ABSTRACT

Title of Dissertation: STUDIES OF IONIZATION BACKGROUNDS
IN NOBLE LIQUID DETECTORS FOR DARK
MATTER SEARCHES

Eli Mizrachi (they/them)
Doctor of Philosophy, 2024

Dissertation Directed by: Dr. Jinkge Xu
Lawrence Livermore National Laboratory

Dark matter is believed to make up almost 85% of the total mass of the universe, yet its

identity remains unclear. Weakly Interacting Massive Particles (WIMPs) have historically been

a favored dark matter candidate, and dual-phase noble liquid time projection chambers (TPCs)

have set the strongest interaction limits to date on WIMPs with a mass greater than several

GeV. However, because no definitive interactions have been observed, the parameter space for

conventional WIMPs is highly constrained. This has sparked greater interest in new sub-GeV dark

matter models. At this mass scale, dark matter interactions with xenon or argon target media

may still produce detectable signals at or near the single electron limit. However, these signals are

currently obscured by delayed ionization backgrounds (“electron trains”) which persist for seconds

after an ionization event occurs. Electron trains have been observed in many different experiments

and exhibit similar characteristics, but their cause is only partially understood.

This work examines the nature of electron trains in various contexts, as well as possible



strategies to mitigate them. First, a characterization of electron trains in the LZ experiment is

presented, including new evidence of a dependence on detector conditions. The characterization

also informed the development of an electron-train veto for LZ’s first WIMP search, which set

world-leading limits on the spin-independent and spin-dependent WIMP-nucleon cross-sections for

medium and high-mass WIMPs.

Next, to complement the analysis of LZ data, hardware upgrades were performed in XeNeu,

a small xenon TPC at Lawrence Livermore National Lab. These included replacing plastics with

low-outgassing metal and machinable ceramic components, as well as a replacement of XeNeu’s

photomultiplier tube array with silicon photomultipliers. The resulting reduction in the intensity

of electron-trains and better position resolution from the respective upgrades will improve future

studies of low energy interactions and phenomena. Concurrent with this work, XeNeu was used

to perform a nuclear recoil calibration and a search for the Migdal effect, the latter of which can

substantially enhance an experiment’s low-mass dark matter sensitivity.

Finally, the development of CoHerent Ionization Limits in Liquid Argon and Xenon

(CHILLAX), is reported. CHILLAX is a new xenon-doped, dual-phase argon test stand that

has the potential to have a higher sensitivity to low-mass dark matter interactions and lower

backgrounds than current liquid xenon TPCs. The system is designed to handle high (percent

level) xenon concentrations in liquid argon that can enable a range of ionization signal production

and collection benefits. CHILLAX demonstrated the feasibility of such concepts by achieving a

world record xenon doping concentration with stable operation.



STUDIES OF IONIZATION BACKGROUNDS IN NOBLE LIQUID DETECTORS FOR
DARK MATTER SEARCHES

by

Eli Mizrachi (they/them)

Dissertation submitted to the Faculty of the Graduate School of the
University of Maryland, College Park in partial fulfillment

of the requirements for the degree of
Doctor of Philosophy

2024

Advisory Committee:
Professor Carter Hall, Chair/Advisor
Research Professor Anwar Bhatti
Assistant Professor Brian Clark
Professor Sarah Penniston-Dorland
Dr. Jingke Xu, Co-Advisor



© Copyright by
Eli Mizrachi

2024
Source Code: https://gitlab.com/emiz/thesis_public

https://gitlab.com/emiz/thesis_public


Acknowledgements

The phrase “standing on the shoulder of giants” always comes to mind when I think of the

people who have made my work and life possible. I’ve tried to be fairly comprehensive in who

I mention, but If I missed you, I apologize. I am incredibly fortunate in having been so well-

supported; there are many more people I’ve interacted with through the years who have left their

mark on me in some way.

I would like to thank Carter Hall, Anwar Bhatti, Brian Clark, Sarah Penniston-Dorland, and

Jingke Xu for serving on my dissertation committee. Carter, thank you for your support as my on-

campus advisor. When the time came for me to find a home in LZ, your assistance was instrumental.

Jingke, thank you for your unending technical guidance and financial support through my years at

LLNL. I’m grateful you granted me the opportunity to contribute to such a prolific group, which

has been responsible for so many critical measurements and interesting projects within the noble

liquid community.

Next, I would like to thank the rest of our group members at LLNL. Teal, thank you for your

mentorship within the LZ S2-only group and our lab. Joining LZ at the height of the pandemic

and MDC3 was an immense challenge, and I can’t imagine navigating my early analysis efforts

without you. Ethan, it was a pleasure to finally have a chance to work with you after only seeing

ii



you in passing at LBNL. I’ll always treasure our mutual appreciation for neat hardware hacks and

niche machining knowledge. To Jimmy (AKA “bigjim”), thank you for always stepping up to lend

a hand, whether on XeNu or CHILLAX. I’m proud to leave the lab in your hands. Godspeed, and

always remember that Gielinor Awaits. Same to your Alec. Speaking of which: Alec and Jianyang,

I look forward to seeing the fruits of your labor within the noble liquid group. To Rachel, I consider

myself fortunate to have been able to work alongside you at both SLAC and LLNL. Our lunchtime

conversations have been nothing less than illuminating, and you always seem to be asking the

questions I wish I had about detectors and lab management. Finally, I owe my sincere thanks to

Danny Naim. I cherished our time together at Berkeley, and when I was scrambling to find a group

in LZ to work with, you had my back in helping me join you at LLNL. I’m not sure how I can ever

repay the favor, but I’m sure someday I’ll be reading headlines about your success.

In no particular order, I would also like to thank some of the other folks in the LLNL RED

Group. Thank you to Sergey for your efforts as an architect of the XeNu high voltage system.

Thank you Tom for always being a friendly presence in our office before the pandemic made offices

irrelevant. To Sean and Angelique, thank you for helping create all of the parts our group has

needed over the years. It’s been enlightening hearing from both of you about your experiences, and

I have taken many of our conversations to heart about how to design components. To Felicia and

Tomi, you both exude a level of brilliance and care that I can only hope to someday match. I look

forward to remaining friends over the years.

Over the past 9 (!!) years, I’ve been fortunate to be connected to and contribute to LZ in a

variety of capacities. Starting at LBNL, I have Peter Sorensen, Kate Kamdin, and James Morad to

thank for helping me get off the ground with high vacuum cryogenic systems. To Lucie Tvrznikova,

iii



thank you for your help with COMSOL when I was struggling to build meshes and simulate electric

fields. I would also like to thank Mani Tripathi for hosting me during the summer of 2015 at UC

Davis, and Aaron Manlaysay for mentoring me during that time. Next, I want to give a shoutout

to Michael Williams, Yue Wang, Jose Soria, Vetri Velan, and Ryan Gibbons for joining me at the

Berkeley grad student beer hours. I’m looking forward to getting more chances to hang out with

you all and everyone else in the LBNL LZ group.

At SLAC, I have Dan Akerib and Tom Shutt to thank for their skills as hosts and chefs,

and for leading an incredible group of people that developed the LZ System Test and Krypton

removal system. My time in the group instilled values and principles in me that I have fiercely

clung onto. Of course, I would not have had that experience without contributions from the rest of

the group’s members. To TJ, I consider myself an expert in designing and configuring high vacuum

cryogenic systems because I managed to absorb a mere 1% of your knowledge. Someday, I might

also understand Crusader Kings to the same extent. To Kelly, thank you for getting LZ GETUP off

the ground. It’s one thing to be so instrumental to so many technical facets of LZ, but to be equally

instrumental to the wellbeing of the collaboration is a tremendous accomplishment that places you

above many of LZ’s members. It’s been empowering to see you become a rising-star physicist who

is also a force for good within the field. Ryan, I will forever curse your name for sucking me into

working on electron trains, but I still managed to enjoy our time together in the S2-only group.

You also have my eternal gratitude for not only appreciating my puns, but contributing your own.

Alden, I still struggle to grasp the magnitude of your contributions to LZ, and I can only hope to

fill a sliver of your shoes on the offline team. Tomasz, LZ would be a frozen pile of rubble if it

were not held together by your collaboration-spanning efforts and zip-ties. Christina, thank you

iv



for being such a gracious host over the years, for donating your liquor collection to me, and for

your boundless expertise in overseeing the development of the krypton removal system. Steffen,

thank you for deftly managing LZ’s software infrastructure. I also have you to thank for what is

possibly the best professional headshot of my life. Shaun, your cheerful presence IR2 was always

appreciated, and I always felt like I could approach you if I needed to bounce ideas off of someone.

Eric, I commend you for seeing the krypton removal process to its completion. Drew, thank you

for your friendship, muon veto, and hotspot veto. I always seem to notice you laugh harder than

anyone else at my jokes, and I really appreciate it. Ann, thank you for being such a great postdoc

to work with in the S2-only group. Without you, I would still probably be tearing my hair out

over tracking livetime following S2s. I can say the same for Micah, who I could always ping when

I had questions about python or problems at NERSC. Nicole, even though you were only with us

for a brief rotation, it was always a joy to see you in the lab and hang out with you after work. I

consider it a privilege that I was able to bring some of your ice cream dreams to fruition and share

them with everyone. Finally, I consider myself incredibly blessed to be rejoining the group under

Maria Elena Monzani, where I look forward to attaining a similar level of success and knowledge

with software and computing. I’m optimistic that our shared enthusiasm for best practices and

sense of humor will lead to a productive and memorable time for the benefit of LZ.

Within LZ, I have other members I would like to thank. Anh, I can’t think of a better person

to share a meal with. I treasure our friendship and the time we spent together in South Dakota, the

Bay Area, and Edinburgh. You’ve made important contributions to LZ, as well as to me personally.

