Narrownamd Stoke Photons Based on DLCZ Protocol

National Cheng Kung University
Department of Physics
Master’s Thesis
基於 DLCZ 機制產生窄頻史托克光子之研究
Narrowband Stoke Photons Generation Based on DLCZ Protocol
學生：楊智閔 Student： Chih-Min Yang

指導老師：陳泳帆博士 Advisor： Dr. Yong-Fan Chen
August 2021
要
這些年來長距離的量子通訊已受到廣泛的關注，在這篇論文中，我們一開始會先介
紹糾纏態的互換特性與量子中繼器並說明如何以 DLCZ 的機制實現長距離的量子通
訊。在第二章中，我們會先半古典的方式介紹量子光學的基本理論: 電磁波引發透
明，而這也正是 DLCZ 模型的第二步作法。在第三章中，我們會以全量子的方式結
合海森堡朗之萬方程以及馬克斯威爾薛丁格方程計算出在 DLCZ 機制下第一步所產
生的史托克光子穩態產生率的解析解、頻譜解析解以及二階的自我關聯函數，並分
析了基態去同調率以及拉曼增益對系統達到穩態所需時間與產生率的影響。在我們
系統下解出史托克光的頻寬極窄，幾乎只受限於基態去同調率，藉由增加拉曼增益
後，甚至可以得到幾乎為單頻光的頻譜。由此，史提克光子可以得到極長的同調時
間。在第四章中，我們會直接從頻率域出發來解出史托克光，這個作法相較前者簡
潔很多，而這個方法的結果與第一種方法幾乎一致且能揭露出史塔克效應造成的影
響。
關 : 拉曼散射, DLCZ 機制, 電磁波引發透明, 量子中繼站
i
Abstract
In recent years, large-scale quantum communication earns widespread attention. In the be-
ginning, we will introduce entanglement swapping, purification, and quantum repeater to
elaborate on how to realize quantum communication over long distance. In chapter two,
we briefly illustrate the semi-classical theory of electromagnetically induced transparency,
which denotes the second step of DLCZ scheme. In chapter three, we utilize the Heisenberg-
Langevin equation and Maxwell-Schrödinger equation to derive the time-dependent Stoke
photons generation by Laplace transform. In addition, we provide the analytic solution of
spectrum and the second-order autocorrelation function of Stoke photons. After that, we will
discuss the influence of two ground-states decoherence on generation rate and the time it
takes to attain steady-state condition. The spectrum of Stoke photons is extremely narrow,
which is constricted by two ground-states decoherence. The analytic solution of the spectrum
is Cauchy–Lorentz distribution. (owing to the collision of particles) By increasing Raman
gain, the spectrum is similar to the spectrum of single-mode photons. In consequence, Stoke
photons possess a long coherent time. In chapter four, we exploit an alternative and con-
venient approach to resolve the spectrum, which reveals the Stark effect that is arduous to
be carried out by the first method. The results of several theoretical predictions are almost
consistent with the first method.
Keyword: Raman Scattering, DLCZ protocol, Electromagnetically induced transparency,

Quantum repeater
ii
謝
很快地就要從碩士畢業了，這兩年下來需要感謝的人很多，首先，我想感謝我的指
導教授，陳泳帆老師，十分感謝陳泳帆老師願意收我做學生，在每次與老師討論與
meeting 的過程中，都讓我對於量子光學的物理意涵有更進一步的了解，除此之外，
當我們實驗結果並不如預期時，透過老師分析問題所在以及謹慎地去思索未來光路
怎樣架設會導致問題產生，都讓我對於實驗物理之謹慎感到欽佩，也感謝老師在忙
碌之中抽空與我討論，幫助我對於模型的延伸以及理解，我才能完成完成這篇碩士
論文。
此外，我想感謝實驗室的每個人，首先，十分感謝濬璿學長在我們剛進實驗室時便
指導我們半古典的量子理論，讓我們盡早進入狀況，在每次與學長的討論中都可以
得到新的想法，在實驗上，由於搬新實驗室，我們的實驗進度較趕，再加上我大學
做的是理論物理，很多時候我對實驗上的考量就只是盡快完成進度而已，也因此走
了很多冤枉路，藉由觀察濬璿學長在每個步驟的謹慎想法，以及有邏輯的去嚴謹考
量每個架設這樣架的好處有什麼？代價是什麼？讓我很深刻的學習到怎樣有效率的
處理事情。也很感謝濬璿學長在我們碩二剛開學時便自己整理好雙光子對的全量子
理論給我們，並提醒我們要在實驗開始忙以前把雙光子對理論趕緊全部自己算過一
次，正因為有學長當時的提醒與整理，我才能在 4 月時順利將雙光子對的理論套用
到 DLCZ 的模型中。並且也十分感謝敬堯學長，在實驗上敬堯學長主要是負責大光
桌的實驗，不過在我們小光桌遇到任何問題時都很熱心的協助我們，甚至還為了幫
我們解決 EIT 的問題時陪我們做到快半夜三點，在理論上，雖然學長接觸的模型與
我的不一樣，不過學長還是願意了解我遇到的問題並給我建議，藉由每次與學長討
論，都可以讓我吸收到很多學長對於全量子理論的獨到見解。最後也很感謝兩位學
長在我們要口試前犧牲自己的時間，聽我們數次的口試預演並給我們建議。
也感謝子瑜學長協助小光桌的實驗並幫助我繪圖軟體的安裝，使我節省很多時間; 感
謝鈴淳處理實驗室的事務，以及徐晧在口試前幫忙聯絡口試老師與處理口試手續; 感
謝實驗室的所有人參加我們的口試預演讓我們增加臨場感; 也感謝實驗室以前的學長
姊打下這樣的實驗與理論基礎讓我們讓我們可以順著路往下走下去，此外也很感謝
我在大學時的指導教授，何俊麟老師，在大學時帶我做理論物理，給我足夠的經驗
讓我在碩論改做理論論文時較好上手。最後希望小光桌未來可以成功完成 DLCZ 跟
光子對的實驗、大光桌可以完成 Diamond，還有實驗室的大家未來可以一切順利!
iii
Table of Contents
要 i
Abstract ii
謝 iii
Table of Contents iv
List of Figures vi
Chapter 1. Introduction 1
1.1 The Swapping of Quantum Entanglement . . . . . . . . . . . . . . . . . . . 2
1.2 Quantum Repeater . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.3 DLCZ Scheme . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
Chapter 2. Theoretical Model of Electromagnetically Induced Transparency 11

2.1 Electromagnetically Induced Transparency . . . . . . . . . . . . . . . . . . 11
2.2 Optical-Bloch Equation . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.1. Density Matrix Expression . . . . . . . . . . . . . . . . . . . . . . 13
2.2.2. The Equation of Motion . . . . . . . . . . . . . . . . . . . . . . . . 13
2.3 Steady-State Solution . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
2.3.1. Brief Review of Perturbation Theory . . . . . . . . . . . . . . . . . 17
2.4 Maxwell-Schrödinger Equation . . . . . . . . . . . . . . . . . . . . . . . . 20
2.5 Light Shift (AC Stark Effect) . . . . . . . . . . . . . . . . . . . . . . . . . 23
2.6 Figures of Electromagnetically Induced Transparency . . . . . . . . . . . . 24
2.7 Slow Light Effect . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
2.8 Interpretation of EIT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.8.1. Coherent Population Trapping . . . . . . . . . . . . . . . . . . . . 31
Chapter 3. Stoke Photons Generation Based on DLCZ Protocol 33

3.1 Brief Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Stimulated Raman Scattering . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2.1. Collective Atomic Operator . . . . . . . . . . . . . . . . . . . . . . 35
3.2.2. Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
3.3 Heisenberg-Langevin Equation . . . . . . . . . . . . . . . . . . . . . . . . 37
3.5 Calculation of Stoke Field . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
3.5.1. Generation Rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41
3.5.2. Raman Gain . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
3.5.3. Spontaneous Raman Scattering Process . . . . . . . . . . . . . . . 46
3.5.4. Transient Raman Scattering . . . . . . . . . . . . . . . . . . . . . . 46
3.5.5. Steady-State Solution . . . . . . . . . . . . . . . . . . . . . . . . . 47
iv
3.6 The Diagram of Generation Rate . . . . . . . . . . . . . . . . . . . . . . . 47
3.7 Spectrum of Stoke Photons . . . . . . . . . . . . . . . . . . . . . . . . . . 55
3.7.1. Stochastic Process . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
3.7.2. Calculation of Spectrum . . . . . . . . . . . . . . . . . . . . . . . 58
3.7.3. Low Gain Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . 60
3.7.4. High Gain Bandwidth . . . . . . . . . . . . . . . . . . . . . . . . . 63
3.8 Second-Order Correlation Function . . . . . . . . . . . . . . . . . . . . . . 65
Chapter 4. An Alternative Approach to Resolve Stoke Photons 67

4.1 The Hamiltonian . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
4.3 Computation of Stoke Field . . . . . . . . . . . . . . . . . . . . . . . . . . 70
4.4 Spectrum of Stoke Photons . . . . . . . . . . . . . . . . . . . . . . . . . . 72
4.5 Generation Rate of Stoke Photons . . . . . . . . . . . . . . . . . . . . . . . 73
4.6 Second-Order Autocorrelation Function . . . . . . . . . . . . . . . . . . . . 73
Chapter 5. Conclusion 78
5.1 Summary and Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78
References 80
Appendix A. Heisenberg-Langevin Equation 83
Appendix B. Fluctuation-Dissipation Theorem 88
Appendix C. Quantization of Electromagnetic Field 91
Appendix D. Maxwell-Schrödinger Equation 94
Appendix E. Correlation Function 96
v
List of Figures
1.1 Schematic illustration of entanglement swapping by employing Bell-state

measurement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.2 Schematic illustration of the decrease of fidelity after entanglement swapping 5
1.3 Schematic illustration of the increase of fidelity after entanglement purification. 5
1.4 Schematic illustration of nested quantum repeater. The black arrows and gray
arrows denote entanglement swapping and entanglement purification respec-
tively; F is the fidelity, and F2 > F1 . . . . . . . . . . . . . . . . . . . . . 7
1.5 Schematic illustration energy level, write beam, and Stoke photon . . . . . 9
1.6 Schematic illustration of generating entanglement. The blue light denotes
writing beam, and purple line denotes Stoke photon . . . . . . . . . . . . . 9
1.7 Schematic illustration of entanglement swapping in DLCZ scheme . . . . . 10
2.1 Three typical types of EIT scheme: Λ-type, V-type, and cascade-type . . . . 12
2.2 Schematic diagram of a Λ-type EIT. The Ωp and Ωc are the Rabi frequency
of probe field and coupling field respectively, where ∆p is the one-photon
detuning, δ is the two-photon detuning, γ31 is the total decay rate of excited
state |3⟩, and γ21 is the dephasing rate of ground states . . . . . . . . . . . 12
2.3 Schematic illustration of the dephasing caused by vacuum fields and colli-
sions of atoms. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.4 Schematic diagram of dressed state picture when one photon detuning ∆c =
0. The left-hand side reveals that we can integrate atoms and coupling fields
into another new atomic system. For this reason, probe field will be absorbed
as one photon detuning ∆p equals to ± Ω2c . . . . . . . . . . . . . . . . . . 24
2.5 Schematic diagram of transmission with different Ωc . The parameter setting:
OD= 5, γ21 = 0, γ31 = Γ, ∆c = 0, and Ωc = 2Γ and Ωc = 6Γ for Figure (a)
and (b) respectively . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.6 Schematic diagram of phase shift with parameters : γ31 = 1Γ, Ωc = 1Γ,
OD= 10, δ = ∆p , γ21 = 0 . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.7 Schematic diagram of transmission with different γ21 . The parameter setting:
γ31 = 1Γ, Ωc = 2Γ, OD= 20, δ = ∆p , γ21 = 0 and γ21 = 0.1Γ for Figure
(a) and (b) respectively . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
2.8 Schematic diagram with different optical depth. The parameter setting: γ31 =
1Γ, Ωc = 5Γ, ∆c = 0, γ21 = 0, OD= 10 and OD= 100 for Figure (a) and (b)
respectively . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26
2.9 Schematic diagram of transmission. The parameter setting: γ31 = 1Γ, Ωc =
5Γ, OD= 30, γ21 = 0, ∆p = 0 and δ = 0 for Figure (a) and (b) respectively 26
2.10 Schematic diagram of probe pulse in EIT experiment. The parameter setting:
γ31 = 1Γ, Ωc = 0.5Γ, δ = 0, γ21 = 0.0002, and OD= 30 and OD= 150 for
Figure (a) and (b) respectively . . . . . . . . . . . . . . . . . . . . . . . . 29
vi
2.11 Schematic diagram of slow light. The red spot in the middle means parts
of the ensemble of atoms. At t1 , probe field was transformed as spin-wave
coherence by coupling field. At t2 , another coupling field readout the co-
herence. Next, the probe field passes through another atom and repeats the
previous process. Therefore, probe fields get retarded . . . . . . . . . . . . 29
2.12 Illustration of transparency . . . . . . . . . . . . . . . . . . . . . . . . . . 30
2.13 Schematic diagram of coherent population trapping . . . . . . . . . . . . . 32
3.1 Schematic diagram of energy level of stimulated Raman scattering, where

∆p = ωp − ω31 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34
3.2 Theoretical diagram of generation rate respective to local time, the parameter
setting: Γ32 = Γ/2, Ωp = 1Γ, γ21 = 0.001Γ, ∆p = 500Γ, OD= 50 and
OD= 100 for figure (a) and (b) respectively, and Γ = 6 × 2π × 106 Hz . . . 47
setting: Γ32 = Γ/2, , γ21 = 0.0002Γ, ∆p = 350Γ, OD= 1000, and Ωp =
0.7Γ and Ωp = 0.9Γ for figure (a) and (b) respectively, and Γ = 6 × 2π × 106
Hz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48
setting: Γ32 = Γ/2, , γ21 = 0.0002Γ, Ωp = 1Γ, OD= 1000, and ∆p = 350Γ
and ∆p = 500Γ for figure (a) and (b) respectively, and Γ = 6 × 2π × 106 Hz 49
setting: Γ32 = Γ/2, , γ21 = 0.0005Γ, Ωp = 1Γ, ∆p = 300Γ, and OD= 500
and OD= 1000 for figure (a) and (b) respectively, and Γ = 6 × 2π × 106 Hz 49
setting: Γ32 = Γ/2, Ωp = 1Γ, ∆p = 200Γ, OD= 800 and γ21 = 0.001Γ for
figure (b); OD= 400 and γ21 = 0.0005Γ for figure (a) Γ = 6 × 2π × 106 Hz 50
3.7 Theoretical diagram of Generation rate with respect to time for different gain
length. The parameter setting(from top to bottom): Red line: gr z = 50,
Orange line: gr z = 40, Yellow line: gr z = 30, Green line: gr z = 20, Blue
line: gr z = 10, Indigo line: gr z = 5, Purple line: gr z = 3, Black line:
gr z = 1, Gray line: gr z = 0.1, Pink line: gr z = 0.01 . . . . . . . . . . . . 51
3.8 Theoretical diagram of Generation rate with respect to pumping power. The
parameter setting: OD=50, ∆p = 500Γ, Γ32 = Γ/2, γ21 = 0.001Γ, The gain
lengths is 0.0023 when pumping power is 0.3Γ, and gain length is 0.1 when
pumping power is 2Γ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52
3.9 Theoretical diagram of Generation rate with respect to optical depth. The
parameter setting: Ωp = 1Γ, ∆p = 500Γ, γ21 = 0.001Γ, Γ32 = Γ/2 . . . . . 52
parameter setting: Ωp = 1Γ, ∆p = 500Γ, γ21 = 0.0001Γ, Γ32 = Γ/2 . . . . 53
3.11 Theoretical diagram of Generation rate with respect to one photon detuning.
The parameter setting: Ωp = 1Γ, OD=50, γ21 = 0.0001Γ, Γ32 = Γ/2 . . . . 53
3.12 Theoretical diagram of generation rate respect to pumping power for different
dephasing rate of two ground states, the parameter setting: Γ32 = Γ/2, ∆p =
300Γ, OD= 50 and γ21 = 0.01Γ and γ21 = 0.0001Γ for figure (a) and figure
(b) respectively. Γ = 6 × 2π × 106 Hz . . . . . . . . . . . . . . . . . . . . 54
vii
3.13 Theoretical diagram of generation rate respect to pumping power for different
dephasing rate of two ground states, the parameter setting: Γ32 = Γ/2, ∆p =
1000Γ, OD= 50 and γ21 = 0.01Γ and γ21 = 0.0001Γ for figure (a) and figure
(b) respectively. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
3.14 Theoretical diagram of non-ergodic process and ergodic process . . . . . . 57
3.15 Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30,
γ21 = 0.001Γ, Γ32 = Γ/2, ∆p = 500Γ and gain lengths=0.0038 . . . . . . . 60
3.16 Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30,
3.17 Schematic diagram of the Stoke photon pulse construction. The tiny red spot
doesn’t represent a single atom. Instead, it denotes the individual ensemble of
atoms, and the oscillating purple line is the Stoke field produced by pumping
and σ21 interaction. The purple bar chart represents the probability of Stoke
creation with respect to time, and the blue bar chart is the Fourier transform. 61
3.18 Theoretical diagram of spectrum. The parameter setting: Ωp = 1Γ, OD=1000,
3.19 Schematic diagram of the Stoke photon pulse construction in high gain regime. 64
4.1 Theoretical diagram of Generation rate with respect to pumping power. The
parameter setting: OD=100, ∆p = 500Γ, Γ32 = Γ/2, γ21 = 0.0001Γ, The
gain lengths is 0.125 when pumping power is 0.5Γ, and gain length is 2 when
pumping power is 2Γ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74
parameter setting: Ωp = 0.5Γ, ∆p = 300Γ, Γ32 = Γ/2, γ21 = 0.0001Γ, The
gain lengths is about 3.5 when OD=1000 . . . . . . . . . . . . . . . . . . . 74
4.3 Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30
∆p = 300Γ, Γ32 = Γ/2, γ21 = 0.0005Γ, The gain lengths is about 0.02 . . . 75
4.4 Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30
∆p = −300Γ, Γ32 = Γ/2, γ21 = 0.0005Γ, The gain lengths is about 0.02 . . 75
4.5 Schematic illustration to realize the shift in spectrum by dressed state picture 76
4.6 Schematic illustration second-order autocorrelation function. The parameter
setting: Ωp = 1Γ, OD=50, γ21 = 0.001Γ, Γ32 = Γ/2, ∆p = 500Γ and gain
lengths=0.025 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
A.1 Schematic diagram of pollen’s particle and stochastic noise from environment 83
E.1 Schematic illustration of autocorrelation function. . . . . . . . . . . . . . . 96

