S1c response was evaluated using a modi-
searches. fied NEST response that includes the measured perturba-
tion of the L-shell contribution [55]. The resulting num-
ber of counts in the SR1 WS ROI was predicted to be
5.0 ± 1.4.
 20

136
2. Xe 2νββ Decay side of the WIMP ROI and occurring at a sufficiently
high rate to form a useful sideband. As only those L-
136
Xe 2νββ decay has been reported in multiple ex- and M-shell events which survive the FV selection were
periments [56, 57], and the half-life reported by EXO-200 of interest to the WS, the FV cut was applied to the
is 2.165 ± 0.016stat ± 0.059sys × 1021 years [58]. Given K-shell events to ensure no extra systematic was intro-
an isotopic abundance of 8.9% [54], the rate of 136 Xe is duced in calculating the veto efficiency. Candidate K-
4.14 ± 0.12 µBq/kg in the TPC. The underlying energy shell events were identified via a selection in the observ-
spectrum was also updated following the prescription in able log10 (S2c)-S1c space. Electron capture decays of 125 I
Ref. [44]. The final counts in the SR1 WS ROI were and 124 Xe, specifically the (35.5+4) keV 125 I gamma ray
evaluated as 15.1 ± 2.4. plus L-shell cascade and K- plus L-, M- or N- shell con-
tributions of 124 Xe 2νDEC, can leak into this selection.
Prior to fitting the veto-untagged populations in recon-
3. Solar Neutrinos structed energy space, the 124 Xe rate was fixed based on
the literature values discussed in Section V D 1. The fit
Low energy ER events are induced by solar neutri- was also performed for veto-tagged events: the ratio of
nos undergoing electroweak interactions with the active the semi-naked 127 Xe K-shell events determined in the
liquid xenon volume. The spectra are dominated by two fits was used to establish the veto efficiency in the
pp chain neutrinos, with small contributions from 7 Be fiducial volume at 78 ± 5%.
and CNO neutrinos. The low energy portion of the The fraction of semi-naked L- and M-shell events in
recoil spectra was calculated using the relativistic ran- the fiducial volume and WIMP ROI was determined us-
dom phase approximation (RRPA) to account for atomic ing the Monte Carlo simulations described in Section II
binding effects, as described in Ref. [59]. The recoil spec- and cross-checked against data. The rate of 127 Xe rel-
tra in the WIMP ROI are approximately flat and the rate evant to the WIMP analysis was ultimately calculated
was calculated based on measurements from Refs. [60– from the product of this factor with the SR1-averaged ac-
62]. 27.1 ± 1.6 solar neutrino ER events in the SR1 WS tivity and the veto inefficiency. The SR1 WS counts were
ROI were expected. thus estimated to 9.2 ± 0.8 events. For the 127 Xe PDF
Solar neutrinos can additionally interact via coherent used in the profile likelihood ratio analysis, the L- shell
elastic neutrino-nucleus scattering (CEνNS) in the case cascade response was simulated with modified recombi-
of 8 B and, to a much lesser extent, hep solar, diffuse nation fractions, as empirically measured in Ref. [55].
supernova (DSN) and atmospheric neutrinos [63]. The
prediction for the CEνNS rate is described in Refs. [60–
62], but was heavily suppressed in the SR1 WIMP ROI F. Accidental Backgrounds
due to the S2raw > 600 phd threshold. Only the domi-
nant 8 B contribution was considered, giving a prediction An instrumental background was induced by the ac-
of 0.14 ± 0.01 events in the SR1 WS ROI, with only a cidental coincidences of pulses identified as S1s and S2s
small flux uncertainty. that have unrelated origins, which were classified as SS
events by the LZ event reconstruction framework. These
127
so-called accidental events can mimic standard S1-S2
E. Xe pairs at a variety of energies, but predominantly those
of low energy recoils in the WIMP ROI.
Cosmogenically-activated 127 Xe was present during Pulses that contribute to accidental events can origi-
SR1 at an exposure-averaged rate of 32.8 µBq/kg, as de- nate from several regions in the detector. S1s unpaired
rived by the extrapolation of the activity determined at with S2s can arise from charge-insensitive regions of the
the start of SR1 through energy spectrum fitting (Ta- TPC, such as the reverse field region below the cathode.
ble II, Section III C). 127 Xe can be problematic for the S2s unpaired with S1s can come from regions with poor
WS if the de-excitation gamma ray(s) following the elec- light collection or from events where the S1 is below de-
tron capture are not detected. This can happen when tection threshold. Additionally, lone S2s can occur for
the 127 Xe atoms are located near the edges of the active interactions towards the top of the detector, where the
liquid xenon volume, where gamma rays can escape the short drift time means the S1 is subsumed within or can-
TPC, leaving behind semi-naked vacancy cascades that not be separated from the S2 pulse. Non-xenon processes
can fall into the WIMP ROI. This would be the case for can also yield pulses that can be misidentified as S1s and
5.2 keV L- and 1.1 keV M-shell cascades, which account S2s. For example, Cherenkov light in the quartz window
for 13.1 % and 2.9 % of all 127 Xe electron captures, re- of a PMT or dark noise pile-up from PMTs can appear
spectively [64]. as S1s, whilst electron emission from the cathode or gate
Most of these semi-naked events can be vetoed by the electrodes could be tagged as S2s.
Skin and OD systems through detection of the gamma Analysis cuts were developed to specifically target
rays that elude the TPC. Semi-naked K-shell events were many of the aforementioned sources of pulses classified as
used to determine the veto efficiency in SR1, being out- S1s and S2s. The cuts used properties of the individual
 21

