Detecting Single Photons with Graphene-Based Josephson Junctions
The Harvard community has made this article openly available. Please share how this access benefits you. Your story matters
Citation Walsh, Evan Daniel. 2020. Detecting Single Photons with Graphene- Based Josephson Junctions. Doctoral dissertation, Harvard University Graduate School of Arts and Sciences.
Citable link https://nrs.harvard.edu/URN-3:HUL.INSTREPOS:37368983
Terms of Use This article was downloaded from Harvard University’s DASH repository, and is made available under the terms and conditions applicable to Other Posted Material, as set forth at http:// nrs.harvard.edu/urn-3:HUL.InstRepos:dash.current.terms-of- use#LAA HARVARD UNIVERSITY Graduate School of Arts and Sciences
DISSERTATION ACCEPTANCE CERTIFICATE
The undersigned, appointed by the
Harvard John A. Paulson School of Engineering and Applied Sciences have examined a dissertation entitled:
“Detecting Single Photons with Graphene-Based Josephson Junctions” presented by: Evan Daniel Walsh
Signature ______Typed name: Professor P. Kim
Signature ______Typed name: Dr. K. C. Fong
Signature ______Typed name: Professor D. Ham
Signature ______Typed name: Professor D. Englund
October 2, 2020 Detecting Single Photons with Graphene-Based Josephson Junctions
a dissertation presented by Evan Daniel Walsh to The School of Engineering and Applied Sciences
in partial fulfillment of the requirements for the degree of Doctor of Philosophy in the subject of Applied Physics
Harvard University Cambridge, Massachusetts October 2020 ©2020 – Evan Daniel Walsh all rights reserved. Dissertation Advisor: Professor Dirk Englund Evan Daniel Walsh
Detecting Single Photons with Graphene-Based Josephson Junctions
Abstract
In this work I present theory, modeling, and experimentation demonstrating that the graphene-based Joseph- son junction (GJJ) is a capable system for the detection of single photons across the electromagnetic spectrum, from the microwave to the infrared. Two different detection mechanisms are exposed: 1) heating of the graphene weak link in the GJJ and 2) quasiparticle generation in the GJJ superconducting contacts. The first relies on graphene’s exceptionally low heat capacity and its decoupled electron and phonon systems. I show in model- ing that these thermal properties can lead to a GJJ photon detector for very low energy microwave photons. Ex- perimentally, I show that a GJJ bolometer with its graphene weak link coupled to a microwave resonator can achieve an energy resolution equivalent to a single 32-GHz photon. The second detection mechanism reveals it- self in the illumination of a GJJ with near-infrared light. I experimentally demonstrate single-photon detection in this system and show that the data is well fit by a model where photon-induced quasiparticles in the supercon- ducting contacts cause the GJJ phase particle to escape. I give an overview of the single-photon, graphene, and
Josephson-junction physics required to arrive at these results before presenting the experimental evidence for single-microwave-photon sensitivity and single-infrared-photon detection with the GJJ.