E.2 Schematic illustration of autocorrelation function for regular signals immersed
in noise and white noise. . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
E.3 Schematic illustration of autocorrelation function for stochastic signals im-
mersed in noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
E.4 Schematic illustration of cross-correlation function for Stoke and anti-Stoke
photons. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98
viii
E.5 Schematic illustration of Hanbury Brown and Twiss experiment . . . . . . 99
E.6 Schematic illustration of antibunched, coherent, and bunched photons . . . 100
ix
Chapter 1
Introduction
With the development of quantum technology in recent years, quantum communica-

tion becomes a potential field; the target of quantum communication is the transmission of
quantum states. Such transmission can assist us in realizing several quantum applications
and solving network vulnerability. For instance, one can combine the unduplicated property
of quantum states and error correction to construct an information security system. In 1984,
Bennett and Brassard developed a quantum key distribution (QKD), which is the so-called
BB84 protocol. In 1991, Artur Ekert introduced entangled pairs as ingredients and proposed
the E91 protocol to encrypt and decrypt information safely and thereby protect messages from
eavesdropping. [1, 2, 3] Furthermore, In 2008, Kimble presented quantum networks, which
surpasses classical networks at dealing with certain issues.[4, 5] In consequence, the distri-
bution of quantum information, such as entangled states or qubits, over long distance has
always been a prospective but challenging task. To be more specific, photonic information in
any quantum channels, such as photonic channels or free-space transmission, are incredibly
susceptible to environmental disturbance, e.g., for typical fibers, the loss is about 0.2 dB/km.
[6] Namely, the photonic information will disappear exponentially over space. For example,
if we deliver a photon from Seoul to Tokyo (≃1000km) with a rate of 10 GHz, it would take
us more than hundreds of years to detect mere one single photon, which is impressively low
efficiency.
In principle, to acquire entangled state with high fidelity, one can extract high fidelity
entangled pairs from a large ensemble of low fidelity pairs; this method has been known as
entanglement purification or entanglement distillation. [7, 8] However, this approach can be
accomplished only if the minimum fidelity is preserved, which often diminish exponentially
along with distance, not to mention a large ensemble of entangled states.
In the classical realm, if we desire to amplify signals, we may exploit amplifier to boost
signals. In the quantum realm, nonetheless, no matter which kinds of operations we operate
on quantum states, the characteristics of quantum states will be altered. “no-cloning theo-
rem indicates that it is impossible to duplicate unknown quantum states.”[9] For instance,
quantum entanglement might be obliterated and collapse to a specific quantum state individ-
1
ually even though we just slightly harass it. So far, it seems unlikely to carry out quantum
communication. However, quantum problems require quantum solutions!
Fortunately, in 1993, Charles H. Bennett proposed a novel scheme called entanglement

swapping, which can connect two unrelated states and entangle them to each other by operat-
ing on the other two irrelevant quantum states. [[10, 11]] Five years later, Hans J Briegel ex-
tended this technique and proposed a new concept named quantum repeater (BDCZ scheme)
[12], which can support us to deliver quantum state with less loss. In 2001, Luming Duan and
collaborators followed the concept of the quantum repeater and further introduce atoms as an
information carrier to perform quantum memory to resolve the difficulty in BDCZ scheme,
which is known as DLCZ protocol. [13, 14, 15] Next, we will briefly illustrate the idea of
quantum swapping, quantum repeater, and DLCZ scheme.
1.1 The Swapping of Quantum Entanglement
As known, quantum entanglement has always been a fundamental and counter-intuitive

characteristic in quantum mechanics. Throughout the past one hundred years, quantum en-
tanglement, as a popular topic, gives rise to several significant debates and studies among
scientists. For instance, Albert Einstein (As he described it“entanglement is a spooky action
at a distance”) proposed EPR paradox (Einstein-Podolsky-Rosen paradox) [16] to question
the completeness of quantum mechanics and had a great debate with Niels Bohr in 1935.
In 1964, John Bell proposed Bell theorem to clarify that the hypothesis of locality in EPR
paradox is invalid. [17] Furthermore, in recent years, entanglement is regarded as a crucial
ingredient to realize quantum communication. In fact, the essential point is the swapping
property of entanglement. [10, 11]
To realize entanglement swapping, we may firstly consider two unrelated entanglement pairs
|Φ, +⟩AB and |Φ, +⟩CD
[ ]
1
|Φ, +⟩AB = √ |0⟩A ⊗ |0⟩B + |1⟩A ⊗ |1⟩B (1.1)
2
[ ]
1
|Φ, +⟩CD = √ |0⟩C ⊗ |0⟩D + |1⟩C ⊗ |1⟩D (1.2)
2
In Equation (1.1) and Equation (1.2), A, B, C, D denotes the particle Alice, Bob, Charlie, and
David possesses; Alice and Bob share a entanglement pairs and Charlie and David share the
2
other one. After that, we can regard these particles as one state |Ψ⟩ on the whole.
|Ψ⟩ =|Φ, +⟩AB ⊗ |Φ, +⟩CD (1.3)

[ ] [ ]
1 1
= √ |00⟩AB + |11⟩AB ⊗ √ |00⟩CD + |11⟩CD (1.4)
2 2
[ ]
1
= |0000⟩ABCD + |0011⟩ABCD + |1100⟩ABCD + |1111⟩ABCD (1.5)
2
In addition to this expression, we can transform them into Bell-state basis. The relation
between Bell-state and above states are
[ ]
1
|00⟩ = √ |Φ, +⟩ + |Φ, −⟩ (1.6)
2
[ ]
1
|01⟩ = √ |Ψ, +⟩ + |Ψ, −⟩ (1.7)
2
[ ]
1
|10⟩ = √ |Ψ, +⟩ − |Ψ, −⟩ (1.8)
2
[ ]
1
|11⟩ = √ |Φ, +⟩ − |Φ, −⟩ (1.9)
2
Hence, we can modify Equation (1.5) as follows

[
1 1 [ ]
|Ψ⟩ = |0⟩A ⊗ √ |Φ, +⟩ + |Φ, −⟩ ⊗ |0⟩D
2 2
1 [ ]
+|0⟩A ⊗ √ |Ψ, +⟩ + |Ψ, −⟩ ⊗ |1⟩D
2
1 [ ]
+|1⟩A ⊗ √ |Ψ, +⟩ − |Ψ, −⟩ ⊗ |0⟩D
2
]
1 [ ]
+|1⟩A ⊗ √ |Φ, +⟩ − |Φ, −⟩ ⊗ |1⟩D (1.10)
2
by extracting bell-state individually and rearranging them, we obtain
1 [ ]
|Ψ⟩BCAD = √ |Φ, +⟩ ⊗ |00⟩AD + |11⟩AD
2 2
1 [ ]
+ √ |Ψ, +⟩ ⊗ |01⟩AD + |10⟩AD
2 2
1 [ ]
+ √ |Φ, −⟩ ⊗ |00⟩AD − |11⟩AD
2 2
1 [ ]
+ √ |Ψ, −⟩ ⊗ |01⟩AD − |10⟩AD (1.11)
2 2
which implies that if we perform Bell-state measurement on Bob and Charlie’s particles,
Alice and David’s particle will collapse into certain entanglement pairs. For example, when
3
the result of Bell measurement is |Φ, +⟩, and Alice measures her particle, if it turns out with
spin up, then David’s particle will collapse into spin up immediately, and vice versa.
Initial preparation
Alice Bob Charlie David

Station 1 Station 2 Station 3
Bell-state measurement
Alice Bob Charlie David

Station 1 Station 2 Station 3
Entanglement
Alice David
Station 1 Station 3
Figure 1.1: Schematic illustration of entanglement swapping by employing Bell-state mea-

surement
4
1.2 Quantum Repeater
Even though entanglement swapping may be one of the decisive notions to achieve quan-
tum communication if we divide the channel of length l between Alice and David into N
sections of length l0 and make sure each successive sections share an entangled pair with
fidelity F1 > 0.5. Nevertheless, after each operation of entanglement swapping, which may
not be a faultless operation, the fidelity of the following entanglement will decline. Thus, it
is necessary to harness entanglement purification and swapping in turns to perform quantum
repeater.
𝐹1 𝐹1
Entanglement swapping
Fidelity< 𝐹1
Figure 1.2: Schematic illustration of the decrease of fidelity after entanglement swapping
The idea of quantum purification, as its name implies, is to enhance the quality of en-
tanglement. In general, one can utilize bilateral controlled-NOT gate on certain entangled
pair to determine whether we should keep this entangled pair or not. (For more detail about
entanglement purification, one can follow the paper by Charles Bennett, 1996.) [7] If we
acquire adequate entangled pairs with high fidelity, we will exploit these entangled pairs as
ingredients for the subsequent entanglement swapping.
𝐹1
𝐹1
𝐹1
Entanglement purification
Fidelity> 𝐹1
Figure 1.3: Schematic illustration of the increase of fidelity after entanglement purification.
5
To possess an entanglement with high fidelity over a long distance, one should adopt
a scheme with hierarchically nested entanglement purifications and swapping to circumvent
the exponential growth of the number of original ingredients (entangled pairs). [12, 18] The
procedure of quantum repeater is as follows: Firstly, we divide the distance l between Alice
and David by N sections (N = 8 for the diagram on the next page; the distance between
adjacent sections is l0 ) and create N pairs of entanglement in parallel (The fidelity of each
pair is F1 ). At this stage, if we just manipulate entanglement swapping on them directly, the
fidelity of entanglement is horrible. Therefore, we patiently prepare X number of entangled
pairs for distillation instead. Here one should note that only one-Mth of entangled pairs would
be preserved after entanglement purification. (we suppose that the swapping of two entangled
pairs with fidelity=F2 would result in an entangled pair with fidelity=F1 , and we require 2M
pairs of entanglement with fidelity=F1 as materials to be purified as 2 entangled pairs with
fidelity=F2 , where F2 > F1 . The reason we mention this is that we expect to possess an
entangled pair with fidelity=F1 over long distance ultimately) Besides, if we perform another
purification in the next step, the number of entangled pairs would be left by one-Mth again.
To sum up, if we require n times of entanglement swapping to entangle Alice and David’s
particles, we require n times of purification as well. Consequently, the number of initial
ingredients for nth level nested quantum repeater would be X = (2M )n . (The M is three for
the diagram on the next page and n=3) One can modify X as follows
[ ]
log2 X =log2 N log2 (2M ) (1.12)
=log2 N log2 (2M ) (1.13)
⇒X = N log2 M +1 (1.14)
where N = 2n , and this result indicates that the number of initial ingredients only grows
polynomially with distance. Although we have to spend more time on checking the fidelity
and waiting for the progress of entanglement purification, the payback is worthy.
6
A 𝑙 D
𝐹1 27
𝑙0
𝐹2 9
𝐹1 2𝑙0 9
𝐹2 3
𝐹1 3
𝐹2 1
𝐹1 1
Figure 1.4: Schematic illustration of nested quantum repeater. The black arrows and gray
arrows denote entanglement swapping and entanglement purification respectively; F is the
fidelity, and F2 > F1
7
1.3 DLCZ Scheme
So far, we have reviewed several major elements to achieve quantum communication.

However, there are few obstacles we have to overcome. For instance, it is unrealistic to suc-
cessfully perform entanglement purifications for all pairs at one time. Hence, it is compulsory
to introduce quantum memory to protect those purified entangled pairs to sustain their entan-
glements until all purification processes are done. In 2001, Luming Duan proposed DLCZ
protocol (LuMing Duan, Mikhail Lukin, Ignacio Cicac, Peter Zoller). The novel design of
DLCZ scheme is that they utilize excitation of atomic ensemble to accomplish entanglement
instead of a single particle. This improvement effectively enhances the generation efficiency.
Furthermore, the entangled pairs are generated by straightforward linear optical manipula-
tion, which is intrinsically robust enough to exclude the obstruction from noise. The DLCZ
scheme is as follows: At first, we force all the populations of atoms to remain in its ground
state |g⟩ (one of the Zeeman sublevels) and inject a strong off-resonance writing beam into
atoms to induce spontaneous Raman scattering. If we successfully obtain a forward Stoke
photon, the information of the writing beam will be stored in the atomic ensemble as spin-
wave coherence, which is highly correlated to the Stoke photon. The long lifetimes of ground
states coherence have been observed in trapped atoms. [19, 20, 21] And we can write down
the excitation of collective atomic mode Ŝ † as below
1 ∑
Ŝ † = √ |s⟩⟨g|i (1.15)
Na i
where Na is the number of all atoms in that ensemble, |s⟩ is the other ground state, and i is to
sum over all atoms. After that, we suppose that the generation rate is extremely low so that
the mean number of generated Stoke photons is much smaller than one, which indicates most
of the atoms remain at the ground state. By considering Stoke photon and collective atomic
ensemble as a state overall, we can express it as
√ [ † ][ ]
|Ψ⟩ = |0⟩a ⊗ |0⟩p + pc Ŝ ⊗ â† |0⟩a ⊗ |0⟩p + O(pc ) (1.16)
where pc is the probability to generate Stoke photon, |0⟩a is the ensemble ground state of
all atoms, |0⟩p denotes zero number state of Stoke photon, and O(pc ) denotes the minuscule
probability of generating a greater number of Stoke photon. To produce entangled pair, one
can prepare two atomic ensembles (Left and Right ensemble with a distance) as the previous
description and inject writing beam into ensemble. Next, we need to make sure the emitted
Stoke photon from these two ensembles will interfere at a beam splitter (50% − 50%) and
arrive at two single-photon detectors respectively. If we only receive a click among two
detectors, we cannot assure the Stoke photon is emitted from ensemble L or R. Namely, one
cannot guarantee which ensemble is excited, and these two ensembles are entangled.
8
Δ𝑝
ȁeۧ
Stoke
Write
ȁ sۧ
Spin-wave
ȁ gۧ
Figure 1.5: Schematic illustration energy level, write beam, and Stoke photon
L Filter Channel Detector 1
BS
R Filter Channel Detector 2
Figure 1.6: Schematic illustration of generating entanglement. The blue light denotes writing
beam, and purple line denotes Stoke photon
It seems like that this scheme is just a method to generate entanglement of atomic ensem-
ble. Nevertheless, the concept of swapping is implicit in itself. The number state of photon
and the excitation of that atomic ensemble are completely correlated. If there is a photon
emitted from left ensemble, then the left ensemble is excited, and vice versa. That is, the
number state of photon is entangled with the ensemble. The reason we attempt to entangle
atomic ensembles is that the atoms possess longer time to preserve coherence. Thus, we ob-
tain spare time to prepare entanglement swapping and purification.
9
Now, if we have already prepared two entangled pairs of l1 ←→ L and l2 ←→ R, we can
inject read beam on both ensemble l1 and l2 to perform the previous entangled procedure and
accomplish entanglement swapping to connect ensemble L and R.
𝑙1 Filter Channel Channel Filter 𝑙2
Entangled BS Entangled
Detector 1 Detector 2
ȁeۧ
𝐿 𝑅
Read
Anti-Stoke
ȁsۧ
ȁgۧ
Figure 1.7: Schematic illustration of entanglement swapping in DLCZ scheme
10
Chapter 2
Theoretical Model of Electromagnetically Induced