Emily, Anna, Amar, and Tom: thank you for being such a great group to hang out with. I think

a lot about how we instantly hit it off, and I’m looking forward to future opportunities to do so

v



again. To Emily, thank you for your efforts to pick up the e-train veto where Ryan left off. Doing

so was a monumental task, and I’m honored to have been able to provide support where I could. I

am unabashedly looking forward to the chill vibes you’ll bring to the Bay Area. I would also like

to give special recognition to the electronics team on LZ. Dev, thank you for teaching me that a

good Manhattan should smell like “death and the French Alps”. Elise, I’m glad we can share in

our enthusiasm for mechanical keyboards. I hope I was able to repay a shred of your team’s efforts

to LZ by refactoring the run spreadsheet. To Georgia and Lindsey, it was a pleasure to meet you

both and have the chance to mentor you in South Dakota. You both show so much promise as new

graduate students, and I know your future advisors will be lucky to have you as students. Nicolas,

thank you for going out of your way to check in on me from time to time. Your sense of fashion

is iconic, as is the level of care you have for the students of LZ. Your sense of hospitality is also

difficult to top, and I appreciate you, Cosmo, and your roommates for welcoming me into your

home. Simran and Issy, thank you for the enlightening conversations about biscuits to go with

coffee during my trip to London. Aiham, I’m sorry we only briefly interacted, but I’m sure we’ll

have more opportunities in the future. Josh, I’ll never forget the time you told me “Don’t quote me

in your acknowledgements, I don’t deserve it.” Your humility is unmatched and I will always look

up to you for that. Makayla, I felt like we immediately became friends at the collaboration meeting

in South Dakota. One of these days, I’ll stop by Santa Barbara so I won’t have to rely on you for

pictures of the sunset down there. Harvey, thanks for being so enthusiastic about VSCode and my

efforts to make work in LZ less painful. I eagerly look forward to the new heights and depths you

continue to (literally) reach. To the underground team, especially Alex, Julia, Doug, and Gavin: it

was a blast (not the mine shaft kind) working with you and building the cabinets for the control

vi



room. I’m glad I was able to leave such a lasting impact on the hardware side of LZ. Lastly, thanks

to the S2-only group, especially Greg, Dan, Kelsey, and Sparshita for your efforts and feedback.

At UMD, I’d like to thank Anwar Bhatti for his efforts to get me started on analysis when

I barely knew how to use a terminal. I’d also like to thank Jon Balajthy for hosting me when I

visited UMD for the first time. Thanks also to John Armstrong for taking on the sampling system,

and to Tim Edberg for hosting poker nights and always being a friendly presence in the group.

To John Silk: thanks for being a great friend while I was at UMD, a great officemate, and letting

me siphon coffee from the Moccamaster, as well as your deep knowledge of RGAs. I’m especially

thankful to Chris Lobb, Fred Wellstood, and Sudeep “Serenity Now” Dutta for supervising me in

the production of transmon qubits during the summer of 2018. To Dave Buehrle, Matt Severson,

and the rest of the UMD Physics Education Research Group: teaching Physics 131/132 was an

incredibly rewarding experience. The skills I developed have been invaluable in my growth as a

mentor. I also want to acknowledge the friends I made at UMD who helped me navigate my first few

years in the program. Ramsey and Milena, thank you for helping me feel at home in the astronomy

department’s parties and events. Even when things seemed like they were falling apart, you both

always made me feel welcome and included. I’m grateful that we’ll be able to keep in touch, and

see where our careers lead us. Luc, I’m so glad we’ve been able to stay in touch! Discussing data

engineering over brunch in Niles is a tradition I would love to keep going. Your perspectives on

work outside of academia were some of the first I encountered, and they were pivotal in shaping

my current trajectory. To Sophie, Jamie, Donny, Anastasia, Jessie, and Sergio: we all parted ways

years ago, for our own reasons. It’s a little absurd that UMD didn’t work out for any of us, and

I’m admittedly still sore about that. Nevertheless, I still find myself caught up in my memories

vii



of sharing lunch and bonding in the PSC lobby with you all. Jamie, I cannot thank you enough

for being by my side throughout the pandemic, and shepherding me through the first few years (!)

of my transition. I’m sure someday we’ll be able to celebrate the energy you selflessly gave to me.

Sophie, it brings me so much joy and peace to know our friendship has endured in these intervening

years. More than anything else, the discovery I’m most proud of is the one you had the biggest

role in: discovering myself. You have my deepest gratitude for that, and so, so much more.

Coming out of the pandemic, I felt incredibly isolated in part because I had left my cohort

behind. Fortunately, I was able to find a home again at Berkeley (Go Bears!) through Igenspectrum.

For that, I’d like to give a huge thanks to Madge, Olive, and everyone else there for advancing the

gay agenda and providing a much-needed safe space for a queer physics graduate student like

myself. The rest of the graduate student community in the Berkeley physics department has also

been wonderfully open and welcoming. Being able to crash the occasional beer hour was crucially

uplifting when I was writing my dissertation, and I feel whole once again knowing that I’ve started

to make new friends (hi Hana!).

To the rest of my friends: I’m so happy you’re all a part of my life. Cheyenne, thank you

for including me in my very first pride month, powwow, and other celebrations. It’s always an

adventure when I get to intersect with your circles of friends. Eric, I’m so glad you’ve been around

these past few years. I envy your drive to work towards a better world, and I’m grateful that you

and Becky have been so close by for lunch, coffee, and support. Radha, your indomitable spirit

is second-to-none. I’m always taking notes on how you’re breaking cycles and blazing your own

path forward. The support and perspectives you shared with me throughout the pandemic were

a beacon of light in an otherwise difficult and isolating time. Lauren: staying with you, Dylan,

viii



and Lorraine was the perfect way to kick off life after my PhD. I’m not bothered being a continent

away, because I know we’ll have the rest of our lives to make up for the distance. I could say the

same for you, Marion. I’m so happy that in the span of a week in Amsterdam, we closed a gap

that was years apart. Nick, if I manage to get a hold of a paper copy of this dissertation I’ll make

sure to send it over to you, as long as you’ll pay for the shipping. To John and the rest of the

game night crew: every opportunity we had to get together was a breath of fresh air. I’m glad I

could always count on playing some great games in exchange for bringing some great snacks. Reina,

Nour, and Rimma: sharing a meal with you on my trips to San Diego has always been something

I’ve looked forward to when I needed to get away from work at home. Because of you, I cherish

every visit to Convoy Street more than the last. To the rest of my friends from San Diego: Zev,

Luz, Bryce, Quinn, Daniel, Andy, Katrina, and so many more! Each one of you has had an impact

on me throughout my PhD. You’ve kept me grounded in a way that physics hasn’t, and I’m happy

for every chance I get to reconnect.

There are too many Mizrachis to list all of you here, so for those of you not mentioned, please

know I appreciate how you contribute to our vast family. Howard, Pam, Naomi, and Jeff: I’m

grateful that we’ve been more in touch these past few years. I felt your encouragement every time

we’ve been around each other. I’ll be sure to look for more opportunities to take a detour to the east

coast, or just come and visit outright. Emily and Michael, thank you for being so accommodating

on my trip after my defense. Kelev is such a good boy and I’m happy I got to help take care of him.

Noah and Kat, Gabe and Sarah, Dan, Alissa, and Brooke: I’m glad I got to see each of you and

have you as my guides for New York. I can’t wait for the next time we get together for Shabbat.

Tia Laura, gracias por cada cena de Shabbat, hasta ahora y siempre. Yosef, Yael, and Grace: thank

ix



you so much for always having my back, from growing up together to making it through graduate

school. Tammy, thank you for being one of my strongest allies and always offering me beauty tips;

I’m glad I finally have the free time to try some of them. Tio Ralphie and Tio Victor: thank you

for all of your intriguing questions about dark matter, and giving me the chance to explain my

work. Tio Jamie, Diane, and Tia Shirley (in memoriam): every visit to Panama, to see you, Eli,

Leo, Danny, Jessica, and the rest of the cousins has been wonderful and memorable because of your

hospitality. Whether for a Seder, Rosh Hashanah, a wedding, or another occasion, a trip there has

always been a welcome respite.

Dan and Shelley, thank you so much for always welcoming me into your home on my visits

and stays near Stanford. Dan, I love being able to talk about my work with you and hear your

stories of growing up with Oma and Grampa. Both you and Shelley have helped me redefine my

idea of what my relationships towards my family and career will look like. Ava and Adele, I’m

always excited to see either of you when I get the chance. You both show so much promise in your

endeavors, and I can’t wait to see where they lead you. Dave: thank you for hosting me in SLO,

showing me your lastest plane builds, explaining HAM radio to me, and taking an interest in my

work. Sue, Robert, Jimmy and Sally (in memoriam): I’m also looking forward to seeing you all in

the future, whether I’m in Tucson, Washington, or Switzerland!

Finally, I’d like to thank my family. Mom, thank you for making sure I was fed, clothed, and

sheltered. Dad, thanks for all of your enthusiastic support over the years. You always gave me

the encouragement and financial support I needed to push myself to greater heights. Lori, thank

you for all of your thoughtful care, and willingness to talk. Every trip home is so much brighter

because of you, and it’s been a gift to spend time with the rest of your family as well. Ben, I’m so

x



glad we got to grow up together for a few years. You and Meghan are always a blast to be around,

and I’m looking forward to shredding some slopes with you both in the future. Nathan, Caitlin,

Lilah, Clyde and Jasper: my love for you is uncountably infinite. You’ve made Portland my second

home, where there’s no greater joy than planning our next meal over the one we’re having. I have

so much to look forward to in all of our lives, and the role I’ll be able to play in them as a sister

and an aunt.

xi



Table of contents

Acknowledgements ii

1. Introduction to Dark Matter 1
1.1. Evidence for Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2

1.1.1. Mass Estimates of Galactic Clusters . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.2. Galactic Rotation Curves . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.1.3. Cosmic Microwave Background (CMB) . . . . . . . . . . . . . . . . . . . . . 5
1.1.4. Big Bang Nucleosynthesis (BBN) . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.1.5. The Bullet Cluster Merger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.1.6. Structure Formation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

1.2. Dark Matter Candidates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.1. Standard Model Neutrinos . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.2.2. Axions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
1.2.3. Weakly Interacting Massive Particles (WIMPs) . . . . . . . . . . . . . . . . . 13
1.2.4. Sub-GeV Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17

1.3. Detection Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
1.3.1. Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3.2. Indirect Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.3.3. Direct Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22

2. Dual-Phase Noble Element Time Projection Chambers 33
2.1. Theory of Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.2. Signal Production . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36

