ABSTRACT

Title of dissertation: SUPERCONDUCTING NANOWIRE
SINGLE-PHOTON DETECTORS FOR DARK
MATTER DETECTION APPLICATIONS

Jamie Shawn Luskin, Doctor of Philosophy, 2024

Dissertation directed by: Dr. Matthew Shaw
NASA Jet Propulsion Laboratory

Superconducting Nanowire Single Photon Detectors (SNSPDs) are a leading detector

technology for time-correlated single-photon counting from the UV to the near-infrared.

Due to their unique combination of low energy thresholds and low intrinsic dark count

rates, SNSPDs have become attractive as sensors for emerging low-mass dark matter (DM)

detection experiments, where they offer the potential to fill existing technology gaps and

enable the exploration of previously unconstrained parameter space. One developing DM

detection concept, sensitive to MeV-scale DM electron recoils, uses n-type GaAs as a cryo-

genic scintillating target instrumented with a large-area SNSPD as the sensor for scintilla-

tion photons.

This thesis focuses on the development and characterization of SNSPDs with mm2-

scale active areas for DM applications in general, and specifically for the detection of scin-

tillation light. This work demonstrates the coupling of n-type GaAs with SNSPDs, and the

design of novel characterization experiments using optical and energy-tagged X-ray exci-

tation to measure the effective light yield and photoluminescence timescales of cryogenic

scintillators using SNSPDs. The isotropic nature of scintillation light and the large active



areas of the devices studied in this work introduce unique challenges for SNSPD design,

nanofabrication, and performance. This thesis provides insights into the current state-of-

the art, limitations, and approaches to scaling SNSPDs to cm2-scale active areas for future

work. The presented findings advance the status of SNSPDs for DM detection and other

emerging High-Energy Physics applications.



SUPERCONDUCTING NANOWIRE SINGLE-PHOTON DETECTORS
FOR DARK MATTER DETECTION APPLICATIONS

by

Jamie Shawn Luskin

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 Yanne Chembo, Chair
Dr. Matthew Shaw, Advisor
Professor Daniel Lathrop, Dean’s Representative
Professor Jeanpierre Paglione
Professor Carlos A. Rı́os Ocampo



© Copyright by
Jamie Shawn Luskin

2024



What we observe is not nature itself, but nature exposed to our method of questioning.

- Werner Heisenberg

ii



DEDICATION

This thesis is dedicated to the memory of my grandfather

Dr. Jerome Lawrence Luskin,

The original Dr. Luskin and my greatest inspiration.

iii



Acknowledgments

As I approach the end of my graduate career, I can’t help but feel overwhelmed with

gratitude for the people who have helped make this journey possible, both directly and

indirectly. It wouldn’t be right to start this section off any other way than to acknowledge

my JPL advisor, Matt Shaw, for offering me the life-changing opportunity to work in our

group. I can’t thank you enough for your guidance, support, and kindness throughout the

past 4.5 years. I also want to express my sincere appreciation to my UMD advisor, Yanne

Chembo, for going above and beyond in providing invaluable guidance throughout this

journey. And to the one and only Professor Maria Spiropulu - thank you for making me

feel at home at Caltech and the INQNET labs, and for treating me like one of your own.

My sincerest gratitude extends to the brilliant JPL SNSPD group members. Andrew

Mueller - thank you for being an exceptional office-mate, friend, and colleague and for

listening to all of my rants (and taking them seriously no matter how inane or dramatic!).

I am constantly amazed by your talents and inspired by the way you think. Ioana Craiciu

- you are truly the best, and I appreciate all of your guidance and support during these last

few chaotic months more than I can ever express. Professor Boris Korzh - you have had

such a profound impact on the way I approach measurements and science. Jason Allmaras -

thank you for being an incredible resource any time, any place, no matter what goes wrong.

I deeply admire and look up to your technical skills, work ethic, and the way you reason

through problems. To the rest of the group- Emma Wollman, Sahil Patel, Emanuel Knehr,

Sasha Sypkens, Bruce Bumble, Andrew Beyer, Fiona Fleming, Jasen Zion and former

members Ekkehart Schmidt and Gregor Taylor - your contributions and support have been

instrumental to my professional development and it is a pleasure to work with you all.

A sincere thank you goes out to my collaborators at Fermilab Cristian Peña, Si Xie, and

Christina Wang for guiding many of the experiments in this thesis, and at LBNL, Maurice

Garcia-Sciveres and Steve Derenzo for the insightful and lively discussions.

iv



Dr. J - I don’t know that I will ever find the words to convey the impact that you and

your class have had on me, but I would not have ended up on this trajectory without you

and your philosophy will always be my foundation. The next (long-overdue) pitcher is on

me!

Keric Morinaga - thank you for sharing your love of yoga with me and for building

the most incredible and supportive community at the Touchstone gyms. Your lessons and

philosophy have played a pivotal role in keeping me mentally grounded throughout this

process.

A pleasant surprise during this PhD was getting to meet so many wonderful people at

Caltech. I am sincerely thankful to Sam Davis, Iraj Umesh, Preeti Bhat, and all of the

INQNET group members for welcoming me and supporting me. Tracy Sheffer - you leave

us all feeling so taken care of, and we simply could not do what we do without you! Rahaf

Youssef - I immensely appreciate your intellect, brilliant insights, kindness, and friendship

throughout the past few years. Claire Ellison, Tommy Sievert, Ray Wynne, and Damian

Wilson - thank you for being the infinite source of joy that you are for everyone around

you. Lautaro Narvaez - thank you for patiently helping me with anything I’ve ever brought

to you. There is truly no problem you can’t solve and your expertise in electronics is truly

wizard-level. Thank you to the Hutzler group members for being the best neighbors in

Downs-Lauritsen and delightful humans to interact with. A tremendous thank you goes

out to the illustrious and incomparable Jason Trevor. Thank you for everything that you

do to keep so many groups running and for teaching me all things high-energy physics and

beyond. You are an absolute legend, one of my favorite people, and the closest thing to a

superhero I’ve ever met.

I would like to thank my family for their support throughout my entire (long!) educa-

tion. I am so fortunate to have all of you in my corner. To my late grandfathers, Jerome

Luskin and Nunzio Brusca - you are profoundly missed and will always be a guiding light

v



for me. I am also eternally grateful for the incredible Leal family for their love and support

during those early years of studying like a maniac.

A posthumous thank you is extended to my late great-uncle, Dr. Burton Kleinman, who

has always been an immense inspiration to me. I deeply admire how your work transcended

the boundaries between mathematics, physics, and chemistry (and I hope I inherited some

of those Kleinman genes!). There is a near-unity probability that we had a lot in common

and I’ll always wish that we could chat about things.

Finally, to my incredible wife, Elaina - I could not have done any of this without your

unconditional support. Thank you for believing in me, helping me grow, and for being the

best partner in the universe.

vi



Table of Contents

Foreword ii

Dedication iii

ACKNOWLEDGEMENTS iv

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

1.2.1 Galactic Rotation Curves . . . . . . . . . . . . . . . . . . . . . . . 2
1.2.2 Cosmological scale . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.2.3 Gravitational Evidence . . . . . . . . . . . . . . . . . . . . . . . . 6

1.3 Dark Matter Candidates . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
1.3.1 Classes of Dark Matter Candidates . . . . . . . . . . . . . . . . . . 10
1.3.2 WIMP Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . . 11
1.3.3 Axions and Axion-Like Particles (ALPs) . . . . . . . . . . . . . . . 13
1.3.4 Dark Photons . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.3.5 sub-GeV Dark Matter . . . . . . . . . . . . . . . . . . . . . . . . . 17

DM-electron scattering . . . . . . . . . . . . . . . . . . . . . . . . 18
DM-electron absorption . . . . . . . . . . . . . . . . . . . . . . . . 19

1.4 DM Detection Strategies . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
1.4.1 Direct Detection of sub-GeV Dark Matter . . . . . . . . . . . . . . 22
1.4.2 n-type GaAs as a Cryogenic Scintillating Target . . . . . . . . . . . 23
1.4.3 DM Detection concept using n-type GaAs . . . . . . . . . . . . . . 26
1.4.4 Sensor Classes for n-type GaAs and Low Mass DM Targets . . . . . 28

2 Superconducting Nanowire Single Photon Detectors 32
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32
2.2 Device Operation Principles . . . . . . . . . . . . . . . . . . . . . . . . . 33
2.3 Detection Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35

2.3.1 Detection Efficiency . . . . . . . . . . . . . . . . . . . . . . . . . 36

vii



2.3.2 Dark Count Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
2.3.3 Timing Resolution . . . . . . . . . . . . . . . . . . . . . . . . . . 38

2.4 Design Tradeoffs for Large-area SNSPDs . . . . . . . . . . . . . . . . . . 39
2.4.1 Active Area vs Timing Performance . . . . . . . . . . . . . . . . . 39
2.4.2 Array Size vs Readout Complexity . . . . . . . . . . . . . . . . . . 40

2.5 Large-active area SNSPDs . . . . . . . . . . . . . . . . . . . . . . . . . . 40
2.5.1 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
2.5.2 Device architecture . . . . . . . . . . . . . . . . . . . . . . . . . . 42

Nanowire vs Microwire Detectors . . . . . . . . . . . . . . . . . . 43
2.6 SNSPDs for Dark Matter Detection and HEP Applications . . . . . . . . . 43

3 Large Active Area SNSPD Design and Characterization 46
3.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46
3.2 Generation 1 Microwire Arrays . . . . . . . . . . . . . . . . . . . . . . . . 49

3.2.1 Detector Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49
3.2.3 Detector Characterization . . . . . . . . . . . . . . . . . . . . . . . 50

Cryogenic set-up and Readout . . . . . . . . . . . . . . . . . . . . 50
Optical Characterization . . . . . . . . . . . . . . . . . . . . . . . 51
Efficiency Measurements . . . . . . . . . . . . . . . . . . . . . . . 55
Tc Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
Timing Resolution Characterization . . . . . . . . . . . . . . . . . 59

3.3 Generation 2 Microwire Arrays . . . . . . . . . . . . . . . . . . . . . . . . 62
3.3.1 Detector Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
3.3.2 Tc Measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
3.3.3 Optical Characterization . . . . . . . . . . . . . . . . . . . . . . . 68
3.3.4 Timing Resolution Characterization . . . . . . . . . . . . . . . . . 70

4 Backgrounds in Microwire Arrays 77
4.1 Extrinsic Background Mitigation . . . . . . . . . . . . . . . . . . . . . . . 78

4.1.1 Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
Dark Packaging . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78

4.2 Intrinsic Backgrounds in Microwire Arrays . . . . . . . . . . . . . . . . . 81
4.3 Scaling with Active Area . . . . . . . . . . . . . . . . . . . . . . . . . . . 81
4.4 Temperature Dependence of Dark Counts . . . . . . . . . . . . . . . . . . 83

5 Optically-excited Scintillation Measurements 87
5.1 Overview of systems measured . . . . . . . . . . . . . . . . . . . . . . . . 88

5.1.1 GaAs Specifications and Analysis . . . . . . . . . . . . . . . . . . 89
5.2 Detector and Sample Information . . . . . . . . . . . . . . . . . . . . . . . 89

5.2.1 System A . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89
Device packaging and Hybridization . . . . . . . . . . . . . . . . . 90

5.2.2 System B . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90

viii



Device packaging and Hybridization . . . . . . . . . . . . . . . . . 90
5.3 Calibration of SNSPDs for Scintillation Light . . . . . . . . . . . . . . . . 91
5.4 Scintillation timescale measurements . . . . . . . . . . . . . . . . . . . . . 94