Transparency
In this chapter we will introduce the fundamental theories in quantum optics, such as
Maxwell-Schrödinger equation, which can address the evolution of field in space and time,
quantum-liouville equation, which governs the state response of atoms when they experi-
ence a specific field, and finally apply perturbation theory to resolve the analytic solution of
transmission rate for steady state. After that, we will briefly discuss several physical inter-
pretations of electromagnetically induced transparency (EIT) theory. In addition, slow-light
effect and Autler-Townes effect will be presented as well. In the end, we will provide several
schematic diagrams of transmission rate and compare the influence of optical-depth, γ21 and
coupling field.
2.1 Electromagnetically Induced Transparency
As is known to all, if we project frequency-tunable light on atoms, light will be absorbed

when its frequency is around the transition frequency. While atoms are transited to an ex-
cited state, they would be released by vacuum fields as time goes by. Because vacuum fields
are spread all around space and disturbing atoms randomly, the information of light cannot
be preserved. In 1990, Stephen Harris and his collaborators proposed a quantum interfer-
ence scheme to create a narrow-band transparent window for the probe field to pass through
medium without loss. [22] This scheme, on the whole, is based on another strong coupling
field acting on |2⟩ → |3⟩ even though there are no population at ground state |2⟩. By virtue
of EIT, the information of probe field will be stored in atoms with spin-wave form, and read
out by coupling field. Hence, we can exploit this scheme to protect the information of pho-
ton. Before we explore this scheme in-depth, we can roughly illustrate this phenomenon in a
classical interpretation. As the strong coupling field will both excite and stimulate the popu-
lation back and forth, the vacuum field is demanding to compete with coupling field and then
suppressed. For this reason, information can be reposited.
11
ȁ3ۧ ȁ3ۧ ȁ3ۧ
Ω𝑐 Ω𝑐
Ω𝑝 ȁ2ۧ
ȁ2ۧ
Ω𝑝
ȁ2ۧ Ω𝑐 Ω𝑝
ȁ1ۧ ȁ1ۧ ȁ1ۧ
Figure 2.1: Three typical types of EIT scheme: Λ-type, V-type, and cascade-type
Considering a three energy level system of Λ-type EIT. The state |1⟩ and |2⟩ are ground
states of atomic energy level and the state |3⟩ is the excited state as shown in Figure (2.2).
All the atoms are initially prepared in the ground state |1⟩, and the transition between these
two ground states is forbidden due to selection rule.
ȁ3ۧ
γ31
Ω𝑝 Ω𝑐
ȁ2ۧ
ȁ1ۧ γ21
Figure 2.2: Schematic diagram of a Λ-type EIT. The Ωp and Ωc are the Rabi frequency of
probe field and coupling field respectively, where ∆p is the one-photon detuning, δ is the
two-photon detuning, γ31 is the total decay rate of excited state |3⟩, and γ21 is the dephasing
rate of ground states
2.2 Optical-Bloch Equation
In this section, we will discuss the time-evolution of states. In quantum mechanics, the
time evolution of states, in general, can be governed by time-dependent Schrödinger equa-
tion; the evolution of operator can be described by Heisenberg equation of motion likewise.
12
Nevertheless, the states of an ensemble of atoms are mostly difficult to be determined because
the state of individual atom might not be identical. Hence, we will address density matrix ex-
pression to characterize the state of the ensemble. Additionally, we will exploit Schrödinger
equation to derive quantum Liouville-Von Neumann equation for density matrix.
2.2.1 Density Matrix Expression
Considering a quantum system composed of several particles. For all particles are in
the same state, this system can be described by an average state vector |ψ⟩, and we call it as
pure state. However, in principle, quantum system usually contains an ensemble of particles
with different states which is known as mixed state or mixture. Thus, the information of this
system is incomplete, and we cannot find an average state vector to represent it. Instead, we
characterize the mathematical formulation of mixed state as density matrix ρ.
The density matrix can be written down as:
∑
ρ= pi |ψi ⟩⟨ψi | (2.1)
i
where pi is the probability to find an individual subsystem of ensemble at state |ψ⟩
Ni
pi = (2.2)
N
where Ni is the number of particle at state |ψi ⟩, and N is the total number of particles. Notice
that we don’t assume |ψi ⟩ as an eigenstate, and the summation of index i is to sum up how
many different states of particles exist.
The element of the density matrix is expressed as:
ρmn = ⟨ϕm |ρ|ϕn ⟩ (2.3)
|ϕm ⟩ and |ϕn ⟩ are one of the eigenstate.
2.2.2 The Equation of Motion
To depict the time-evolution of state when atomic ensemble are interacting with light,
we can start from the time-dependent Schrödinger equation:
∂|ψ⟩
Ĥ|ψ⟩ = i} (2.4)
∂t
The equation of motion for density matrix can be derived as below:
13
∂ρ ∂
= (|ψ⟩⟨ψ|) (2.5)
∂t ∂t
1
= (Ĥ|ψ⟩⟨ψ| − |ψ⟩⟨ψ|Ĥ) (2.6)
i}
1
= [Ĥ, ρ] (2.7)
i}
Equation (2.7) is known as quantum Liouville-Von Neumann equation, and it should be no-
ticed that this equation is different from the Heisenberg equation of motion by a minus sign.
The reason is that density matrix doesn’t correspond to any physical observable which means
it is not an operator.
Even though Liouville-Von Neumann equation is quite general, it is just suitable for
the ideal case. In practice, the atomic ensemble is never an isolated system. For instance,
atoms will be disturbed by vacuum field. As the stimulation of vacuum field, atoms will emit
photons randomly. Moreover, atoms may not only have interaction with the environment due
to unpredictable effects but also have interaction with other atoms. Hence, it is necessary to
introduce relaxation. The equation of motion is modified as:
∂ρ 1 ∂ρ
= [Ĥ, ρ] + { } (2.8)
∂t i} ∂t
This equation is known as Optical-Bloch equation. [23] The second term on the right-hand
side represents the phenomenological results of spontaneous emission and dephasing.
Figure 2.3: Schematic illustration of the dephasing caused by vacuum fields and collisions
of atoms.
In general, vacuum fields with uncountable modes spread all over the universe. Ac-
cordingly, vacuum fields will contribute to the transition of population from excited state to
ground state by decay rate Γ. Besides, it will also lead to the decoherent rate by Γ2 because the
fluctuation from vacuum is random in time-domain. It can be realized as follows: considering
a two-level system, and the ensemble is initially a pure state, as the first atom spontaneously
emits to ground state, the state of the first atom will free evolute. Hence, if all atoms decay
14
randomly, the accumulated phase between excited state and ground states of each atom will
cancel out as we regard them as an ensemble. Also, as a result of the collisions of atoms
and inhomogeneous magnetic field, the coherence of atomic ensemble, corresponding to the
non-diagonal terms of density matrix, will be destroyed.
Considering the Hamiltonian of our system:
Ĥ = Ĥ0 + ĤI (2.9)

 
0 0 0
 
Ĥ0 = Ekinetic energy + 0 }ω21 0  (2.10)
0 0 }ω31
Ĥ0 denotes the Hamiltonian of atoms, and ĤI indicates the interaction part. Owing to what
we concern about is the energy difference, we set ground state |1⟩ energy is zero.
We assume that the electric fields are plane wave, and the coupling field is much stronger
than probe field.
 
0 0 −d⃗13 · E
⃗
⃗ =
ĤI = −d⃗ · E  0 0 ⃗
−d⃗23 · E  (2.11)
−d⃗31 · E
⃗ −d⃗32 · E
⃗ 0
where d⃗ is the electric dipole moment of atoms. The reason of Ĥ12 and Ĥ21 equal zero is that
the ground states transition is forbidden.
The electric field in our consideration is propagating along with ẑ direction.
⃗ =E
E ⃗ probe + E ⃗ p ei(⃗kp ·⃗z−ωp t) + E
⃗ coupling = E ⃗ c ei(⃗kc ·⃗z−ωc t) + c.c. (2.12)
Utilizing the unitary operator U = exp(−iĤ0 t/}) to transform the interacting Hamiltonian
into Dirac picture:
 
1 0 0
 
U = 0 e−i}ω21 t 0  , ĤDirac = U † ĤI U (2.13)
0 0 e−i}ω31 t
Take the ĤDirac,13 and ĤDirac,23 term as an example,
ĤDirac,13 = −d13 e−iw31 t (Ec ei(kc z−ωc t) + Ec∗ ei(−kc z+ωc t) + Ep ei(kp z−ωp t) + Ep∗ ei(−kp z+ωp t) )
(2.14)
ĤDirac,23 = −d13 e−iw32 t (Ec ei(kc z−ωc t) + Ec∗ ei(−kc z+ωc t) + Ep ei(kp z−ωp t) + Ep∗ ei(−kp z+ωp t) )
(2.15)
15
As the frequency of electromagnetic field is approximately equal to the resonant fre-
quency of atomic energy level, we can observe that most of oscillating terms in Hamiltonian
are quite rapid. Hence, we just leave the slowly oscillating term. For instance, in ĤDirac,13 ,
exp(−i(ω31 + ωc )t), exp(−i(ω31 − ωc )t), and exp(−i(ω31 + ωp )t) are rapid oscillating term
comparing to exp(−i(ω31 − ωp )t). In time scale these terms will be average to none. Thus,
we can neglect these terms, and this is so-called rotating-wave approximation. After that, we
transform the Hamiltonian in Dirac’s picture back to Schrödinger picture.
The remaining Hamiltonian is:
 
0 0 −d13 · Ep∗ e−i(kp z−ωp t)
 
ĤIRW A = 0 0 −d23 · Ec∗ e−i(kc z−ωc t)  (2.16)
−d31 · Ep ei(kp z−ωp t) −d32 · Ec e−i(kc z−ωc t) 0
As we known, Rabi frequency indicates the transition rate of population between atomic
energy levels. As the dipole moment or electric field increase, atoms will be more susceptible
to be perturbed by electric field. Namely, the intensity of Rabi frequency reflects the strength
of interaction among atoms and the applied field. Hence, we introduce Rabi frequency to
substitute electric field. The relation between Rabi frequency and the electric field is:
2d⃗ij · E
⃗
Ωij = (2.17)
}
The formation of Hamiltonian will be:
 
0 0 Ω∗p e−i(kp z−ωp t)
−}  
ĤI =  0 0 Ω∗c e−i(kc z−ωc t)  (2.18)
2
Ωp ei(kp z−ωp t) Ωc ei(kc z−ωc t) 0
After the approximation above, the exponential terms in Hamiltonian is still kind of pesky.
To simplify the further calculation, we choose a specific rotating coordinate system which is
oscillating with the electric field. Therefore, the exponential terms will be dispelled, and the
atomic coordinate system will gain oscillating terms.
We define the coordinate-transformation operator:
 
1 0 0
 
Ŝ = 0 e−i(ωp −ωc )t 0  (2.19)
0 0 e−iωp t
Next, we should notice that the transformed Hamiltonian H̃ should still satisfy Schrödinger-
equation as shown below:
16
∂ ∂ ∂ Ŝ † ∂
H̃|ψ̃⟩ = i} |ψ̃⟩ = i} S † |ψ⟩ = i} |ψ⟩ + i}S † |ψ⟩ (2.20)
∂t ∂t ∂t ∂t
∂ Ŝ † ∂ Ŝ †
= i} |ψ⟩ + S † Ĥ|ψ⟩ = i} Ŝ Ŝ † |ψ⟩ + Ŝ † Ĥ Ŝ Ŝ † |ψ⟩ (2.21)
∂t ∂t
∂ Ŝ †
= [i} Ŝ + Ŝ † Ĥ Ŝ]|ψ̃⟩ (2.22)
∂t
where |ψ̃⟩ = Ŝ † |ψ⟩
Finally, we can obtain the formulation of transformation Hamiltonian:
∂ Ŝ †
H̃ = [i} Ŝ + Ŝ † Ĥ Ŝ] (2.23)
∂t
Besides the transformation of Hamiltonian, we should also notice that we are in a specific
oscillating coordinate. Density matrix, in contrast, would gain a exponential term.
 
ρ̃11 (z, t) ρ̃12 (z, t) ρ̃13 (z, t)
 
ρ̃(z, t) = Sˆ† |ψ⟩⟨ψ|Ŝ = ρ̃21 (z, t) ρ̃22 (z, t) ρ̃23 (z, t) (2.24)
ρ̃31 (z, t) ρ̃32 (z, t) ρ̃33 (z, t)
We next write down the final Hamiltonian as below

 
0 0 Ω∗p
−}  
H̃ =  0 2δ Ω∗c  (2.25)
2
Ωp Ωc 2∆p
2.3 Steady-State Solution
We have presented the formation of the Hamiltonian and Optical-Bloch equation above.
In this section, we will utilize the perturbation theory to derive the analytic solution of probe
field under steady-state assumption.
2.3.1 Brief Review of Perturbation Theory
In practice, dynamic problems in quantum mechanics are too complex to be resolved

exactly. However, one can still implement perturbation theory to approach an approximate
result. For instance, as the probe field is much weaker than coupling field, the probe field
17
can be regarded as perturbation term as follow:
   
0 0 0 0 0 Ω∗p
−}   −}  
H̃ = H̃ (0) + λH̃ (1) = 0 2δ Ω∗c  + λ 0 0 0 (2.26)
2 2
0 Ωc 2∆p Ωp 0 0
Similarly, density matrix and relaxation term can be viewed as:

 (0) (0) (0)   (1) (1) (1) 
ρ̃11 ρ̃12 ρ̃13 ρ̃11 ρ̃12 ρ̃13
 (0) (0) (0)   (1) (1) (1) 
ρ̃ = ρ̃(0) + λρ̃(1) = ρ̃21 ρ̃22 ρ̃23  + λ ρ̃21 ρ̃22 ρ̃23  (2.27)
(0) (0) (0) (1) (1) (1)
ρ̃31 ρ̃32 ρ̃33 ρ̃31 ρ̃32 ρ̃33
∂ρ
Relaxation = { } = R(0) + λR(1) (2.28)
∂t
 (0) (0) (0)

  (1) (1) (1)

Γ31 ρ33 − γ221 ρ12 − γ231 ρ13 Γ31 ρ33 − γ221 ρ12 − γ231 ρ13
   
= − γ221 ρ(0)
21
(0)
Γ32 ρ33
(0)
− γ232 ρ23
(1) (1)
 + λ − γ221 ρ21 Γ32 ρ33
(1)
− γ232 ρ23 
γ31 (0) γ32 (0) (0) γ31 (1) γ32 (1) (1)
− 2 ρ31 − 2 ρ32 −(Γ31 + Γ32 )ρ33 − 2 ρ31 − 2 ρ32 −(Γ31 + Γ32 )ρ33
(2.29)
Following Optical-Bloch equation, and extract those terms accompanied with λ0 and λ1 re-
spectively.
d (0) 1
[ρ̃ + λρ̃(1) ] = [H̃ (0) + λH̃ (1) , ρ̃(0) + λρ̃(1) ] + R(0) + λR(1) (2.30)
dt i}
d (0) 1 (0) (0)
ρ̃ = [H̃ , ρ̃ ] + R(0) (2.31)
dt i}
d (1) 1 1
ρ̃ = [H̃ (0) , ρ̃(1) ] + [H̃ (1) , ρ̃(0) ] + R(1) (2.32)
dt i} i}
According to population conservation, the solution of zero-order equation is:
 