2.2.1. Excitation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
2.2.2. Recombination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.2.3. Scintillation and Electroluminescence . . . . . . . . . . . . . . . . . . . . . . . 39
2.2.4. Recoil Discrimination . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.2.5. Recoil Energy Determination . . . . . . . . . . . . . . . . . . . . . . . . . . . 41

2.3. Lowering Detection Thresholds in Xenon . . . . . . . . . . . . . . . . . . . . . . . . . 45
2.3.1. CENNS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
2.3.2. S2 Only Dark Matter Searches . . . . . . . . . . . . . . . . . . . . . . . . . . 50

2.4. Low Energy Electron Backgrounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
2.4.1. Nomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
2.4.2. Overview of Photoionization, Grid Emission, and Electron Trains . . . . . . . 56

xii



3. Characteristics of Delayed Electron Emission in the LZ Experiment 67
3.1. The LZ Experiment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67

3.1.1. First Dark Matter Search Results . . . . . . . . . . . . . . . . . . . . . . . . . 68
3.2. Measuring Pulse Rates with Deadtime . . . . . . . . . . . . . . . . . . . . . . . . . . 75

3.2.1. Live, Dead, and Wall Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
3.2.2. Factors Affecting Deadtime . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79
3.2.3. Pulse Rates vs. Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
3.2.4. Repeated Trials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84

3.3. LZ Data Extraction and Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85
3.3.1. Anatomy of an Electron Train . . . . . . . . . . . . . . . . . . . . . . . . . . 87
3.3.2. ALPACA Data Extraction Algorithm . . . . . . . . . . . . . . . . . . . . . . 89
3.3.3. Livetime and RQ Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . 94

3.4. Electron Train Behavior in LZ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
3.4.1. Position and Time Dependence of SEs . . . . . . . . . . . . . . . . . . . . . . 102
3.4.2. Progenitor Size, Drift Time, and Electron Lifetime Dependence . . . . . . . . 108
3.4.3. Electron Loss Normalization . . . . . . . . . . . . . . . . . . . . . . . . . . . 112

3.5. Evidence of a Drift Field Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . 116
3.5.1. Electron Train Origins . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119
3.5.2. Dependence on TPC Region . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121
3.5.3. Previous Extraction Region and Drift Field Studies . . . . . . . . . . . . . . . 123
3.5.4. Rates in TPC Regions of LZ . . . . . . . . . . . . . . . . . . . . . . . . . . . 126
3.5.5. Fits to Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129
3.5.6. Varying the Extraction Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 135
3.5.7. Varying the Drift Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140
3.5.8. Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142

4. Experimental Studies with a Compact Xenon TPC 145
4.1. The XeNeu Detector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 146

4.1.1. Central Detector Components . . . . . . . . . . . . . . . . . . . . . . . . . . . 146
4.1.2. Detector Top Flange . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 154
4.1.3. Detector Cart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
4.1.4. Circulation Cart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 167
4.1.5. Data Acquisition and Slow Control . . . . . . . . . . . . . . . . . . . . . . . . 170

4.2. Physics Measurements with XeNeu . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
4.2.1. Measurement of High Energy Nuclear Recoils . . . . . . . . . . . . . . . . . . 171
4.2.2. Search for the Migdal Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173

4.3. Plastic Reduction Efforts in XeNeu . . . . . . . . . . . . . . . . . . . . . . . . . . . . 178
4.3.1. Overview of Outgassing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
4.3.2. Design of a Plastic-Free Bottom PMT Holder . . . . . . . . . . . . . . . . . . 182
4.3.3. Single Electron Trigger Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . 187
4.3.4. Electron Background Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 191

4.4. S2 Detection Improvements with Silicon Photomultipliers . . . . . . . . . . . . . . . 202
4.4.1. Introduction to SiPMs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 203
4.4.2. Replacement SiPM Array Design . . . . . . . . . . . . . . . . . . . . . . . . . 206

xiii



4.4.3. Electronics Re-cabling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 207
4.4.4. SPE Gain Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 218
4.4.5. Position Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222

5. Towards Lower Detection Thresholds with Xenon-Doped Argon TPCs 232
5.1. Benefits and Challenges of Xenon Doped Argon TPCs . . . . . . . . . . . . . . . . . 233
5.2. Operating Conditions for a Xenon-Doped Liquid Argon Mixture . . . . . . . . . . . 237

5.2.1. Eutectic Point Estimate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 238
5.2.2. Predicted Xenon Gas Concentrations . . . . . . . . . . . . . . . . . . . . . . . 243

5.3. CHILLAX Operational Concept and Thermosyphons . . . . . . . . . . . . . . . . . . 249
5.3.1. Theory of Operation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 250
5.3.2. Thermosyphon Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252

5.4. Evaluation of Hamamatsu VUV4 SiPM Liquid Argon Scintillation Photon Detection
Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 259

5.5. The CHILLAX Gas Sampling System . . . . . . . . . . . . . . . . . . . . . . . . . . 264
5.5.1. Introduction to Sampling Systems . . . . . . . . . . . . . . . . . . . . . . . . 265
5.5.2. Design and Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . 271
5.5.3. Commissioning and Calibration . . . . . . . . . . . . . . . . . . . . . . . . . . 277

5.6. Demonstration of High Concentration Xenon Doping in CHILLAX . . . . . . . . . . 295

Appendices 305

A. Information on Stripping Kapton Cables 305
A.1. Cable Assembly Procedures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310

A.1.1. Sub-C Connector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 310
A.1.2. MMCX Connector . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 311
A.1.3. D-sub Connectors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 314
A.1.4. Ring Terminal (Bottom PMT) . . . . . . . . . . . . . . . . . . . . . . . . . . 317
A.1.5. Sub-C Connector (FEP Cable) . . . . . . . . . . . . . . . . . . . . . . . . . . 318

B. Supplementary Information for XeNeu 320
B.1. Table of Slow Control Sensors and Instruments . . . . . . . . . . . . . . . . . . . . . 320
B.2. Lessons Learned . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321

B.2.1. The XeNeu Vacuum Box . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
B.2.2. CAD Software and Chain-of-Custody . . . . . . . . . . . . . . . . . . . . . . 323

C. THERANOS Procedures 324
C.1. RGA Stand P&ID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
C.2. XeNeu P&ID . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
C.3. RGA Calibration Procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324

C.3.1. Hardware Connections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324
C.3.2. Pressure Guidelines . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 325

C.4. Initial Vacuum . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
C.4.1. XeNeu Pumpout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327

xiv



C.4.2. RGA Stand Pumpout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 327
C.5. Gas Mixing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328

C.5.1. XeNeu Gas Panel Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 328
C.5.2. Xenon Addition to MV1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
C.5.3. First MV1 Expansion to MV2 . . . . . . . . . . . . . . . . . . . . . . . . . . . 329
C.5.4. Second MV1 Expansion to MV2 and Argon Addition . . . . . . . . . . . . . . 330

C.6. Sampling . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 331
C.6.1. MV2 Pressure Reduction to 10 torr . . . . . . . . . . . . . . . . . . . . . . . . 331
C.6.2. Priming RGA and LV1: MV2 Pressure Reduction to 1.5 torr . . . . . . . . . 332
C.6.3. Finishing a Scan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332
C.6.4. Finishing for the Day . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333

C.7. Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
C.7.1. Dilution Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 333
C.7.2. Expected Concentration - Double Expansion . . . . . . . . . . . . . . . . . . 333
C.7.3. Expected Concentration - Drain Refill . . . . . . . . . . . . . . . . . . . . . . 334
C.7.4. Calibration Factor . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 334

C.8. Recording Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
C.8.1. Recorded Parameters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
C.8.2. Rough Calculations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335

C.9. RGA Software Configuration and Setup . . . . . . . . . . . . . . . . . . . . . . . . . 335
C.9.1. Scan Speed . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 335
C.9.2. Electron Multiplier (CEM or CDEM) Operation . . . . . . . . . . . . . . . . 336
C.9.3. Pressure vs. Time Scans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337
C.9.4. Analog Scans . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 337

C.10.Using the Keithley 6510 for Pressure Readout . . . . . . . . . . . . . . . . . . . . . . 338
C.10.1.Pressure Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 338
C.10.2.Key Front Panel Features . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 339
C.10.3.Real-time Readout . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
C.10.4.Channel Scan . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 340
C.10.5. Scan Options . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
C.10.6. Saving Data from a Channel Scan . . . . . . . . . . . . . . . . . . . . . . . . 341

C.11.Parsing Scan Data with rgaplot . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 341
C.12.Summary of Recommendations for Future Development of THERANOS . . . . . . . 341

D. Historical Review of Electron Trains 344
D.1. 2008 - ZEPLIN-II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 344
D.2. 2009 - Burenkov et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 348
D.3. 2011 - ZEPLIN-III, XENON10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349

D.3.1. XENON10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
D.4. 2012 - Akimov et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 352
D.5. 2014 - XENON100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353

D.5.1. Photoionization Rates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 354
D.5.2. Rate Dependence on Purity and S2 Size, Uncorrelated Rates . . . . . . . . . 356

xv



D.6. 2016 - Akimov et al. (RED-1) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
D.6.1. “S3” Pulses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 358
D.6.2. Trapped SEs After Muons (Power Law) . . . . . . . . . . . . . . . . . . . . . 359

D.7. 2017 - Sorensen . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 361
D.7.1. Trapped Electron Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 362

D.8. 2018 - Sorensen and Kamdin . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 363
D.8.1. Trapped Electrons? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365

D.9. 2020 - Akimov et al. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 367
D.10.2020 - LUX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 368

D.10.1.Photoionization - Neutral Impurities . . . . . . . . . . . . . . . . . . . . . . . 369
D.10.2.Electron Bursts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 370
D.10.3.Delayed Backgrounds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 372
D.10.4.Photon Trains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375
D.10.5.Grids - Multiple Electrons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 375

D.11.2021 - PIXeY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 376
D.11.1.Pre-S1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 377
D.11.2.S1 Photoionization . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378
D.11.3.Post S2 and Post Drift . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 378