5.4.1 Measurement setup . . . . . . . . . . . . . . . . . . . . . . . . . . 96
5.4.2 Results and exponential fitting . . . . . . . . . . . . . . . . . . . . 97

5.5 Effective Light-yield Measurements . . . . . . . . . . . . . . . . . . . . . 98
5.5.1 Measurement Setups . . . . . . . . . . . . . . . . . . . . . . . . . 99
5.5.2 System A effective light yield . . . . . . . . . . . . . . . . . . . . 100
5.5.3 System B Effective Light Yield . . . . . . . . . . . . . . . . . . . . 100

5.6 Future work towards Absolute Light Yield Measurements . . . . . . . . . . 101

6 X-ray Excited Scintillation Measurements 103
6.1 Compton Scattering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105
6.2 X-ray Source . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107
6.3 Calibration of HPGe Detectors . . . . . . . . . . . . . . . . . . . . . . . . 108
6.4 Timescale Measurements . . . . . . . . . . . . . . . . . . . . . . . . . . . 111

6.4.1 Background Measurements of GaAs-SNSPD System . . . . . . . . 113
6.4.2 CAEN Digitizer Readout . . . . . . . . . . . . . . . . . . . . . . . 114
6.4.3 Source Effect on GaAs-SNSPD System . . . . . . . . . . . . . . . 115

6.5 Compton Scattering Analysis and Challenges . . . . . . . . . . . . . . . . 119
6.5.1 Poisson Statistics and Mean Photon Number Fitting . . . . . . . . . 122
6.5.2 Understanding the Observed Distribution . . . . . . . . . . . . . . 124
6.5.3 Understanding System Losses and Secondary X-ray Generation . . 128

7 Designs and Challenges of Improved Microwire Arrays 132
7.1 Optical Stack Design for Microwire Arrays . . . . . . . . . . . . . . . . . 133

7.1.1 Rigorous Coupled Wave Analysis . . . . . . . . . . . . . . . . . . 136
7.1.2 Angled-Incidence RCWA Simulations . . . . . . . . . . . . . . . . 136
7.1.3 DBR Optical Stack Design for Broad-Angle Efficiency . . . . . . . 138

Angular Distribution of Light from GaAs . . . . . . . . . . . . . . 140
DBR Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 148

7.2 Fabrication Challenges and Steps Forward . . . . . . . . . . . . . . . . . . 149

8 Microwire Arrays in Other Experiments 155
8.1 Detection of High Energy Particles with Large Area SNSPDs . . . . . . . . 155

8.1.1 Detector Design and Measurement Infrastructure . . . . . . . . . . 156
8.1.2 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 156

8.2 Broadband Reflector Experiment for Axion Detection . . . . . . . . . . . . 158
8.2.1 Detection Concept and Experiment Design . . . . . . . . . . . . . 158
8.2.2 RCWA Efficiency Simulations . . . . . . . . . . . . . . . . . . . . 160
8.2.3 Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162

9 Conclusions 164

ix



9.1 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 164
9.2 Future Work and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . 165

A Procedure for IV Curve Analysis with DC Coupled Cryogenic Amplifiers 166

B n-type GaAs Sample Information 171
B.1 Sample A - More Information . . . . . . . . . . . . . . . . . . . . . . . . . 171
B.2 Sample B - More Information . . . . . . . . . . . . . . . . . . . . . . . . . 172

C Publications 174

Bibliography 175

x



Chapter 1: Dark Matter

1.1 Introduction

Elucidating the particle nature of dark matter (DM) is one of the largest unsolved prob-

lems in fundamental physics. Though DM has evaded direct detection to date, indirect

probes of its gravitational interactions with celestial bodies at multiple scales indicate that

it comprises nearly 80% of the matter in the Universe and is necessary to explain large-scale

structure formation.

The most mature and large-scale experimental DM search programs have focused on

the Weakly Interacting Massive Particle (WIMP) paradigm. Due to a lack of conclusive de-

tection signal, new theoretical frameworks have emerged in recent years, and there is now

strong motivation to explore a broader set of candidate particles experimentally. Hidden

sector DM candidates are neutral under Standard Model (SM) forces and charged under

new forces that can weakly couple to ordinary matter via portal interactions. The range of

potential mass of hidden sector DM is broader than that of the WIMP DM. Importantly, the

mass range below a GeV, the so-called sub-GeV DM contains theoretically well-motivated

targets that are either below the noise floor or not accessible in WIMP search experimental

programs and thus require the development of novel experimental approaches. Increasing

the sensitivity of direct detection experiments to lower masses of DM requires the develop-

ment of new target materials sensitive to proposed interactions and new low-noise sensors

with low-energy thresholds. As these technologies develop, sub-GeV DM can be explored

1



substantially with smaller-scale table-top direct detection pilot experiments.

This chapter serves as a brief introduction to DM. Section 1.2 briefly summarizes the

experimental evidence for DM, Section 1.3 presents an overview of relevant DM candi-

date particles, and Section 1.4 introduces DM detection strategies to motivate the work

contained in this PhD thesis.

1.2 Evidence for Dark Matter

Evidence for the existence of DM has steadily increased over the last several decades.

Observational support includes its gravitational effects, such as galaxy rotation curves,

gravitational lensing, and the large-scale structure of the universe, all of which cannot be

accounted for by visible matter alone. Further evidence comes from cosmological observa-

tions consistent with predictions from the Lambda cold dark matter model (ΛCDM), which

serves as the standard model of Big Bang cosmology.

1.2.1 Galactic Rotation Curves

The earliest evidence for a cold, non-luminous matter pervading the Universe came

from the study of galactic rotation curves. These were first measured in the Coma Cluster

by Franz Zwicky in the 1930s [1] and later in the 1970s by V. Rubin [2] in the Andromeda

galaxy and others [3]. Obtaining a rotation curve entails measuring the rotational velocity

of galaxy or galaxy clusters via Doppler shifts and subsequently plotting this quantity with

respect to the distance from the center of the galaxy.

The rotational velocity of a galaxy can be expressed using Newtonian mechanics as

νrot(r) =

√
GM(r)

r
, (1.1)

2



where M is the enclosed mass, r is the radial distance from the galactic center, and G is

the gravitational constant. Assuming that visible matter comprises the entirety of the mass

distribution, νrot should decrease as 1√
r2 for distances beyond the visible galactic disc. In

practice, the measured vrot curves flatten at large r (see figure 1.1), implying a constant

mass distribution that can’t be explained by visible matter alone.

Measured galactic rotation curves are all consistent with a spherically symmetric mass

distribution that increases linearly with r [4]. On the contrary, the visible mass, consti-

tuted by stars and gases which can absorb and emit light, is concentrated in the inner disk.

Zwicky, in addition to pioneering these early measurement techniques, coined the term

”dunkle materie” (dark matter) to refer to the non-luminous component of the mass distri-

bution.

Other proposed explanations for this discrepancy include modifications to gravity in-

cluding Modified Newtonian Gravity [5]; however, these lack consistency with phenomena

at other scales [6] that are explained in the following sections.

1.2.2 Cosmological scale

The Cosmic Microwave Background (CMB) is the background radiation in the Uni-

verse comprised of the photons that first decoupled from matter shortly after the Big Bang.

It was initially believed to be isotropic until faint (µK-scale) anisotropies in temperature

and polarization were discovered in the early 1990s. The origins of these anisotropies

are understood to be from the matter-energy density of the early universe. Namely, it is

hypothesized that oscillations in the baryon-photon fluid under the influence of the gravita-

tional potential due to local wells and peaks of cold dark matter produced the temperature

anisotropies, which now manifest in the CMB power spectrum as acoustic peaks [9].

The CMB has a blackbody spectrum with a temperature of 2.73 K, peaked at λ =

3



Figure 1.1: Rotation curve measured for the galaxy NGC 6503 (black dots) and fit to DM-halo
model shown with the solid black line. The dashed black line is the Keplerian prediction based on
visible matter. The dotted-dash is the DM-halo rotation curve that is required to explain the flatness
of the curve at large r. Figure adapted from ref. [7]

4



Figure 1.2: The CMB map as seen by NASA’s COBE (top)and WMAP (middle) missions and
ESA’s Planck mission (bottom). Figure adapted from [8].

5



1.06mm. Atmospheric water molecules absorb this wavelength of light, making it amenable

to study via ground and space-based missions (see figure 1.2 for examples). Missions

such as COBE [10], WMAP [11], and Planck [12] have directly measured the tempera-

ture anisotropies. The power spectrum of these temperature fluctuations offers a means to

validate cosmological models that make predictions about this spectrum.

Lambda-Cold Dark Matter (ΛCDM), a mathematical model of the Big Bang serving as

the cosmological standard model, contains three main components describing the baryonic

matter, dark matter, and dark energy contents of the Universe. As shown in figure 1.3, the

ΛCDM model prediction fits the CMB anisotropy data with high statistical significance.

This alignment reinforces the robustness and accuracy of the ΛCDM model and the impor-

tance of dark matter in our understanding of the Universe and its evolution.

The baryonic and non-baryonic matter densities extracted from the 2018 Planck col-

laboration are Ωbh2 = 0.0224 and ΩDMh2 = 0.120, where h is the dimensionless Hubble

constant [12]. The ratio between dark matter and baryonic matter densities is given by

ΩDM/Ωb ≈ 0.120/0.0224 ≈ 5.4, indicating that dark matter is approximately five times

more abundant than ordinary matter.

1.2.3 Gravitational Evidence

Gravitational lensing measurements have also provided a compelling set of evidence

for the existence of dark matter. Gravitational lensing is a phenomenon whereby a massive

celestial body causes significant perturbations in the space-time metric such that light is

bent from a distant source as it travels toward an observer. Most lines of sight that are ac-

cessible are in the weak gravitational lensing regime, where distortions can’t be measured

from a single source. In composite lensed images, mass distributions can be reconstructed

through correlations in the observed ellipticity of the distorted objects. In these measure-

6



Figure 1.3: Temperature power spectrum of CMB measured by Plank mission – red dots , and the
λCDM model prediction is shown in blue. Figure taken from ref. [12]

ments, the reconstructed mass causing the lensing phenomenon exceeds the visible mass of

the object by a factor of tens to hundreds, pointing to the presence of dark matter.

A well-known example is the bullet cluster [13], which was formed through the colli-

sion of two galaxy clusters roughly 3.8 billion years ago. The Hubble space telescope was

used to obtain weak gravitational lensing measurements, from which the mass distribution

was derived. The luminous mass distribution was also extracted from X-ray spectra of the

cluster measured by the Chandra x-ray observatory. The mass reconstructed from lensing

is a factor of 200 more abundant than the luminous mass extracted from the spectroscopic

measurements. As shown in figure 1.4, there is a clear offset between the visible center

of mass and the center of mass reconstructed from weak gravitational lensing. This off-

set arises because dark matter only interacts conventionally through gravity and not via

traditional electromagnetic forces. Consequently, during the collision, dark matter passed

through largely unaffected, while the visible matter, which interacts electromagnetically,

experienced drag and slowed down, leading to the observed separation between the two.

7



Figure 1.4: Composite image taken from ref. [13]of the galaxy cluster 1E 0657- 56, also known as
the “Bullet Cluster.” Pink depicts the hot gas measured by Chandra X-ray observatory and includes
most of the normal baryonic matter - Blue depicts the concentration of mass determined by weak
gravitational lensing

8



1.3 Dark Matter Candidates

It has has been conclusively shown that none of the particles in the Standard Model

can account for the properties of dark matter, as they either interact too strongly with light

or do not have the necessary mass and behavior. DM thus requires extensions beyond

the Standard Model of particle physics, which describes the known fundamental particles

and forces. This has led to the development of numerous theoretically well-motivated

dark matter candidate particles. Constraints on these candidates are typically calculated

one model at a time, but it is possible that dark matter is composed of multiple species

of particles. This collection of particles is often referred to as the dark sector. The dark

sector may involve complex interactions, similar to the visible sector, but these would be

extremely weak or invisible to current detection methods, making them difficult to study

directly.