pulses, such as their shape or the observed hit patterns of

the light on the PMTs, to distinguish them from typical
S1s and S2s. A subset of cuts also exploited the physical
relationship of certain parameters with the drift time es-
tablished for the given S1-S2 pair. In particular, the S1
TBA is expected to decrease with increasing drift time,
as more S1 light from events lower in the TPC is seen in
the bottom PMT array. Also, the width of an S2 pulse
is expected to increase with drift time, since the ioniza-
tion electron cloud has more time to diffuse as it drifts
towards the extraction region. Further cuts were devel-
oped to target events occurring near the liquid surface,
which showed a characteristic narrow width and square
shape, as well as events happening between the liquid
surface and the anode, in which case the S1 and S2 pulse
were often merged resulting in an atypical, elongated S2
pulse. FIG. 12. The accidentals PDF normalized to
A useful auxiliary dataset to study the accidental back- events/tonne/yr. The 10% and 90% quantiles (dashed)
ground was formed from Unphysical Drift Time (UDT) as well as the median (solid) of the ER and NR bands, as
events. UDT events have reported drift time exceed- reported in Ref. [7], are shown in blue and red, respectively.
ing the maximum value of 951 µs measured in SR1, and SR1 WS events remaining after all data selections are also
must therefore be formed by S1-S2 pairs that were not shown (black dots). The regions outside the WS ROI are
due to standard single interactions in the xenon and/or marked with a shaded grey area. The number of predicted
were not physically correlated, which would indicate that accidental events in the entire ROI is 1.2 (0.2 inside the
these events must be of accidental origin. Several statis- NR band). The top and right panels show the projections
tical tests were conducted to test the independence of on each axis of the ACS events surviving all analysis cuts.
The orange lines represent the functions that were used
the S1 and S2 variables in the UDT population and no
to build the two-dimensional PDF. Regions with larger
significant correlation was found, confirming that UDT data fluctuations, starting from the dashed gray lines, were
events were suitable for investigating accidental coinci- smoothed out.
dences. However, the number of UDT events in SR1 was
too small to comprehensively map out the distribution
of accidental events in the WIMP ROI, especially after
applying accidental-specific analysis cuts. reducing systematic uncertainties associated with finding
and classifying accidental events.
To model the accidental background, a data-driven ap-
To investigate the agreement between the UDT popu-
proach was taken, combining pulses at the waveform level
lation and the manufactured ACS population, the spectra
to manufacture artificial accidental events, called “Acci-
of the UDT and ACS populations in several dimensions
dental ChopStitch” (ACS) events. S1 and S2 pulses were
of interest were compared (e.g. pulse size, S1 TBA, drift
selected from Other classified events, as opposed to those
time). A KS test in each case indicated that the two
classified as SS or MS. These events were required to pass
distributions were not distinct enough to fail the test, re-
the SR1 live time vetoes and a subset of the WS data
quality cuts designed to remove noise-prone events. The inforcing the idea that ACS events were a good proxy for
UDT population that persists after WS cuts was domi- accidentals.
nated by events where a trigger was observed on the S2. 30 million ACS events were generated, with only ap-
Therefore, candidate S1s to create ACS events were re- proximately 22,000 events passing all the analysis cuts.
quired to be in the pre-trigger region, and candidate S2s This number of events were not sufficient to produce a
required to be straddling the trigger point of their re- smooth PDF across the WS ROI. Instead, the S1-S2 dis-
spective events. ACS events were manufactured splicing tribution was projected into its corresponding axes and
together the period before the start of the S2 from the an interpolating function was found for those regions at
candidate S1 event with the period including and after large pulse areas where the number of events was low.
the S2 from the candidate S2 event. The pulse environ- The two-dimensional PDF was constructed by taking the
ment in the vicinity of each pulse was therefore preserved, outer product of the smoothed out versions of the S1 and
which can affect the classification of the combined event. S2 projections. The resulting PDF is shown in Figure 12.
No further criteria were applied in the selection of the S1 The PDF was normalized using an independent calcu-
and S2 pulses, ensuring the eventual decision of whether lation of the accidentals rate. The expected number of
the resultant pair forms a SS lies entirely with the LZ accidental events in the WS ROI was calculated by mul-
event classification algorithm. One of the main advan- tiplying two quantities: the observed number of UDT
tages ACS events offer is that they can be processed as events after applying a basic selection of data quality
real data through the LZ event reconstruction framework, cuts that exclude periods of elevated TPC activity and
electronics interference but before applying any of the
 22