iii Contents
Title Page i
Copyright ii
Abstract iii
Table of Contents iv
Acknowledgements vi
1 Introduction 1
2 Single Photon Detection 4 2.1 Introduction ...... 4 2.2 Quantum Description of Light ...... 5 2.3 Photon Counting Statistics ...... 9 2.4 Detection Mechanisms ...... 12
3 Graphene 16 3.1 Introduction ...... 16 3.2 Thermal Properties ...... 21 3.3 Optical Properties ...... 24 3.4 Coupling Photons to Graphene ...... 27
4 Josephson Junctions 32 4.1 Introduction ...... 32 4.2 The Josephson Tunneling Junction ...... 33 4.3 Macroscopic Description: RCSJ Model and Phase Particle Dynamics ...... 35 4.4 Microscopic Description: Andreev States ...... 40 4.5 Graphene-Based Josephson Junctions ...... 42 4.6 Absorption in Superconducting Contacts ...... 45
5 Modeling of a GJJ SPD 49 5.1 Introduction ...... 50 5.2 Device Concept and Input Coupling ...... 51 5.3 Graphene Thermal Response ...... 54 5.4 Graphene-Based Josephson Junction ...... 60 5.5 Photon Detection Performance ...... 65 5.6 Conclusion ...... 69
iv 6 Another Approach: Johnson Noise Thermometry 70 6.1 Introduction ...... 70 6.2 Johnson Noise Thermometry Basics ...... 71 6.3 Single Photon Detection ...... 72 6.4 Photon-Number Resolving ...... 73
7 GJJ Microwave Bolometry with Single-Photon Sensitivity 76 7.1 Abstract ...... 77 7.2 Introduction ...... 78 7.3 Experimental Setup ...... 78 7.4 GJJ Microwave Response ...... 82 7.5 Bolometer Performance ...... 85 7.6 Methods ...... 86
8 Detecting Single Infrared Photons with a GJJ 93 8.1 Abstract ...... 94 8.2 Introduction ...... 95 8.3 Evidence of Single-Photon Detection ...... 98 8.4 Switching Mechanism ...... 100 8.5 Absorption ...... 101 8.6 Physical Model ...... 103 8.7 Tables of measured devices ...... 104 R 8.8 Calculation of photon from laser power ...... 105 8.9 Single-Photon Detection in Pulsed Measurements ...... 106 8.10 Proportionality of Γmeas in laser power ...... 108 8.11 Sweeping Experiments versus Counting Experiments and their correspondence ...... 109 8.12 Experimental determination of polarization orientation ...... 110 8.13 HFSS simulation ...... 111 8.14 Γmeas of JJ using tetra-layer graphene ...... 112 8.15 Modeling of Γmeas in Fig. 8.3B ...... 113 8.16 Gate Dependence of the GJJ Single-Photon Response ...... 116
9 Conclusion 119
Appendix A Experimental Setup 121 A.1 Device Mounting ...... 121 A.2 Optical Setup ...... 122 A.3 Electronic Setup ...... 125
References 128
v Acknowledgments
God has abundantly blessed me in the 9 years since I started my Ph.D. with a path I would have never expected.
The family and friends who have supported me during that time have been nothing short of astounding and I would not be at this milestone in my life without them. First and foremost, I acknowledge my wife, my forever friend, my love, Beth. Who knew the adventure we would be embarking on when we went to play frisbee 7 years ago? She has been there with me every step of the way, through multiple new career opportunities, through dat- ing, engagement, and marriage, through starting a new church, through four children! Her love, encouragement, and prayers for me have provided me with the strength to continue on in my Ph.D. work. A loving wife, a loving mother, dedicated and strong, I could go on and on! So to Beth — thank you. To my four children, Aubree, Na- talee, Daniel, and my youngest son who I can’t wait to meet — I love you and love watching you grow up. Thank you for knowing how to put a smile on Daddy’s face and being the loving, silly, amazing kids that you are. To my parents, Dan and Joy — your loving support throughout my life has set the foundation for me to succeed.
Your advice of “half work, half life” when I started graduate school stuck with me throughout my Ph.D. studies. I could not ask for more loving parents. To my brother Brady — your love and support of me, and now my family, have been great blessings to me throughout my life.
Of course the work in this dissertation was not just “my” work but the work of a large team of people. First, I would like to thank KC Fong who was the driving force behind the project. Thank you for your hard work and dedication throughout this project and for teaching me all the interesting physics and experimental techniques I needed to be successful. Next, thank you to my advisor Dirk Englund, for the advice and support along the way and for making me feel a part of your lab even though I took a non-traditional path to get there. To Philip Kim — thank you for always having an open door for me. One person who is truly an indispensable member of the team
vi is Gil-Ho Lee. Gil-Ho — thank you for all the devices, all the teaching on Josephson junction physics, and all the overall hard work you put into this project. To Dima Efetov — thank you for showing me the ropes and all the advice you gave me during our time together. Almost all of my work was carried out at BBN Technologies and I want to thank everyone in the Quantum Group there for welcoming me in. The conversations, the help on my experiments, and the general camaraderie made for a great graduate experience.