There are specific properties that dark matter models must satisfy based on obser-

vational evidence. First, dark matter does not emit light, meaning any viable candidate

must have an extremely weak or vanishing electric charge and electric or magnetic dipole

moments. As a result, dark matter cannot radiate energy electromagnetically or ther-

malize through photon emission, making it approximately dissipationless. Additionally,

dark matter is known to be collisionless—if collisions occurred, dark matter halo profiles

would be spherical, but observations show that DM halos are clustered, with prolate shapes

in their centers and triaxial structures toward their outskirts [14]. Finally, dark matter

is non-relativistic (or ’cold’), as indicated by the formation of large-scale cosmic struc-

tures [15, 16].

Dark matter candidates are discussed extensively in reviews [17–20]. The following

sections contain first an overview of the general classes of DM candidates, WIMP dark

matter for context regarding the previous work in the community, and finally sub-GeV

9



dark matter for its relevance to the context of the work contained in this PhD thesis.

1.3.1 Classes of Dark Matter Candidates

The search for DM has expanded beyond Weakly Interacting Massive Particles (WIMPs)

and now includes a wide range of alternative candidates. These candidates can be broadly

classified into two categories: wavelike (or ultralight) and particle-like dark matter. The

classification depends on the de Broglie wavelength, λdB, which is given by:

λdB =
h
p

where h is Planck’s constant, and p is the momentum of the particle. This wavelength

provides insight into whether the particle behaves more like a classical particle or exhibits

wave-like properties. Wave-like behavior occurs when the de Broglie wavelength is com-

parable to or larger than the characteristic scale of the system, for example the the average

inter-particle separation in a galaxy like the Milky Way. Particle-like behavior is exhibited

when the de Broglie wavelength is much smaller than the relevant physical scales, causing

the dark matter to behave classically with trajectories similar to standard particles. These

two classes differ fundamentally in their mass, behavior, and detection strategies.

In general, wavelike dark matter consists of particles with extremely low masses. These

candidates include bosonic fields such as scalars or pseudoscalars [21]. Axions are a

prominent example, originally introduced to resolve the strong CP problem in quantum

chromo-dynamics (QCD) [22]. Their mass range is typically 10−22 eV to 10−5 eV, and

they are promising dark matter candidates due to their ability to form a coherent oscillat-

ing field. Axions can be detected through haloscope experiments like ADMX [23], where

they convert into photons in the presence of strong magnetic fields. Other novel detec-

tion techniques involve nuclear magnetic resonance (NMR) [24] and condensed matter

10



systems [25]. Dark Photons are another example of wave-like dark matter, arising from

a hidden U(1) gauge symmetry [26, 27]. With masses less than 1 eV, dark photons can

oscillate with ordinary photons, leading to observable signatures. Detection methods of-

ten involve photon mixing experiments, such as cavity searches or precise atomic clock

measurements.

Particle-like dark matter candidates arise in extensions of the Standard Model and can

range from keV to solar mass scales. Sterile Neutrinos [28] are one such candidate, with

masses in the range of keV to MeV. Unlike active neutrinos, sterile neutrinos do not interact

via the weak force, making them challenging to detect. Additionally, models proposing a

Dark Sector suggest the existence of dark baryons, analogous to protons and neutrons but

confined to a dark sector. This dark sector may involve its own set of forces and particles,

weakly interacting with Standard Model particles. Detection could arise from, missing

energy signatures in collider experiments, dark sector particles decaying into visible matter,

or direction detection searches [29–31].

1.3.2 WIMP Dark Matter

Of the possible DM candidates, the (particle-like) Weakly Interacting Massive Particle

(WIMP) paradigm has been the most tested experimentally over the past several decades.

Broadly, a WIMP is a proposed fundamental particle which interacts via gravity and an-

other force(s) that is non-vanishing in strength at the weak-scale or below, with a mass

range from a few GeV/c2 to several TeV/c2.

The primary WIMP production mechanism is thermal freeze out. In the early hot and

dense Universe, WIMPs interacted more frequently with ordinary matter due to the high

temperature. As the Universe expanded and cooled, the interactions between WIMPs and

other particles became less frequent due to lower particle densities. Eventually the expan-

11



sion rate of the universe exceeded the annihilation rate of the dark matter, resulting in a

reduced number of WIMPs and leaving a stable relic density.

Through observational evidence from cosmological surveys and insights from particle

physics experiments, the relic abundance of DM has been determined with high precision.

Any DM theory must account for this precise relic abundance in order to be considered

viable. The relic density of WIMP DM can be obtained from the Boltzman equation and

expressed as:

Ωχ ≈ 3×10−27 cm3 s−1

⟨σAν⟩ (1.2)

where ⟨σAν⟩ is the thermally averaged annihilation cross section for DM and h is the

dimensionless Hubble parameter. A particle with electroweak-scale interactions naturally

leads to a relic abundance that is consistent with the relic abundance derived from cosmo-

logical constraints. This has been referred to as the “WIMP miracle”.

WIMP experiments are generally designed to detect nuclear recoils caused by interac-

tions between WIMPs and atomic nuclei. The interaction cross-section for WIMPs with

nuclei is typically larger than that with electrons. Thus, the recoil energy deposited in a

nucleus from a WIMP collision is generally more significant than the energy deposited in

an electron collision.

Targets designed to probe WIMP interactions include noble liquids such as xenon and

argon. The LZ experiment is one of the largest-scale WIMP searches. The characteristic

nuclear recoil energy in liquid xenon is on the order of 10 keV, which is comparable to

the recoil energy from radiogenic backgrounds. Because of this, one of the primary design

considerations in WIMP experiments is the minimization of radiogenic backgrounds.

Another challenge facing WIMP experiments is the the intrinsic scaling of the interac-

12



tion rate. The number of WIMP interactions, nDM, is given by the relation:

nDM =
ρDM

MDM
(1.3)

where ρDM is the dark matter density and MDM is the mass of the dark matter particle.

As MDM increases, nDM decreases, making the interaction rate lower. For this reason,

WIMP experiments need to have extremely large target volumes to significantly probe the

parameter space. For example, the latest iteration of the LZ experiment features a target

mass of 7 tons of liquid xenon.

1.3.3 Axions and Axion-Like Particles (ALPs)

The concept of axions emerged in the 1970s as a solution to the strong CP problem in

quantum chromodynamics (QCD). The strong CP problem refers to the apparent absence

of a significant violation of the CP symmetry (combined Charge Conjugation and Parity)

in strong interactions, despite the fact that the QCD Lagrangian allows for a potential CP-

violating term:

LQCD =
θ

32π2 GµνG̃µν (1.4)

where Gµν is the gluon field strength tensor, and G̃µν is the axion field strength tensor.

The parameter θ is expected to be a small number, but its value is not constrained by

theory, and large values would induce strong CP-violation effects that are not observed

experimentally.

In 1977, Peccei and Quinn proposed a solution to this problem by introducing a new

symmetry (Peccei-Quinn symmetry) that spontaneously breaks, leading to what became

known as the axion. The axion is a light scalar particle that provides a dynamic mechanism

13



to relax the value of the θ -term to near zero, effectively solving the strong CP problem.

Axion-like particles (ALPs) are generalizations of the axion. These are hypothetical

light scalar particles that do not necessarily arise from the Peccei-Quinn mechanism but

share similar properties, such as interactions with photons and other particles.

The axion field a(x) can be described by the following Lagrangian density:

Laxion =
1
2

∂µa∂
µa− 1

2
m2

aa2 +gaγγaFµν F̃µν (1.5)

Here, ma is the axion mass, Fµν is the electromagnetic field strength tensor, and F̃µν is

its dual. The coupling constant gaγγ describes the interaction strength of the axion with two

photons. This term allows axions to couple to photons and can lead to observable effects,

such as axion-photon conversion in external magnetic fields.

Axion-like particles have similar interactions, but their Lagrangian can also include

interactions with other particles, depending on their origin:

LALP =
1
2

∂µa∂
µa− 1

2
m2

aa2 +gaγγaFµν F̃µν +gaψψaψ̄ψ (1.6)

where ψ represents a fermion field, and the second term introduces the possible inter-

action of ALPs with fermions.

RF resonant-cavity experiments, such as ADMX and HAYSTAC (Haloscope At Yale

Sensitive To Axion CDM), aim to detect axion dark matter by exploiting axion-photon con-

version in a strong magnetic field. The axion signal is expected to manifest as a narrow-

band electromagnetic wave at a frequency determined by the axion mass, resonantly en-

hanced by a tunable microwave cavity. ADMX and HAYSTAC focus on axion masses in

the range ∼ 1–40 µeV, corresponding to frequencies in the GHz range. HAYSTAC has

probed masses in the region of ∼ 4–20 µeV, complementing ADMX at slightly higher fre-

quencies. These experiments contribute to systematically excluding portions of the viable

14



axion parameter space.

Resonant-cavity experiments face several limitations that constrain their ability to ex-

plore the full axion parameter space. These setups are inherently narrow-band, sensitive to

a small range of axion masses at a time. Scanning a broader range requires precise tuning

of the cavity frequency, which is time and resource intensive. Additionally, the resonant

frequency of a cavity is determined by its dimensions, making it challenging to probe very

small axion masses (< 1 µeV) due to the prohibitively large cavity sizes required, or very

large axion masses (> 40 µeV), where resonant enhancement becomes inefficient. Noise

further limits sensitivity, with thermal noise dominating at lower frequencies and quan-

tum noise at higher frequencies, though the HAYSTAC collaboration has made significant

progress in overcoming the latter through quantum squeezing techniques. Finally, the weak

axion-photon coupling strength (gaγγ ) necessitates long integration times to detect potential

signals, adding to the challenges.

Despite significant progress by experiments like ADMX [23, 32–36] and HAYSTAC [37,

38], large regions of the axion mass parameter space remain unexplored. The low-mass

range (< 1 µeV) and high-mass range (> 40 µeV) are particularly challenging due to the

constraints of cavity design and sensitivity. Exploring these regions will require innovative

approaches and new technologies, as resonant-cavity designs are fundamentally limited in

their scalability. Developing broadband haloscopes is essential to overcome these chal-

lenges, enabling simultaneous searches over wide mass ranges and reducing the time re-

quired to explore the uncharted axion landscape. The Broadband Reflector Experiment for

Axion Detection (BREAD) collaboration introduces an innovative approach to search for

axion and dark photon dark matter across a wide range of masses, spanning approximately

µeV to eV, without requiring tuning. The experiment utilizes a cylindrical dish resonator

designed to fit within solenoidal magnets. In this setup, axions (in the presence of the mag-

netic field) or dark photons interact with the metallic surface of the dish, converting into

15



photons independent of their mass. These photons are subsequently directed by a parabolic

focusing reflector onto a highly sensitive single-photon counting detector.

For the pilot dark photon experiment within the BREAD program, the large-area mi-

crowire arrays developed and characterized in this thesis are currently being integrated as

the single-photon detector. Chapter 8 describes the characterization and integration of these

sensors into the experiment and plans for the future.

1.3.4 Dark Photons

Dark photons are hypothetical particles that are introduced as the gauge bosons (force

carriers) of a new U(1) symmetry associated with DM interactions. The idea behind dark

photons is that they could mediate interactions between dark matter particles and possibly

visible matter, but with a very weak coupling to the standard model particles. This idea is

motivated by the dark sector in theoretical physics, which posits the existence of a hidden

sector of particles that interact very weakly with ordinary matter.