pulse-based cuts (310 events), and the rejection efficiency log-likelihood, was 0.0+0.2 .
evaluated on the ACS population after applying all the
analysis cuts (99.6% efficiency). This method combines
the best characteristics of each data sample: the UDT
population, which is anchored on an experimental mea-
Solar ν ER Det. NR 127 Xe Accidentals
surement, and the ACS population, which contains an 136 Xe 37 Ar β Decays & Det. ER
arbitrarily large number of manufactured events. The 4.50
9.8 keV 60 keV 13.4 keV
nr ee
predicted number of accidental events in the SR1 live- 45 keV ee
4.25 nr
time following this method was 1.2 ± 0.3. Both statis-
tical and systematic uncertainties were accounted for in 4.00
this calculation, with the dominant ones being the sys-

log 10 (S2c [phd])

tematic uncertainties arising from the difference observed 3.75
when S1 and S2 shape cuts were applied to either UDT
or ACS data. Nevertheless, an overall good agreement 3.50
was observed when the full list of analysis cuts was ap-
plied to both the UDT and ACS populations, resulting 3.25
in a similar rejection efficiency.
3.00 0.9 keV ee 2.9 keV ee 5.1 keV ee 7.4 keV ee
5 keV nr 15 keV nr 25 keV nr 35 keV nr
TABLE V. Expected and fit values for each contribution in 2.75
the fit to the OD-tagged event sample. The detector NR
0 10 20 30 40 50 60 70 80
S1c [phd]
population was left unconstrained.

FIG. 13. Events that passed all WS cuts, but were tagged by
Source Expected Events Fit Result
the OD (i.e. fail the WS OD veto cut) are shown in log10 (S2c)-
Solar ν ER 1.44 ± 0.03 1.43 ± 0.03
S1c space. Each data point is represented as a pie chart, with
Detector neutrons 0.8 0.0 ± 0.8
37 sectors representing the likelihood it originated from the given
Ar 2.9 ± 0.5 2.8 ± 0.5
136 background. 1σ and 2σ contours for background and signal
Xe 0.79 ± 0.12 0.79 ± 0.12 (neutrons in this case) are overlaid.
127
Xe 1.6 ± 0.2 1.6 ± 0.2
β decays + Det. ER 10.7 ± 2.6 11.3 ± 2.2
Accidentals 0.09 ± 0.03 0.10 ± 0.03
Total 18.2 ± 2.7 18.1 ± 2.4
A secondary calculation was made using the results in
Section IV B. To convert from the 10 observed MS events
to the number of SS events that pass all cuts, several fac-
G. Neutrons tors needed to be applied. Systematic effects were diffi-
cult to quantify for the initial neutron event observation
The constraint on the neutron background level for the and for some of these factors, thus uncertainties are not
WS was derived using an auxiliary fit to events which quoted here. The 48.2% OD veto efficiency established
failed the OD veto cut, but which otherwise passed all for an OD threshold of 400 phe and a coincidence window
selection cuts. The fit was performed analogously to that of 400 µs was unfolded and an 11.5% OD veto inefficiency
for the WS (Section V H), with the expected rate of each for the WS veto cut of 37.6 phe threshold and 1200 µs
background set to 5% of that determined for the WS sam- TPC S1-OD time separation was factored in to give the
ple. This 5% accounted for the chance of an accidental number of OD-untagged MS counts. Then a simulation-
OD-TPC coincidence that would cause the event to fail based MS:SS ratio of 2.3:1 was used to convert to SS
the OD veto cut, in which an unrelated, above-threshold counts. None of the calibration energy spectra closely
OD pulse would have occurred within the 1200 µs veto approximate those of radiogenic neutrons, thus the fac-
window following the TPC signal. An exception was tor was derived from simulations as it was expected to
made for 127 Xe, where OD coincidences were anticipated, be dependent on energy. Nevertheless, the number com-
and thus the expected value was set to 1.6 ± 0.2 based on pares favourably to those determined from DD (2.0:1)
its veto efficiency derived in Section V E. The OD-tagged and AmLi (1.3:1) calibrations, which were not as high in
events are shown in Figure 13. The best-fit number of energy as the radiogenic neutrons on average. Finally,
neutrons in this sample was 0.0+0.8 , as illustrated in the the survival fraction for FV and ROI cuts was calculated
fit results in Table V. The resulting shape of the log- from simulations and applied. After all these considera-
likelihood profiled in the amount of neutron background tions, 0.15 events were estimated for the WS events sur-
was well-characterized by a 4th order polynomial and was viving the OD veto cut, and 1.2 events were estimated for
used for the shape of the constraint on the neutron pop- the number of OD-tagged WS events. These values are
ulation in the WS fit. The predicted neutron background comparable to the 0.0+0.2 WS and 0.0+0.8 OD-tagged fit-
in the WIMP fit region, derived from the same profiled derived neutron background contributions, respectively.
 23