Beyond those directly involved in the work, so many other friends and family helped me get to this point. To my Aunt Sylvia, Grandpa Armenak (how I wish you could see this day!), Grandma Winnie, Uncle Yervant, Uncle
Peter, and my cousin Ben — I treasure your love and support. Steve, Jordan, Chen, Demetrios, Paul, Nishant, and Alexei — your friendship throughout the years has obviously been a huge blessing to me and I’ve learned actually that good friends like you can become family. Michael Rabenberg — my Ph.D. would not have been possible without all the deep philosophical discussions we shared over burgers. To my old lab mates, Tony, Jesse,
Rui, Tommi, Daniel, Danny, and Xi — it was great working together and supporting one another. To Yi Lin — thank you for being a great advisor and making my time at NASA so enjoyable. To my new colleagues at STR — thank you for the support while I finished this degree and for creating such a great team atmosphere to work in.
And to my church family — your prayers were invaluable to me and my family throughout my graduate career.
vii And God said, “Let there be light,” and there was light.
And God saw that the light was good. And God separated the light from the darkness.
Genesis 1:3-4 ESV 1 Introduction
Light in its many forms is an integral part of daily life, from the visible sunlight that illuminates our world to the radio waves that are the basis for cell phones, WiFi, and even microwave ovens. The smallest constituent of light, a single photon, has such low energy that a photon in the visible regime is barely perceptible (if at all) to the human eye.1 The limitation of the human eye to seeing these photons is two-fold: 1) the efficiency of the receptors in our eyes is not high enough for the photon to reliably trigger a response and 2) there are other back- ground noise processes that trigger false detections, or “dark counts,” which prevent us from perceiving true single-photon events above this background. Efficiency and dark count rate are the two most fundamental pa- rameters that will determine if any device is able to detect single photons or not. These parameters become even
1 more important for radio-frequency photons, whose energy is 1000 times less than that of the photons visible to humans.
Why might someone want to detect a single photon in the first place? Near-infrared photons are the basis of many quantum cryptography schemes, with the promise of secure communications over thousands of kilome- ters.2–4 In the far-infrared/terahertz regime, small numbers of photons from astrophysical sources are contin- uously arriving at Earth — being able to detect them could reveal further information about our universe.5–8
Some theorize that elusive dark matter could interact with matter to produce terahertz photons that could be detected with single-photon detectors.9,10 Radio-frequency detectors can be used in quantum computing as an alternate readout scheme for quantum states.11,12 Amazingly, the device investigated in the current work shows potential for detection of light across the electromagnetic spectrum, with demonstrated energy resolution on the scale of single radio-frequency photons and verified detection of single infrared photons. The device isa graphene-based Josephson junction (GJJ).
Graphene is a two-dimensional material comprised of a single layer of carbon atoms in a honeycomb hexago- nal lattice while a Josephson junction is a device in which two superconductors are separated by a non-superconducting material (in this case, graphene) but that still allows a supercurrent to flow through it. We initially chose the
GJJ for the task of single-photon detection due to the unique thermal properties accompanying graphene’s two- dimensional character, properties that allow graphene-based detectors to have high efficiency and low dark count rates. Modeling and experimental results in the radio-frequency regime confirmed this to be true. The JJ was originally viewed merely as a useful tool for capitalizing on graphene’s exceptional properties. However, as ex- periments to detect infrared photons with the GJJ began, it soon became evident that the JJ itself had its own set of properties desirable for the detection of photons. In these experiments, the graphene played the supporting role while the interesting single-photon physics belonged to the JJ. In the end, the GJJ proved to be a fascinating device with the combined properties of the graphene and JJ being enablers for single-photon detection.
2 This dissertation is laid out as follows. In Chapter 2, I lay out what is meant by a single photon (a more con- troversial term than one may think) and briefly discuss detection in general. In Chapters 3 and 4, I explain the physical properties of graphene and Josephson junctions, respectively, and how those properties are desirable for detecting single photons. In Chapters 5 and 6, I present the modeling for graphene-based detection of single pho- tons using two different detection readout schemes: 1) the JJ and 2) Johnson noise thermometry, respectively.
The discussion of Chapter 5 is reproduced from Ref. [13] and is the method I ultimately pursued experimentally.
Chapter 7 presents the experimental demonstration of the GJJ’s ability to resolve energy down to the level of a single radio-frequency photon as originally presented in Ref. [14] and Chapter 8 presents the experimental direct detection of single infrared photons. Finally, the Conclusion is presented in Chapter 9.