In its theoretical framework, a dark photon can mix with the standard photon, leading to

the possibility of dark matter-photon interactions. This interaction is particularly interesting

for explaining anomalies observed in certain experiments, such as the g-2 anomaly in the

muon magnetic moment, and for providing a candidate for dark matter.

The Lagrangian for the dark photon can be written as:

Lγ ′ =−1
4

FµνFµν − 1
4

F ′
µνF ′µν +

ε

2
FµνF ′µν +m2

γ ′A
′
µA′µ (1.7)

where Fµν is the field strength tensor for the standard photon, F ′
µν is the field strength

tensor for the dark photon, and ε is a small parameter that represents the mixing between

the standard photon and the dark photon. The term m2
γ ′ corresponds to the mass of the dark

photon, and the mixing term enables the possibility for photon-dark photon oscillations.

16



The interaction between the dark photon and standard model particles is typically mod-

eled as a kinetic mixing term:

Lint = εFµνF ′µν (1.8)

This term leads to weak couplings between ordinary matter (via the electromagnetic

interaction) and dark matter (via the dark photon).

1.3.5 sub-GeV Dark Matter

The absence of a DM signal in existing WIMP direct detection experiments and the lack

of supporting non-Standard Model physics in collider experiments have led to interest in the

development of other candidates for DM. The lack of sizable interaction between DM and

ordinary matter motivate the hypothesis that dark matter is neutral under all standard model

forces but interacts through new forces that to-date have been undiscovered. This is referred

to as hidden sector DM. In several frameworks, the sub-GeV scale arises naturally and via

production mechanisms beyond the standard freeze-out. Each mechanism implies targets

in the DM parameter space that are accessible experimentally. The DM relic density can be

accounted for by mechanisms that describe portal interactions between hidden sector forces

and ordinary, for example via an exchange of a dark photon. Importantly, the parameter

space for these models is unexplored experimentally and also, in part, beyond the scope of

astrophysical bounds.

The higher mass region of the hidden sector parameter space is well-probed by WIMP

search experiments (few GeV to 100 TeV). The lower region of the hidden sector DM

parameter space below a few GeV is not well-probed by WIMP search experiments and

requires new detection strategies. This is referred to as sub-GeV dark matter.

One of the challenges in probing light dark matter via traditional nuclear recoil methods

17



lies in the kinematics; for lighter masses of DM, the total kinetic energy transferred to the

nucleus during a recoil is suppressed. Low nuclear recoil energies are a challenge to detect

and necessitate sensors with drastically reduced energy thresholds and background rates.

Another avenue to search for DM is through interaction with electrons in target ma-

terials as opposed to nuclei. This is more kinematically favorable, as a higher fraction

of kinetic energy can be transferred from a light DM particle to an electron via inelastic

scattering processes, for example. Due to the inverse relationship between interaction rate

and the DM mass shown in Equation 1.3, interactions can be probed with smaller target

volumes and exposure for lighter dark matter masses. For this reason, pilot programs for

sub-GeV DM searches are able to utilize table-top sized experiments.

The measurable signatures of hidden sector DM are dictated by the type of new force

coupling the DM to ordinary matter. As are described in benchmark models [39], a new

force can be mediated by scalar of vector bosons which couple to leptons or hadrons. The

bench-marked coupling models are described in detail in this reference for hidden sector

DM in general, and also the low-mass (sub-GeV) parameter space for hidden sector dark

matter .

For concreteness, sub-GeV DM that interacts with electrons through a dark-photon

mediator is described in the following section.

DM-electron scattering

Using the notation in reference [40], a DM particle χ can couple to the standard model

via a U(1) gauge boson A
′

(dark photon). A′ kinetically mixes with the SM U(1) field via

the following interaction:

L ⊃ ε

2cosθ
Fµν

Y F
′
µν (1.9)

where θ is the mixing angle, ε is the coupling strength, and Fµν

Y and F
′
µν are the field

18



strengths.

The scattering cross-section can be written as

σe =
16πµ2

χeαε2αD

(m2
A′ +α2m2

e)
2 ≈

{16πµ2
χeαε2αD

m4
A′

, mA′ ≫ αme

16πµ2
χeαε2αD

(αm4
e)

, mA′ ≪ αme

(1.10)

where µχe is the DM-electron reduced mass and α = g2
D (the U(1)D coupling.

The corresponding form factor can be written as

FDM(q) =
m2

A′ +α2m2
e

m2
A′ +q2 ≈

{1, mA′ ≫ αme

α2m2
e

q2 , mA′ ≪ αme

(1.11)

where q is the DM-electron momentum transfer. The DM form factor describes how the

interaction between dark matter particles and ordinary matter varies with the momentum

transfer in a scattering event which which is essential for calculating event rates in dark

matter detection experiments and understanding the interaction mechanisms. In a DM-

electron scattering interaction, the DM can transfer at most its kinetic energy 1/2mA′ν2,

where ν = 10−3c is the galactic DM velocity and c is the speed of light. The parameter

space for both regimes is shown in figure 1.6.

DM-electron absorption

Dark matter (DM) candidates such as axion-like particles (ALPs) and dark photons can

interact with ordinary matter via absorption, especially through interactions with bound

electrons. The absorption of dark photons A′ can be modeled as the absorption of a non-

relativistic particle with mass m with coupling ε to electrons [26, 27]. In absorption other

DM transfers its entire rest mass energy mA′c2.

19



The absorption of dark photons or axion-like particles occurs when these particles inter-

act with electrons bound in atoms. The process is commonly referred to as the axio-electric

effect for axions and axion-like particles. This interaction leads to the absorption of the

dark-matter particle, which transfers its energy to an electron, causing it to be ejected from

its atomic orbit. For a dark photon, this effect arises due to kinetic mixing with the photon

field, while for axions, the interaction is based on the axion’s coupling to the electromag-

netic field.

For dark photons, the absorption cross section can be expressed as:

σγ ′(E) =
ε2

m4
γ ′

(
E2

∆E2 +E2

)
(1.12)

where ε is the kinetic mixing parameter, mγ ′ is the mass of the dark photon, E is the

energy of the incoming dark photon, and ∆E is the energy width of the absorption process,

typically related to the material’s band gap or the electron binding energy.

For axion-like particles (ALPs), the absorption cross section is given by:

σa(E) =
g2

aγγ

m4
a

(
E2

∆E2 +E2

)
(1.13)

where gaγγ is the axion-photon coupling constant, ma is the mass of the ALP, and the

other terms are analogous to those for dark photons.

Current direct detection experiments such as XENON100 [41–43], CDMSlite [44], and

SuperCDMS [45–47] are sensitive to these types of interactions, especially with electron

recoil events. The cross section can be probed by observing the energy deposited by dark-

matter particles in the form of electron recoils, allowing for the constraining of the coupling

constants (ε for dark photons, gaγγ for ALPs) in regions of parameter space that were

previously unexplored.

However, the sensitivity of these experiments to lower mass range for dark-matter parti-

20



cles, especially below a few MeV, remains poorly probed. This region is difficult to explore

due to the relatively low recoil energies associated with light dark-matter candidates, which

are often below the detection thresholds of existing experiments. Moreover, many direct

detection experiments have better sensitivity at higher masses (e.g., GeV-scale DM), leav-

ing a significant gap in their ability to probe lighter candidates such as dark photons or

ALPs with small masses.

To overcome these limitations, future experiments are under development with the goal

of lowering the energy thresholds and increasing the sensitivity of detectors. These next-

generation experiments aim to probe the unexplored regions of parameter space where dark

photons and axions may exist. Theoretical studies have already begun to investigate the

potential for detecting bosonic DM absorption in materials with lower energy thresholds,

such as semiconductors and superconducting materials. These materials probe potentially

weakly coupled interactions of low-mass dark photons and axions, and thus could enable

the detection of these particles in new mass ranges [48, 49].

1.4 DM Detection Strategies

As shown schematically in figure 1.5, there are the three general categories of exper-

iments that provide complementary insight into the nature of dark matter. Indirect dark

matter detection strategies focus on the measurement of the products of dark matter inter-

actions (i.e. gravitationally) rather than the dark matter itself. Accelerator experiment aim

to produce dark matter directly or via decays of heavier new particles. Direct DM detection

experiments aim to directly measure the collisions between dark matter and a target ma-

terial in terrestrial apparatuses. Importantly, the verification of an excess as a dark matter

signal would require concurrent study and verification between all of the above channels

as well as consistency with expected daily and annual modulation of the signal [50, 51].

21



Figure 1.5: Schematic showing the possible channels of dark matter detection. Indirect DM detec-
tion involves searching for the standard model (SM) products of dark matter interactions and con-
straining annihilation cross sections. Collider searches seek to produce DM through high-energy
collisions. Direct Detection probe interactions of DM with a target material through a measurable
signal of the energy deposition.

Direct detection experiments search for DM interactions with nuclei or electrons in target

materials. Large-scale, mature experiments have placed exclusions on much of the DM pa-

rameter space above 1 GeV. With the theoretical developments and evidence from collider

experiments that DM might have a lighter mass (>GeV) than once previously thought, it is

important to devise novel detection strategies to explore the sub-GeV DM parameter space.

1.4.1 Direct Detection of sub-GeV Dark Matter

Many new experiments spanning a diversity of fields from atomic [52] and molecu-

lar [53] physics to condensed matter physics [54], [55], [56], [57] have been proposed to

probe sub-GeV DM via both electron and nuclear recoils. Figure 1.6 shows the approxi-

22



mate mass range of sub-GeV regime, along with some of the proposed experiments that are

under development.

WIMPAxion DM

(a)

�����

�$"0- сѷ ��-$*0. (�/ -$�'. !*- � / �/$)" .0�Ҋ� � ��-& (�// - �- 
$) 0. җ.*'$� '$) .Ҙ *- 0)� - �*).$� -�/$*) !*- /# ) �- / -( *- '*)"
/ -( җ'*)"Ҋ��.# � �)� .#*-/Ҋ��.# � '$) .Ѷ - .+ �/$1 '4Ҙѵ �# 
#*-$5*)/�' .��' $)�$��/ . /# (�..Ҋ. ).$/$1$/4 -�)" ѵ
�- �$/ѷ �ѵ �..$"ҝ�/*)4 �-**& �)$1 -.$/4Ѹ ���ҝ��-$) ��$)

.*Ҋ��'' � �*.*)$� ��-& (�// -Ѷ .0�# �. � ��-& +#*/*)Ѷ �)
�3$*)'$& +�-/$�' Ѷ *- � .��'�- +�-/$�' Ѷ 2$/# (�.. . �*2) /* р
( �ѵ 
) +�-/$�0'�-Ѷ $) .$'$�*)Ѷ � �*.*)$� ��-& (�// - +�-/$�' �.
'$"#/ �. р  �ҝc2 ��) � ��.*-� � җ$)./ �� *! .��// - �ҘѶ
� '$1 -$)" /* /#  ' �/-*) �)  ) -"4 �*-- .+*)�$)" /* /# 
+�-/$�' ҁ. (�.. �)�  3�$/$)" /#  ' �/-*) ��-*.. /# �-4./�' ��)�
"�+ѵ