8B 124 Xe 85 Kr Sys. rate unc.

Solar ν ER 127 Xe Det. ER Sys. & stat. rate unc. TABLE VI. Number of events from background components
136 Xe 214 Pb Accidentals Data in the WIMP ROI in the 330 ± 12 tonne-days SR1 exposure.
37 Ar 212 Pb Total background The middle column shows the predicted number of events
102 with uncertainties as described in Sections V B to V G. The
uncertainties were used as constraint terms in a combined fit
of the background model plus a 30 GeV/c2 WIMP signal to
Events / keV ee / tonne / year

the selected data, the results of which are shown in the right
101 column. 37 Ar and detector neutrons used non-Gaussian prior
constraints and are totaled separately. Values at zero have no
lower uncertainty.

100 Source Expected Events Fit Result

214
Pb 164 ± 35 -
212
Pb 18 ± 5 -
85
Kr 32 ± 5 -
10-1 Det. ER 1.4 ± 0.4 -
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 β decays + Det. ER 215 ± 36 222 ± 16
Reconstructed Energy [keV ee ] ν ER 27.1 ± 1.6 27.2 ± 1.6
127
Xe 9.2 ± 0.8 9.3 ± 0.8
124
Xe 5.0 ± 1.4 5.2 ± 1.4
FIG. 14. Background model before fitting to the SR1 data 136
Xe 15.1 ± 2.4 15.2 ± 2.4
(except for the 37 Ar component, for which the post-fit nor- 8
B CEνNS 0.14 ± 0.01 0.15 ± 0.01
malization is used). The total model is shown in dark blue,
Accidentals 1.2 ± 0.3 1.2 ± 0.3
and SR1 data after all WS cuts have been applied are denoted
Subtotal 273 ± 36 280 ± 16
by the black points. This represents a background event rate
of (6.3 ± 0.5) × 10−5 events/keVee /kg/day.
37
Ar [0, 288] 52.5+9.6
−8.9
Detector neutrons 0.0+0.2 0.0+0.2
30 GeV/c2 WIMP – 0.0+0.6
H. Background Expectations Summary and Fit Total – 333 ± 17