1 0 0
 
ρ̃(0) = 0 0 0 (2.33)
0 0 0
This result is trivial since the coupling field will impel the population back to ground state |1⟩,
we believe that population are mostly deposited in |1⟩. After that, we can calculate first-order
equation with boundary condition.
(1) (1) (1)

B.C. ρ̃11 + ρ̃22 + ρ̃33 = 0 (2.34)
18
And the first order system of equation is:
dρ˜21 −γ21 Ω∗
=( + iδ)ρ˜21 + c ρ˜31 (2.35)
dt 2 2
dρ˜31 iΩp γ31 iΩc
= + (i∆p − )ρ˜31 + ρ˜21 (2.36)
dt 2 2 2
Recall that we consider steady state condition ρ̃˙ ij = 0. Accordingly the solution is as fol-
lowed:
Ω∗c Ωp
ρ˜21 = (2.37)
(iγ21 + 2δ)(iγ31 + 2∆p ) − |Ωc |2
(iγ21 + δ)Ωp
ρ˜31 = (2.38)
(γ21 − 2iδ)(γ31 − 2i∆p ) + |Ωc |2
In experiment, the intensity of coupling field is much more powerful than probe field, which
will compel the population remains in |1⟩. As a result, coupling field is barely interacting
with atoms, and the strength of coupling can be deemed as a constant independent to space
and time.
19
2.4 Maxwell-Schrödinger Equation
Up to present, one can employ OBE to demonstrate the evolution of density matrix
under a specific Hamiltonian. Nevertheless, Probe fields and atomic dipole moment vary at
different locations as time goes by. In consequence, it is compulsory to introduce Maxwell-
Schrödinger-equation, which can describe the time-space-evolution of electric field. [24] To
begin with, we start from the set of Maxwell-equations in matters:
The Gauss law for electric field : ∇ · D = ρfree (2.39)

The Gauss law for magnetic field : ∇ · B = 0 (2.40)
∂B
The maxwell Faraday equation : ∇ × E = − (2.41)
∂t
∂D
The revised Ampere′ s law : ∇ × H = Jfree + (2.42)
∂t
Where, M is the magnetization, ρf ree is free charge, and D is the displacement field. Secondly,
we can look at the curl of Faraday’s equation.
∂
∇×∇×E=− (∇ × B) (2.43)
∂t
(2.44)
In our experimental arrangement, there are no magnetic material M = 0, current Jfree = 0,

and free charge ρf ree = 0.
∂
∇ × ∇ × E = −µ0 ∇×H (2.45)
∂t
∂ ∂D
= −µ0 (Jfree + ) (2.46)
∂t ∂t
∂ 2D
= −µ0 2 (2.47)
∂t
where D = ϵ0 E + P, and P is the electrical polarization. Associated with the relation: ∇ ×
∇ × E = ∇(∇ · E) − ∇2 E We assume that electric field is a plane wave E = E0 ei(kz−wt)^ x.
Thus, ∇ · E = 0. And the equation is simplified as
1 ∂ 2E ∂P
∇2 E − = µ 0 (2.48)
c2 ∂t2 ∂t
20
~
The electrical polarization can also be expressed as P = n⟨d⟩
⃗
P = nTr(ρd) (2.49)
∑
=n ⃗ ii
(ρd) (2.50)
i
∑
=n ρij d⃗ji (2.51)
ij
notice that d⃗ij = d⃗12 = d⃗21 = 0 as i = j

Transform ρ to ρ̃ to extract exponential term.
R.H.S of MSE:
∂ 2P
−µ0 = µ0 nρ̃31 ωp2 d13 ei(kp z−ωp t) + µ0 nρ̃32 ωc2 d12 ei(kc z−ωc t) + c.c. (2.52)
∂t2
Also, we consider the amplitude of electric field vary slowly with respect to time and space
∂Ep,c (z, t) ∂Ep,c (z, t)

≪ |ωp,c Ep,c (z, t)|, ≪ |kp,c Ep,c (z, t)| (2.53)
∂t ∂z
L.H.S of MSE can be simplified as:
1∂ ∂ 1∂ ∂
−2ikc eikc z−ωc t ( + )Ec (z, t) − 2ikp eikp z−ωp t ( + )Ep (z, t) + c.c. (2.54)
c ∂t ∂z c ∂t ∂z
As mentioned above, the coupling field is a constant, and we can extract each exponential
|d |2 ωp2
term. With the substitution of γ31 = 31
3πϵ3
, the MSE for probe is:
0
1∂ ∂ iγ31 ODρ̃31
( + )Ωp = (2.55)
c ∂t ∂z 2L
2
L is the length our atomic ensemble OD= 3λ2πnL represents the optical depth, and it indicates
the absorption rate of incident light. Considering steady state solution:
∂ iγ31 ODρ̃31
Ωp = (2.56)
∂z 2L
Substitute the solution ρ̃31 from OBE into MSE, we can obtain the formula of Ωp (z)
ODzγ31 (γ21 −2iδ)
−
2L((γ21 −2iδ)(γ31 −2i∆p )+|Ωc |2 )
Ωp (z) = Ωp (0)e (2.57)
In an isotropy and linear dielectric medium, the electrical polarization is :
P = ϵ0 χe E (2.58)
21
where χe is the electric susceptibility. Associated with the Equation (2.4) and Weisskopf-
|dij |2 ω 3
Wigner theory, Γij = 3πϵ 0 }c
3 , we can attain the relation as followed:
ndij ρ̃ji Γij 3nλ3 ODΓij ρ̃ji

χe = = 2
= (2.59)
ϵ0 E0 Ωji 4π kΩji L
We consider χe ≪ 1. The dispersion relation is

ω √ √ χe
k= = ω ϵµ ≈ k0 1 + χe ≈ k0 (1 + ) (2.60)
v 2
Im(χe ) χe
eikz = e− 2 k0 z ei(1+Re( 2 ))k0 z (2.61)
The transmission rate and phase shit will be
Ωp (L) 2
Transmission rate : Tp = e−Im(χe )k0 L =| | (2.62)
Ωp (0)
−ODγ31 (γ212 γ +4γ δ 2 +γ |Ω |2 )
31 31 21 c
2 δ 2 +γ 2 (γ 2 +4∆2 )+2γ γ31|Ω |2 +(|Ω |2 −4δ∆ )2
4γ31
=e 21 31 p 21 c c p
(2.63)
1 ODΓ31 Re(ρ̃31 )
Phase shift : ∆ϕ = Re(χe )k0 L = (2.64)
2 2Ωp
(γ21 ∆p + δ(4δ∆p − |Ωc | ))
2 2
=− 2 2 (2.65)
4γ31 δ + γ21 (γ31 + 4∆2p ) + 2γ21 γ31 |Ωc |2 + (|Ωc |2 − 4δ∆p )2
2 2
22
2.5 Light Shift (AC Stark Effect)
Afterwards, we will discuss how the intensity of coupling field alters the effect of trans-
parency. To start with, we firstly ignore the probe field and normalize the Hamiltonian of
field and atomic level, which can be expressed as follows
[ ]
} 0 Ω∗c
H̃ = − (2.66)
2 Ωc 2∆c
where ∆c denotes one photon detuning. (one should remind that the energy level of |2⟩ and
|3⟩ are shifted even though we don’t apply coupling field; the reason is that these states are
under slowly varying amplitude transformation, thus, the energy levels are different from the
energy level in original basis.) By diagonalizing the Hamiltonian, we can deem it as a new
atomic system with new eigen energy. The eigenvectors and eigenvalues are
} √
E1 = [−∆c − ∆2c + |Ωc |2 ] (2.67)
2
} √
E2 = [−∆c + ∆2c + |Ωc |2 ] (2.68)
2 √
∆c − ∆2c + |Ωc |2
|E1 ⟩ = − |2̃⟩ + |3̃⟩ (2.69)
Ωc
√
∆c + ∆2c + |Ωc |2
|E2 ⟩ = − |2̃⟩ + |3̃⟩ (2.70)
Ωc
After that, we can consider two types of situations; ∆c = 0 and ∆c ≫ Ωc .

For ∆c = 0 eigenstates and eigenvalues will be
−}
E1 = Ωc |E1 ⟩ = |2̃⟩ + |3̃⟩ (2.71)
2
}
E2 = Ωc |E2 ⟩ = −|2̃⟩ + |3̃⟩ (2.72)
2
As ∆c = 0, the atomic energy levels will be modified as |E1 ⟩ and |E2 ⟩ with energy difference
}Ωc . One can interpret this phenomena as follows: Atoms are just like wearing coupling field
when we regard coupling field and atom into a new atomic system, which is so-called dressed
state picture. For ∆c ≫ Ωc , eigenstates and eigenvalues will be
}[ ( |Ωc |2 ) 12 ] Ω2c
E1 = − ∆c − ∆c 1 + ∼ −∆ c } − } |E1 ⟩ ∼ |3̃⟩ (2.73)
2 |∆c |2 4∆c
}[ ( |Ωc |2 ) 12 ] Ω2c
E2 = − − ∆c + ∆c 1 + ∼ − } |E2 ⟩ ∼ |2̃⟩ (2.74)
2 |∆c |2 4∆c
23
intensity
ȁE 2 ۧ
ȁ 3ۧ Ω𝑐 ȁ3ۧ
ȁE 2 ۧ ȁE1 ۧ
Ω𝑐
ȁE1 ۧ Ω𝑝
ȁ 2ۧ ȁ 2ۧ
ȁ 1ۧ
Figure 2.4: Schematic diagram of dressed state picture when one photon detuning ∆c = 0.
The left-hand side reveals that we can integrate atoms and coupling fields into another new
atomic system. For this reason, probe field will be absorbed as one photon detuning ∆p
equals to ± Ω2c
As coupling field gets stronger, the splitting effect enhance. Therefore, probe will not be
absorbed near resonance. Instead, absorption peak will appear at another detuning ∆p =
± 12 Ωc
2.6 Figures of Electromagnetically Induced Transparency
In this section, we will demonstrate the role of coupling field, optical depth, the dephas-
ing rate of two ground states, one photon and two-photon detuning in electromagnetically
induced transparency.
To begin with, let’s follow the last topic we consider about, which is AC Stark effect.
Figure 2.5: Schematic diagram of transmission with different Ωc . The parameter setting:
OD= 5, γ21 = 0, γ31 = Γ, ∆c = 0, and Ωc = 2Γ and Ωc = 6Γ for Figure (a) and (b)
respectively
24
As the increase of coupling strength, the transparency window is open out. The energy
levels are reorganized into two shift-energy levels owing to AC Stark effect mentioned above.
The width of the maximum absorption peak is about the strength of coupling Rabi frequency.
In experiment, we can utilize this width to determine the coupling strength.
0 .8
0 .4
∆Φ ( π)
0 .0
-0 .4
-0 .8
-4 -2 0 2 4
∆p ( Γ)
Figure 2.6: Schematic diagram of phase shift with parameters : γ31 = 1Γ, Ωc = 1Γ, OD=
10, δ = ∆p , γ21 = 0
The phase varies dramatically around resonant frequency. The abrupt phase shift will
be reflected in the dispersion relation and lead to slow light effect. Slow-light effect will be
review at the end of this chapter.
Figure 2.7: Schematic diagram of transmission with different γ21 . The parameter setting:
γ31 = 1Γ, Ωc = 2Γ, OD= 20, δ = ∆p , γ21 = 0 and γ21 = 0.1Γ for Figure (a) and (b)
respectively
It is worth noting that as γ21 = 0, the transmission may achieve 100% if two-photon
detuning equals zero. As demonstrated above, the probe field transmission rate was reduced
25
when γ21 arises. We can realize this effect as follows. The probe field, in fact, doesn’t pass
through atoms directly without interaction. Instead, it is involved in a process; the informa-
tion of probe field, including the phase, was initially stored in atoms by spin-wave coherence.
After that, the probe field was read out by coupling field. This progress, practically, is sus-
ceptible to γ21 because γ21 will result in a decoherence process. If the coherence of these two
ground states is wiped out, the probe field will be read as fluorescence. Therefore, the inten-
sity of probe field decreases, In fact, those read-out probe fields are uncorrelated to original
ones.
Figure 2.8: Schematic diagram with different optical depth. The parameter setting: γ31 =
1Γ, Ωc = 5Γ, ∆c = 0, γ21 = 0, OD= 10 and OD= 100 for Figure (a) and (b) respectively
With the increase of optical depth, the width of transmission spectrum is compressed. Be-
cause optical depth enhances the delay time. If the pulse of the read-out probe field spread
wider in time domain, the width of its frequency will be narrower by Fourier transformation.
Figure 2.9: Schematic diagram of transmission. The parameter setting: γ31 = 1Γ, Ωc = 5Γ,
OD= 30, γ21 = 0, ∆p = 0 and δ = 0 for Figure (a) and (b) respectively
The left-hand side indicates that transmission rate will decay by increasing or decreasing
26
two-photon detuning. As two-photon detuning is not zero, the result of quantum interference
between two fields may be destructive. On the other hand, if two-photon detuning is zero,
the transmission rate achieves 100% whether one photon detuning is large or not. Conse-
quently, one may only alter two-photon detuning to adjust to maximum transmission rate in
the experiment.
27
2.7 Slow Light Effect
It is well known that speed of light is the upper limit of velocity. However, people
might run faster than light in a sense. In EIT experiment, the group velocity of light may be
decreased since light pulse will stop by and spend time on interacting with atoms. We can
also obtain this result by the definition.
dω c
vg = = (2.75)
dk n + ωp ( dn(ω)
dω
)
where n = 1 + 12 Re(χe )
As dn/dω > 0, the group velocity will decrease. Also, as dn/dω < 0, the group velocity
may have a chance to exceed the speed of light. Assuming that γ21 = 0 and δ = 0. Delay
time will be simplified as:
L L
∆T = Tgroup − Tvacuum = − (2.76)
vg c
d 1 d
= [ Re[χe ]kL] = ∆ϕ (2.77)
dω 2 d∆p
1 ODΓ31 1 ODΓ31 2Ωp
= Re[ρ̃31 ] = (2.78)
2 Ωp 2 Ωp |Ωc |2
ODΓ31
= (2.79)
|Ωc |2
According to the above formula, we can conclude that the increase of optical depth and de-
crease of coupling field both extend the delay time.
We may interpret this phenomenon in a classical way. At t1 , the ensemble absorbs a tiny
portion of probe field, and creates spin-wave coherence between two ground states. At t2 ,
as the ensemble absorbs coupling field, the spin-wave coherence was readout. Likewise, the
probe field will be absorbed as it reaches another atom, and so on. It seems like atoms are
dragging the probe field to pass through the ensemble of atoms. The role of coupling field is,
on the other hand, a read-out procedure. As coupling field declines, the chance of reading out
spin-wave decreases as well. Moreover, if we turn off coupling in large optical depth regime,
the probe field would even be stored in atoms. Thus, quantum memory can be accomplished.
[25]
Next, we provide the retard of probe pulse in EIT experiment.
28
(
a) (
b)
Figure 2.10: Schematic diagram of probe pulse in EIT experiment. The parameter setting:
γ31 = 1Γ, Ωc = 0.5Γ, δ = 0, γ21 = 0.0002, and OD= 30 and OD= 150 for Figure (a) and
(b) respectively
ȁ3ۧ ȁ3ۧ
𝑡1
Ω𝑐 Ω𝑐
Ω𝑝 Ω𝑝
ȁ2ۧ ȁ2ۧ
ȁ1ۧ ȁ1ۧ
ȁ3ۧ ȁ3ۧ
𝑡2
Ω𝑐 Ω𝑐
Ω𝑝
ȁ2ۧ Ω𝑝 ȁ2ۧ
ȁ 1ۧ ȁ1ۧ
Figure 2.11: Schematic diagram of slow light. The red spot in the middle means parts of
the ensemble of atoms. At t1 , probe field was transformed as spin-wave coherence by cou-
pling field. At t2 , another coupling field readout the coherence. Next, the probe field passes
through another atom and repeats the previous process. Therefore, probe fields get retarded
29
2.8 Interpretation of EIT
ȁ3ۧ ȁ3ۧ ȁ3ۧ ȁ3ۧ
Ω𝑐 Ω𝑐 Ω𝑐 Ω𝑐
Ω𝑝 = Ω𝑝 + Ω𝑝 + Ω𝑝
ȁ2ۧ ȁ2ۧ ȁ2ۧ ȁ 2ۧ
ȁ1ۧ ȁ1ۧ ȁ1ۧ ȁ1ۧ
Figure 2.12: Illustration of transparency
In addition to above calculation to obtain analytic solution of transparency, one can still
realize this phenomenon in a physical description. The probability amplitude to absorb probe
field in EIT scheme can be regarded as the superposition of several transition amplitude
with different path. For instance, the absorption of probe might be: one transition from
|1⟩ → |3⟩, |1⟩ → |3⟩ → |2⟩ → |3⟩, or |1⟩ → |3⟩ → |2⟩ → |3⟩ → |2⟩ → |3⟩...and
so on. As the transition amplitudes are out of phase, the total transition amplitude will be
zero(|A1 + A2 + A3 + ...| = 0). Thus, we can deem it as a result of quantum interference.
After that, we will utilize coherent population trapping to comprehend this phenomenon on
the next page.
30
2.8.1 Coherent Population Trapping
In addition to the interpretation of EIT discussed above, we can also apply a clear con-
cept, CPT, to realize this phenomenon. [26, 27] To begin with, review the Hamiltonian
 