D.12.2021 - ASTERiX . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 380
D.12.1.Extraction Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 382
D.12.2.Drift Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
D.12.3.Drift Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 385
D.12.4.Infrared Photoionization of Impurities . . . . . . . . . . . . . . . . . . . . . . 385
D.12.5.Conclusions - Liquid Surface Electrons? . . . . . . . . . . . . . . . . . . . . . 385
D.12.6.Vibration Tests of ASTERiX . . . . . . . . . . . . . . . . . . . . . . . . . . . 387

D.13.2021 - XENON1T . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 387
D.13.1.Position Dependence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 388
D.13.2.Progenitor Size . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 389
D.13.3.Delay Time Since Progenitor . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
D.13.4.Extraction Field . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 391
D.13.5.Drift Time . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
D.13.6.Purity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 392
D.13.7.Conclusion - Liquid Impurities . . . . . . . . . . . . . . . . . . . . . . . . . . 393

E. Repositories and Resources 394
E.1. Repositories . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 394

E.1.1. This repository . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
E.1.2. LZ Gitlab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395
E.1.3. LLNL Gitlab . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 395

E.2. Software . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
E.2.1. CAD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
E.2.2. Zotero . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
E.2.3. VSCode . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 396
E.2.4. Inkscape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399

xvi



E.2.5. Notion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
E.2.6. Draw.io . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399
E.2.7. Python . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 399

References 402

xvii



List of Tables

1.1. Nomenclature for Equation 1.8 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

2.1. Light emission properties of liquid xenon and argon. Overall scintillation response
times are dependent on the proportion of states created with excitation and recom-
bination, which are themselves dependent on the concentration of impurities, initial
interactions, and applied electric field. . . . . . . . . . . . . . . . . . . . . . . . . . . 39

2.2. Table of electron train power law exponents from previous studies. . . . . . . . . . . 62

3.1. Table of child pulse rate normalization factors. . . . . . . . . . . . . . . . . . . . . . 102
3.2. Definitions of position-correlated and uncorrelated child pulses. . . . . . . . . . . . . 103
3.3. Coordinates used to determine bounds for averaging fields within the TPC. Regions

were chosen by eye to be representative of values in the bulk of each region. . . . . . 118
3.4. Grid voltages and corresponding nominal field values from simulations. For some

groups the cathode voltage difference was only nominally 28 kV, leading to slightly
varied drift fields of 197, 194, and 191 kV/cm when the gate voltage was adjusted.
For the sake of this analysis, these three drift fields are grouped together as nominally
194 kV/cm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 118

3.5. Table of runs, grouped by voltage settings. Tuples indicate consecutive sets of runs
e.g (9199, 9201) corresponds to 9199, 9200, 9201. LZ Users: see Fields Investi-
gation Data spreadsheet for more information and sv_SEDecayTime_lr_EdV_7_8_9_DdV_16_28_35.list
in ALPACA module [89] inputs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 119

3.6. Dependencies of fast and slow regions in a summed waveform of alpha events [99].
As an example, the amplitude of the fast component increased at a higher purity
and decreased with an increasing applied electric field. . . . . . . . . . . . . . . . . . 124

3.7. Table of electron train power law exponents from previous studies. . . . . . . . . . . 125
3.8. Table of values shown in Figure 3.48. A 15% systematic uncertainty has been added

to the errors for data taken at the 3900 and 3410 kV/cm (7 and 8 kV) fields to com-
pensate for uncertainties introduced in the choice of fit regions and cuts. Reported
field values for this work were lowered slightly from those calculated in Table 3.4 to
account for a lower potential on the gate grid as a result of deflection between the
gate and anode grids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144

xviii



4.1. DataFrame.head() of merged trigger and pulse dataframe. The join is performed on
event_id and event_timestamp which uniquely identifies events. peak_time is the
peak time of a trigger signal relative to a main event trigger, while npe, start_time,
and end_time are quantities belonging to a pulse in the same event. The first three
rows correspond to pairings of three different pulses with the same trigger pulse.
Subsequent steps detailed in the text explain how distinct pulses with triggers are
selected. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190

4.2. Key operating parameters for data taken during Runs 11 and 12. Run 11 had a
slightly smaller gas gap, leading to a higher field. . . . . . . . . . . . . . . . . . . . . 192

4.3. Fit values and uncertainty for each value shown in Figure 4.34. Uncertainties are
statistical only. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199

4.4. Fit values from Equation 4.2 applied to each rate curve shown in Figure 4.35. Fits
were performed in the range of 40 ms-1 s. Uncertainties are statistical only. Because
the fit does not account for the presence of a possible decaying exponential, the
magnitudes of 𝛽 may be elevated by ~10%. There is an additional ~10% systematic
error on 𝛽 in the Run 11 rates that stems from the choice of the left edge of the fit. . 200

5.1. Gas flow regimes as defined by their Knudsen number. . . . . . . . . . . . . . . . . . 265
5.2. Selected sampling system components . . . . . . . . . . . . . . . . . . . . . . . . . . 273
5.3. Tables of pumps studied for gas recirculation at low pressures. Helium leak rates

were independently checked and generally found to be in agreement with quoted
leak rates. Some pumps were modified from their stock configuration in an attempt
to lower leak rates. Of the three modified pumps, only the ACP-15’s modification
produced a substantial improvement by eliminating detectable leaks through the
ballast port. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 289

A.1. Strippers tested on AWG 30 Kapton coaxial cable. Note that stripper 6 has a
continuously adjustable blade which accommodates wires up to 0.125” in diame-
ter. Stripper 4 has the same operating mechanism as stripper 5, but uses custom
components for gripping and cutting Kapton insulation. . . . . . . . . . . . . . . . . 306

B.1. Table of XeNeu Sensors and Instruments . . . . . . . . . . . . . . . . . . . . . . . . . 320

C.2. RGA scan speed table from SRS [186]. Note how the scan speed and noise floor are
inversely related. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 336

C.4. Possible testing matrix. Rows are xenon concentrations in ppm. . . . . . . . . . . . . 342

D.1. Dependencies of fast and slow ROIs in a summed waveform of alpha events observed
by [99]. As an example, the amplitude of the fast component increased at a higher
purity and decreased with an increasing applied electric field. . . . . . . . . . . . . . 365

xix



List of Figures

1.1. Rogstad and Shoshak compared observed visible densities of galaxies with their ro-
tational velocities. A few years later, Albert Bosma demonstrated the flat rotation
curves of 25 different galaxies which added to the consensus surrounding dark matter
at galactic scales. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4

1.2. 2018 Angular power spectrum constructed from Planck CMB data [14]. The first
three peaks provide information on the critical density, baryon density, and dark
matter density respectively. The blue line is a fit from dark energy-dark matter
models, and the red points are observed data. The dotted gray line marks a change in
the scale of the x-axis. The CMB is extremely uniform, with temperature fluctuations
Δ𝑇
𝑇 ∼ 10−5 only visible at small scales. . . . . . . . . . . . . . . . . . . . . . . . . . . 6

1.3. Primordial abundances of light elements predicted by BBN, from [15]. Yellow boxes
correspond to observed abundances. The narrow vertical band corresponds to con-
straints from CMB measurements, while the wider band is a limit set by observations
of deuterium and 4He abundances. Relative abundances of 3He have yet to be reliably
measured, and the disagreement between the measured abundance of 7Li and that
predicted by CMB measurements and standard BBN is known as the cosmological
lithium problem. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

1.4. A comparison of optical and x-ray images confirmed that galaxies had passed through
each other relatively unimpeded, while the ICM had lagged behind due to higher
interaction rates. Additional microlensing calculations resulted in contours (green
lines) which increased towards the centers of the galactic distributions [18]. . . . . . 9

1.5. A simplified snapshot of constraints on the axion-photon coupling strength [36].
Other couplings are possible (e.g. axion-electron) and plots for those couplings can
be found at the same source. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13

1.6. A numerical solution of Equation 1.5 demonstrating the freeze-out process where
𝑌 = 𝑛𝜒/𝑠 and 𝑥 = 𝑚𝜒/𝑇 . Solid curves approach a relic density as the temperature
of the universe 𝑇 decreases. An increasing dark matter annihilation rate Γ (indicative
of a stronger SM coupling) lowers this relic density. The dashed curve tracks the
abundance of dark matter particles that remain in thermal equilibrium. From [40]. . 16

1.7. A numerical solution of the Boltzmann equation demonstrating the freeze-in process
where 𝑌 = 𝑛𝜒/𝑠 and 𝑥 = 𝑚𝜎/𝑇 . Solid curves approach a relic density as the
temperature of the universe 𝑇 decreases. An increasing dark matter production rate
Γ increases the relic density. The dashed curve tracks the abundance of SM particles
that remain in thermal equilibrium. From [40]. . . . . . . . . . . . . . . . . . . . . . 19

xx



1.8. Detection methods for particle dark matter rely on three distinct classes of particle
interaction; each are denoted with a corresponding arrow which signifies the flow of
time for that process [46]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

1.9. Observations of specific processes at the LHC can help set limits on various dark
matter models, and can be compared to results from direct detection experiments.
Furthermore, the model-independent nature of collisions grants the ability to sweep
larger mass ranges of parameter space in comparison to more targeted searches [48] . 22

1.10. Fermi-LAT limits on the dark matter annihilation cross-section from a gamma ray
survey of 15 dwarf spheroidal galaxies. Limits are set by searching for gamma rays
from the decay of products of specific DM annihilation channels. Observed spectra
are compared to simulated spectra, which are swept over a range of DM masses that
roughly correspond to WIMP models [52]. The thermal relic cross-section is shown
as a dashed gray line for the sake of comparison. . . . . . . . . . . . . . . . . . . . . 23

1.11. Prominent dark matter direct detection experiments, grouped by the detection tech-
nology used and measured signals. Single phase detectors and dual-phase time pro-
jection chambers (TPCs) utilize noble elements such as xenon or argon. Measuring
exclusively charge signals, DAMIC uses silicon wafers while NEWS-G uses a mix-
ture of neon and methane gas. Gas TPCs are constructed to search for a direc-
tional WIMP “wind”. Scintillating bolometers, cryogenic detectors, and scintillating
crystals all utilize crystals of various compositions to probe different WIMP mass
ranges. Each experiment under these categories has unique hardware suitable for
reading out the desired quantities. PICO and DRIFT are notable for using fluori-
dated compounds to set the best spin-dependent WIMP interaction limits. Figure
from [54]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24