The complete background expectations, as per the dis-

cussions in Sections V B to V G, can be found in Table VI.
Figure 14 illustrates these in reconstructed energy space
within the WIMP ROI, using the simulated events for
each component, scaling each as per its assessed contri- from the auxiliary measurements presented in this paper
bution in the table. This is with the exception of 37 Ar, (νb ).
which did not have a tight prior constraint (Section V C), For all components, except the 37 Ar and neutron back-
and for which the normalization derived from the follow- grounds, the constraints were Gaussian with standard
ing fit procedure was used. deviation corresponding to the systematic uncertainty
Events in the WS ROI which pass all data quality and on the expectation of each background. The 37 Ar back-
physics cuts were fitted using an extended, unbinned like- ground was constrained using a uniform distribution be-
lihood containing both signal (WIMP) and background tween 0 and 288 events, the latter being three times the
components: expected number of events in the exposure based on pre-
dictions for its production in the LZ xenon payload while
L(µs , θ ) =Pois(N0 |µtot ) on the surface [50] (Section V C). A constraint on the
N0 Nb
! number of neutron NR events was derived from a fit to
Y 1 X
× xe ) +
µs fs (x xe )
µb fb (x events tagged in the OD, as described in Section V G.
µ
e=1 tot (2) The results of the fit are listed in Table VI, and all 335
b=1
Nb
Y events passing data quality and physics cuts are shown
× fb (µb |νb ) , in Fig. 15. Pie charts are used for the events in the NR
b=1 band, showing how they have been attributed to different
PNb background components as a result of the fitting.
where µtot = µs + b=1 µb is the sum of signal and back- It can be seen from Table VI that, excluding the spe-
ground levels and e is an index which runs up to the cial case of 37 Ar, the likelihood fit does not provide bet-
total number of observed events, N0 . Both signal (fs ) ter constraints for the background components than our
and background PDFs (fb ) are functions of the analysis pre-fit assessments in the vast majority of cases. In other
parameters xe = (S1c, log10 (S2c)). The set of nuisance words, our background model, determined without using
parameters θ is the set of counts for each background the WS data, is consistent with the fit to the WS data,
component {µb }. Constraint functions, fb (µb |νb ), limit with better precision than the WS data alone can pro-
the value of each nuisance parameter to that expected vide.
 24

124 Xe 30 GeV WIMP Solar ν ER Det. NR 124 Xe β Decays & Det. ER

30 GeV WIMP Solar ν ER Det. NR β Decays & Det. ER
8B 136 Xe 37 Ar 127 Xe 8B 136 Xe 37 Ar 127 Xe Accidentals
Accidentals
4.50
9.8 keV 60 keV 13.4 keV 140
nr ee
45 keV ee
4.25 nr
120
4.00 100

Corrected z [cm]
log 10 (S2c [phd])

3.75 80
3.50 60

3.25 40

3.00 20
0.9 keV ee 2.9 keV ee 5.1 keV ee 7.4 keV ee
5 keV nr 15 keV nr 25 keV nr 35 keV nr 0
2.75 0 2 20 2 30 2 40 2 50 2 60 2 70 2 75 2
0 10 20 30 40 50 60 70 80
S1c [phd] Corrected r 2 [cm 2 ]

FIG. 15. Left: Low-energy data after all data quality and physics cuts in log10 (S2c)-S1c space. Contours enclose 1σ and 2σ of
the best-fit background model (shaded grey), the 37 Ar component (orange ellipses), a 30 GeV/c2 WIMP (purple dashed lines),
and 8 B solar neutrinos (shaded green regions). The solid red line shows the NR median, and the red dotted lines indicate the
10% and 90% quantiles. Right: Corrected z and r2 positions of the same data. Dashed black lines outline the active liquid
xenon volume and dashed gray lines represent the fiducial volume. In both figures, events falling below the 2σ contour of the
best-fit background model are shown as pie charts for which the size of each wedge is determined by the relative weight of each
of the background components in the fit.