0 0 Ω∗p
−}  
H̃ =  0 2δ Ω∗c  (2.80)
2
Ωp Ωc 2∆p
To simplify calculation, we set δ = ∆p = 0, the eigenvalues and eigenvectors are:
λ1 = 0 |e1 ⟩ = cosθ|1⟩ − sinθ|2⟩ (2.81)

−} 1
λ2 = Ω |e2 ⟩ = √ {sinθ|1⟩ + cosθ|2⟩ + |3⟩} (2.82)
2 2
} 1
λ3 = Ω |e3 ⟩ = √ {−sinθ|1⟩ − cosθ|2⟩ + |3⟩} (2.83)
2 2
√
where Ω = |Ωc |2 + |Ωp |2 , and tanθ = Ωp /Ωc .
The chance of population in first eigenstate |e1 ⟩ to be excited to |3⟩ is ⟨3|H̃|e1 ⟩ = 0; we call
this state as dark state because this state won’t interact with light. It is noteworthy that as the
coupling field is much stronger than probe field, the dark state |e1 ⟩ ∼ |1⟩, which indicates
that |1⟩ in our system is a dark state. On the other hand, the bright state |B⟩ is defined as
sinθ|1⟩+cosθ|2⟩. Next, we transform the Hamiltonian into |dark⟩, |Bright⟩, and |3⟩ basis.
1 1
H̃eigenbasis = λ2 [|B⟩ + |3⟩][⟨B| + ⟨3|] + λ3 [−|B⟩ + |3⟩][−⟨B| + ⟨3|] (2.84)
2 2
The transformed Hamiltonian element is H̃new,ij = ⟨i|H̃eigenbasis |j⟩, where i, j represents

dark state, bright state, and |3⟩.
 
0 0 0
} 
H̃new = − 0 0 Ω (2.85)
2
0 Ω 0
In the new Hamiltonian, there exists a field with Rabi-frequency Ω operating on bright
state and excited state. A non-related dark state. As populations are transferred to excited
state and then spontaneous emit to bright state and dark state, the population deposited in
dark state will be build up. Finally, as population are totally trapped at a dark state; the Ω
field can pass through atoms. That is to say, the absorption and emission of light cancel out
due to certain coherent superposition of state. As a result of that dark state is a coherent
superposition of two ground states, this phenomena is called coherent population trapping.
[28, 29]
31
ȁ3ۧ
ȁ𝐵𝑟𝑖𝑔ℎ𝑡ۧ
ห𝑒1 ۧ = ȁ𝐷𝑎𝑟𝑘ۧ
Figure 2.13: Schematic diagram of coherent population trapping
32
Chapter 3
Stoke Photons Generation Based on DLCZ Protocol
As mentioned in the introduction, quantum communication has always been an essential

topic in quantum information. However, quantum states are quite vulnerable to the environ-
ment. Hence, scientists are dedicated to exploring solutions to protect quantum states. For-
tunately, a concept named quantum repeater was proposed by Hans J Briegel to overcome
the difficulty of long-distance quantum communication. In general, quantum repeater can be
achieved by biphoton scheme and DLCZ (Duan-Lukin-Cirac-Zoller) scheme. [13]
DLCZ and biphoton schemes [30] are just slightly different from each other. In DLCZ
scheme, the generation of photons involves two procedures. The first step: we inject a far-
detuned laser pulse, write, which induces a Raman scattering process, and Stoke photons will
be generated. The second step: we illuminate atoms by a resonant laser pulse (|2⟩ → |4⟩),
which resembles an EIT process, to read out anti-Stoke photons. On the other hand, biphoton
scheme can be achieved if we inject the write beam and read beam simultaneously. Interest-
ingly, the generation of Stoke beams based on these two schemes is quite different.
In this chapter, we will mainly discuss the generation rate of Stoke photons based on
DLCZ scheme with two methods. The first method was based on Laplace transform and
moving frame transformation to obtain Stoke photons generation. [31, 32] Furthermore, we
are inspired by Dr.Kolchin, [30, 33] who solved Biphoton generation, and we employ Fourier
transform to resolve Stoke photon generation in DLCZ scheme. Finally, we compare the
generation rate based on these two approaches. In addition, we provide the spectrum and
second-order autocorrelation function of Stoke photons.
33
3.1 Brief Review
In our far-detuning arrangement, the emitted photons are generated by Raman scattering.
If the detuning is not tremendous enough, two problems will occur. The first one is that some
of the populations will be impelled into the other ground state, which enhances the difficulty
a lot for us to compute Stoke photon generation. The second one is that if detuning is not far
enough, some population will stay at excited state, and then emit by spontaneous emission.
Those spontaneously emitted photons are not correlated to the corresponding anti-Stoke pho-
tons.
Because our writing beam and a specific mode of vacuum field interact with atoms simul-
taneously by Raman process, atoms with a tiny population will be excited to a virtual state
and then scatter to the other ground state |2⟩, which is so-called spontaneous Raman scat-
tering. This process is a two-photon process, which indicates that the absorption pumping
field and the emission of Stoke photons occur simultaneously. If Stoke photons are abundant
enough, then it is possible that these Stoke photons will induce the generation of other Stoke
photons. We call this process stimulated Raman scattering.
3.2 Stimulated Raman Scattering
ȁ3ۧ
γ31
𝑎ො𝑠
Ω𝑝
ȁ2ۧ
γ21
ȁ1ۧ
Figure 3.1: Schematic diagram of energy level of stimulated Raman scattering, where ∆p =
ωp − ω31
At first, we prepare the population of a collection of identical atoms in ground state |1⟩.
These atoms ensemble are assumed to be pencil-shape, and the detune ∆p = ωp − ω31 is
34
extremely huge, allowing us to make ground state approximation reasonably. The electric
fields of our pumping beam and Stoke photon are expressed as:
⃗ p = Ep (z, t)e−i(kp z−ωp t) êp + E ∗ (z, t)ei(kp z−ωp t) êp

E (3.1)
p
⃗ −i(ks z−ωs t) ∗
Es = Es (z, t)e ês + Es (z, t)ei(ks z−ωs t) ês (3.2)
In semi-classical regime, electric field can be written in classical expression. The square
of electric field’s amplitude corresponds to the intensity of electric field, which can be de-
termined without measurement. Nevertheless, in quantum model, we will regard field as an
operator. The quantization of electric field will be discussed in appendix.
In experiment, the pumping field is quite strong, and it is not our target field. Hence, we still
treat it in classical regime. The expression of Stoke field:
√ √
⃗s = }ωs }ωs † −i(ks z−ωs t)
E âs ei(ks z−ωs t) ês + â e ês (3.3)
2ϵ0 V 2ϵ0 V s
⃗ =E
E ⃗p + E
⃗s (3.4)
Recall that âs and a†s is the annihilation operator and creation operator of Stoke photon, re-
spectively. ⟨â†s âs ⟩ represents the number of photon.
3.2.1 Collective Atomic Operator
In semi-classical treatment, we usually utilize density matrix to represent the state of

atoms directly. Yet, in quantum model, the atoms behaviour are depicted in terms of collective
atomic operator σ̂jk
The definition of collective atomic operator:
1 ∑
Nz
σ̂jk = |j⟩i ⟨k|i (3.5)
Nz i
where Nz is the number of atoms around z − δz to z + δz, and the index i means ith atom.
The summation is to average the collective behaviour of atoms.
It should be notice that σ̂jk is an operator; one cannot obtain any information about atoms
without operation. For instance, the population in ground state is ⟨ψ|σ̂11 |ψ⟩ rather than σ̂11 .
The relation between atomic operator and density matrix:
ρjk = ⟨j|ψ⟩⟨ψ|k⟩ = ⟨σ̂kj ⟩ (3.6)
35
3.2.2 Hamiltonian
Same as the formation in semi-classical regime, the Hamiltonian can be expressed as:
 
0 0 −⟨1|d⃗ · E|3⟩
⃗
 ⃗ 
Ĥ =  0 }ω21 −⟨2|d⃗ · E|3⟩ (3.7)
⃗ ⃗ ⃗
−⟨3|d · E|1⟩ −⟨3|d · E|2⟩
⃗ }ω31
Recall that the ground states transition is forbidden in our consideration. d⃗ is the electric
dipole moment; we define the coupling constant of Stoke field and dipole as gs , and the Rabi
frequency is Ωp .
√
d32 Es }ωs d32
gs = = (3.8)
} 2ϵ0 V }
2d31 Ep
Ωp = (3.9)
}
ˆ
where dij = ⟨i|d|j⟩. The coupling constant, which is similar to Rabi-frequency, denotes the
strength of interaction between Stoke field and atomic system. After that, we can write down
the Hamiltonian under rotating wave approximation and slowly varying amplitude.
[
−}
H̃ = 0 × σ̃11 (z, t) + 0 × σ̃12 (z, t) + Ω∗p σ̃13 (z, t)
2
+0 × σ̃21 (z, t) + 0 × σ̃22 (z, t) + 2gs∗ â†s σ̃23 (z, t)
]
+Ωp σ̃31 (z, t) + 2gs âs σ̃32 (z, t) + ∆p σ̃33 (z, t) (3.10)
Notice that the Hamiltonian H̃ is merely the collective Hamiltonian around z. The general
expression of Hamiltonian including all atoms is
∫ [
−} N L
H̃sum = 0 × σ̃11 (z, t) + 0 × σ̃12 (z, t) + Ω∗p σ̃13 (z, t)
2 L 0
+0 × σ̃21 (z, t) + 0 × σ̃22 (z, t) + 2gs∗ â†s σ̃23 (z, t)

]
+Ωp σ̃31 (z, t) + 2gs âs σ̃32 (z, t) + ∆p σ̃33 (z, t) (3.11)
where L is the length of atoms’ distribution.

The atomic operator after the transformation due to slowly varying amplitude is
1 ∑
Nz
σ̃jk (z, t) = |j⟩i ⟨k|i ei(kjk z−ωjk t) (3.12)
Nz i
36
3.3 Heisenberg-Langevin Equation
After introducing the collective atomic operator and Hamiltonian, we will address the
equation of motion, which can be governed by Heisenberg-Langevin equation.
In the beginning, to demonstrate the time-evolution of operator, one can write down Heisen-
berg equation of motion:
dÔ 1
= [Ô, Ĥ] (3.13)
dt i}
The Hamiltonian here includes not only the system’s Hamiltonian but also the environment’s
disturbance. We can express it as:
Ĥtotal = ĤS + ĤR + ĤSR (3.14)
where ĤS , ĤR , and ĤSR is the Hamiltonian of system which includes our experimental field
and atoms, reservoir, and system-reservoir interaction, respectively.
Hence, the Heisenberg equation can be modified as
dσ̂ij 1
= [σ̂ij , Ĥ] + R̂ij + F̂ij (3.15)
dt i}
where R̂ij indicates the relaxation terms, and F̂ij is the result of Langevin noise, which will
be discussed in detail in Appendix A.
γ21 γ31
R̃11 = Γ31 σ̃33 (z, t) R̃12 = − σ̃12 (z, t) R̃13 = − σ̃13 (z, t) (3.16)
2 2
γ21 γ32
R̃21 = − σ̃21 (z, t) R̃22 = Γ32 σ̃33 (z, t) R̃23 = − σ̃23 (z, t) (3.17)
2 2
γ31 γ32 [ ]
R̃31 = − σ̃31 (z, t) R̃32 = − σ̃32 (z, t) R̃33 = − Γ31 + Γ32 σ̃33 (z, t) (3.18)
2 2
The tilde notation indicates it is under transformation because of slowly varying amplitude.
Γ terms indicate the result of spontaneous emission; γ terms describe the decoherence influ-
ence, which might occurf when magnetic field is inhomogeneous or the collisions of atoms.
The collective Langevin noise are
1 ∑
Nz
(k)
F̃ij (z, t) = F̃ij (t) (3.19)
Nz k
The Langevin noise, in general, accompanies dissipation of the system to compensate for the
uncertainty, which will be reviewed in Appendix A.
37
One can write down six Heisenberg-Langevin equations as below:
∂ σ̃11 1 i
= F̃11 + Γσ̃33 + (σ̃13 Ω∗p − σ̃31 Ωp ) (3.20)
∂t 2 2
∂ σ̃21 1 iσ̃23 Ω∗p
= F̃21 − γ21 σ̃21 − iâs gs σ̃31 + (3.21)
∂t 2 2
∂ σ̃31 Γ
= F̃31 − ( + i∆p )σ̃31 − iΩ∗p − iσ̃21 gs∗ â†s (3.22)
∂t 2
∂ σ̃22 Γσ̃33
= F̃22 + iâ†s gs∗ σ̃23 − iâs gs σ32 + (3.23)
∂t 2
∂ σ̃23 Γ iσ̃21 Ωp
= F̃23 + (− + i∆p )σ̃23 + iâs gs (σ̃22 − σ̃33 ) + (3.24)
∂t 2 2
∂ σ̃33 i
= F̃33 − iâ†s gs∗ σ̃23 + iâs gs σ̃32 − Γσ̃33 + (σ̃31 Ωp + σ̃13 Ωp∗ ) (3.25)
∂t 2
It should be noticed that I have assumed γ31 = γ32 = Γ, Γ31 = Γ32 = Γ/2 in above formulas.
Next, we apply the method of integral factor to retrieve the solution of σ̃31 , σ̃23 , and σ̃21 .
[ ]
2F̃31 − 2iâs gs σ̃21 − iΩ∗p −( Γ
+ σ̃31 (0)e−( 2 +i∆p )t
Γ
σ̃31 = 1−e 2 +i∆ p )t
(3.26)
Γ + 2i∆p
[ ]
2F̃23 + iΩp σ̃21 −( Γ −i∆p )t
+ σ̃23 (0)e−( 2 −i∆p )t
Γ
σ̃23 = 1−e 2 (3.27)
Γ − 2i∆p
The exponential terms involving ∆p are negligible because the contribution can be average
to none if the detuning is considerable. Besides, the initial condition of σ̃31 (0) and σ̃23 (0) is
zero. We neglect F̃31 and âs in σ31 comparing to Ωp .
After substituting σ̃31 and σ̃23 into equation (3.21), we can express it as
[ ]
∂ σ̃21 −|Ωp |2 γ21 Ω∗p gs âs iΩ∗p F̃23
= − σ̃21 − + + F̃21 (3.28)
∂t 2Γ − 4i∆p 2 Γ + 2i∆p Γ − 2i∆p
It is noteworthy that HLE can only provide the information of how atoms response under
the operation of fields. To acquire the spatial and time-evolution of field, one must utilize
Maxwell-Schrödinger equation. The detail derivation of MSE will be discussed in Appendix
D.
[ ]
∂ 1∂ iN gs∗
+ âs (z, t) = σ̃23 (z, t) (3.29)
∂z c ∂t c
38
3.5 Calculation of Stoke Field
To reduce the two-variable differential equation, we define τ = t − z/c and z = z ′ to

transform the coordinate. That is to say, σ̃21 is perturbed until t = z/c. In this new frame,
MSE can be modified as
∂ iN gs∗
âs (z, τ ) = σ̃23 (z, τ ) (3.30)
∂z c
by substituting equation (3.27) into equation (3.30)
[ ]
∂ iN gs∗ 2F̃23 + iΩp σ̃21
âs (z, τ ) = (3.31)
∂z c Γ − 2i∆p
′
After that, we will operate Laplace transformation on âs (z, τ ), âs (z, τ ), and Langevin noise.
∫ ∞
′ ′
L [âs (z, τ )] = âs (z, τ )e−sz dz (3.32)
0
∫ ∞
−sz ∞
=e 0
− (−s) âs (z, τ )e−sz dz (3.33)
0
= sL [âs (z, τ )] − âs (0, τ ) = se(s, τ ) − âs (0, τ ) (3.34)
Similarly, the transformation of F̃21 (z, τ ), F̃23 (z, τ ), and σ̃21 (z, τ ) is
L [âs (z, τ )] = e(s, τ ) L [σ̃21 (z, τ )] = q(s, τ ) (3.35)