�*( /�-" / (�/ -$�'. !*- � / �/$)" .0�Ҋ� � ��-& (�// -Ѷ .0�#
�. '$,0$� 3 )*)Ѷ '$,0$� �-"*)Ѷ .$'$�*)Ѷ �)� " -(�)$0(Ѷ #�1 
�  ) 0. � +- 1$*0.'4 !*- �
�� . �-�# .ѵ �/# - +-*+*. � /�-" /
(�/ -$�'. 2*0'� � ) 2 /* /# ��-& (�// - . �-�# �*((0)$/4ѵ
�# . $)�'0� ) 2 .�$)/$''�/*-.Ѷ .0+ -�*)�0�/*-.Ѷ .0+ -!'0$�
# '$0(Ѷ �$-�� (�/ -$�'.Ѷ +*'�- (�/ -$�'.Ѷ �$�(*)�Ѷ .$'$�*)
��-�$� Ѷ �-*(�/$� *-"�)$� /�-" /.Ѷ �)�(*' �0'�- "�. . җ�$"ѵ сҘѵ
�#�/ҁ. )  � �Ѷ #*2 1 -Ѷ $. � !0'' � .�-$+/$*) *! /# .$")�'
+-*�0� � �4 � ��-& (�// - +�-/$�' $)/ -��/$)" 2$/# /# . 
(�/ -$�'. $) / -(. *! /# +�-/$�' ҁ. +-*+ -/$ .ѵ 
) ���$/$*)Ѷ /# . 
.$")�'. )  � /* � ��'$�-�/ �  3+ -$( )/�''4ѵ �#$. 2*-&Ѷ 2#$�#
$. $) .*( ��. . �'- ��4 0)� -2�4Ѷ - ,0$- . � %*$)/  Ȃ*-/
� /2  ) - . �-�# -. $) +�-/$�' Ѷ �./-*+�-/$�' Ѷ �*)� ). �
(�// -Ѷ �)� �/*($�Ҋ(*' �0'�-Ҋ*+/$��' +#4.$�.ѵ

�-*($.$)" � / �/*-.
�)  Ȃ �/$1 . �-�# !*- '*2Ҋ(�.. ��-& (�// - - ,0$- . �
� / �/$*) ( �$0( �)� � . ).*- 2$/# �/ ' �./ /#-  +-*+ -/$ .ѵ
�$-./Ѷ /# ( �$0( #�. /* - ��/ /* /# � +*.$/$*) *! /$)4 �(*0)/.
*!  ) -"4ѵ � �*)�Ѷ /# ( �$0((0./ � .��'��' /* � '�-" 
)0(� - *! �/*($� ҂/�-" /.҃ /* $)�- �. /# �#�)� *! ��+/0-$)"
� -�- $)/ -��/$*) 2$/# � ��-& (�// - #�'* +�-/$�' ѵ �$)�''4Ѷ /# 
. ).*- �)�( �$0((0./ #�1 �)  3,0$.$/ '4 '*2 -�/ *!
� / �/*-Ҋ$)�0� � ���&"-*0)�. җ҂��-& �*0)/.҃Ҙѵ �#4.$�$./. #�1 
(��  )*-(*0. +-*"- .. *1 - /# +�./ ! 2 4 �-. $) �0$'�$)"
/# !$-./ " ) -�/$*) *! 0'/-�. ).$/$1 � / �/*-. /#�/ �- ��' /*
. �-�# !*- ��-& (�// - 2$/# (�.. . �. '*2 �. р � �ҝc2Ѷ �0/
(0�# - (�$). /* � �*) ѵ

�# !$-./ � (*)./-�/$*) *! . ).$/$1$/4 /* .$)"'  ' �/-*). ��( 
2$/# ��/� !-*( � -*0"#'4 рпҊ&" /�-" / �*).$./$)" *! '$,0$� 3 )*)Ѷ
2#$�# #�. �)  3�$/�/$*)  ) -"4 *! �-*0)� рс  � ҙтѶ рчҚѵ �*/#
�-"*) �)� '�-" - 3 )*)Ҋ��. � � / �/*-. #�1 .0�. ,0 )/'4
.#*2) $(+- ..$1 . ).$/$1$/4 /* � -/�$) ��-& (�// - ��)�$��/ .
�. '$"#/ �. �-*0)� ф � �ҝc2 ҙршѶ спҚѶ �0/ !0-/# - +-*"- .. $.
�0-- )/'4 #�(+ - � �4 ���&"-*0)�.Ѷ �.  3+'�$) � � '*2ѵ

�) � 1$� /#�/ #�. .  ) $)�- �$�' +-*"- .. - � )/'4Ѷ $)�'0� .
� .+ �$�' ҂�&$++ -҃ �#�-" Ҋ�*0+' � � 1$� җ���Ҙ ҙсрѶ ссҚѶ
�*).$./$)" *! ($''$*). *! .$'$�*) +$3 '. /#�/ �- рф (( × рф ((
× хцф (( $) .$5 ѵ �#$. � 1$� $. ��' /* +- �$. '4 ( �.0- /# 
)0(� - *!  ' �/-*). $)  ��# +$3 'Ѷ +-*1$�$)" .0Ȃ$�$ )/
. ).$/$1$/4 /* .+*/ � �*''$.$*) � /2  ) � �ҝc2 ��-& (�// - �)� �
.$'$�*) �-4./�' /#�/ +-*(*/ . �)  ' �/-*) /* /# �*)�0�/$*)
��)� җ�$"ѵ тѶ ' ȅҘѵ ���# �&$++ - ��� #�. � (�.. *! *)'4 ��*0/ с
"ѵ � !$-./ ��-& (�// - . ).$/$1$/4 / ./ 2$/# � .�$ )� Ҋ"-�� ���
#�. �  ) �*)�0�/ � �4 /# �����
 �*''��*-�/$*) �/ � -($'��
ҙссҚ җ.  �4)*+.$.ѷ ��-& ��// - � / �/*- �-*1 . $/.
� ).$/$1$/4Ҙѵ �# ) 3/ ./ +Ѷ +'�)) � !*- спсрѶ $. /* $)./�'' �
-*0"#'4 рппҊ" � / �/*-Ѷ �*).$./$)" *! ��*0/ фп �&$++ - ���.Ѷ �/
��)���ҁ. 0)� -"-*0)� '��*-�/*-4 ���
��ѵ �# ���
�Ҋ�
�*''��*-�/$*) $. �'.* +'�))$)" /* $)./�'' � рҊ&" � / �/*- �/ /# 
�*��) �)� -"-*0)� 
��*-�/*-4 $) �-�)� ҙстҚѶ �)� /# ��
� +�-/( )/ *! �) -"4 җ���Ҙ $. !0)�$)" /# �о� *! � (0'/$Ҋ&"
� / �/*-Ѷ ��'' � �.�0-�Ѷ 2#$�# �*0'� /�& ��/� '�/ - /#$. � ��� 
ҙсуҚѵ �# �*)./-0�/$*) *! /# . '�-" Ҋ(�.. � / �/*-. $.
�#�'' )"$)"ѵ �# ���. )  � /* � .0Ȃ$�$ )/'4 '*2 �*./Ѷ /# 
- ��Ҋ*0/  ' �/-*)$�. (0./ - �� *0/ (�)4 ���. �/ *)� �)� 2$/#
'*2 )*$. Ѷ �)� /# -��$*��/$1 ���&"-*0)�. (0./ � �*)/-*'' �
/* $)�- �$�' +- �$.$*)ѵ

+#4.$�.ѵ�+.ѵ*-" Ҟ Ҷ спсп �( -$��) �#4.$��' �*�$ /4 Ҟ � � (� - ррѶ спсп Ҟ �#4.$�. ртѶ рцс Ҟ ��
ѷ рпѵррптҝ�#4.$�.ѵртѵрцс ��" т

(b)

sub-GeV DM

Figure 1.6: (a) Parameter space of dark matter (DM) mass showing the sub-GeV regime between
WIMP and axion DM predictions. (b) Experiment proposals for sub-GeV dark matter searches.
Materials in use are indicated by solid lines, while those under consideration for the near-term
and long-term are shown by solid-dashed and light-dashed lines, respectively. The horizontal scale
indicates the mass-sensitivity range. Figure adapted from ref. [58].

1.4.2 n-type GaAs as a Cryogenic Scintillating Target

Scintillators are a widely adapted experimental target for direct dark matter searches [59].

In direct DM searches, when dark matter particles interact with electrons in a scintillating

material, they deposit energy, causing the material to emit photons through radiative de-

23



Acceptor Energy Level Wavelength
Shallow defects 1.44 eV 861 nm

Ionized boron on arsenic site 1.33 eV 930 nm
Si˙GaV˙Ga complex 1.16 eV 1069 nm

Si˙GaV˙GaSi˙Ga complex 0.93 eV 1333 nm

Table 1.1: Radiative acceptors in n-type GaAs and their associated energy levels and wavelengths.

cays. These photons are then detected by sensitive detectors, providing a measurable signal

that can be analyzed to infer the presence of dark matter. As a novel cryogenic scintillator,

n-type Gallium Arsenide (GaAs) features a band gap of 1.52 eV and desirable properties

as a target for the MeV/c2 mass range, which falls within the poorly explored sub-GeV

regime [60]. An n-type GaAs target coupled to a single photon detector is the system of

interest that this work will focus on. N-type GaAs doped with silicon (Si) and boron (B)

was first proposed as a cryogenic scintillator for the detection of sub-GeV DM via elec-

tron recoils due to its high luminosity and the commercial availability of large, high-purity

crystals[61–63]. Scintillation photons are emitted from n-type GaAs via the following pro-

cess:

(1) Electronic excitation leaves one or more holes in the valence band

(2) Acceptors trap the valence band holes

(3) The acceptor holes recombine radiatively with delocalized donor band electrons and

scintillation photons are emitted

The emission wavelengths are thus dependent on acceptor energy levels. Known accep-

tors and their associated energy level/corresponding scintillation wavelength are listed in

table 1.1 [64]. Scintillation in the 930 nm band is the most dominant. Figure 1.7 adapted

from ref [62]shows the x-ray excited luminescence spectra of n-type GaAs doped with

silicon and boron measured with silicon and InGaAs CCDs at 10K, where the contribu-

tions from these different radiative decay processes can be seen. The relative intensities of

24



> REPLACE THIS LINE WITH YOUR PAPER IDENTIFICATION NUMBER (DOUBLE-CLICK HERE TO EDIT) < 
 

3 

doped GaAs crystals. But in the framework of this work, such 
features will not be considered. For simplicity, these emission 
peaks were collected in one group – band A. The nature and 
specific of these short wavelength emissions can explains the 
position and asymmetric shape of band A in GaAs:Si and 
GaAs:(Si, B) (see Fig. 2). As can be seen from the experimental 
and literature data, the shape, intensity and peak position of the 
bands in 800-900 nm region in GaAs crystals drastically 
depends on purity, type of dopants and defects [11-15]. And it 
also depends, as shown in Reference [13], on the type of 
conductivity. 

 The transitions from shallow silicon donors to boron 
acceptors on an arsenic site leads to emissions at 930 nm (band 
B) [10, 16, 17]. 

The energy relaxation on gallium vacancy-donor center leads 
to emission at ~1070 nm [18]. This center (band C) is a gallium 
vacancy (VGa) bound to a donor (like Si, Ge, Sn, C, S, Se, Te). 
The peak position of band C depends on the type of activator 
impurity. 

The nature of the broad band D at ~1335 nm is associated 
with defects or complex centers like (SiGa + VGa + SiGa) or 
(SiGa + SiAs) and other [11, 12]. It is important to note that the 
formation of long-wavelength centers at ~1377 nm was 
observed in undoped GaAs crystals with neutron dose. The 
intensity of band D increases with neutrons dose rate 
increasing, which indicates the important role of vacancies in 
the formation of this emission center [12]. 

 

B. Light Yield Estimation 
The photocurrent of all samples under X-ray was measured 

using the InGaAs or Si photodiodes with or without filters and 
shown on Fig. 3. Table III compiles the total light output and 
the contribution of each emission band separately which were 
obtained in this work based on PIN detectors photocurrent. 