3 [...] it is the nature of all bodies, as they have been formed by the coming together of particles, that those same bodies also give off particles. These particles are emitted inthe shape of the objects from which they come, and thus we call them “images.” These images are far too fine to be perceptible in themselves, but the evidence supports the conclusion that these images exist, and that as they move they preserve, in some degree, their respective positions that they held in the solid bodies from which they came. [...] these images move with unsurpassable speed. Epicurus, in Letter to Herodotus, c. 300 BC 2 Single Photon Detection
2.1 Introduction
Before discussing how a graphene-based Josephson junction can detect single photons, it must be discussed what exactly constitutes a single photon. The most basic definition of a photon is a quanta of light, i.e. a quantized excitation of the electromagnetic field. That light has a particle nature was hypothesized as far back as theancient
Greek philosophers but its wave nature, as famously captured by Maxwell’s equations in the mid to late 19th cen- tury, was more heavily emphasized (with the notable exception of Isaac Newton’s corpuscular theory of light based on the work of Pierre Gassendi) until the groundbreaking work of Max Planck in 1900 and Albert Ein-
4 stein in 1905. With that said, describing light as a particle is somewhat controversial — Lamb, for instance, gives a thorough history of the term photon and then proceeds to argue it should be abandoned altogether (or at least require a license)15 while Loudon acknowledges the term’s usefulness, its pitfalls, and cautiously gives a defini- tion to start his book on Quantum Electrodynamics.16 In this chapter we discuss more thoroughly the quantum description of light, how one can count single photons, and mechanisms of detection.
2.2 Quantum Description of Light
The following discussion follows closely that of Schiff who gives a thorough exposition starting from the princi- ple of least action, classical Lagrangian field theory, and classical Hamiltonian field theory leading to quantized versions of these theories.17 To show what is meant by a single photon, I start with Maxwell’s equations, show how to arrive at a Hamiltonian from Maxwell’s equations, and then show how a quantized version of the Hamil- tonian leads to a quantum harmonic oscillator solution corresponding to a photon picture of electromagnetism.
The vacuum (i.e. sourceless) Maxwell’s equations are:
1 ∂⃗B ∇ × ⃗E + = 0 (2.1a) c ∂t
∇ · ⃗E = 0 (2.1b)
1 ∂⃗E ∇ × ⃗B − = 0 (2.1c) c ∂t
∇ · ⃗B = 0 (2.1d)
5 where ⃗E is the electric field, ⃗B is the magnetic field, and c is the speed of light. Recall that ⃗E and ⃗B can be written in terms of a vector potential A⃗ and scalar potential φ as:
1 ∂A⃗ ⃗E = − − ∇φ (2.2a) c ∂t
⃗B = ∇ × A⃗ (2.2b)
Now Maxwell’s equations must be consistent with the Lagrange (field) equations given by:
" # ∂L X ∂ ∂L − = 0 (2.3) ∂ψ ∂x ∂ ∂ /∂ i j j ψi xj
where L is the Lagrangian density, ⃗ψ = A⃗ , φ are the field components, and⃗x is a four-vector consisting of the three spatial dimensions,⃗r, and one time dimension, t. It can be shown that the appropriate choice of L for consistency with Equation 2.1 is:
! 2 1 1 ∂A⃗ 1 2 L = + ∇φ − ∇ × A⃗ (2.4) 8π c ∂t 8π
To go from this Lagrangian description to a Hamiltonian description, we introduce the canonical conjugate mo- mentum density, ∂L ⃗P = (2.5) ∂ ∂⃗ψ/∂t which for 2.4 is: ! 1 1 ∂A⃗ ⃗P = + ∇φ (2.6) 4πc c ∂t
6 (Note I’ve been lax in notation here — in Equation 2.5, ⃗P is a four-vector while in Equation 2.6 it is only a three- vector, this decision made because the φ component of Equation 2.5 is zero. A similar note applies to Equations
2.7 and 2.8). Now the Hamiltonian density in general is given by:
∂⃗ψ H = ⃗P · − L (2.7) ∂t which for electromagnetism is:
1 H = 2πc2|⃗P|2 + |∇ × A⃗ |2 − c⃗P · ∇φ (2.8) 8π
Taking the integral of Equation 2.8 to get the full Hamiltonian and noting that the third term when integrated is zero: Z 1 H = d⃗r 2πc2|⃗P|2 + |∇ × A⃗ |2 (2.9) 8π
So far the discussion has been completely classical. Here we shift to a quantized picture: we replace the vectors ˆ ˆ ⃗P and A⃗ with the vector operators ⃗P and A⃗ . We can expand these operators in terms of an orthonormal basis.