1.12. The above plot shows simulated scattering rates for a 100 GeV WIMP with a spin-
independent cross section of 10-45 cm2 and the kinematic properties of a typical
thermal relic. At lower nuclear recoil energies, xenon is an excellent choice of target
material due to its high mass. However at higher nuclear recoil energies, scattering
becomes incoherent as a WIMP interaction would only probe a portion of the xenon
nucleus. Thus, lower recoil thresholds enable detectors to benefit from coherent
scattering amplification of the nuclear interaction cross-section. Circles on each
curve indicate typical detection thresholds. Taken from [58]. . . . . . . . . . . . . . . 28

1.13. Differential ionization rates versus electron recoil energy for different target materials
[59]. Notable here is that for events below a few hundred eV, the irreducible solar
neutrino background is not expected to meaningfully obscure signals. . . . . . . . . . 30

1.14. Recent limits set by WIMP direct detection experiments [60]. The neutrino fog is
where an irreducible background caused by coherent scattering of solar, atmospheric,
and supernova neutrinos (Section 2.3.1) is expected to become dominant. Searches in
this region require a 10𝑛 increase in exposure for in order to obtain a corresponding
factor of 10 increase in cross-section sensitivity. . . . . . . . . . . . . . . . . . . . . . 31

1.15. A chart highlighting the increasing levels of sensitivity reached by contemporary
WIMP search experiments. Different colors indicate different technologies used. Note
that some experiments pictured are more sensitive to WIMPs in a significantly lower
mass range. See [61] for the figure shown here and [62] for a previous version. . . . . 32

xxi



2.1. Mockup of a TPC and event with S1 and S2 pulses. A particle interaction produces
light (𝛾) which is detected as the S1 pulse at time 𝑡 = 0 in an event window.
Some electrons are also ionized in the initial interaction; an external electric field

⃗𝐸 prevents some of them from recombining by drifting them upwards. After some
drift time Δ𝑡, the electron cloud reaches the liquid surface where the electric field
extracts them from the liquid. Upon extraction, the electrons generate light via
electroluminescence in the gas, which is detected as the S2 pulse in an event window.
Original image courtesy of C. Faham and D. Malling. . . . . . . . . . . . . . . . . . . 34

2.2. Comparison of a simple two-grid TPC (Figure 2.2a) and a three-grid TPC (Fig-
ure 2.2b). The three-grid TPC has independently tunable drift and extraction fields.
These are more suitable for TPCs with larger drift lengths where a high extraction
field may be otherwise difficult to maintain. . . . . . . . . . . . . . . . . . . . . . . . 35

2.3. The liquid-gas interface in a dual-phase TPC forms a potential barrier that prevents
electrons from being extracted. The extraction field modifies the shape of the poten-
tial barrier and increases the energy of drifting electrons so they are able to overcome
it. This results in a higher electron extraction efficiency. . . . . . . . . . . . . . . . . 45

2.4. Overview of liquid xenon nuclear recoil charge and light yield measurements from
[74] at nominally similar electric field strengths. Charge yield results from [73] are
shown in the bottom figure as purple triangles. . . . . . . . . . . . . . . . . . . . . . 47

2.5. Predicted CENNS rates for a detector with a 10 m standoff from two sources: a 3
GWth nuclear reactor, and the ISIS neutron spallation source. In general, scattering
rates for a neutron spallation source are lower due to the smaller neutrino flux,
but are detectable at higher threshold energies because they are produced at higher
energies than in a nuclear reactor [58]. . . . . . . . . . . . . . . . . . . . . . . . . . . 49

2.6. Event rate as a function of electrons produced by dark matter-electron scatters in
xenon [78]. Colored spectra correspond to contributions from individual electron
shells and the gray band represents the overall spectrum when varying the effect of
higher-order processes. Left axes on the plot correspond to the maximum allowable
cross-sections for benchmark models, while right axes are cross-sections constrained
by the correct relic abundance. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51

2.7. Low energy S2 spectrum in the 2019 XENON1T light dark matter search [79]. The
analysis threshold was cut off at the vertical dotted line where events below ~4-5 elec-
trons in size were not considered due to the presence of “unmodeled backgrounds”.
In a 2021 reanalysis (Section D.13), an empirical model of electron trains was con-
structed which enabled limits to be set on light dark matter candidates based on
the observed single-to-few electron background rate. Additional backgrounds in this
plot include the expected CENNS rate (shown in red) from 8B solar neutrinos and
a flat ER background from 214Pb decays (blue). The expected signal model from 4
GeV and 20 GeV spin-independent DM models are also shown in orange and purple
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53

xxii



2.8. Hardware-related electron background signals in dual-phase xenon TPCs can be
grouped by the timescales in which they occur relative to an event. Each relates to
unique mechanisms and sources, some of which are speculative. Dependencies are
factors which have been investigated in studies of the respective backgrounds. Solid
lines indicate a strong connection has been observed. Dotted lines indicate a weak
dependence, no dependence, or a suspected dependence that requires further study. . 55

2.9. Distinct timescales in a single scatter event where low energy electron backgrounds
are most prominent. Prior to an S1, low energy backgrounds are typically dominated
by spontaneous pulses (dashed outline) which are not associated with a event in the
TPC. These are often due to grid emission. After the S1 and S2, prompt photoion-
ization backgrounds (solid outline and fill) are most common up to the maximum
drift time in a TPC 𝑡𝑑𝑟𝑖𝑓𝑡. As both the S1 and S2 can act as progenitors, they
each have their own timeframes for prompt backgrounds. Note also that prompt
backgrounds can be intense enough that single electron (SE) pulses often combine
and pile-up to form smaller S2 pulses; this is shown after the progenitor S2. After
prompt backgrounds from the S2 pulse have died off, electron trains in the form of
delayed SEs and small S2s (solid outline and no fill) are still visible. It has been
shown that delayed small S2s are not well-explained by SE pile-up [84]. . . . . . . . 57

3.1. The TPC (1) sits at the center of the LZ experiment. Segmented acrylic tanks (2),
shown in green, form the OD which encircles the titanium outer vessel. 8” PMTs (3)
are mounted on a frame to detect light from the OD and water shielding. The frame
is lined with Tyvek to improve light collection. Water shielding (4) is contained by
the stainless steel water tank. High voltage for the TPC grids is provided to the
TPC via a feedthrough (5) which connects to the cathode. Neutron calibrations are
conducted using two conduits (only one shown, 6) which provide a clear path to the
TPC. Figure taken from [90]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 69

3.2. Calibration events plotted in log10(S2c)−S1c space. CH3T events are represented by
dark blue points, and the median of their simulated distribution is represented by the
solid blue line. The dotted blue lines represent 10% and 90% quantiles. DD events are
represented by orange points and the same properties of their simulated distributions
are shown with red lines. Contours of constant NR and electron-equivalent recoil
energies are shown in gray. From [92]. . . . . . . . . . . . . . . . . . . . . . . . . . . 70

3.3. Map of the TPC in reconstructed depth and radial coordinates. The dotted black
line represents the maximum extent of the active volume, which is slightly distorted
because of electric field nonuniformities near the walls. The solid black line represents
the fiducial cut, which is used to reject events which are too close to the walls. Events
marked with gray points fail the fiducial cut, while events with black points pass it.
Red crosses and blue circles represent events which were rejected by skin or OD veto
detectors, respectively. From [92]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72

3.4. Data from Figure 3.3 plotted in log10(S2c) − S1c space. Events in the WIMP ROI
are fairly sparse. From [92]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73

3.5. SR1 ER energy spectrum passing all cuts, superimposed with background estimates.
From [92]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

xxiii



3.6. Limits on the spin-independent cross-section for WIMPs with masses between 101

and 104 GeV. Results from DEAP-3600, LUX, XENON1T, and PandaX-4T are also
shown. From [92]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74

3.7. An LZ event after an S2 pulse; the S2 pulse is not shown. Single electron (SE) pulses
are marked with red dots. The taller pulses at 4200 and 6600 µs are multiphoto-
electron (MPE) pulses, which are ignored here. The time axis in the event window
is relative to the trigger, which was fired at random; t=0 on this axis is when the
trigger was fired. In this trigger configuration, 1 ms of data was captured before the
trigger and 11 ms was captured after. An additional axis is shown at the bottom
denoting time since the end of the S2 pulse. . . . . . . . . . . . . . . . . . . . . . . . 82

3.8. Counts of single electron pulses between 15 and 25 ms after an S2 pulse. Bins are 1
ms wide. the last two SE pulses from the event shown in Figure 3.7 are captured in
this histogram. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83

3.9. Livetime between 15 and 25 ms after an S2 pulse. Bins are 1 ms wide. . . . . . . . . 83
3.10. Pulse count and livetime histograms for a progenitor S2 different from the one shown

in Figure 3.7. The same period of time since the progenitor S2 is shown as in
Figure 3.8 and Figure 3.9. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85

3.11. Pulse count and livetime histograms for pulses following thousands of different pro-
genitors. Counts of single electrons pulses appear to decrease exponentially more
than 100 ms after a progenitor (Figure 3.11a). This decrease is mostly an artificial
result of livetime also decreasing exponentially (Figure 3.11a), and should serve as
a motivating example for livetime corrections. . . . . . . . . . . . . . . . . . . . . . . 86

3.12. Overview of data extraction and processing structure . . . . . . . . . . . . . . . . . . 87
3.13. Mockup of an electron train captured with a random trigger. The sequence of events

1 and 2 illustrate how a trigger holdoff prevents events from overlapping as triggers
are blocked until the holdoff expires. Livetime is only within the bounds of event
windows and is further limited to the time between vetoing pulses (pink triangles).
This is because the trigger efficiency for any pulse outside of an event is manifestly
zero. If a pulse-based trigger is used, then livetime includes time inside of events
between vetoing pulses and outside of events when a trigger could have occurred.
This is shown with a dotted line extending the rightmost livetime arrow. . . . . . . . 88