VI. CONCLUSIONS and 127 Xe. Work is ongoing to understand the back-
ground events observed in the veto detectors themselves,
A backgrounds model for LZ was developed with analy- which can in turn further inform the TPC model.
sis of data from the first science run, and was successfully The background rate in the WIMP ROI was estab-
employed in the inaugural WIMP results reported by the lished as (6.3±0.5)×10−5 events/keVee /kg/day: this rep-
experiment in Ref. [7]. The pre-fit model agrees well resents a 57 times reduction over the background rate of
with the observed data in the WIMP ROI (Figure 14). (3.6±0.4)×10−3 events/keVee /kg/day reported by LUX
Sources outside the WIMP ROI were well-characterized, after their WS criteria were applied in Ref. [28]. This
which enables their inferred activities to inform addi- rate is likely to improve further as the components in the
tional physics searches in a broader ROI. Fitting of the model evolve with time. The cosmogenically-activated
gamma-ray sources and radon alpha decays was achieved xenon and 37 Ar will decay to subdominant levels. On
to good precision: gamma-ray activities were found to be the other hand, the state of the detector becomes more
compatible with assay expectations, whereas radon lev- variable with longer exposures, which could, for example,
els were found to be higher than expected. The results lead to enhancement of sources contributing to accidental
inform strategies for radon emanation and assay mea- coincidences. The WS ROI definition, cuts, and FV may
surements for future experiments. also change in the future, which would alter the back-
The model is comprehensive across a wide range of ground profile in consideration for a next WIMP analy-
energies, and can be easily adapted for other physics sis. A more sophisticated profile-likelihood ratio analy-
searches. The backgrounds in the higher energy regions sis involving more parameters, such as time dependence,
up and beyond the ROI relevant for 136 Xe 0νββ have could be developed to better utilize the background infor-
been characterized, ready for blinded searches with fu- mation detailed in this paper. Analyses presented here,
ture science runs. For WIMPs and low energy searches, such as the radon alpha movement studies, demonstrate
these background results can be used as the basis for un- that position dependence is also viable for a next physics
derstanding how to optimize the detector conditions for analysis.
later science data-taking: for instance, investigating how
to push analysis thresholds whilst maintaining a work-
able accidental coincidence rate. All three detectors were ACKNOWLEDGMENTS
leveraged in the determination of contributions to the
model, with the assessed veto performance in the OD
and Skin being instrumental to estimates for neutrons The research supporting this work took place in part
at SURF in Lead, South Dakota. Funding for this work
 25

is supported by the U.S. Department of Energy, Office of at Imperial College London and additional support by
Science, Office of High Energy Physics under Contract the University College London (UCL) Cosmoparticle
Numbers DE-AC02-05CH11231, DE-SC0020216, DE- Initiative. We acknowledge additional support from the
SC0012704, DE-SC0010010, DE-AC02-07CH11359, Center for the Fundamental Physics of the Universe,
DE-SC0012161, DE-SC0015910, DE-SC0014223, Brown University. K.T. Lesko acknowledges the support
DE-SC0010813, DE-SC0009999, DE-NA0003180, of Brasenose College and Oxford University. The LZ
DE-SC0011702, DE-SC0010072, DE-SC0015708, DE- Collaboration acknowledges key contributions of Dr.
SC0006605, DE-SC0008475, DE-SC0019193, DE- Sidney Cahn, Yale University, in the production of
FG02-10ER46709, UW PRJ82AJ, DE-SC0013542, calibration sources. This research used resources of the
DE-AC02-76SF00515, DE-SC0018982, DE-SC0019066, National Energy Research Scientific Computing Center,
DE-SC0015535, DE-SC0019319, DE-AC52-07NA27344, a DOE Office of Science User Facility supported by the
& DOE-SC0012447. This research was also sup- Office of Science of the U.S. Department of Energy un-
ported by U.S. National Science Foundation (NSF); der Contract No. DE-AC02-05CH11231. We gratefully
the UKRI’s Science & Technology Facilities Council acknowledge support from GitLab through its GitLab
under award numbers ST/M003744/1, ST/M003655/1, for Education Program. The University of Edinburgh
ST/M003639/1, ST/M003604/1, ST/M003779/1, is a charitable body, registered in Scotland, with the
ST/M003469/1, ST/M003981/1, ST/N000250/1, registration number SC005336. The assistance of SURF
ST/N000269/1, ST/N000242/1, ST/N000331/1, and its personnel in providing physical access and
ST/N000447/1, ST/N000277/1, ST/N000285/1, general logistical and technical support is acknowledged.
ST/S000801/1, ST/S000828/1, ST/S000739/1, We acknowledge the South Dakota Governor’s office,
ST/S000879/1, ST/S000933/1, ST/S000844/1, the South Dakota Community Foundation, the South
ST/S000747/1, ST/S000666/1, ST/R003181/1; Por- Dakota State University Foundation, and the University
tuguese Foundation for Science and Technology (FCT) of South Dakota Foundation for use of xenon. We also
under award numbers PTDC/FIS-PAR/2831/2020; the acknowledge the University of Alabama for providing
Institute for Basic Science, Korea (budget number IBS- xenon. For the purpose of open access, the authors
R016-D1). We acknowledge additional support from the have applied a Creative Commons Attribution (CC BY)
STFC Boulby Underground Laboratory in the U.K., the licence to any Author Accepted Manuscript version
GridPP [65, 66] and IRIS Collaborations, in particular arising from this submission.