Choose plane waves of the form:
1 i⃗k·⃗r ⃗u⃗ (⃗r) = √ ε⃗ e (2.10) kξ V kξ
⃗ where ε⃗k1 and ε⃗k2 are two unit polarization vectors perpendicular to each other and to the propagation vector k and V is the volume of a large cubical box used to set boundary conditions for⃗k (V will not affect our calculations ˆ ˆ and disappears when limits to infinity are taken). Now expanding ⃗P and A⃗ :
X h i ˆ † ∗ ⃗P(⃗r, t) = ˆp⃗ (t)⃗u⃗ (⃗r) + ˆp (t)⃗u (⃗r) (2.11a) kξ kξ ⃗kξ ⃗kξ ⃗kξ
7 X h i ⃗ˆ † ∗ A(⃗r, t) = ˆq⃗ (t)⃗u⃗ (⃗r) + ˆq (t)⃗u (⃗r) (2.11b) kξ kξ ⃗kξ ⃗kξ ⃗kξ
Here, ˆp and ˆq are expansion coefficient operators, † denotes Hermitian adjoint, and the sum is over the half k plane. Now inserting Equation 2.11 into Equation 2.9, making use of the commutation relations:
h i h i † † ~ ′ ′ ˆq⃗ (t), ˆp ′ ′ (t) = ˆq (t), ˆp⃗′ ′ (t) = i δkk δ (2.12) kξ k⃗ ξ ⃗kξ k ξ ξξ
(with all other commutations zero), and noting the integrals will be trivial due to the presence of delta functions, we have for the Hamiltonian operator:
X 2 2 † k † Hˆ = 4πc ˆp⃗ (t)ˆp (t) + ˆq⃗ (t)ˆq (t) (2.13) kξ ⃗kξ 4π kξ ⃗kξ ⃗kξ
Other than normalization, this is the Hamiltonian of a quantum harmonic oscillator. The usual steps can be taken to rewrite Equation 2.13 in terms of creation and annihilation operations. Briefly, in this case the annihila-
ˆ ( ) ˆ ( ) tion operator (which can be derived through calculating the time evolution of p⃗kξ t and q⃗kξ t in the Heisenberg
~ dF = [ , ] picture of quantum mechanics via i dt F H ) is given by:
2 1 4πc iωt ˆa⃗ (t) = ˆq⃗ (t) + i ˆp⃗ (t) e (2.14) kξ 2 kξ ω kξ
† where we’ve used ω = ck. The corresponding creation operator is ˆa (t). These operators correspond to annihi- ⃗kξ ⃗ lating or creating, respectively, a plane-wave mode with propagation vector k and polarization ε⃗kξ. Defining the number operator, which has non-negative-integer eigenvalues, as:
ω † Nˆ⃗ (t) = ˆa (t)ˆa⃗ (t) (2.15) kξ 2π~c2 ⃗kξ kξ
8 we can rewrite Equation 2.13 as: X 1 Hˆ = ~ω Nˆ⃗ (t) + (2.16) kξ 2 ⃗kξ
| ⟩ It can be shown that the eigenvectors of this Hamiltonian are the number states, n⃗kξ , which are also the eigen- vectors of the number operator, with
ˆ | ⟩ = | ⟩, = , , , ... N⃗kξ n⃗kξ n⃗kξ n⃗kξ n⃗kξ 0 1 2 (2.17)
and X 1 Hˆ|n⃗ ⟩ = ~ω n⃗ + |n⃗ ⟩, n⃗ = 0, 1, 2, ... (2.18) kξ kξ 2 kξ kξ ⃗kξ
This equation says that the energy measured in an electromagnetic field must be an integer number of ~ω plus some zero-point energy. To review, we started with Maxwell’s equations, calculated the corresponding Hamil- tonian through the use of the Langrange field equations, and expanded the field components in terms ofplane waves to show the Hamiltonian could be cast as that of a quantum harmonic oscillator with energy quantized as integer multiples of ~ω. Equation 2.18 captures exactly what is meant by a single photon: it is a single quantum excitation of an electromagnetic field mode. Although the current discussion focused on plane wave modes inan infinite vacuum, the ideas extend to modes in a cavity or other systems aswell.