The obtained results indicate a very high efficiency of the 

scintillation process in activated crystals. It is seen that the LY 
of the GaAs:(Si, B) is very high: 125 ph/keV. The Si doped 
GaAs crystal has LY=71 ph/keV. It is important to note that 
about 65% of the total output of activated GaAs is contained in 
the region of 1000-1700 nm, i.e. bands C and D. This feature is 
dictated by the specificity of the radiative relaxation of energy 
in the co-doped GaAs crystal. The results show how important 
the addition of the acceptor boron is. Also, this indicates the 
need to use photo detectors that are sensitive in a fairly wide 
range (like Ge, InGaAs). Silicon-based photodiodes are only 
partially suitable as photodetectors, since 35 to 60% of the light 
is lost in case of GaAs:(Si, B) and GaAs:Si respectively. 

It should be noted that the experimental approach used to 
determine the light output does not take into account the 
refractive index. GaAs has a refractive index of 3.55 at 930 nm 
and 3.4 at 1300 nm. Since the scintillation light is emitted 
isotropically, only a small fraction is able to exit the crystal 
directly. In this experiment, scattering within the crystal or on 
the roughened surface allowed a large fraction of the light to 
exit. A coating with a graded index of refraction will allow 

 
Fig. 2. The X-ray luminescence spectra of undoped GaAs, GaAs:Si and 
GaAs:(Si, B) crystals obtained by Si (left column) and InGaAs (right column) 
CCDs at 10K. 
 

TABLE II 
THE PARAMETERS OF RADIATIVE CENTERS IN GAAS CRYSTALS 

Band λ, nm E, eV Comments 

A 800-870 1.42-1.54 
Transition from CB to 
VB, exciton and shallow 
levels (defects) 

B 933 1.33 
The boron in arsenide site 
(BAs center) 

C 1050-1090 1.14-1.18 Center like (SiGa + VGa) 

D 1305-1335 0.93-0.95 
Complex center like 
(SiGa + VGa + SiGa) or 
(SiGa + SiAs) 

 

 
Fig. 3. The InGaAs (Top) and Si (Bottom) PIN photodiodes current 
dependence on X-ray filament current with/without the different shortpass 
(FES) and longpass (FEL) optical filters obtained on the GaAs:(Si, B) (Left) 
and GaAs:Si (Right) sample at 10 K. 

Figure 1.7: X-ray excited luminesence spectra of n-type GaAs doped with silicon and boron mea-
sured with a silicon CCD (left) and an InGaAs CCD (right) showing the near-infrared emission
bands. Figure adapted from ref.[62].

the emission bands depend on the dopant and free carrier concentrations, as do the decay

timescales.

One of the most desirable features of n-type GaAs compared to other cryogenic scintil-

lators is its lack of afterglow [60], the emission of scintillation photons at long time scales

after a particle interaction has occurred. In order to avoid these afterglow effects in GaAs,

the samples must be doped above the so called Mott transition [65]. In cooling the crystals

from room temperature to below 1 K, some of the free carriers bind to individual silicon

atoms at an energy level 2.3 meV below the conduction band, and repulsion then forces

the additional electrons into the next higher available energy level, which is in the conduc-

tion band. Electrons in the conduction band remain highly mobile down to 0 K, and they

serve two purposes: (1) annihilation of any meta-stable radiative states that would result

in unwanted afterglow, and (2) efficient combination with electron holes to maximize the

prompt scintillation light (the desired signal).

To demonstrate this, figure 1.8 adapted from reference [63] shows the intensity of the

scintillation light emitted from a GaAs sample at 10K of sample before, during, and after

25



as the system returns to its ground state. A closer look at the
GaAs emission curve [Fig. 8(b)] shows that it is indistinguish-
able from the instrumental background. The apparent absence
of thermoluminescence in n-type GaAs can be explained by the
annihilation of metastable radiative states by the delocalized
conduction band electrons. A search of the literature found only
one paper describing thermally stimulated luminescence from a
conductive semiconductor.32 This paper described thermolumi-
nescence from an n-type ZnSe crystal using UV (365 nm) exci-
tation. This was later explained as the absorption of the UV
exciting radiation (attenuation length 10!4mm) by a layer of
non-conductive surface defects.33

IV. CONCLUSIONS

(1) Electron excitations "1.44 eV can produce 1.33 eV
photons.

(2) The most luminous GaAs sample studied (#13316) has
an observed scintillation luminosity of 43 photons/keV,
below the theoretical limit of about 200 photons/keV.

(3) Ionization from background radiation (e.g., muons and
gamma rays) is not expected to cause afterglow, pro-
vided that the free carrier concentration is above the
Mott transition, where even near 0 K conduction band
electrons are not frozen out and can annihilate any meta-
stable radiative states.

These measurements show that n-type GaAs is a promis-
ing cryogenic scintillator for DM particle detection in the
MeV/c2 mass range in that it can be grown as large, high-
quality crystals, and has good scintillation luminosity, a
threshold sensitivity at the 1.52 eV bandgap, and potentially
no afterglow. However, more work is needed to optimize the
doping concentrations, to reduce hole traps that compete
with boron acceptors, and to develop anti-reflective coatings

FIG. 7. (a) Optical emission of sample 13 316 at 10 K before, during, and after exposure to a 50 keVp X-ray beam, measured in 2 s time bins. (b) Enlargement
of emission around the time the beam was turned off.

FIG. 8. (a) Thermally stimulated luminescence of GaAs sample #13330 and crystals of NaI(Tl), NaI, and CsI as a function of temperature after a 30-min expo-
sure to a 50 keVp X-ray beam at 10 K. All crystals were similar in size. The vertical scale is the same for all samples. (b) Comparison between the GaAs curve
and the instrumental background with vertical scale enlarged 20x.

114501-5 Derenzo et al. J. Appl. Phys. 123, 114501 (2018)

 09 N
ovem

ber 2024 19:55:39

Figure 1.8: Figure adapted from ref [63] showing (a) the normalized intensity of emitted scintillation
light before, during, and after irradiation with 50 keVp X-ray beam and (b) an inset showing the
time during which the shutter closed. Of note is the sharp return to a baseline around 0 intensity
after irradiation.

a 600 second exposure to the 50 keVp X-ray beam. The scintillation intensity is cleanly

correlated with the opening and closing of the excitation beam shutter with no lingering

emission after irradiation, as shown by the return to a baseline around 0 intensity.

1.4.3 DM Detection concept using n-type GaAs

In the detection concept that is the focus of this work, photons are emitted from n-

type GaAs after excitation by DM-electron scattering or absorption of DM by an electron

and subsequent radiative recombination. The near-infrared scintillation photons are then

detected by an SNSPD. The signal associated with a DM – electron interaction depends on

the mass of the DM particle and the interaction. In a DM-electron scattering event, for the

case that the DM particle is heavier than an electron (the case relevant for n-type GaAs as a

target material), the maximum energy transferred is the kinetic energy of the DM. Given a

galactic DM velocity vχ = 10−3 c, where c is the speed of light, a 10 MeV/c2 DM particle

transfers

26



Figure 1.9: Schematic depiction of the detection concept: a DM particle scatters off (or absorbs
onto) an electron in n-type GaAs, exciting it to a higher-energy level; scintillation photons are
emitted through radiative recombination processes; these are subsequently detected by the SNSPD
array

27



Ee ≤
1
2

mχv2
χ ≲

1
2
(10MeV)2 (10−3c

)2
≲ 5eV [39]. (1.14)

A 5 eV energy deposit into n-type GaAs leads to the emission of 3–4 scintillation pho-

tons given its 1.52 eV band-gap.

In a DM-electron absorption event, the DM transfers all of its rest mass to the electron.

The energy transferred to an electron is thus given by simply the DM mass [39]:

Ee ≈ mχc2
χ ≈ mχ . (1.15)

In principle, using n-type GaAs as a target, bosonic DM (dark photons, scalars, and axion-

like particles) can be probed via absorption onto electrons in n-type GaAs for DM masses

down to 1 eV [48].

Figure 1.10 shows projections for GaAs as a scintillating target for dark photon absorp-

tion with sensitivity down to one or more photons.

Figure 1.11 shows projected sensitivities of the dark matter - electron scattering cross

section as a function of dark matter mass for GaAs targets, given exposures of 1g-month to

10kg-years for two different form factors.

1.4.4 Sensor Classes for n-type GaAs and Low Mass DM Targets

The sensor requirements for the scintillation photons emitted from n-type GaAs are

high sensitivity in the near-infrared, low dark count rate, and cryogenic operation. Given

this, superconducting devices are natural candidate to develop detection concept. It should

be noted that the development of SNSPDs for this detection concept will extend to many

other experiments in the dark matter detection landscape in the future.

SNSPDs are typically operated at sub-kelvin cryogenic temperatures and have been

28



J
H
E
P
0
6
(
2
0
1
7
)
0
8
7� �� ��� ��� ���

��­��

��­��

��­��

��­��

��­��

��� [��]

ϵ

��
��
��
� �
��
���
��
��
������� ���������

�����������

��� �������

���������
��

��������� ��

Figure 3. Constraints (shaded regions) and prospective sensitivities (solid colored lines) for axion-
like particle (ALP) dark matter (left) and dark-photon (A′) dark matter (right), assuming that
the ALP/A′ constitutes all the dark matter. Colored regions show constraints from XENON10,
XENON100, and CDMSlite, as derived in this work, as well as the DAMIC results for A′ from [85].
Shaded bands around XENON10 and XENON100 limits show how the bound varies when changing
the modeling of the secondary ionization in xenon. Deep- and light-purple solid lines show projected
90% C.L. sensitivities for SuperCDMS SNOLAB HV using either Ge (20 kg-years) or Si (10 kg-
years) targets, respectively. Yellow, orange, and green solid lines show projected sensitivities for
hypothetical experiments with the scintillating targets CsI, NaI, and GaAs, assuming an exposure
of 10 kg-years. All projections assume a realistic background model discussed in the text, but zero
dark counts to achieve sensitivity to low-energy electron recoils. In-medium effects are included for
all A′ constraints and projections. Shaded gray regions show known constraints from anomalous
cooling of the Sun, red giant stars (RG), white dwarf stars (WD), and/or horizontal branch stars
(HB), which are independent of the ALP or A′ relic density. Also shown (left) are the combined
bounds from XENON100 [22], EDELWEISS [21], CDMS [20], and CoGeNT [19]; and (right) a
bound derived in [35] based on XENON100 data from 2014 [22]. Shaded orange region in left plot
is consistent with an ALP possibly explaining the white dwarf luminosity function.

– 12 –

Figure 1.10: Constraints (shaded regions) and projected sensitivities (solid colored lines) for dark
photon dark matter. The yellow, orange, and green solid lines show projected sensitivities for hypo-
thetical experiments with the scintillating targets CsI, NaI, and GaAs, assuming an exposure of 10
kg-years. Figure adapted from ref. [48].

29



� �� ��� ���
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��

�χ [���]

σ
�
[�
�
�
]

�������

�
�
�
�
�
���

�
����

����
�

�
�
�
�-
�
�
��

�
�
�
��

�������
����@�

������

�������
����@�

����

�
�
�
�
�
��

���
���������

���=�
�����

�� ��-
���������

� ��-�
���

����� ��� �-������

����� � �-������

�����
�@���

��

��� ��� ��� ��� ���
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��
��-��

�χ [���]

σ
�
[�
�
�
] �������

���
���

��

���
���

�����
�����

�

�����

�����
����

��@�
�����

�

���
�-�

���

�����
���

���@
���

��

���=(�/�)
�

��� �
��������

������@�����

���
�� ��

��-�
������

�� � �
�-��

��

���
�� ���

�-�
����

�

���
�� � �

-��
����

Figure 1.11: Projected sensitivities (blue lines) to dark-matter-electron scattering mediated by a
heavy mediator (left) or light mediator (right) of a GaAs target assuming for different exposures
(1g-month, 100g-month, 1 kg-year, 10 kg-year). We assume zero background events for events
with one or more photons, a radiative efficiency of 1, and a photon detection efficiency of 1. Ex-
isting constraints are shown in gray from SENSEI, DAMIC at SNOLAB, XENON10, XENON100,
XENON1T, DarkSide-50, and CDMS-HVeV [66–75]. Orange regions labelled “Key Milestone”
are from [76].

demonstrated with near-unity system detection efficiency in the NIR and ultra-low dark

count rates. The main challenge with SNSPDs in the context of instrumenting large-scale

scintillating targets is scaling the active area of SNSPDs to the cm2-regime while preserving

these detection metrics to the greatest extent possible.