[1] D. S. Akerib et al. (LZ Collaboration), Phys. Rev. C 102, M. Woods, JINST 6, P10002 (2011).
014602 (2020). [15] M. Szydagis, A. Fyhrie, D. Thorngren, and M. Tripathi,
[2] D. S. Akerib et al. (LZ Collaboration), Phys. Rev. C 104, JINST 8, C10003 (2013).
065501 (2021). [16] “Supplementary material for the LZ SR1 WIMP search
[3] D. Akerib et al. (LZ Collaboration), Phys. Rev. D 104, result,” https://tinyurl.com/LZDataReleaseRun1.
092009 (2021). [17] D. S. Akerib et al. (LZ Collaboration), Astroparticle
[4] D. S. Akerib et al. (LZ Collaboration), (2015), Physics 116, 102391 (2020).
arXiv:1509.02910 [physics.ins-det]. [18] M. Szydagis, G. A. Block, C. Farquhar, A. J. Flesher,
[5] B. J. Mount et al. (LZ Collaboration), (2017), E. S. Kozlova, C. Levy, E. A. Mangus, M. Mooney,
arXiv:1703.09144 [physics.ins-det]. J. Mueller, G. R. C. Rischbieter, and A. K. Schwartz,
[6] D. S. Akerib et al. (LZ Collaboration), Nucl. Instrum. Instruments 5 (2021), 10.3390/instruments5010013.
Meth. A 953, 163047 (2020). [19] S. Das, (2016), arXiv:1603.08591 [hep-ex].
[7] J. Aalbers et al. (LZ Collaboration), (2022), [20] J. B. Albert et al. (EXO-200 Collaboration), Phys. Rev.
arXiv:2207.03764 [hep-ex]. C 92, 045504 (2015).
[8] J. Heise, in 2022 Snowmass Summer Study (2022) [21] N. Chott and R. Schnee (for the LZ Collaboration), in
arXiv:2203.08293 [hep-ex]. Workshop on Low Radioactivity Techniques: LRT2022
[9] D. S. Akerib et al. (LZ Collaboration), Eur. Phys. J. C (2022) arXiv:2211.11857 [physics.ins-det].
80, 1044 (2020). [22] S. Lindemann, H. Simgen, and G. Zuzel, AIP Conference
[10] M. Arthurs, D. Huang, C. Amarasinghe, E. Miller, and Proceedings 1338 (2011), 10.1063/1.3579574.
W. Lorenzon, Journal of Instrumentation 16, P07047 [23] M.-M. Bé, V. Chisté, C. Dulieu, M. Kellett, X. Mougeot,
(2021). A. Arinc, V. Chechev, N. Kuzmenko, T. Kibédi, A. Luca,
[11] D. S. Akerib et al. (LZ Collaboration), Phys. Rev. D 101, and A. Nichols, Table of Radionuclides, Monographie
052002 (2020). BIPM-5, Vol. 8 (Bureau International des Poids et
[12] D. S. Akerib et al. (LZ Collaboration), Astroparticle Mesures, Pavillon de Breteuil, F-92310 Sèvres, France,
Physics 125, 102480 (2021). 2016).
[13] S. Agostinelli et al. (GEANT4 Collaboration), Nucl. In- [24] A. R. Yu. Khazov, I. Mitropolsky, Nuclear Data Sheets,
strum. Meth. A506, 250 (2003). Vol. 107, 2715 (2006).
[14] M. Szydagis, N. Barry, K. Kazkaz, J. Mock, D. Stolp, [25] J. Timar and B. S. Zoltan Elekes, Nuclear Data Sheets,
M. Sweany, M. Tripathi, S. Uvarov, N. Walsh, and Vol. 121, 143 (2014).
 26