L [F̃21 (z, τ )] = f21 (s, τ ) L [F̃23 (z, τ )] = f23 (s, τ ) (3.36)
By applying Laplace transformation on equation (3.31), we can rewrite it as

[ ]
1 N gs∗ Ωp 2iN gs f23
e(s, τ ) = âs (0, τ ) − q(s, τ ) + (3.37)
s c(Γ − 2i∆p ) c(Γ − 2i∆p )
Now, one can substitute equation (3.37) into equation (3.21) after coordinate transformation.
It yields
[ ] [
∂q −|Ωp |2 γ21 Ωp∗ gs N gs Ωp
= − q− âs (0, τ ) −
∂τ 2Γ − 4i∆p 2 s(Γ + 2i∆p ) c(Γ − 2i∆p )
] ∗
2if23 N gs iΩp f23
+ + + f21 (3.38)
c(Γ − 2i∆p ) Γ − 2i∆p
Employ the integrating factor method on equation 3.38, the solution of q is

∫ τ [ ]
−I H1 H3 ′
q(s, τ ) = e H2 f23 + f21 − âs − f23 eI dτ + q(s, 0)e−I (3.39)
0 s s
39
To simplify further calculation and notations, I define several new notations just for conve-
nience. The new notations are as below:
γ21 |Ωp |2 N |gs |2 |Ωp |2

M= + G= (3.40)
2 2Γ − 4i∆p c(Γ2 + 4∆2p )
Ω∗p gs iΩ∗p
H1 = H2 = (3.41)
Γ + 2i∆p Γ − 2i∆p
2iN |gs |2 Ω∗p N g s Ωp
H3 = J= (3.42)
c(Γ2 + 4∆2p ) c(Γ − 2i∆p )
where integrating factor I(τ ) = [M − Gs ]τ

Substituting the result of q(s, 0) into equation (3.37), one can obtain:
âs (0, τ ) U f23 J

e(s, τ ) = + − q(s, 0)e−I(τ )
s[ s
∫ τ s ]
J −I(τ ) [ H1 H3 ] I(τ ′ ) ′
− e H2 f23 + f21 − âs (0, τ ) − f23 e dτ (3.43)
s 0 s s
Finally, by taking the inverse Laplace transform with formulas below, we have the desired
solution of Stoke field âs (z, τ ). Here, we provide the formulas one may require by looking
up the Laplace transform table.
[ z n √ ]
L ( ) 2 In ( 4az) = s−(n+1) e s
a
(3.44)
a
[ a] z n √ 1
L −1 s−(n+1) e s = ( ) 2 In ( 4az) L [1] = (3.45)
a s
In (x) represents the modified Bessel function of first kind.
∑
∞
1 x
In (x) = i−n Jn (ix) = ( )2k+n (3.46)
k=0
k!Γ(k + n + 1) 2
In addition, to carry out Stoke field, one have to utilize convolution theorem on account of
the products terms of Laplace transform in e(s, τ ).
Convolution theorem indicates that
L [f (t) ∗ g(t)] = F (s)G(s) (3.47)
Hence, we have
∫ t
−1
L [F (s)G(s)] = f (t) ∗ g(t) = f (τ )g(t − τ )dτ (3.48)
0
where f (t) ∗ g(t) denotes the convolution of f (t) and g(t)
40
The exact solution of Stoke field is
∫ z ∫ z √
′ ′ ′ −M τ
âs (z, τ ) = âs (0, τ ) + U F̂23 (z , τ )dz − Je I0 ( 4Gτ (z − z ′ ))σ̃21 dz ′
∫ τ √ 0 √
0
z ′
+J H1 âs (0, τ ) I1 ( 4G(τ − τ ′ )z)eM (τ −τ ) dτ ′ −
G(τ − τ ) ′
∫ τ ∫0 z √
′
J I0 ( 4G(τ − τ ′ )(z − z ′ ))eM (τ −τ ) [H2 F̂23 (z ′ , τ ′ ) + F̂21 (z ′ τ ′ )]dz ′ dτ ′
0 0
∫ τ∫ z √
′ z − z′ √
+J eM (τ −τ ) I1 ( 4G(τ − τ ′ )(z − z ′ ))H3 F̂23 (z ′ , τ ′ )dz ′ dτ ′ (3.49)
0 0 G(τ − τ ′)
3.5.1 Generation Rate
To start with, the definition of generation rate here denotes the number of photons pass-
ing through a cross-section per second. Recall that there is a physical quantity, Poynting
vector, which is pretty similar to this definition. The generation rate indicates photon flux.
Poynting vector, on the other hand, represents energy flux per area. As a result, we can firstly
employ surface integral to connect these two physical quantities.
⃗= 1E
S ⃗ ×B
⃗ (3.50)
µ0
The energy flux can be expressed as

∫ ∫
1
S · dA =
⃗ ⃗ ⃗ × B)
(E ⃗ · dA ⃗ (3.51)
µ0
∫
⃗ × k̂ × E ) · dA
1 ⃗
= (E ⃗ (3.52)
µ0 c
⃗ × k̂ × E
As we know, E ⃗ remains k̂ direction
[ ] ∑
⃗ × (⃗k × E)
E ⃗ = ϵijk Ej (⃗k × E)
⃗ k (3.53)
i
jk
∑∑
= ϵijk ϵkmn Ej km En (3.54)
jk mn
∑∑
= −ϵkji ϵkmn Ej km En (3.55)
jk mn
∑∑[ ]
= δjn δim − δjm δin Ej km En (3.56)
j mn
41
[ ] ∑[ ]
⃗ × (⃗k × E)
E ⃗ = Ej ki Ej − Ej kj Ei (3.57)
i
j
⃗ · E)k
= (E ⃗ i − Ei (E
⃗ · k̂) = |E|
⃗ 2 k̂ (3.58)
where ϵijk represents Levi-Civita symbol, and δij is Kronecker delta function. Finally, we
can connect electric field and energy flux as below
∫ ∫
1
S · dA =
⃗ ⃗ |E|
⃗ 2 dA (3.59)
cµ0
Remind that we have quantized the electric field

√ √
⃗s = }ωs }ωs † −i(ks z−ωs t)
E âs ei(ks z−ωs t) ês + â e ês (3.60)
2ϵ0 V 2ϵ0 V s
We ignore highly oscillating terms, which will be averaged to none in time scale, and recall
that }ω is the energy of a photon. Thus, the generation rate will be
1 1 [ }ω ]( † )
Rs = â â + ââ† A (3.61)
}ω cµ0 2ϵV
c( † 1)
= â â + (3.62)
L 2
Because the the generation rate of Stoke field is much more than 1, it can be further simplified
as
c †
Rs = ⟨â (z, t)âs (z, t)⟩ (3.63)
L s
To calculate the generation rate, it is necessary to introduce fluctuation-dissipation theorem
[34] to carry out diffusion coefficients. Here, we consider Mokovian process, that is, white
noise will not influence another noise at a different location or different time. As a result, we
can write down the delta-correlated function as below.
1
⟨F̂â† (z, t)F̂â (z ′ , t′ )⟩ = 2Dâ† ,â δ(t − t′ )δ(z − z ′ ) (3.64)
ρ
† 1
⟨F̂21 (z, t)F̂21 (z ′ , t′ )⟩ = γ21 δ(t − t′ )δ(z − z ′ )σ̃11 (3.65)
ρ
†
⟨F̂23 (z, t)F̂23 (z ′ , t′ )⟩ = 0 (3.66)
† 1
⟨F̂21 (z, t)F̂21 (z ′ , t′ )⟩ = γ21 δ(t − t′ )δ(z − z ′ )σ̃22 (3.67)
ρ
† 1 γ21
⟨F̂23 (z, t)F̂21 (z ′ , t′ )⟩ = δ(t − t′ )δ(z − z ′ )σ̃31 (3.68)
ρ 2
† 1 γ21
⟨F̂21 (z, t)F̂23 (z ′ , t′ )⟩ = δ(t − t′ )δ(z − z ′ )σ̃13 (3.69)
ρ 2
42
1 ∑ 1 ∑
Nz Nz
†
⟨σ̃21 (z, 0)σ̃21 (z, 0)⟩ =⟨ |1⟩i ⟨2|i |2⟩j ⟨1|j ⟩ (3.70)
Nz i Nz j
∑
Nz
( 1 )2 1 1
= ⟨ |1⟩i ⟨1|i ⟩ = = (3.71)
i
Nz Nz ρ
where ρ denotes the linear density of atoms. In addition, the expectation value of Langevin
noise is zero on account of the stochastic fluctuation.
⟨F̂ij† (z, t)⟩ = ⟨F̂ij (z, t)⟩ = 0 (3.72)
The detailed derivation of the diffusion-coefficient will be reviewed in Appendix B.
It is noteworthy that the generation of Stoke field is zero at z = 0. Thus, we retain contributive
terms. Also, on account of that atoms at z remain unperturbed when τ = 0, the expectation
value ⟨σ̃21 (z, 0)⟩ = 0. The generation rate of Stoke photon can be reduced as follows
[ ∫ z
c −Aτ
√
Rs = |J| e2
I02 ( 4Gτ (z − z ′ ))dz ′ +
Lρ
∫ τ∫ z √
0
′ [ γ21 ∗ ]
eA(τ −τ ) I02 ( 4G(τ − τ ′ )(z − z ′ )) (γ21 + (H3 σ̃31 + H3 σ̃13 ) dz ′ dτ ′ −
2
0
∫ τ
0
√ ]
A(τ ′ −τ ) z − z′ √ √ (H3∗ σ̃31 + H3 σ̃13 )γ21
e I0 ( 4G(τ − τ )(z − z ))I1 ( 4G(τ − τ )(z − z ))
′ ′ ′ ′
0 G(τ − τ ′ ) 2
(3.73)
where A = M + M ∗ . To acquire an analytic solution, we will discard the third term owing
to the tiny contribution which can be realized by performing Riemann integral. Furthermore,
there are few integration results of Bessel function can assist us in simplifying mathematical
structure.
∫ z √ [ √ √ ]
I02 ( a(z − z ′ ))dz ′ = z I02 ( az) − I02 ( az) (3.74)
0
∫ ∞
[ √ √ ] ea/2m [ a a ]
e−mt I02 ( at) − I12 ( at) dt = I0 ( ) − I1 ( ) (3.75)
0 m 2m 2m
we define that H = H3∗ σ̃31 + H3 σ̃12 . The generation rate can be carried out as
gr γ21 −Aτ [ 2 √ √ ]
Rs = e z I0 ( 4Gτ z) − I12 4Gτ z
4
2 ∫ τ [ √ √ ]
gr γ21 ′ H
+ eA(τ −τ ) z I02 ( 4G(τ − τ ′ )z) − I12 ( 4G(τ − τ ′ )z) [1 + ]dτ ′ (3.76)
4 0 2
43
by changing the variable, τ − τ ′ = τ ′′ , we have
gr γ21 −Aτ [ 2 √ √
Rs = e z I0 ( 4Gτ z) − I12 4Gτ zbig]
4
2 ∫ 0 √ √
gr γ21 ′′ [ ] H
− e−Aτ z I02 ( 4Gτ ′′ z) − I12 ( 4Gτ ′′ z) [1 + ]dτ ′′ (3.77)
4 τ 2
gr γ21 −γ21 τ [ 2 √ √ ]
= e z I0 ( gr γ21 τ z) − I12 ( gr γ21 τ z)
4
2 ∫ τ √ √
gr γ21 ′ [
+ e−γ21 τ z I02 ( gr γ21 τ ′ z) − I12 ( gr γ21 τ ′ z)dτ ′ (3.78)
4 0
where gr is defined as Raman gain, and we suppose that ∆p is really far-detune, which will
be addressed as below. we suppose that ∆p is really far-detune and H can be ignored. Asso-
ciating with Weisskopf-Wigner theorem[35], we have
N |gs |2 ODΓ32 ODΓ32 |Ωp |2

= gr = (3.79)
c 4L 4∆2p γ21 L
3.5.2 Raman Gain
In fact, the generation rate we derived above includes both spontaneous Raman scatter-
ing and stimulated Raman scattering (SRS) processes. Raman gain coefficient, in general,
can be regarded as a quality that can describe how strong the stimulated Raman scattering
is and how Stoke photons are amplified as they propagate in the medium. Its mathematical
structure is expressed as
∂
Is (z) = gr Is (z) (3.80)
∂z
where Is denotes the intensity of stoke photon; we can replace it with the generation rate de-
rived above. Raman gain is just similar to the chain reaction. While Stoke photons propagate
through the medium, Stoke photons will be induced by other Stoke photons produced ahead
of themselves.
Seeing that the solution of equation (3.80) is
Is (z) = Is (0)egr z (3.81)
one can firstly consider the steady-state solution in high-gain limit

gr γ21 z gr z [ gr z gr z ]
Rs = e 2 I0 ( ) − I0 ( ) → gz ≫ 1 (3.82)
4 2 2
44
In addition, there is a property of modified Bessel function of first kind we can utilize
[ ]
ex 4n2 − 1
In (x) = √ 1− (3.83)
2πx 8x
Hence, the generation rate for steady state is

gr z [[ ] [ ]]
γ21 gr z gr z e 2 1 3
Rs (z) = e 2 √ 1+ − 1− (3.84)
4 πgr z 4gr z 4gr z
gr z
γ21 e
= √ (3.85)
4 πgr z
Hence, by comparing the coefficient of equation (3.81) and coefficient (3.85), we can write
down the Raman gain. To simplify the mathematical structure, the derivation above is started
from the real form of Raman gain, but one can still derive it from equation (3.76).
(Remind that the Raman gain we derive is for steady state.)
Besides, Raman gain is proportional to the linear density of atoms, and the generation rate
will increase as z increases. Thus, we prefer utilizing gr z to characterize the ability of am-
plification.
45
3.5.3 Spontaneous Raman Scattering Process
Spontaneous Raman scattering photons, generally, are the outcome of the interaction
between pumping field and vacuum field or fluctuation noise. Hence, to extract the result of
spontaneous Raman scattering photons, let us consider the low gain case (if Raman gain is
weeny, the stimulated photons can be negligible). Following the equation (3.78). While
γ21 τ
gr γ21 τ z ≪ (3.86)
2
For all τ , the exponential terms in integration of equation (3.78) is extremely sharp comparing
to Bessel function, allowing Bessel functions to be extracted out of integration, and equation
(3.78) can be revised as
gr γ21 z −γ21 τ [ 2 √ √ ] gr γ212

z [e−γ21 τ − 1]
Rs ≃ e I0 [ gr γ21 τ z] − I12 [ gr γ21 τ z] + (3.87)
4 4 −γ21
[ ]
gr γ21 z −γ21 τ [ 2 √ √ ] gr γ21 z
≃ e I0 [ gr γ21 τ z] − I12 [ gr γ21 τ z] − 1 + 1 ≃ (3.88)
4 4
the derivation above was simplified by

x
I0 [x] ≃ 1 I1 [x] ≃ as x ≪ 1 (3.89)
2
3.5.4 Transient Raman Scattering
While time is short enough comparing to the decoherence time, the second term can be
ignored. by utilizing the property of modified Bessel function,
[ ]
ex 4n2 − 1
In [x] = √ 1− + .... as x≫1 (3.90)
2πx 8x
The generation rate is approximately