3.14. Flow chart illustrating the event loop of the BigDEB algorithm. . . . . . . . . . . . . 90
3.15. Flow chart illustrating the pulse loop of the BigDEB algorithm. . . . . . . . . . . . . 91
3.16. General overview of steps in livetime and RQ post processing. . . . . . . . . . . . . . 95
3.17. The first two events of an electron train in LZ, showing the presence of SEs for

multiple milliseconds after an S2 pulse. . . . . . . . . . . . . . . . . . . . . . . . . . . 99
3.18. PMT hitmaps from the first and second events shown in Figure 3.17. Hitmaps on

the left correspond to the top PMT array, while hitmaps on the right correspond to
the bottom PMT array. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100

3.19. The rate of SEs after S2s, per square centimeter of liquid surface in the TPC. The
x-axis is the radial distance between between the child SE pulse and the most recent
progenitor S2 pulse. The stack axis denotes the time since the end of the progenitor
S2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103

xxiv



3.20. Diagram showing geometric constraints needed to correct for the lack of pulses ob-
served relative to a progenitor when pulses are observed at a radius 𝑅𝑆𝐸 which
extends beyond the walls of the TPC. A child pulse may be observed anywhere in-
side the TPC along the circle defined by 𝑅𝑆𝐸, Counts of pulses along this arc are
then used to correct for the lack of pulses observed along the arc outside of the TPC. 105

3.21. Rates of position-correlated and uncorrelated SE pulses versus time since the end
of their progenitor S2. The rates prior to 1 ms are a known artifact of pulse pile-
up from photoionization (see also Figure 3.22). Position-uncorrelated rates drop
off quickly after the full drift time has elapsed, in comparison to the rates from
position-correlated pulses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

3.22. Rates of position-correlated SE and S2 pulses versus the size of their respective pro-
genitors. For the smallest progenitors, SE rates appear to decay monotonically. As
the progenitor size increases, pulses lose their SE classification, causing a distortion
in the rate prior to 1 ms. This feature is not present when selecting for S2 pulses
because pulse classification is conserved for S2s undergoing pile-up. . . . . . . . . . . 108

3.23. Rates of SEs versus the size of their progenitor, 𝑒𝑅, in the number of electrons
extracted. A time window of 3-300 ms was chosen to avoid integrating any residual
photoionization. The position-uncorrelated rates do not exhibit a dependence on
the size of the progenitor; this behavior was also observed in XENON1T [80]. Most
progenitors are between 103 and 104 𝑒𝑅 in size; SR1 data was used in this plot to
maximized the available statistics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

3.24. Rates of SEs versus the drift time of their progenitor. The position-correlated rate
appears to linearly depend on the drift time of the progenitor even after normalizing
by 𝑒𝐼 , the size of the progenitor at the interaction site. The position-uncorrelated
rates do not exhibit a dependence on the drift time of the progenitor. This behavior
was also observed in XENON1T [80], and highlighted as an indication that uncorre-
lated delayed electrons may be correlated with previous progenitors. SR1 data was
used in this plot to maximized the available statistics. . . . . . . . . . . . . . . . . . 110

3.25. Drift time dependence of SE rates versus time since the progenitor S2. The position-
uncorrelated rates do not exhibit a clear drift time dependence at any timescale.
Position-correlated photoionization rates have a counterintuitive dependence on the
drift time. At shorter drift times, S2s should lose fewer electrons and photoionize the
TPC more. However, lower photoionization rates are observed in this case, and this
phenomenon was also observed in LUX [81]. This behavior remains unexplained,
and should be the subject of future work. . . . . . . . . . . . . . . . . . . . . . . . . 112

xxv



3.26. Rates of SEs verses the electron lifetime recorded at the time of the progenitor pulse.
The rate at higher electron lifetime values appears to taper off, in agreement with
behavior observed by XENON1T [80] and their explanation that the impurity or im-
purities responsible for electron trains does not perfectly track the concentration of
impurities which affect the electron lifetime. A sharp drop in the position-correlated
rate is apparent after the electron lifetime exceeds 5 ms. This may be an artifact of in-
terpolating electron lifetime data, as explained in the text. The position-uncorrelated
rate also exhibits a slight dependence on the electron lifetime, which would be con-
sistent with the explanation from XENON1T [80] that these backgrounds are mostly
from prior electron trains. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113

3.27. Electron lifetime dependence of SE rates versus time since the progenitor S2. A weak
dependence on the electron lifetime is visible for both the position-correlated and
uncorrelated rates at all timescales after the full drift time. The position-uncorrelated
rate in the 2-4 ms electron lifetime bin is noticeably uneven due to wrongly associated
pairs of SEs and progenitor S2s. This occurs for all other position-uncorrelated rates
in both this figure and Figure 3.25, but the effect is enhanced with lower statistics. . 114

3.28. The same rates from Figure 3.25 and Figure 3.27. instead normalized by 𝑒𝐿, the
number of SEs lost by the progenitor as it drifts through the liquid. Both the drift
time dependence and electron lifetime dependence that were visible in Figure 3.25
and Figure 3.27 are negated, indicating that impurities which capture electrons in
the liquid are primarily responsible for electron trains. . . . . . . . . . . . . . . . . . 115

3.29. Simulated field strengths in the extraction region of the LZ TPC. Key features such
as the anode and cathode are indicated in the plot. The cathode is not pictured; it
is located at Z = 0 mm and is the lower boundary for the drift liquid. The TPC
walls are located at R ≈ 740 mm; only the first 50 mm are shown here to highlight
the scale of the grid spacing in R. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117

3.30. Electron trains from progenitors in the gas, drift, and below cathode regions of
XENON1T [80]. A 200 ms holdoff was imposed after interactions to avoid overlap-
ping electron trains between progenitors. The SE rate during this 200 ms holdoff
period is shown with the solid purple line. It served to verify that the holdoff period
was adequate in allowing SE rates to return to a baseline level before SEs from pro-
genitor pulses were tracked. It also showed that rates following progenitors in the gas
and reverse field regions returned fairly quickly to the pre-holdoff rates, indicating
that electron trains are linked to drift region interactions. . . . . . . . . . . . . . . . 122

3.31. Power law in SE rate observed by Akimov et al. after muon S2s. . . . . . . . . . . . 123
3.32. Low-pass filtered summed waveform of electron trains from alpha events with a “fast”

and “slow” exponential component. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124
3.33. SE rates following single scatter events in the gas above and below the anode, extrac-

tion liquid above the gate grid, and drift liquid between the gate and cathode of LZ.
Delayed electron rates in the gas and extraction liquid are much less intense than
those in the drift liquid. This plot combines the rates from both position-correlated
and uncorrelated pulses, which is indicated by No Cut (NC) to the relative pulse
positions in the legend. No area normalization is used in this plot in order to enable
direct comparisons to Figure 3.30. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 127

xxvi



3.34. Rate of SEs shown in Figure 3.19, projected in time since the progenitor S2. Nor-
malizing rates by the area subtended by each interval in Δ𝑟 reveals that the flux of
position-uncorrelated SEs is fairly uniform for Δ𝑟 > 20 cm. . . . . . . . . . . . . . . 128

3.35. SE rates from Figure 3.33, split into the position-correlated and uncorrelated com-
ponents. The position-correlated rates in the gas are largely from drift liquid events
which leaked into the gas event selection. The relatively large fluctuations in these
rates over time compared to the rates in other regions also indicate mis-associated
S2 normalizations. Position-correlated rates in the extraction liquid appear to have
a power law component, but with a much lower amplitude than those in the drift liq-
uid. The exponents in these two regions are visually similar, but fits (Section 3.5.5)
indicate that the power law in the extraction liquid may be slightly steeper. . . . . . 130

3.36. Power law fit to the position-correlated SE rate following events originating in the
drift liquid shown in Figure 3.35. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131

3.37. Power law fit to the position-correlated SE rate following events in the extraction
liquid shown in Figure 3.35. A constant term in the fit approximates the presence
of position-uncorrelated backgrounds at long timescales, but does not account for
residual photoionization backgrounds at short timescales. As a consequence, there
is a large systematic uncertainty in 𝛽 associated with the left edge of the fit. . . . . . 132

3.38. Subtracting the position-uncorrelated SE rate in Figure 3.35 from the position-
correlated rate in Figure 3.35 removes many of the photoionization backgrounds
in the position-correlated SE rate between 100us and 1ms. It also removes the back-
ground component after 100ms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 133

3.39. Fitted value of 𝛽 in Figure 3.38 as a function of fit limits. Subtracting the position-
uncorrelated rate leaves behind some discontinuities in the rate. First, rates before
1ms are still slightly elevated. Second, rates after ~20ms are highly dependent on
the intensity of the position-uncorrelated flux, which increases with the field. . . . . 133

3.40. Power law fits to position-correlated rates in Figure 3.25 show minimal variation in
the exponent and a clear trend in the amplitude that increases with drift time. . . . 134

3.41. Rate of position-correlated and uncorrelated single electrons from extraction liquid
progenitors, at three different extraction voltages. Rates are normalized by 𝑒𝑆, the
size of the progenitor S2 at the liquid surface. . . . . . . . . . . . . . . . . . . . . . . 136

3.42. S2-normalized rate flux of position-correlated single electrons following extraction
liquid progenitors, at three different extraction voltages. The background flux from
position-uncorrelated electrons has been subtracted, revealing a slightly clearer power
law in each dataset. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 136

3.43. S2-normalized rate flux of position-correlated SEs following extraction liquid pro-
genitors, without background subtraction. A power law is fit with the addition of a
constant term as an alternative to account for the underlying position-uncorrelated
background. The power law exponents are only marginally affected. The left edge of
the fit is set to 2ms because of residual photoionization backgrounds at 1ms which
are not well-modeled by the constant terms. . . . . . . . . . . . . . . . . . . . . . . . 138

xxvii



3.44. S2-normalized rate flux of position-correlated single electrons following drift liquid
progenitors, at three different extraction voltages. The position-uncorrelated back-
ground has been subtracted. The electron lifetime in the 8 and 9kV datasets was
identical, and lower than the electron lifetime in the 7kV dataset. This resulted
in a steeper slope from photoionization backgrounds between ~1-5ms. Because the
position-correlated rates from the drift liquid are so intense, subtracting the position-
uncorrelated background has has no effect on the features in the rates at short (few
ms) and long (few hundred ms) timescales. . . . . . . . . . . . . . . . . . . . . . . . 139