As an alternative sensor, TES-based technologies are under development to instrument

n-type GaAs. TES are somewhat maturely developed for nuclear recoil DM experiments.

In large collaborations such as SuperCDMS [45], large-area cryogenic detectors, with a to-

tal area of 45cm2, utilizing TES-based athermal phonon sensors have achieved 4 eV energy

sensitivity [45]. Given that σE ∝
√

A, where A is the total instrumented area, resolutions of

600meV in 1cm2 detectors are feasible given the current state-of-the-art. Extending ather-

mal phonon sensors to lower mass electron recoil DM searches will entail design features

such as lowering the critical temperature (Tc) of the TES and optimizing sensor geometry

30



to improve athermal phonon collection efficiency. This would enable further enhancements

in sensitivity and expand the instrumented active area per channel beyond 1cm2.

In GaAs, both phonons and NIR photons are produced after excitation. The TESSER-

ACT collaboration is developing both nuclear recoil and electron recoil experiments with

the approach of exploring multiple target materials and developing multiple detection strate-

gies. One of the developing ER experiments involves n-type GaAs instrumented with TES-

based athermal phonon sensors. The prototype design includes a 1 cm3 GaAs target in-

strumented with athermal phonon sensors on its surface, and inside an optical cavity with a

1 cm3 x 1 mm thick Ge crystal that serves as an infrared photon collector [77]. TES readout

offers the possibility to use the same detector to measure both scintillation and phonons in

coincidence which can give background discrimination capabilities. The main challenge to

TES readout for low energy threshold DM experiments is the low energy excess (LEE), the

observed increase in event rates or signals at low energy thresholds that are not accounted

for by known backgrounds or expected detector noise. Nearly all experiments instrumented

with phonon sensors with sufficiently low energy thresholds see a this steeply rising event

rate [78–81]. Efforts to understand the origin of [82] and mitigate the LEE are ongoing.

31



Chapter 2: Superconducting Nanowire Single Photon Detectors

2.1 Introduction

Since their first demonstration by Golt’sman et al. in 2001 [83], superconducting nanowire

single photon detectors (SNSPDs) have emerged as detector for time correlated single pho-

ton counting from the UV to the near-infrared. Specialized devices have shown detection

efficiency of 98% at 1550 nm [84], timing jitter as low as 2.7 ps [85], intrinsic dark count

rates of 6×10−6 counts per second [86], and energy thresholds as low as 42.8 meV (wave-

length 29 µm) [87]. Many of these demonstrations have involved detectors that are opti-

mized for one performance metric at the expense of others. As a relatively new detector

technology, one of the challenges for SNSPDs is to realize devices that perform well across

multiple metrics. This continues to develop as our understanding of the fundamental device

physics and intrinsic trade-offs becomes more complete.

SNSPDs have made their way into a wide variety of applications, including quantum

optics [88], [89] quantum information science [90–95], imaging [96, 97] and free-space

laser communication [98]. Most of these demonstrations involve fiber-coupled light in-

cident on a small, circular active area. Driven by imaging [96] and astronomy[99] ap-

plications, the active areas of SNSPDs have grown steadily over the past several years

from the order of the cross section of a single-mode optical fiber core (10s of µm2) to ar-

rays with active areas in the mm2 regime [100]. Large-scale arrays up to kilopixel [101]

and now megapixel-scales [102] have been developed using row-column [103] or ther-

32



mal row-column multiplexing techniques[104] that provide spatial resolution, and time-of-

flight readout bus multiplexing for large scale cameras [102].

Recent developments in nanofabrication, device characterization, and modeling of fun-

damental device physics have pushed forward the performance of SNSPDs in their his-

torical strengths as single photon detectors, revealing novel applications in fundamental

physics. The project described in this paper will focus on the development of SNSPDs

for dark matter searches using cryogenic scintillating targets. With sufficient research and

development of the technology, these SNSPDs have the ability to provide new capabilities

for low-mass dark matter searches and across many areas of High Energy Physics.

2.2 Device Operation Principles

SNSPDs are constructed from current-carrying nanoscale wires. Typical devices have

wire widths from 10s to 100s of nm and thicknesses of 2-10 nm. A common technique to

achieve high optical coupling efficiency and absorption is to meander the nanowire over the

active area with a high fill factor. SNSPDs are usually operated from 1–4 K depending on

the superconducting materials used (WSi, MoSi, NbTiN, NbN, etc.) with a DC bias current

applied through a bias tee. In parallel with the nanowire is the readout arm which consists

of a load impedance, usually the input to a readout amplifier. The readout portion can be

DC or AC coupled. A schematic of the photon detection mechanism in SNSPDs is shown

in figure 1. The detection cycle begins with the nanowire in the current-carrying supercon-

ducting state. When a photon is absorbed in the cross section of the nanowire, it triggers a

cascade of events that ultimately leads to a detection signal. Qualitatively, this cascade can

be understood as follows: absorption of a photon leads to the formation of electron-hole

pairs, which through a down-conversion process transfer their energy into a hot-spot of ex-

cited quasi-particles and high energy phonons. The evolution of this hot-spot leads to the

33



4

Figure 1.1: (a) Schematic of the SNSPD detection process. The device begins in
the current-carrying superconducting state (1) until a photon is absorbed (2) and
generates a hotspot of excited quasiparticles and high-energy phonons through a
process of downconversion shown in (b). The evolution of these quasiparticles
leads to the instability of the superconducting state, resulting in the formation of
vortices (vortex-antivortex pair generation if the hotspot is in the center of the
nanowire or vortex entry from the edge if the absorption is near the edge) as shown
in (c). As the vortices move due to the current flow (3), they dissipate energy,
leading to the formation of a normal domain across the entire cross-section of the
nanowire. Once formed, the normal domain grows along the length of the wire
due to Joule heating (4), increasing the impedance of the device. This change in
impedance diverts current from the nanowire to the readout circuitry, leading to
a voltage transient (center). Once current is diverted to the readout, the nanowire
cools, recovers to the superconducting state, and current returns to the device. (d)
A typical SNSPD is operated using a bias-tee with the DC port carrying the bias
current and the RF port coupled to a low-noise amplifier.

Figure 2.1: (a) Schematic depiction of the SNSPD photon detection mechanism reproduced
from [105]. The device begins in the current-carrying superconducting state (1) until a photon is
absorbed (2) which generates a hot-spot of excited quasi-particles and high-energy phonons through
a down-conversion process depicted in (b). The evolution of this hot-spot destabilizes the supercon-
ducting state and leads to the formation of vortices, which are depicted in (c). Current flow causes
the vortices to move (3), and they dissipate energy, causing the formation of a normal domain across
the entire cross-section of the nanowire. The normal domain propagates along the length of the wire
(4), which increases the impedance of the device. This change in impedance diverts current from
the nanowire to the readout circuitry, leading to a voltage transient (center). Once current is diverted
to the readout, the nanowire cools, recovers to the superconducting state, and current returns to the
device. (d) A typical SNSPD is operated using a bias-tee with the DC port carrying the bias current
and the RF port coupled to a low-noise amplifier.

34



instability of the superconducting state, and vortices form. The dissipative vortex evolution

leads to the formation and growth of a normal domain which spreads along the length of

the nanowire, increasing the impedance of the device, until the nanowire transitions fully

into the normal state. At this time, the bias current gets diverted into the readout electron-

ics. The nanowire then cools and recovers to its superconducting state, and its impedance

decreases until the bias current returns fully to the device. In typical SNSPDs, the photon

absorption and hot-spot formation phase typically occurs on a picosecond timescale, and

the recovery phase timescale which is set by the kinetic inductance of the nanowire, is on

the order of nanoseconds[105].

Traditional SNSPDs are comprised of nanowires with widths O(100 nm) meandered

over the photoactive area and are fabricated using electron-beam lithography (EBL) to

achieve the required resolution and optimal performance. Recent advances in the prepara-

tion of thin superconducting films have enabled the fabrication of a novel class of single-

photon detectors with wire widths of several micrometers. Microwire devices offer many

advantages over traditional nanoscale devices for applications such as dark matter searches

where large active areas area required, as they are amenable to fabrication via optical lithog-

raphy - a more scalable fabrication process than electron beam lithography which is used

to fabricate traditional devices [106].

2.3 Detection Metrics

This section reviews some of the detection metrics of SNSPDs, following ref [107] and

the current state-of-the-art for each metric.

35



2.3.1 Detection Efficiency

The system detection efficiency (SDE) is the probability of registering an electrical out-

put signal produced by a photon once the photon has entered into the the photon detection

system. The SDE can be thought of as being comprised of multiple components:

SDE = ηcoupling ·ηabsorb ·ηinternal ·ηtrigger, (2.1)

where ηcoupling is the efficiency of the delivery of light to the active area of the device,

ηabsorb is the probability that a photon incident on the detector’s active area gets absorbed

into the material, ηinternal is the probability that an absorbed photon results in an observable

output pulse, and ηtrigger is the efficiency of the readout electronics in registering the output

signal as a detection event.

While conceptually useful, it is often difficult to accurately measure each separate com-

ponent of SDE. Instead, it is conventionally measured by using the following relation:

SDE =
Rlight −Rdark

Rincident
, (2.2)

where Rlight is the recorded count rate when the device is under illumination, Rdark is the

count rate recorded when the device is not under illumination (background count rate), and

Rincident is the incident rate of photons delivered to the detector.

The current state-of-the-art result for the highest SDE is 98% for the TE polarization

(electric field parallel with the nanowire grating) at 1550 nm in a MoSi SNSPD developed

by the NIST group [84]. This was achieved using an optical stack with distributed Bragg

reflector (DBR) mirrors and a 50-µm diameter active area.

36



2.3.2 Dark Count Rate

A dark count in a photodetector is defined as an output pulse that registers as a detection

event in the absence of an incident photon. The dark count rate (DCR) can be expressed as:

DCR = Rintrinsic +Rbackground +Relectronic. (2.3)

The intrinsic dark counts, Rintrinsic, in SNSPDs result from spontaneous resistive state

formation. At the temperatures at which SNSPDs operate (1–4 K), it has been shown that

the primary phenomenon underlying intrinsic dark counts is thermally activated vortex-

anti-vortex pair unbinding [108] or vortex entry and crossing from the nanowire’s edge.

This effect is exacerbated by geometric factors [109, 110], such as constrictions in the

nanowire width or sharp bends at the ends of meanders, where it is observed that dark

count rates increase [110]. In these regions, increased current density from current crowd-

ing [111–114] lowers the barrier for both vortex entry and vortex-anti-vortex unbinding.

Background counts, Rbackground, primarily result from spurious photons reaching the

detector’s active area, originating from black-body radiation outside or within the cryo-

stat. Spurious photons can also arise from Cherenkov radiation in materials surrounding

the detector or from transition radiation from charged particles passing through bound-

aries between media with different dielectric properties. The background count rate in a

given system, Rbackground, can be reduced through appropriate shielding around the detec-

tor, the implementation of light-tight packaging, the use of mid-IR absorbing paint on areas

surrounding the detector in the cryostat, and spectral filtering to attenuate the black-body

radiation reaching the detector. This is a particularly important consideration for rare-event

search applications such as dark matter detection.