[ √ √ ]
gr γ21 z e2 gr γ21 zτ ( 1 )2 e2 gr γ21 zτ ( 3 )2
Rs (τ ) = √ 1+ √ − √ 1− √
4 2π gr γ21 zτ 8 gr γ21 zτ 2π gr γ21 zτ 8 gr γ21 zτ
(3.91)
√
e2 gr γ21 zτ
= (3.92)
8πτ
One should remind that this approximate outcome is based on high-gain limit and short time.
In addition, for high-gain transient Raman scattering, the number of stimulated photons will
exceed the number of spontaneous photons as time evolutes.
46
3.5.5 Steady-State Solution
When time tends to infinity, the first term of generation rate can be abandoned. By utilizing
the integration formula above, the steady solution of Stoke is given by
gr γ21 z gr z [ gr z gr z ]
Rs = e 2 I0 ( ) − I0 ( ) (3.93)
4 2 2
3.6 The Diagram of Generation Rate
Even though we have to assume local time tends to infinity to obtain exact solutions, we can
estimate how long it will take to reach steady-state condition by equation (3.78).
Figure 3.2: Theoretical diagram of generation rate respective to local time, the parameter
setting: Γ32 = Γ/2, Ωp = 1Γ, γ21 = 0.001Γ, ∆p = 500Γ, OD= 50 and OD= 100 for figure
(a) and (b) respectively, and Γ = 6 × 2π × 106 Hz
One should notice that the time denotes local time τ , and the local time that we finally con-
sider is located at the end of atomic medium. Since the atomic medium is especially tiny, it is
viable to deem the local time as real time, namely, every atom experiences the same duration
of the illumination from pumping field.
According to Figure (3.2), one may realize that the generation rate achieves steady-state con-
dition immediately. Moreover, the generation rate in Figure (3.2) (b) is about twice as the
generation rate in Figure (3.2) (a). The reason is that the optical depth in Figure (3.2) (b) is
twice larger than the optical depth in Figure (3.2) (a) This indicates that the generated Stoke
photons are entirely based on the spontaneous Raman scattering process. By equation (3.79),
the Raman gain times the length of atomic ensemble (gr L) are 0.025 and 0.05 for Figure (a)
and (b) respectively.
47
As we have known, the system attains steady-state condition immediately in low gain cases
because γ21 dispels ground-state coherence quickly. Next, let us discuss which parameters
determine the time system needs to attain steady-state condition in high gain cases. In gen-
eral, induced photons are accumulated through space. However, if Stoke photons are abun-
dant enough, these Stoke photons would possess a chance to rebuild those disappearing lo-
cal ground-state coherence. For this reason, local ground-state coherence is accumulated
through time. Hence, the quantity of Stoke photons increases as time evolutes. Because the
destruction of γ21 is proportional to the amount of ground-state coherence, the enhancement
of ground-state coherence and the damage of ground-state coherence get a balance eventu-
ally. Therefore, system achieves a steady-state.
setting: Γ32 = Γ/2, , γ21 = 0.0002Γ, ∆p = 350Γ, OD= 1000, and Ωp = 0.7Γ and Ωp = 0.9Γ
for figure (a) and (b) respectively, and Γ = 6 × 2π × 106 Hz
where gr z = 2.5 and gr z = 4.1 for figure (a) and (b), respectively. As shown, it only takes
system 1 ms to reach steady-state condition for Figure 3.3 (a), and it takes system roughly
2 ms to attain steady-state condition for Figure 3.3 (b). The reason is that the gain-length
is about larger in Figure 3.3 (b) than Figure 3.3 (a). There is more preserved ground state
coherence for the latter generation.
48
setting: Γ32 = Γ/2, , γ21 = 0.0002Γ, Ωp = 1Γ, OD= 1000, and ∆p = 350Γ and ∆p = 500Γ
where gr z = 5.1 and gr z = 2.5 for figure (a) and (b), respectively.
setting: Γ32 = Γ/2, , γ21 = 0.0005Γ, Ωp = 1Γ, ∆p = 300Γ, and OD= 500 and OD= 1000
where gr z = 1.4 and gr z = 2.8 for figure (a) and (b), respectively. As mentioned above,
the time that system needs to attain steady-state condition arises while gain-length increases.
By increasing optical depth, the quantity or the density of atoms increases. That is, Stoke
photons are more likely to induce other Stoke photons while passing through other atoms. In
addition, if detuning declines, the coupling effect of atoms and field enhances. As a result,
the productivity of Stoke photons escalates. Both of these two adjustments will slow down
the disappearance of local coherence. Thus, the number of Stoke photons gets amplified with
time.
49
As shown above, the difference in optical depth barely alters the time that system needs to
reach steady-state condition. However, the steady condition is susceptible to the dephasing
rate γ21 of two ground states.
setting: Γ32 = Γ/2, Ωp = 1Γ, ∆p = 200Γ, OD= 800 and γ21 = 0.001Γ for figure (b);
OD= 400 and γ21 = 0.0005Γ for figure (a) Γ = 6 × 2π × 106 Hz
To compare the influence of γ21 with identical gain, we adjust optical depth to maintain this
condition. The gain of figure (a) and (b) are both 2.5, and the stimulated Raman process
will appear. As figures demonstrate, it takes roughly 200µs and 400µs to attain a steady
state for figure (a) and figure (b), respectively. Hence, we can conclude that the decrease
of dephasing rate will lead to a longer time to reach steady. The reason is that while the
dephasing rate decreases, the local coherence between two ground states, which was built up
by stimulated Raman scattering, would not be destroyed rapidly. Therefore, it will provide
stimulated Raman scattering process more room and more time to accumulate the coherence
of two ground states. On the other hand, if dephasing rate becomes more extensive, it will
destroy the ground-states coherence immediately. Thus, dephasing rate and Stoke photons
generation will attain a balanced condition quickly.
Besides, the ratio of time to achieve steady state is similar to the ratio of the inverse of de-
phasing rate. The reason for the identical ratio is that the exponential term in equation (3.78)
dominates the behaviour of generation rate for low gain. Alternatively, we can realize it by
considering the Stoke photons generation based on pumping and induced dipole interaction
which is limited by ground states decoherence. Yet, this result may not be suitable for a large
gain.
50
2 2 2 2
2 0 2 0
1 8 1 8
1 6 1 6
1 4 1 4
( z , τ) / γ2 1 ]
1 2 1 2
s
1 0 1 0
[2 R 1 0
8 8
lo g
6 6
4 4
2 2
0 0
-2 -2
-3 -2 -1 0 1 2 3
lo g 1 0
[ τγ2 1 / 2 ]
Figure 3.7: Theoretical diagram of Generation rate with respect to time for different gain
length. The parameter setting(from top to bottom): Red line: gr z = 50, Orange line: gr z =
40, Yellow line: gr z = 30, Green line: gr z = 20, Blue line: gr z = 10, Indigo line: gr z = 5,
Purple line: gr z = 3, Black line: gr z = 1, Gray line: gr z = 0.1, Pink line: gr z = 0.01
Here, we provide several theoretical simulations for the relation between gain-length and
generation rate. As shown, while gain-lengths are low (gr z < 0.1), Stoke photons are en-
tirely produced by spontaneous Raman scattering, thereby reaching steady state instantly.
When small times (log10 ( γ212 τ ) < 2), the accumulation of local coherence have not been con-
structed. The generations of Stoke photons are mainly based on spontaneous Raman process
as well. However, it will be enhanced by the accumulation of local coherence as time goes by.
When gain lengths become larger, the stimulating effect will enhance generation rate rapidly.
51
Next, we will demonstrate the relation between generation rate and pumping power, optical
depth, detuning, and the decoherence rate of two ground states for steady-state conditions.
1 0 0 0
G e n e r a tio n r a te ( 1 /s )
8 0 0
6 0 0
4 0 0
2 0 0
0
0 .4 0 .6 0 .8 1 .0 1 .2 1 .4 1 .6 1 .8 2 .0
Ωp ( Γ)
Figure 3.8: Theoretical diagram of Generation rate with respect to pumping power. The
parameter setting: OD=50, ∆p = 500Γ, Γ32 = Γ/2, γ21 = 0.001Γ, The gain lengths is
0.0023 when pumping power is 0.3Γ, and gain length is 0.1 when pumping power is 2Γ
With the increase of pumping power, it is more likely that more populations would be pumped
to a virtual state, and then vacuum field and noise would be more likely to bring the population
into the other ground state by the two-photon process. Hence, more Stoke photons will be
created.
2 5 0
2 0 0
1 5 0
1 0 0
5 0
0
0 5 1 0 1 5 2 0 2 5 3 0 3 5 4 0 4 5 5 0
O D
Figure 3.9: Theoretical diagram of Generation rate with respect to optical depth. The pa-
rameter setting: Ωp = 1Γ, ∆p = 500Γ, γ21 = 0.001Γ, Γ32 = Γ/2
As shown on the last page, the generation rate is practically proportional to optical depth.
52
The reason is that the gain lengths in the previous figure are about 0.025 at OD = 50, which
is too low to bring about chain effects to generate other photons. Accordingly, spontaneous
Raman scattering will dominate the result of generation rate. Next, let’s consider larger OD
5 0 0 0 0
4 0 0 0 0
3 0 0 0 0
2 0 0 0 0
1 0 0 0 0
0
0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0
O D
rameter setting: Ωp = 1Γ, ∆p = 500Γ, γ21 = 0.0001Γ, Γ32 = Γ/2
2 5 0 0
2 0 0 0
1 5 0 0
1 0 0 0
5 0 0
0
2 0 0 2 5 0 3 0 0 3 5 0 4 0 0 4 5 0 5 0 0
∆p ( Γ)
Figure 3.11: Theoretical diagram of Generation rate with respect to one photon detuning.
The parameter setting: Ωp = 1Γ, OD=50, γ21 = 0.0001Γ, Γ32 = Γ/2
As detuning increases, the vacuum field would be more arduous to interfere with the system.
Therefore, populations are barely transferred into the other ground state, and the generation
rate will decline. Nonetheless, one should recall that our theoretical model is based on ground
state approximation. Thus, the detuning possesses its own lower limit.
53
(
a) (
b)
Figure 3.12: Theoretical diagram of generation rate respect to pumping power for different
dephasing rate of two ground states, the parameter setting: Γ32 = Γ/2, ∆p = 300Γ, OD= 50
and γ21 = 0.01Γ and γ21 = 0.0001Γ for figure (a) and figure (b) respectively. Γ = 6 × 2π ×
106 Hz
The gain lengths are 0.0278 and 2.778 while pumping power is 2Γ for figure (a) and (b)
respectively. It should be noted that Raman gain is related to decoherence rate, if dephasing
rate rises, the build-up coherence between two ground states will be destroyed gradually, and
stimulated photons are demanding to be produced. In consequence, the upper limit of two
ground-state coherence and the intensity of Stoke photons will decline.
(
a) (
b)
Figure 3.13: Theoretical diagram of generation rate respect to pumping power for different
dephasing rate of two ground states, the parameter setting: Γ32 = Γ/2, ∆p = 1000Γ, OD= 50
and γ21 = 0.01Γ and γ21 = 0.0001Γ for figure (a) and figure (b) respectively.
The gain lengths are 0.0025 and 0.25 while pumping power is 2Γ for figure (a) and (b) re-
spectively. That is to say, most photons are brought about by spontaneous Raman scattering.
Therefore, the generation rate is scarcely influenced by the dephasing rate.
54
3.7 Spectrum of Stoke Photons
Up to present, we have demonstrated how to determine the annihilation operator of Stoke

field. To characterize the spectrum, we can exploit Fourier transform on annihilation operator
and evaluate the ensemble average of ⟨â†s (−ω)âs (ω)⟩. However, the Fourier transform of our
solution involves complicate mathematical structure. For this reason, we will briefly review
the temporal first order autocorrelation function G(1) (τ ), and introduce Wiener-Khinchin the-
orem to connect the relation between spectrum and G(1) (τ ). In theoretical calculation, one
deliver G(1) (τ ) as the ensemble average of annihilation and creation operator with a delay
time as follows
G(1) (τ ) = ⟨a∗ (t1 )a(t1 + τ )⟩ (3.94)
To carry out the spectrum of Stoke photon, one might attempt to harness Fourier transform on
the pulse. Nevertheless, Stoke photons are produced randomly, so the pulses are immersed
in stochastic signals and noise. In this situation, one can utilize autocorrelation function to
rebuild the information of pulse and retrieve the spectrum. (note that the information denotes
the autocorrelation of signal rather than original pulse) To start with, let’s start from Parseval
theorem to define power spectral density. Parseval theorem [36] denotes that one can write
down the total energy in time or frequency domain as below
∫ ∞ ∫ ∞ ∫ ∞ ∫ ∞
′
|a(t)| dt =
2
A∗ (ω)A(ω ′ )eiωt e−iω t dωdω ′ dt (3.95)
−∞
∫−∞
∞ ∫−∞
∞
−∞
= 2πδ(ω − ω ′ )|A(ω)|2 ωdω ′ (3.96)

∫−∞
∞
−∞
= 2π|A(ω)|2 dω (3.97)
−∞
where a(t) is the signal, and A(f ) is the Fourier transform of signal. Note that the 2π factor is
kind of different to familiar Parseval theorem definition because the Fourier time-frequency
transformation I define is following Chapter 4, equation (4.17). One can define the power
spectral density by considering random process as
2π
S(ω) = lim ⟨|A(ω)|2 ⟩ (3.98)
T −→∞ T
which indicates the power per frequency bandwidth.
55
According to the Einstein-Wiener-Khinchin theorem, [37] we can relate it to autocorrelation
function as follows
∫ T ∫ T
2π 2 2 ′
S(ω) = lim ⟨a∗ (t)a(t′ )⟩e−iω(t−t ) dtdt′ (3.99)
T −→∞ 4π 2 T − T2 − T2
∫ T ∫ T
1 2 2 ′
G(1) (t′ − t)eiω(t −t) dt′ dt
W SS
=⇒ lim (3.100)
T −→∞ 2πT − T2 − T2
∫ 0 ∫ T
+
τ1
1 2 2
= lim dτ2 G(1) (τ1 )eiωτ1 dτ1 (3.101)
T −→∞ 2πT −T − T2
τ
− 21
∫ T ∫ T τ
− 21
1 2
+ lim dτ1 dτ2 G(1) (τ1 )eiωτ1 dτ1 (3.102)
T −→∞ 2πT 0 − T2
τ
+ 21
∫ T
1
= lim dτ1 G(1) (τ1 )[T − τ1 ]eiωτ1 dτ1 (3.103)
T −→∞ 2πT −T
∫ ∞
1
⇒S(ω) = dτ1 G(1) (τ1 )eiωτ1 dτ1 (3.104)
2π −∞
∫
1 ∞
= dτ1 G(1) (τ1 )eiωτ1 dτ1 (3.105)
π 0
′
where τ1 = t′ − t and τ2 = t 2+t ; WSS indicates what we consider is a wide sense stationary
process, which we will discuss on the next page.
56
3.7.1 Stochastic Process
Here we just provide several levels of stochastic process to support previous discussion. For
a strict stationary process, the probability density functions for any order do not dependent
on the shift in time. [38, 36, 39, 40]
P (x1 , x2 , ...xk ; t1 , t2 , ...tk ) = P (x1 , x2 , ...xk ; t1 + T, t2 + T, ...tk + T ) (3.106)
Yet, this condition is kind of difficult to achieve. Thus, what we consider is the wide-sense
stationary process; it only requires that the mean value is time-invariant, and autocorrelation
function only depends on τ . Hence, we can revise the two-variable function G(1) (t′ , t′′ ) with
G(1) (τ ), and conclude that the Fourier transform of autocorrelation function is the power
spectral density for wide-sense stationary stochastic process. Furthermore, the definition of
autocorrelation function is the expectation value of a∗ (t)a(t + τ ), which can be evaluated by
ensemble average in theoretical calculation. Nevertheless, it is unrealistic to prepare numer-
ous samples to obtain G(1) (τ ).
As a result, we will introduce a strict condition, ergodic process. [38, 36, 40] Ergodicity
exists if the result of the ensemble average is equal to the time average of a sample. That is
to say, if the process is ergodic, we can retrieve the result by performing a time average on a
single sample.
∫ ∞
∗
G (τ ) = ⟨a (t)a(t + τ )⟩ =
(1)
a∗ (t)a(t + τ )dt (3.107)
−∞
one can distinguish the non-ergodic process and ergodic process as follows
Ensemble A(non-ergodic) Ensemble B (ergodic)