3.45. Extraction field dependence of the delayed electron fraction from drift region electron
trains in XENON1T [80]. The delayed electron fraction was calculated by summing
the number of electrons observed between 2-200ms after a progenitor S2 in the drift
region, then normalizing by the size of the extracted S2. . . . . . . . . . . . . . . . . 140

3.46. S2-normalized rate flux of position-correlated and uncorrelated single electrons from
drift liquid progenitors, at three different drift voltages. . . . . . . . . . . . . . . . . 141

3.47. S2-normalized rate flux of position-correlated single electrons following drift liquid
progenitors, at three different drift voltages. The background flux from position-
correlated extraction liquid electrons at the same extraction field has been sub-
tracted. The power laws only appear to be marginally affected. . . . . . . . . . . . . 142

3.48. Drift field variation of the electron train power law exponent in past studies, com-
pared to results from this work. The 9kV dataset is excluded due to a combination
of low statistics and qualitative similarity to gas events. Data from this figure is
reproduced in Table 3.8. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 143

4.1. XeNeu (left) and its accompanying circulation cart (right). The detector is housed in
a cylindrical can which hangs underneath the main outer vacuum insulation system.
The circulation cart holds the gas circulation pump, getter, and xenon gas bottles.
The gas bottles and getter are on the backside of the cart and not visible in this
picture. Note that this picture is not up to date with the P&IDs shown in Figure 4.13
and Figure 4.16. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147

4.2. XeNeu detector components. At the bottom of the detector assembly is a large
cylindrical PTFE piece which houses the bottom PMT (Figure 4.5a). The field cage
(Figure 4.4a) is mounted to the top of the bottom PMT holder. The weir reservoir
is machined from a single block of PEEK and sits in a dedicated cutout on the
side of the bottom PMT holder. The weir height is independently adjustable with
clamps on two of the vertical support rods. Directly above the field cage is the top
PMT assembly (Figure 4.6), which clamps to all three support rods. High voltage
feedthroughs are welded to the detector top flange, and descend into slots machined
into the top of the bottom PMT holder. . . . . . . . . . . . . . . . . . . . . . . . . . 148

xxviii



4.3. 2D schematic (drawn to scale) of XeNeu showing dimensions of the active volume
and labels of key components. Several components are configurable and the subject
of this work. The PMT holder has a PTFE variant (Figure 4.5a, pictured here) as
well as one which is constructed from low-outgassing materials (Figure 4.23). The
top photosensor array originally used PMTs (Figure 4.6), which were replaced with
a SiPM array (Figure 4.38, pictured here). The field cage (Figure 4.4) can also be
assembled in two configurations, either with a PTFE reflector (pictured here) to
improve S1 light collection, or without one to reduce outgassing rates. . . . . . . . . 150

4.4. Field cage components for the XeNeu TPC. The field cage is held together by three
nylon screws cut to the appropriate length such that they fully engage with tapped
holes in the base ring, which is made of PEEK. . . . . . . . . . . . . . . . . . . . . . 151

4.5. Two views showing key features of the bottom PMT holder assembly. . . . . . . . . 153
4.6. The XeNeu top PMT assembly. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 155
4.7. Ports on the top flange of the detector. Clockwise from the top of the detector flange:

PMT high voltage feedthrough (see also Figure 4.40), PMT signal feedthrough, liq-
uid outlet port, gate HV feedthrough, and liquid inlet port. Continuing clockwise:
pumpout port, RTD feedthrough, cathode HV feedthrough tube, and another RTD
feedthrough. The gate and cathode feedthroughs, as well as the pumpout and cen-
tral ports are accessible from the vacuum box mounting flange. This flange and
the central port on the detector top flange are transparent for visibility of other
components. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

4.8. Cutaway of tee showing capillary and thermosyphon paths. The capillary is nested
inside the tee and protrudes into the detector space through the 1/4” tube which is
welded to the detector top flange. The capillary is used to draw liquid from the weir
into the heat exchanger. The second branch of the tee (bottom) is used to volume
share the detector gas with the condenser, forming a crude thermosyphon. . . . . . . 157

4.9. XeNeu detector with HV feedthroughs made visible. The HV feedthrough tubes are
mated to the ceramic feedthroughs via a copper reducer. End caps are attached to the
gate (left) and cathode (right) feedthroughs which contain sockets for running wires
to the gate and cathode. The brass end cap belongs to the cathode feedthrough, and
the copper end cap belongs to the gate feedthrough. The caps are different because
of an incident where a solder joint on the cathode feedthrough failed, which leaked
silicone oil over the TPC components. Ethan Bernard (LLNL) performed the repair
and created the replacement brass cap, which is fixed in place with a set screw. . . . 159

4.10. Schematic for high voltage delivery and monitoring circuit. 𝑉𝐼𝑁 is the connection
point for the high voltage power supply, Precursors to a full electrical breakdown are
monitored with an oscilloscope at the output of a high-pass filter, 𝑉𝑀𝑂𝑁 . 𝑉𝑀𝑂𝑁 and
the high voltage output 𝑉𝑂𝑈𝑇 are fed out of the box through coaxial connections.
𝑅𝐶𝐿 is a current-limiting resistor, which protects components in the event of an
electric short. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 160

xxix



4.11. Schematic showing three different designs for a high voltage resistor chain on a field
cage. At the left is a naive design that is unsafe because a short from the gate or
cathode grid would have an unrestricted current. In the middle is a safer design
that installs current-limiting resistors before the gate and cathode. A tradeoff of
this design is a reduced voltage drop between the gate and cathode, requiring higher
input voltages to match the unsafe design. At the right is the design of the resistor
chain in XeNeu. Current-limiting resistors are in series with gate and cathode grids,
however the grids are disconnected from other components. Consequently, the gate
and cathode voltages are equal to the voltage of their respective electrodes. All
resistors are 2 GΩ and the resistors between rings plug directly into holes machined
into the outer diameter of the rings. Resistors adjacent to the gate and cathode plug
into ports on the caps of the respective feedthroughs. . . . . . . . . . . . . . . . . . . 162

4.12. Rendering of XeNeu with portions of the vacuum box and the outer can made trans-
parent. Electrical connections are fed out through the flange on the far right side,
which has 5 SHV, 5 BNC, and 5 DB15 connectors. Directly above this flange, the
condenser and heat exchanger (Figure 4.15) are mounted in the volume contained
by an 11 in ISO-K nipple. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163

4.13. P&ID for components on the XeNeu detector cart. Arrows on the heat exchanger
lines denote the normal circulation path. The blue dotted line demarcates the vac-
uum vessel. The dotted line from the condenser to the detector represents a con-
nection which joins the detector and condenser gas spaces (Figure 4.8), acting as a
simple thermosyphon. PT: pressure transducer, PI: pressure (dial) indicator. . . . . 164

4.14. Rear view of XeNeu cart. The condenser and heat exchanger (Figure 4.15) are housed
in an 11” ISO-K nipple, offset from the axis of the detector. The flange on this nipple
has connections to the cryocooler as well as gas inlet and outlet connections to the
heat exchanger. In the foreground, oil filled plastic boxes are mounted on top of the
detector and connect to the high voltage feedthroughs. . . . . . . . . . . . . . . . . . 166

4.15. Hardware used for the thermal management in XeNu is housed in an 11” ISO-K
flange offset from the axis of the detector. Pre-cooled gas exits the heat exchanger
and enters the condenser. The gas is then condensed, where it enters the detector
space as a liquid. Liquid exits the weir reservoir via a capillary line, which enters the
heat exchanger. The condenser also has a third port which couples to the detector gas
space. The gas exchange line is capable of cooling the detector volume if circulation
is halted. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168

4.16. P&ID for the circulation cart. The SRV is not mounted on the cart, but included
here for clarity. V32 and V31 are intentionally dead ends which lead to a volume
that previously contained an Ar37 calibration source. . . . . . . . . . . . . . . . . . . 169

4.17. Drawing showing top view of shielding and background detectors arranged around
XeNeu for measuring high energy nuclear recoils. A mockup of the XeNeu detector
cart is visible at the right, surrounded by backing detectors (blue rectangles). Back-
ing detectors at high scatter angles were positioned closer to XeNeu to make up for
the decreased elastic scattering cross section of 14.1 MeV neutrons at those angles. . 172

xxx



4.18. Results from the XeNeu recoil energy measurement compared to previous values.
Values at the two highest recoil energies were averaged over their electric fields due
to low statistics. From [106]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 174

4.19. Illustration of the Migdal effect. In a nuclear recoil with a dark matter particle, the
nucleus is displaced from its electron shells. As a result, some of these electrons may
escape, producing a coincident electron recoil. From [118]. . . . . . . . . . . . . . . . 175

4.20. Scattering cross sections as a function of scattering angle, and expected Migdal ER
spectrum. From [107]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 176

4.21. S2 spectrum from interaction at a scattering angle of 16°. Three background com-
ponents are shown in the spectrum: single scatters, passive scatters, and multiple
scatters. These are overlaid with predicted counts of Migdal interactions as well as
the measured counts from the best-fit model. Fit residuals are shown in the bottom
plot for the best-fit model (black points) and the predicted model (dotted line). The
shaded grey band represents the systematic uncertainty in the NR background models.177

4.22. Diagram showing key processes for outgassing. The bulk material is represented
by the grid of larger green circles. These processes are not mutually exclusive.
For example a species which is permeating the material may adsorb and desorb
repeatedly as it diffuses into the vacuum space. Figure adapted from [119]. . . . . . 180

4.23. Exploded view of the final design of the new bottom PMT holder. From top to
bottom: the Shapal retaining ring, the grid contact ring, two Shapal high voltage
pockets, and the aluminum base. White components are made from Shapal. The
grid contact ring is colored green here to improve its visibility, but is made from
aluminum like the rest of the PMT holder. Screws used were made of nylon due to
their proximity to high voltage components. . . . . . . . . . . . . . . . . . . . . . . . 183

4.24. View of PMT holder assembly from above, with the field cage removed except for
the PEEK base ring. Dark tracks are visible in the cathode (left) and gate (right)
pockets. These tracks are formed from rubbing of metal surfaces on Shapal components.185