The electronic noise component, Relectronic, arises from electronic noise which causes

37



the readout electronics to register a detection click that did not originate from the detector

itself. This can be mitigated by using low-noise cryogenic amplifiers, or if the waveforms of

an SNSPD are measured on an oscilloscope, pulse shape discrimination between an actual

SNSPD output pulse and a detection event caused by electronic noise can be implemented.

The current record for the lowest DCR observed in an SNSPD was measured to be

6× 10−6 cps in a WSi SNSPD with an active area of 400 µmby 400 µm, measured at 300

mK [115].

Chapter 4 of this thesis details the study of the dark count rates in large-area microwire

arrays and efforts.

2.3.3 Timing Resolution

Because SNSPDs are used for time-correlated single-photon counting, timing resolu-

tion is an important detection metric. The timing resolution is typically characterized by

the instrument response function (IRF), which is measured by comparing the detector re-

sponse to a reference. Typically, the full width at half max (FWHM) of the IRF is taken

to be the timing jitter of the device. The measured IRF has contributions from intrinsic

and extrinsic IRFs. Extrinsic contributions from electrical and amplifier noise can broaden

the IRF [116, 117], and a lot of work has been done in the SNSPD community to engineer

low-noise cryogenic amplifiers and readout to minimize this. Intrinsic contributions include

contributions from the stochastic nature of energy down-conversion in the superconductor

after photon absorption and those from geometric jitter, geometric inhomogeneities, and

constrictions in the device. These dominate in large-scale meander arrays. Chapter 3

details these effects and presents jitter measurements in large-area microwire arrays.

Sub-3-ps jitter has been demonstrated in specialized 5-µm-long niobium nitride (NbN)

devices that were designed to minimize intrinsic (down-conversion) and geometric jit-

38



GND

AMP

SNSPD

photon 
arrival

pulse delays

longitudinal
geometric jitter

time

a) b)

Figure 2.2: The origin of longitudinal geometric jitter is shown in (a), where the microwave signals
associated with photon detection events along different segments of the nanowire have different
lengths to travel. This leads to different delay times relative to the photon arrival, shown in (b).

ter [85].

2.4 Design Tradeoffs for Large-area SNSPDs

2.4.1 Active Area vs Timing Performance

There is an intrinsic trade-off between the active area and timing performance in SNSPDs.

Increasing the active area of an individual pixel also increases the length of the nanowire,

which in turn increases the kinetic inductance. The recovery timescale in SNSPDs is gov-

erned by the relationship Tr =
Lk
RL

, where Lk is the kinetic inductance and RL is the load

resistance. Because of this direct proportionality, the recovery timescale (Tr) will be slower

in SNSPDs with longer wire lengths, which will also limit the maximum count rate achiev-

able by the detector. Additionally, longer nanowires have greater longitudinal geometric

jitter, which arises from propagation delays due to detection events at varying locations

along the nanowire’s length. The origin and effect of longitudinal geometric jitter on delay

times in the onset of the SNSPD detection pulse are shown in figure 2.2. This effect can be

39



mitigated using differential readout, in which both ends of the nanowire are connected to

readout circuits. In this configuration, when a photon is absorbed at a particular position in

the nanowire, the counter-propagating positive and negative voltages pulses counter prop-

agate, and the arrival times of the signals at the two ends are which are slightly offset from

each-other can be analyzed such that the propagation delay effects are canceled out.

Larger SNSPD pixel areas have more bends and a greater probability of constrictions

occurring along the nanowire length. Constrictions have the effect of locally increasing

current density and causing suppressed Iswitch . This prevents operation at higher Ibias, at

which the jitter performance is better due to enhanced signal-to-noise ratio (SNR).

2.4.2 Array Size vs Readout Complexity

One way to alleviate the constraints described in the previous subsection, where larger

SNSPDs have the intrinsic trade-off of lower timing resolution, and a reduced maximum

count rate (slower recovery time), is to design an SNSPD array in which multiple pixels

operate in parallel. Direct readout of individual pixels increases thermal load in the cryostat

and readout complexity. Given this, the mm2-scale arrays studied in this work have been

designed in linear array formats. These arrays, with pixel counts of 4, 8, and 16, are feasible

for direct readout. The direct, single-ended readout scheme employed in this work has been

scaled to 64-pixel arrays [118]; however, in the pursuit of cm2-scale devices with thousands

of pixels, this approach isn’t scalable. Multiplexing techniques reduce the number of coax

lines required to bias and readout arrays.

2.5 Large-active area SNSPDs

In recent years, the active areas of SNSPDs have steadily increased from the scale of a

single-mode optical fiber diameter (∼100 µm2) to the mm2 regime [119]. As new applica-

40



tions for SNSPDs that leverage these large active areas emerge, there is a growing interest in

exploring novel materials and device architectures. For applications in high energy physics

such as dark matter (DM) searches, it is necessary to realize SNSPDs with larger active

areas than the current state-of-the-art, by at least an order of magnitude. Scaling devices

beyond the mm2 regime remains an open question that presents many challenges in terms

of design, fabrication, and readout.

2.5.1 Materials

The question of which materials are best for large-active area SNSPD fabrication re-

mains an open topic in the community. For scaling to mm2 active areas and beyond with

large pixels sizes, it is an open question which approach to take as far as the material selec-

tion. The challenge is to yield a thin-film superconductor with high material homogeneity

over a large active area. WSi and MoSi are promising materials for large active-area de-

tectors since the films are amorphous and are, thus, expected to be very homogeneous over

large areas compared to polycrystalline superconductors such as NbN or NbTiN. Producing

a thin-film superconductor with a large area is generally easier with an amorphous mate-

rial than with a polycrystalline one. Amorphous superconductors lack the grain boundaries

present in polycrystalline materials, which can interrupt superconducting pathways and

create weak links between grains. This uniformity makes amorphous films less prone to

structural defects over large areas, which is beneficial for achieving consistent supercon-

ducting properties across the entire film. In polycrystalline films, grain boundaries intro-

duce stress and dislocations that are amplified over large areas. Amorphous films, being

structurally homogeneous, have fewer internal stresses and are more resilient over larger

areas. Additionally, amorphous films are easier to deposit over large areas because their

lack of well-defined crystal structure makes them less sensitive to the surface on which

41



they’re deposited. This allows for more flexibility in deposition conditions and substrates,

making it simpler to produce large, uniform films without needing the precise conditions

required for well-formed crystal growth in polycrystalline films. The substrate versatility of

amorphous superconductors was shown with MoSi in reference, wherein MoSi deposited

on a passivated GaAs surface was observed to have the same critical temperature as MoSi

deposited on silicon, and the detection metrics of the devices fabricated from both of these

were very similar [120].

2.5.2 Device architecture

In traditional devices, nanowires are often patterned in a meander geometry to enhance

optical absorption. This design allows the nanowire to cover a larger area by folding back

and forth in a continuous, closely packed pattern, thereby increasing the likelihood that an

incoming photon will interact with the detector.

The fundamental unit enabling superconducting meanders is the hairpin bend, which

consists of a single 180° turn connecting two wire segments. However, one of the main

factors limiting the performance of meander-based SNSPDs is the current crowding ef-

fect, particularly in the bending regions [111, 112]. In SNSPDs, current crowding oc-

curs when the super-current density increases at sharp bends, corners, or discontinuities in

the nanowire. This localized increase in current density can surpass the superconductor’s

switching current, Iswitch, causing the SNSPD to transition to the normal state prematurely.

As a result, the performance of the detector is degraded.

Additionally, the higher current density in the bends lowers the energy barrier for vortex

entry, which is a primary cause of dark counts in SNSPDs. Thermally activated vortices

can lead to an increase in dark count rate, further degrading performance by decreasing the

signal-to-noise ratio [113]. The impact of current crowding becomes more pronounced in

42



meanders with higher fill factors.

To mitigate these effects, several strategies can be employed. One approach is to use

bend thickening techniques, which help reduce the localized current density in the bending

regions [113, 114]. Additionally, the design of the bends can be optimized to minimize cur-

rent crowding and the associated enhancement of dark counts. Specifically, sharper bends

with smaller inner radii tend to degrade performance, while rounded bends with smoother

transitions are more effective [109, 111]. Another technique to improve current crowding

effects involves helium ion irradiation of the straight segments of the meander [121].

Nanowire vs Microwire Detectors

In 2020, single photon sensitivity in micrometer-wide superconducting wires was first

demonstrated in small microstrip devices [122]. Near-IR photons were detected in a device

made from a 3.3 nm-thick MoSi film with 2 µm-wide wires. This led to increased interest

in developing microwire devices with non-trivial active areas and an exploration of new

materials. Near-IR photon sensitivity in wires as wide as 4 µm and with active areas as large

as 1 mm2 have since been reported [123], with similar results seen in NbN devices[124].

2.6 SNSPDs for Dark Matter Detection and HEP Applications

SNSPDs are the single-photon detector of choice for quantum information science and

technology applications. However, their unique combination of high timing resolution, low

dark count rates, and a response over a broad range of wavelengths makes the technology

attractive for HEP applications. In particular, SNSPDs can be used in low-mass DM ex-

periments where the primary sensor requirements are (1) (vanishingly) low intrinsic dark

count rates, (2) a high photon detection efficiency and (3) low energy thresholds for infrared

single photon detection. For these metrics, SNSPDs are the highest performing detectors

43



available in separate demonstrations. To mature the technology for DM detection, it is nec-

essary to combine these performance metrics into one device and understand the inherent

trade-offs. This is a central focus of the work contained within this thesis. Additionally, it

is important to investigate the fundamental limits of the aforementioned detection metrics

for SNSPDs.

There are now multiple pilot and proof-of-concept phase low mass dark matter detec-

tion experiments using SNSPDs in the detection scheme. The development of large-active

area SNSPDs to instrument n-type GaAs targets is one of the primary focuses of the work

contained in this thesis. Light A’ Multilayer Periodic Optical SNSPD Target (LAMPOST)

is a dark photon search targeting DM in the eV mass range, via coherent absorption in a

multilayer dielectric haloscope [115]. The layered dielectric target is designed to capture

the conversion of dark photons to regular photons which are then focused onto an SNSPD.

In this detection concept, the sensitivity is ultimately governed by the limits of the SNSPD

and in principle as the energy thresholds of SNSPDs are lowered, versions of this exper-

iment that probe lower mass dark photon DM can be implemented to constrain more of

the parameter space at lower masses. Of interest to this future direction are mid-infrared

SNSPDs whose current status and challenges are summarized in the next section.

The Broadband Reflector Experiment for Axion Detection (BREAD) experiment fea-

tures a cylindrical metal barrel to convert dark matter into photons and parabolic reflectors

to focus signal photons onto a photosensor [125, 126]. This science program contains two

pilot experiments. GIGABread is the RF pilot axion and dark photon experiment targeting

DM in the tens of GHz regime and focuses them onto a coaxial horn antenna. INFRABread

is the program’s pilot dark photon search at IR frequencies. The detectors developed and

characterized in this thesis are being deployed as the photo sensors for this pilot experiment.

More on the current status of this implementation is described in chapter 8.

In addition to their use in low mass DM searches, the mm2-scale arrays from this work

44



are also large enough to be potentially interesting technology for broader HEP applications.

One of the 8 pixel 2 mm x 2mm SNSPD arrays of this work was deployed in the Fermi-

lab Test Beam Facility and exposed to various high energy particles [127]. The results

are d