2.3 Photon Counting Statistics
In practice, number states, including single-photon states, are hard to prepare. Instead, coherent states as pre- pared by a laser are often used. A coherent state, |α⟩, (sometimes referred to as a Glauber state) is an eigenstate of the annihilation operator and can be written in terms of the number states, |n⟩ (dropping subscripts for simplic-
9 ity), as:18 ∞ X n − 1 |α|2 α |α⟩ = e 2 √ |n⟩ (2.19) ! n=0 n where α is the eigenvalue of the annihilation operator acting on |α⟩. This state |α⟩ itself is not an eigenstate of the
Hamiltonian but instead is a superposition of the energy eigenstates (number states). This fact means there is not a well-defined number of photons in a coherent state but instead an expected (or average) number ofphotons.
Taking the expectation value of the number operator in a coherent state gives:
⟨α|Nˆ |α⟩ = |α|2 (2.20)
meaning |α|2 is the average number of photons in the state |α⟩. Furthermore, if we project a coherent state into the number state eigenbasis, we can see the average occupation of each number state is given by:
2n |α| −| |2 |⟨n|α⟩|2 = e α (2.21) n!
This is the equation of a Poisson distribution with the parameter of the distribution being the average number of photons in the state, |α|2. In the experiments of Chapter 8, we used a laser as our photon source so we when we say we’ve detected a single photon, we really mean we’ve projected a coherent state into an energy eigenbasis where the resulting measurement collapsed the wavefunction into a single-photon state (but the measurement could have instead yielded zero, two, three, etc. photons with probability given by a Poisson distribution).
It is easiest to think of the Poisson statistics of a coherent source in terms of a finite laser pulse. A finite (in space and time) coherent pulse when measured will yield a finite number of photons. If many such identical pulses were measured, the measured number of photons in the pulses would follow a Poisson distribution. Some experiments in Chapter 8 were pulsed. Most however, were performed using a continuous wave (CW) laser. Fur-
10 Figure 2.1: Poisson distribution of a coherent light source (A) Whether using an ensemble of pulses from a pulsed laser, dividing a CW laser into many time bins, or dividing either into many spatial bins, the number of photons per pulse, time bin, or spatial bin will follow a Poisson distribution. (B) The Poisson distribution for an average of 1 photon (blue) or 0.13 photons (red) per pulse, time bin, or spatial bin. Note that a coherent state with on average 1 photon still has a relatively high probability (∼26%) of yielding a measurement of more than 1 photon. For the experiments of Chapter 8, an average number of photons per pulse of 0.13 was chosen to ensure a very low chance of measuring multiple photons in a given pulse (∼0.8%) . thermore, all the experiments used a laser spot size much larger than the GJJ detector itself. In the CW, large spatial extent case, one can think of the number of photons being detected by the GJJ as follows. Divide the in- coming laser into many pulses in both time and space (see Figure 2.1A). The number of photons measured in these pulses will also follow a Poisson distribution with the caveat that the distribution will be weighted by the intensity profile of the laser. For the single-mode-fiber-coupled laser of Chapter 8, the intensity is constantin time and has a Gaussian spatial profile. Spatially, we only sample the laser in one location (the location of theGJJ) so the Poisson statistics in this dimension are not observable. In time, however, we continuously monitor the detector for trigger events. If the detector is truly detecting single photons, then we expect that if we divide our measurement into many time bins, that the number of trigger events per bin should follow a Poisson distribution.