4.25. Screenshot of slow control readout during condensation of xenon with redesigned
bottom PMT holder. When liquid xenon made contact with the aluminum bottom
PMT holder, over the course of two minutes the detector temperatures rose almost
10 K. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186

4.26. Trigger schematic for studies of SE rates. Signals needed to be amplified twice
in order to exceed the minimum discriminator threshold. Output from the 3-fold
coincidence module was recorded at the digitizer for trigger efficiency studies. The
1-fold coincidence module provided the main event trigger. A pulse from the 3-fold
coincidence would also fire both the 1-fold coincidence and a gate generator. A long
cable was used for the gate generator signal to stagger the arrival times of the 3-fold
coincidence and vetoing pulses at the 1-fold coincidence so the first coincidence pulse
in an event would not veto itself. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188

4.27. Plot of AFT 75-25 vs pulse area showing counts of pulses. Single electron pulses are
roughly within the closed red dashed curve. The band of pulses below the dashed
orange line corresponds to S1-like pulses, which have a very low AFT 75-25. . . . . . 189

xxxi



4.28. Example of a trigger efficiency histogram. The first peak at a pulse area of roughly
60 photoelectrons (npe) corresponds to SEs. The trigger efficiency approaches 100%
for pulses larger than four electrons in size. . . . . . . . . . . . . . . . . . . . . . . . 191

4.29. Rates of child pulses following progenitors in Runs 11 and 12. Rates are normalized
following the convention established in Equation 3.8. Child pulses are additionally
binned by their size along the y-axis. Pulse sizes are normalized by the size of the
single electron pulse in each run. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 193

4.30. Total integrated rates from Figure 4.29 over pulses up to 10 SE in size. For this
plot, rates were integrated from 100 µs to 1 s. At all circulation rates, approximately
60% of the pulse rate in Run 11 is due to SEs, whereas SEs make up almost 80%
of the child pulse rate in Run 12. With low-outgassing components in Run 12, the
circulation rate had a much smaller impact on reducing background rates. This
suggests that plastic components in the TPC are a dominant source of impurities
responsible for ionization backgrounds. . . . . . . . . . . . . . . . . . . . . . . . . . . 195

4.31. Projection of Figure 4.29 showing integrated rates of child pulses up to 10 SE in size
at different circulation rates. For this plot, rates were integrated from 100 µs to 1 s.
Rates for pulses of all sizes are affected by the circulation rate in Run 11. This is in
contrast with rates for Run 12 in Figure 4.31, where only SE rates appear to have a
slight circulation rate dependence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 195

4.32. Projection of Figure 4.29 examining rates of single electron pulses over time at differ-
ent circulation rates. Rates prior to 100 µs are excluded due to the high prevalence
of merged pulses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

4.33. Example of a double exponential fit between 400 µs and 30 ms in the single electron
rate time dependence. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 197

4.34. Fit values for each fit parameter and rate curve shown in Figure 4.32. Error bars
represent statistical uncertainties. Values are plotted along the x-axis according to
their run number and circulation speed; they are also printed in Table 4.3. . . . . . . 198

4.35. Single electron rates from Run 11 and Run 12. The rates in both runs appear to
converge on a power law with a constant component (Equation 4.2) after 30-40 ms.
Fit parameters for these power laws are presented in Table 4.4. . . . . . . . . . . . . 201

4.36. Typical dimensions of a SiPM and equivalent circuit for a single SiPM. . . . . . . . . 204
4.37. Rendering of the new XeNeu SiPM array. Each S13371 unit has four cells which can

be read out individually, but in this design the cells in each unit were connected in
parallel. A PEEK spacer was also designed to sit on the perimeter of the board and
reduce wear between the edge of the board and the metal components in the tray
and bracket assembly Figure 4.38. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 206

xxxii



4.38. Exploded view of the SiPM bracket and tray assembly. The tray is in the foreground,
and has a cutout in the center that accepts the SiPM array and PEEK array spacer.
Two L-shaped brackets on opposing ends of the tray are screwed into place, capturing
the SiPM array. Two additional holes at the rear corners of the tray are designed
to be concentric with tapped holes in the bracket so the tray can be securely mated
to the bracket. The bracket is in the background and shares a similar design as
the previous iteration. Two new holes in the same plane as the holes for the rear
mounting clamp are included as optional electrical ground points at the corners of the
bracket. Cutouts for the forward mounting posts on the bracket were intentionally
asymmetric so the bracket alone could be installed by sliding it laterally onto the
mounting posts. Through holes were used in all cases to ensure the absence of any
trapped volumes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208

4.39. Pieces of metal were observed on the sealing face of an o-ring (orange) belonging to
a leaking feedthrough. The metal was believed to a originate from threads on the
feedthrough, which was likely overtightened. At the bottom left, a feedthrough as it
was normally installed can be seen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 210

4.40. Right: Rendering of the 7-pin Mil-spec feedthrough. Left: Pinout of the 7-pin
feedthrough, as used on XeNeu. Numbers and colors corresponded to labels used on
the XeNeu vacuum box. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 211

4.41. Power delivery cable assembly. At the left, crimp on terminals are held together with
a 2-56 screw and two nuts. These terminals offered a simplified way to temporarily
and securely attach multiple ground connections throughout the detector. The main
ground connection (visible at the top) initially used a stainless steel braid; this was
later replaced with a copper braid. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212

4.42. Example of a custom PMT pin connection on a coaxial cable. The pin is soldered
to a ring terminal, which is crimped to the inner dielectric of the cable. The inner
conductor is inside the crimp sleeve, which is then flooded with solder. The solder-
covered tip of the inner conductor can be seen in this picture emerging from the
body of the crimp sleeve, to the left of the base of the PMT pin. Not shown here
are the copper wick to extend the outer shield, and heatshrink around the body of
the ring terminal and cable in order to provide additional strain relief. . . . . . . . . 213

4.43. 23-pin Sub-C connector cutaway and pin mapping. This connector was used to
feed signals from the SiPM array and bottom PMT through the detector top flange
(Figure 4.7). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 215

4.44. Comparison of top mounting bracket designs used to suspend components in the
XeNeu detector space. The new bracket (Figure 4.44b) has a profile which occupies
almost 50% less space than the old bracket (Figure 4.44a). . . . . . . . . . . . . . . . 217

4.45. Trigger configuration for SPE calibrations. When the bottom PMT (CH10) crosses
a threshold set at the discriminator, the discriminator fires a pulse which is picked
up by the coincidence module. If the coincidence module is not being vetoed, it fires
a pulse (TRIG) to the trigger channel on the DAQ and to the gate generator. The
gate generator vetoes the coincidence module for a set period of time after receiving
a pulse from the coincidence module. . . . . . . . . . . . . . . . . . . . . . . . . . . . 219

xxxiii



4.46. Example waveform of an SPE captured during preliminary SiPM tests at the time
indicated with the orange “T”. The SiPM was biased to 55 V and the signal was
filtered with a 5 MHz low-pass filter. A voltage and time scale of 2.00 mV and 2
µs per division were used; these are printed at the top and bottom of the figure,
respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220

4.47. Example of an ROI integral spectrum from a single SiPM at a single voltage. The
sum of four gaussians was fit to the spectrum. This fitting procedure was repeated for
each SiPM channel at each bias voltage in order to determine the minimum voltage
needed to resolve SPEs. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221

4.48. Gain curves for each SiPM channel. “ALL” represents the gain curve obtained after
fitting to the sum of the integral spectra for all channels. Note that CH 8 is missing:
this channel was broken on the DAQ so CH 9 was used instead. . . . . . . . . . . . . 223

4.49. Coordinate system used for positioning sources around XeNeu. 𝑅, 𝑍, and 𝑑𝑙𝑎𝑠𝑒𝑟
were measured with a standard measuring tape or calipers, where possible. . . . . . 223

4.50. Trigger for single electrons using the bottom PMT. The bottom PMT signal was
doubled using two FIFO modules to make use of the full range of the discriminator.
The main event trigger was configured such that Discriminator 1 would need to be
coincident with a delayed copy of itself. This would ideally happen for a pulse with
a width greater than the delay time. . . . . . . . . . . . . . . . . . . . . . . . . . . . 225

4.51. Example waveform showing the sum of the bottom PMT and SiPM outputs, with
two SE pulses. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226

4.52. Comparison of reconstructed pulse areas for electroluminescence signals in the SiPM
array vs. the bottom PMT. SE pulses are prominently featured in this plot between
20 to 80 npe on the PMT axis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227

4.53. Reconstructed pulse positions using the CoM approach. The square distribution
is unrealistic and a consequence of pulses at the edge of the TPC being misrecon-
structed towards the center. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 228

4.54. Comparison of positions generated by the CoM algorithm and true positions from
the simulation. The CoM algorithm is accurate for pulses with a true position within
1 cm of the TPC center, but misreconstructs positions thereafter. . . . . . . . . . . . 228

4.55. Reconstructed pulse positions using the LRF. . . . . . . . . . . . . . . . . . . . . . . 230
4.56. Comparison of positions generated by CoM and LRF. . . . . . . . . . . . . . . . . . 231

5.1. Emission spectra from xenon-doped argon gas mixtures generated via electron beam
excitation. Light production from xenon excimers centered around 174 nm virtually
saturates at concentrations as low as 10 ppm. From [141]. . . . . . . . . . . . . . . . 235

5.2. Emission spectra from xenon-doped argon gas mixtures generated via heavy ion beam
excitation. Subplots are arranged in columns of the same xenon concentrations and
rows of the same total mixture pressures. Rows a through e correspond to total mix-
ture pressures of 400, 780, 1000, 1200, and 1400 mbar respectively. Efthimiopoulos
et al. note that emission at concentrations of 100 ppm is predominantly from atomic
transitions. From [145]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 236

5.3. Examples of solid xenon formation resulting from excessive distillation of xenon in
a xenon-doped liquid argon mixture. . . . . . . . . . . . . . . . . . . . . . . . . . . . 237

xxxiv



5.4. Plot of temperature composition curves for xenon and argon (Equation 5.5) at at-
mospheric pressure. This plot shows the eutectic point of the binary mixture, which
is the intersection po