[26] A. Naylor, Background studies for LUX and LZ: acti- tion in preparation,” (2022).
vation, gammas from rock and krypton removal, Ph.D. [46] K. C. Oliver-Mallory, Backgrounds in the LUX and LZ:
thesis, University of Sheffield (August 2022). Extending the Sensitivity of LUX to Low-mass Dark Mat-
[27] W. C. Taylor, Neutron Calibrations and Activation Sig- ter, Ph.D. thesis, University of California, Berkeley (Fall
nals in Low-Background Liquid Xenon Dark Matter De- 2021).
tectors, Ph.D. thesis, Brown University (May 2022). [47] S. Zhu and E. McCutchan, Nuclear Data Sheets 175, 1
[28] D. Akerib et al. (LUX Collaboration), Astroparticle (2021).
Physics 62, 33 (2015). [48] K. Auranen and E. McCutchan, Nuclear Data Sheets
[29] J. J. Back and Y. A. Ramachers, Nucl. Instrum. Meth. 168, 117 (2020).
A 586, 286 (2008). [49] V. V. Kuz’minov et al., Atomic Energy 73, 1010 (1992).
[30] L. Baudis, A. Kish, F. Piastra, and M. Schumann, Eur. [50] J. Aalbers et al. (LZ Collaboration), Phys. Rev. D 105,
Phys. J. C 75, 485 (2015). 082004 (2022).
[31] J. Katakura, Nuclear Data Sheets, Vol. 112, 495 (2011). [51] F. Larkins, Atomic Data and Nuclear Data Tables 20,
[32] “Observations of double electron capture in 124 Xe in LZ, 311 (1977).
publication in preparation,” (2022). [52] E. Aprile et al. (XENON Collaboration), Nature 568,
[33] “Constraints on effective field theory couplings in LZ, 532 (2019).
publication in preparation,” (2022). [53] E. Aprile et al. (XENON Collaboration), Phys. Rev. C
[34] S. Baker and R. D. Cousins, Nucl. Instrum. Meth. 221, 106, 024328 (2022).
437 (1984). [54] J. R. de Laeter, J. K. Böhlke, P. D. Bièvre, H. Hidaka,
[35] C. Silva, J. P. da Cunha, A. Pereira, V. Chepel, M. I. H. S. Peiser, K. J. R. Rosman, and P. D. P. Taylor, Pure
Lopes, V. Solovov, and F. Neves, Journal of Applied and Applied Chemistry 75, 683 (2003).
Physics 107, 064902 (2010). [55] D. J. Temples, J. McLaughlin, J. Bargemann, D. Baxter,
[36] A. Kolmogorov, Giornale dell’Istituto Italiano degli At- A. Cottle, C.E. Dahl, W.H. Lippincott, A. Monte, and
tuari , 83 (1933). J. Phelan, Phys. Rev. D 104, 112001 (2021).
[37] N. Smirnov, Bulletin Moscow University , 3 (1933). [56] A. Gando et al. (KamLAND-Zen Collaboration), Phys.
[38] N. Abgrall et al. (MAJORANA Collaboration), As- Rev. C 85, 045504 (2012).
tropart. Phys. 93, 70 (2017). [57] L. Si et al., (2022), arXiv:2205.12809 [nucl-ex].
[39] “Muon and muon-induced backgrounds in LZ, publica- [58] J. B. Albert et al. (EXO Collaboration), Phys. Rev. C
tion in preparation,” (2022). 89, 015502 (2014).
[40] M. Carson, J. Davies, E. Daw, R. Hollingworth, [59] J.-W. Chen, H.-C. Chi, C. P. Liu, and C.-P. Wu, Phys.
V. Kudryavtsev, T. Lawson, P. Lightfoot, J. McMillan, Lett. B 774, 656 (2017).
B. Morgan, S. Paling, M. Robinson, N. Spooner, and [60] M. Agostini et al. (Borexino Collaboration), Phys. Rev.
D. Tovey, Astroparticle Physics 21, 667 (2004). D 100, 082004 (2019).
[41] S. J. Haselschwardt, S. Shaw, H. N. Nelson, M. S. With- [61] B. Aharmim et al. (SNO Collaboration), Phys. Rev. C
erell, M. Yeh, K. T. Lesko, A. Cole, S. Kyre, and D. T. 88, 025501 (2013).
White, Nucl. Instrum. Meth. A 937, 148 (2019). [62] N. Vinyoles et al., Astrophys. J. 835, 202 (2017).
[42] S. Shaw, Dark Matter Searches with the LUX and LZ [63] D. Akimov et al. (COHERENT Collaboration), Science
Experiments, Ph.D. thesis, University College London 357, 1123 (2017).
(November 2016). [64] D. S. Akerib et al. (LUX Collaboration), Phys. Rev. D
[43] A. Bradley, D. Akerib, et al., Physics Procedia 61, 658 96, 112011 (2017).
(2015), 13th International Conference on Topics in As- [65] P. Faulkner et al., J. Phys. G 32, N1 (2005).
troparticle and Underground Physics, TAUP 2013. [66] D. Britton et al., Philos. Trans. R. Soc. A 367, 2447
[44] S. J. Haselschwardt, J. Kostensalo, X. Mougeot, and (2009).
J. Suhonen, Phys. Rev. C 102, 065501 (2020).
[45] “Grid radiogenic backgrounds in LUX and LZ, publica-