In Chapter 8 we show this is indeed the case. I note that caution must be taken when interpreting the observed
Poisson distribution — any independent, random events will follow Poisson statistics. If the GJJ were only sen- sitive to 2-photon states, for example, the detections of these 2-photon states would still be Poisson distributed in time. However, in Chapter 8 we show that the number of detection events is also proportional to the num-
11 ber of photons thus demonstrating sensitivity to 1-photon states (sensitivity to 2-photon states would result in a number of detection events proportional to the square of the the number of photons). A helpful discussion of photon-counting statistics with a slightly different derivation is given in Chapter 5 of Ref.[19].
Intuitively, one might expect that to execute a single-photon experiment with a laser one should set the average number of photons in the coherent states to be 1. However, as illustrated by the blue line in Figure 2.1B, with an average of 1 photon in a coherent state, there is still a fairly large chance of measuring multiple photons, ∼26%.
Instead, single-photon experiments with coherent states should choose average photon numbers well below 1.
For the pulsed measurements of Chapter 8, we chose an average of 0.13 photons per pulse (Poisson distribution shown in red in Figure 2.1B). For this average number of photons, the chances of a measurement yielding multi- ple photons is only ∼0.8%, 15 times smaller than the chance of the measurement yielding 1 and only 1 photon.
In CW experiments, the choice of photon rate (i.e. average number of photons per time bin) takes the place of av- erage number of photons per pulse. The photon rate should be chosen such that the average number of photons reaching the detector during its response time is much less than 1. In Chapter 8, we estimate the response time of the GJJ infrared photon detector to be on the order of 1 ns, suggesting the photon rate should be kept well below
1 GHz to guarantee single photon events. This condition was easily met, with typical photon rates less than 100
Hz. In practice, the limiting factor to the photon rate was instead heating of the GJJ.
2.4 Detection Mechanisms
Now that a single photon has been defined, I discuss how one could detect these low energy excitations. InChap- ters 5 and 6 I will discuss two specific detection mechanisms: threshold detection with a Josephson junction
(JJ) and thermal detection with Johnson-noise thermometry (JNT), respectively. Threshold detection is fairly straightforward: the photon causes some parameter to change and once that parameter surpasses a threshold (de- termined by the detector properties), a photon count is registered. In a JJ, the photon causes the voltage to jump
12 from zero to nonzero (i.e. the JJ to switch from superconducting to resistive). These jumps are fairly large (Figure
2.2A) so the choice of a threshold voltage is somewhat trivial — there is a large range of voltages sufficiently above the noise floor and sufficiently below the peak voltage upon photon absorption to limit the number offalseposi- tives (i.e. measurement dark count rate) and false negatives (i.e. measurement detection inefficiency), respectively.
Note the inclusion of the word “measurement” to emphasize this dark count rate and inefficiency are due to the measurement setup and not the intrinsic properties of the detector itself. The threshold detection mechanism is similar to that used in superconducting nanowire single-photon detectors (SNSPDs).20,21 JNT could also be used in a threshold detector although its strength lies more in acting as a calorimeter, being able to determine how much energy was deposited as opposed to simply whether or not energy was deposited (see Chapter 6). In the
JNT graphene single-photon detector setup, a photon raises the temperature of the graphene which in turn raises its voltage noise. Measuring the power of the voltage noise, one could set a threshold between the base voltage noise power and the elevated voltage noise power upon absorption to register photon detections. However, the measured voltage noise is proportional to the temperature, unlike in a JJ where the measured voltage is a constant determined solely by the bias current with no relationship to the temperature. To take full advantage of JNT, one can then use the measured value of voltage noise to estimate the temperature as a function of time and from that deduce how much energy was deposited into the detector. This could provide much more information than a simple threshold detection, such as the frequency of the photon being detected or how many photons were being detected. Photon-number resolving with JNT is discussed in Section 6.4.
So far in this section, it was assumed that the signal-to-noise ratio (SNR) was sufficient to either set a thresh- old that would limit dark counts without limiting efficiency or in the case of JNT, determine a temperature with enough resolution to draw conclusions about the amount of energy the detector absorbs. However, high SNR in the raw voltage or voltage noise signal is not a necessary condition for single photon detection. It was pointed out in a helpful review of calorimeters by McCammon that filtering the signal can drastically improve SNR.22 Specif-
13 Figure 2.2: Detection of high SNR versus low SNR signals (A) A voltage versus time trace from the GJJ of the experiments in Chapter 8. The voltage signal from single-photon events in this case was high above the noise floor simplifying the measurement scheme (B) In cases where the SNR is low, matched filtering can be used to extract a signal out of the noise floor. Shown is a simulated measurement withSNR of 1 (blue), the noiseless signal (red), and the output of a matched filter acting on the measured signal using the noiseless signal asthe template (black). The matched filter changes a barely discernible signal into one well above the noise floor. Such a technique couldprove especially useful in JNT.
ically, for a known signal s0(t) (often called a template) with spectrum˜s0(f) and known noise spectral density
SN(f), an optimal linear filter (a “matched filter”) will be of theform:
∗ ˜s (f) h(f) = 0 (2.22) SN(f)
∗ where the denotes complex conjugate. If we have a measured signal smeas(t), then the filtered version, sfilt(t), will be: Z ∞ ∗ ˜s0(f)˜smeas(f) sfilt(t) = (2.23) −∞ SN(f)
This filtered version of the measured signal will have peaks at times where the original signal resembled thetem- plate. I show an example in Figure 2.2B. A noisy signal (blue) is formed by adding white Gaussian noise to a noiseless signal (red). The peak signal value is 1 as is the root-mean-square (RMS) noise value, yielding an SNR of 1. As is clear, a threshold detector would not work on this raw signal. However, using a template of the signal to match filter the noisy signal, a peak rises out of the noise at the time where the signal appears (blackline).The
14 filtered signal is normalized so the RMS noise is still 1 and it can be seen that theSNRisnow ∼7. A threshold detector could now operate effectively on this filtered signal. The level of the peak will also be proportional to the energy in the original signal so JNT could be used as well. As clearly illustrated, the advantage of this method is that it can drastically increase the SNR compared to the raw measurement. However, the disadvantage is that both the shape of the signal and the noise must be known to form the optimal filter. Typically, the noise can be measured so determining SN(f) is straightforward. However, determining the proper template can be more difficult. Even so, a matched filter with errors in the template will still often increase the SNR. Alternatively,a bank of matched filters could be used on the measured signal to increase the probability of picking outlow-SNR signals if a range of templates is expected prior to the measurement. In the experiments of Chapter 8, the SNR was high enough that matched filters were unnecessary. However, future experiments, especially in JNT, could benefit from such a technique.
15 If N classical particles in two dimensions interacting through a pair potential Φ(⃗r) are in equilibrium in a parallelogram box, it is proved that every⃗k ≠ 0 Fourier component of the density must vanish in the thermo- dynamic limit [...]. This results excludes conventional crystalline long-range order in two dimensions [...].
N. David Mermin, 1968 3 Graphene
3.1 Introduction
Graphene, a single layer of carbon atoms held together by sp3 hybridized bonds in a honeycomb lattice and orig- inally thought to be thermodynamically unstable,23–25 was first isolated and observed in 2004.26 Since its ex- perimental realization, graphene has proven its worth in a myriad of applications, from sensors of light,27–30 heat,31–33 and pressure34 to electrochemical energy storage35–37 to mechanical lubrication.38 The reason for graphene’s such widespread use is the unique properties it possesses, including being the strongest material by weight (in terms of Young’s modulus), having the largest specific surface area, and having a relativistic band struc-
16 Figure 3.1: Graphene’s band structure ture. It is this last property that produces the exceptional electronic, optical, and thermal properties in graphene that render the material an ideal candidate for single photon detection and that will be the focus of the current chapter.
Graphene’s band structure was first calculated more than half a century before the material itself wasiso- lated.39 Graphene’s honeycomb structure can be thought of as a triangular lattice with two atoms per unit cell √ with the lattice spacing being 3 times the carbon-carbon spacing of a ∼1.42 Å (i.e. the Bravais lattice is formed by carbon atoms that are second-nearest neighbors instead of nearest neighbors). The reciprocal lattice is also triangular, yielding a hexagonal Brillouin zone. Using a tight-binding model only including nearest-neighbor interactions, it can be shown that the dispersion relation is given by:39–41
s √ √ 3 3 E± (k) = ±t 3 + 2 cos 3k a + 4 cos k a cos k a (3.1) y 2 y 2 x