Time-based statistics of a single sample Time-based statistics of a single sample
Amplitude
Amplitude
time time
Amplitude
Amplitude
time time
Amplitude
Amplitude
time time
ensemble-based statistics of samples ensemble-based statistics of samples
Figure 3.14: Theoretical diagram of non-ergodic process and ergodic process
57
In ensemble A, even though these three samples may exist stationarity, the individual be-
haviour is quite different. Considering the simple case, we evaluate the mean value of signal
from ensemble average at t might not be equivalent to the time average of a specific sample.
However, for ensemble B, the ensemble average is the same as the time average of any cho-
sen sample. Thus, we call this ensemble to exist ergodicity.
In summary, we can realize this strategy as follows: If we want to acquire the spectrum, we
can exploit Fourier transform on signal and square it
[ ] [ ]
F.T . a(t) = A(f ) F.T . a∗ (t) = A∗ (f ) (3.108)
[ ] [ ] { [ ]}
|A(f )|2 = F.T . a(t) F.T . a∗ (t) = F.T . F.T .−1 A∗ (f )A(f ) (3.109)
[ ]
= F.T . a∗ (t) ∗ a(t) (3.110)
alternatively, we can reverse the procedure. That is, the Fourier transform of the convolution
of a(t) will be the spectrum as well.
Heretofore, that’s all about theoretical computation. In experiment, to measure the first order
correlation function, one can inject a light beam toward a beam splitter and make sure these
two splitting light beams arriving the same photon detector with a delay time τ difference.
3.7.2 Calculation of Spectrum
To compare the spectrum we will discuss in next section, we define a τ related function as
below
c †
H(s) = ⟨â (t)âs (t + s)⟩ (3.111)
L s
It should be noted that the H(s) is not an exactly autocorrelation function, but the behaviour
is identical.
∫ ∫ ∫ ∫
c 2 s+τ L τ L γ21 (τ ′′ −τ ) γ21 (τ ′ −(s+τ ) γ21 ′
H(s) = lim |J| e 2 e 2 δ(z − z ′′ )δ(τ ′ − τ ′′ )
τ −→∞ L ρ
√ 0 0 0 0
√
I0 ( 4G(τ − τ )(L − z ))I0 ( 4G(τ + s − τ ′ )(L − z ′ ))dz ′′ dτ ′′ dz ′ dτ ′ (3.112)
′′ ′′
∫ ∫
c 2 γ21 τ L γ21 (s) γ21 (τ ′ −τ )
≃ lim |J| e 2 e 2
τ −→∞ L ρ 0 0
√ √
I0 ( 4G(τ + s − τ ′ )x)I0 ( 4G(τ − τ ′ )x)dτ ′ dx x = L − z′ t = τ − τ ′
2 ∫ ∞∫ L √ √
gr γ21 γ21
= e−γ21 t e 2 s I0 ( gr γ21 x(t + s))I0 ( gr γ21 xt)dtdx (3.113)
4
2 ∫ ∞
0 0
gr γ21 γ21 2 [√ √
= e−γ21 t e− 2 s gr γ21 L(t + s)I1 ( gr γ21 L(t + s))
4 gr γ21 s
√0 √ √ √ ]
× I0 ( gr γ21 Lt) − gr γ21 LtI1 ( gr γ21 tL)I0 ( gr γ21 L(t + s)) dt (3.114)
58
∫
γ21 ∞ [√ γ21 √
H(s) = e−γ21 t e− s
gr γ21 L(t + s)I1 ( gr γ21 L(t + s))
2
2s 0
√ √ √ √ ]
× I0 ( gr γ21 Lt) − gr γ21 LtI1 ( gr γ21 tL)I0 ( gr γ21 L(t + s)) dt (3.115)
†
In above calculation, owing to what we consider is steady state, we just leave the ⟨F̂21 (z, t)F̂21 (z, t)⟩
γ21
contribution term, and remind that M is about 2 for far detuning assumption, and 4G =
gr γ21 . Finally, we can employ the Fourier transform on H(s) to obtain the spectrum.
∫ ∫
γ21 [√ √
∞ ∞
2 iωs γ21
R(ω) = e e−γ21 t e− 2 s gr γ21 L(t + s)I1 ( gr γ21 L(t + s)) (3.116)
2π 2s 0
0
√ √ √ √ ]
× I0 ( gr γ21 Lt) − gr γ21 LtI1 ( gr γ21 tL)I0 ( gr γ21 L(t + s)) dtds (3.117)
For low gain regime, the spectrum can be reduced as

∫ ∫
γ21 ∞ iωs ∞ −γ21 t − γ21 s gr γ21 Ls 1
R(ω) = e e e 2 dt ds (3.118)
2π 0 2 s
0
∫ ∞
γ21 −1 gr γ21 L γ
eiωs e− 2 s ds
21
= (3.119)
2π −γ21 2 0
γ21
gr γ21 L
= × 2
γ2
(3.120)
4π 2
ω + 21 4
here, I exploit I0 (x) = 1, I1 (x) = x/2 for x ≪ 1. The spectrum has a Lorentzian line shape,
and the half width is γ21 /2.
For high gain, the spectrum can be approximated as
2
1 gr L −gr L 4ω
R(ω) ∼
2
γ21
e e (3.121)
2π
For more detail of this approximation, please read reference [31]
59
3.7.3 Low Gain Bandwidth
2 5 0 0 0
( 1 / s ) ( 1 / Γ)
2 0 0 0 0
1 5 0 0 0
1 ×10−3Γ
S p e c tru m
1 0 0 0 0
5 0 0 0
-0 .0 1 0 -0 .0 0 5 0 .0 0 0 0 .0 0 5 0 .0 1 0
F r e q u e n c y ( Γ)
Figure 3.15: Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30,
γ21 = 0.001Γ, Γ32 = Γ/2, ∆p = 500Γ and gain lengths=0.0038
2 5 0 0 0
( 1 / s ) ( 1 / Γ)
2 0 0 0 0
1 5 0 0 0
S p e c tru m
1 0 0 0 0
5 0 0 0 2 ×10−3Γ
-0 .0 1 0 -0 .0 0 5 0 .0 0 0 0 .0 0 5 0 .0 1 0
Figure 3.16: Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30,
As depicted, the spectrum bandwidth is nearly equivalent to the decoherence rate of two
ground states in a low gain regime. One can comprehend this outcome as follows: From the
perspective of spectrum, as decoherence rate grows, more information of atoms will dissi-
pate to environment, which will lead to a vigorous noise response and pollution in our system.
60
Therefore, the Stoke photons, which are created by pumping field and ground states coher-
ence interaction, will possess more modes and a broader spectrum. Moreover, we can realize
it from the angle of pulse construction. (This is just a personal interpretation, and I start
from classical picture) While vacuum field harasses atomic system, it would not only create
Stoke photons but also produce coherence between ground states (The coherence just exists
in the local region where photons are generated, and this kind of coherence is stochastic over-
all. Namely, the ensemble average of ground-state coherence ⟨σ̃21 ⟩all = 0) In addition, one
should recall that photon propagates through space in wave packet form. The square of am-
plitude denotes the probability of photon appearance. While pumping field passes through
certain atoms which have possessed ground state coherence, another Stoke photon will be
created. However, owing to the damage of γ21 , the local coherence will be obliterated within
1/γ21 time which denotes the duration of Stoke pulse.
Pumping field
γ21
γ21
F.T.(P(t))
P(t) of Stoke
frequency
Time
Figure 3.17: Schematic diagram of the Stoke photon pulse construction. The tiny red spot
doesn’t represent a single atom. Instead, it denotes the individual ensemble of atoms, and
the oscillating purple line is the Stoke field produced by pumping and σ21 interaction. The
purple bar chart represents the probability of Stoke creation with respect to time, and the blue
bar chart is the Fourier transform.
In summary, by virtue of vacuum field disturbance, it is possible that local coherence might
61
exist at a certain moment though they will cancel out on the whole ensemble. After that,
γ21 , which destroys ground-states coherence“exponentially”, will eliminate the probability
of producing Stoke photon rapidly. The evolution of the Stoke photon amplitude could be
interpreted as wave packet of a single Stoke photon in time domain. In consequence, the
bandwidth of Stoke spectrum, which can be retrieved by employing Fourier transform on
Stoke photon pulse, is γ21 .
62
3.7.4 High Gain Bandwidth
2 0
2 .4 x 1 0
2 0
2 .0 x 1 0
( 1 / s ) ( 1 / Γ)
2 0
1 .6 x 1 0
2 0 1 . 4 8 ×1 0 -5
Γ
1 .2 x 1 0
S p e c tru m
1 9
8 .0 x 1 0
1 9
4 .0 x 1 0
0 .0
-4 -4 -4 -4 -4 -4
-3 .0 x 1 0 -2 .0 x 1 0 -1 .0 x 1 0 0 .0 1 .0 x 1 0 2 .0 x 1 0 3 .0 x 1 0
Figure 3.18: Theoretical diagram of spectrum. The parameter setting: Ωp = 1Γ, OD=1000,
As the diagram demonstrates, the strength of Stoke photon is built up apparently in central
frequency, which resembles single-mode photon comparing to Figure(3.15). We can appre-
hend this result as follows: In a high gain regime, Stoke photons can be generated by a stim-
ulated process. As stimulated Stoke photon goes through another coherence-disappearing
ensemble, it would induce not only another Stoke photon but also rebuild the coherence of
the ensemble. Accordingly, the Stoke photon, which is produced by pumping field and σ̃21
interaction, possesses a longer duration of the pulse, bringing about the compression of the
spectrum.
If gain increases, more stimulated photons will assist in repairing ground states coherence,
and the pulse will be extended. Thus, bandwidth is inversely proportionally to gain in equa-
tion (3.121).
Nevertheless, one should be aware that this result is just an approximation. In reality, while
optical depth is enormous enough, pumping field might be absorbed. We cannot suppose
that pumping field is a constant in previous calculations thereby. Besides, to achieve high
gain, we reduce one photon detuning to 200Γ, which may lead to difficulty in maintaining
population to stay in the ground state. Hence, this approximation is just suitable for certain
cases.
63
Stoke field produced earlier
Pumping field
γ21
γ21
F.T.(P(t))
P(t) of Stoke
frequency
Time
Figure 3.19: Schematic diagram of the Stoke photon pulse construction in high gain regime.
On account of the protection by stimulated Stoke photon, ground states coherence acquires a
longer time on generating new Stoke photon, leading to narrowband spectrum.
64
3.8 Second-Order Correlation Function
Notice that the first order autocorrelation is the correlation between field; the second order
correlation function, on the other hand, can describe the intensity-intensity correlation. The
definition of Glauber second order correlation function G(2) (τ ) is expressed as
G(2) (τ ) = ⟨â†s (L, t)â†s (L, t + τ )âs (L, t + τ )âs (L, t)⟩ (3.122)
Yet, the expectation value of the G(2) is quite complicated to resolve directly. Hence, one
could exploit Wick’s theorem to decompose it into several second-order correlation compu-
tations. Wick’s theorem denotes that
⟨â†s (L, t)â†s (L, t + τ )âs (L, t + τ )âs (L, t)⟩ (3.123)
= ⟨â†s (L, t)â†s (L, t + τ )⟩⟨âs (L, t + τ )âs (L, t)⟩
+ ⟨â†s (L, t)âs (L, t + τ )⟩⟨â†s (L, t + τ )âs (L, t)⟩
+ ⟨â†s (L, t)âs (L, t)⟩⟨â†s (L, t + τ )âs (L, t + τ )⟩ (3.124)
The first term on the right-hand side is zero apparently, and the third term is pretty similar
to the square of generation rate. Likewise, the second term resembles H(τ ) which we have
demonstrate above.
In fact, ⟨â†s (L, t)âs (L, t + τ )⟩ = G(1) (τ ), and G(1) (τ ) = G(1)∗ (−τ )
The normalized second order correlation function can be written as
⟨â†s (L, t)â†s (L, t + τ )âs (L, t + τ )âs (L, t)⟩

g (2) (τ ) = (3.125)
⟨â†s (L, t)âs (L, t)⟩⟨â†s (L, t + τ )âs (L, t + τ )⟩
by multiplying c2 /L2 on both denominator and numerator
Rs2 + H 2 (τ )
g (2) (τ ) = (3.126)
Rs2
65
( τ)
2 .0
(2 )
A u to c o r r e a tio n fu n c tio n g 1 .5
2 7 µs
1 .0
0 .5
0 .0
0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 0 4 5 0 5 0 0
τ ( µs )
( τ)
2 .0
(2 )
A u to c o r r e a tio n fu n c tio n g
1 .5
5 5 µs
1 .0
0 .5
0 .0
0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 0 4 5 0 5 0 0
τ ( µs )
As mentioned before, if γ21 increases, the noise will harass system intensely, leading to the
detriment of ground-state coherence. Consequently, the coherence time of Stoke photon will
reduce as well. The coherence time of Figure (3.20) is almost equivalent to 1/γ21 = 26.5µs.
This result suggests that Raman Stoke photons possess a long coherent time.
66
Chapter 4
An Alternative Approach to Resolve Stoke Photons
In the last chapter, we have demonstrated how to obtain stoke photon generation by Laplace
transform. Now we will propose another approach to characterize stoke photon generation.
Similarly, the spectrum and second-order correlation will be presented.
4.1 The Hamiltonian
As mentioned before, the Hamiltonian after slowly varying amplitude is

 
0 0 Ω∗p
−}  
H̃ = 0 0 2gs∗ â†s  = H̃ (0) + λH̃ (1) (4.1)
2
Ωp 2gs âs ∆p
In last section, we operate this Hamiltonian directly on Heisenberg equation. This time,
however, we will apply perturbation theory, and separate the Hamiltonian into two parts; the
influence of Stoke field is regarded as perturbation
   
0 0 Ω∗p 0 0 0
−}   −}  
H̃ (0) = 0 0 0 H̃ (1) = 0 0 2gs∗ â†s  (4.2)
2 2
Ωp 0 ∆p 0 2gs âs 0
In zero order perturbation, we can extract six independent equation as below:
Γ (0) i ( (0) ∗ )
(0)
∂ σ̃11 (0)
= F̃11 + σ̃33 + σ̃13 Ωp − σ̃31 Ωp (4.3)
∂t 2 2
iσ̃23 Ω∗p
(0) (0) (0)
∂ σ̃21 γ21 σ̃21
= F̃21 − + (4.4)
∂t 2 2
(0)
∂ σ̃31 (Γ ) (0) i ( (0) (0) )
= F̃31 − + i∆p σ̃31 − σ̃11 − σ̃33 Ω∗p (4.5)
∂t 2 2
(0)
∂ σ̃22 Γ (0)
= F̃22 + σ̃33 (4.6)
∂t 2
Γ (0) i ( )
(0)
∂ σ̃32
= F̃32 − σ̃32 − 2∆p σ̃32 + σ̃12 Ω∗p
(0) (0)
(4.7)
∂t 2 2
i ( (0) )
(0)
∂ σ̃33
= F̃33 − Γσ̃33 + σ̃31 Ωp − σ̃13 Ω∗p
(0) (0)
(4.8)
∂t 2
67
Before proceeding further calculation, we should notice that the Heisenberg Langevin equa-
tion in first order is
(1)
dσ̃ij 1 (0) 1 (1)
= [σ̃ij , H̃ (1) ] + [σ̃ij , H̃ (0) ] + R(1) + F (1) (4.9)
dt i} i}
By taking a look at this equation, we operate commutator between the Hamiltonian of first
order and collective atomic operator in zero order. Because the solutions from zero order
involves Langevin noise, they will accompany annihilation operators or creation operators in
first order HLE. For this reason, the contribution of Langevin noise can be neglected com-
paring to other terms.
In addition, we make a strong assumption, ground state approximation, to continue further
computation, or the populations will be impelled to another ground state |2⟩ as time goes by.
Finally, we have zero order solution
 −iΓΩ∗p −2∆p Ω∗p

1 0
 Γ2 +4∆2p

σ̃ (0) = 
 0 0 0 
 (4.10)
iΓΩp −2∆p Ωp
Γ2 +4∆2p
0 0
By substituting equation 13 into first order Heisenberg-Langevin equation, we can obtain
Γ (1) i ( (1) )
(1)
∂ σ̃11
= F̃11 + σ̃33 + σ̃13 Ω∗p − σ̃31 Ωp
(1)
(4.11)
∂t 2 2
âs gs Ω∗p iσ̃23 Ω∗p
(1) (1) (1)
∂ σ̃21 γ21 σ̃21
= F̃21 − − + (4.12)
∂t 2 Γ + 2i∆p 2
Γ + 2i∆p (1) i ∗ ( (1) (1) )
(1)
∂ σ̃31
= F̃31 − σ̃31 − Ωp σ̃11 − σ̃33 (4.13)
∂t 2 2
(1)
∂ σ̃22 Γ (1)
= F̃22 + σ̃33 (4.14)
∂t 2
Γ (1) i ( )
(1)
∂ σ̃32
= F̃32 − σ̃32 − 2∆p σ̃32 + σ̃12 Ω∗p
(1) (1)
(4.15)
∂t 2 2
i ( (1) )
(1)
∂ σ̃33
= F̃33 − Γσ̃33 + σ̃31 Ωp − σ̃13 Ω∗p
(1) (1)
(4.16)
∂t 2
In general, we can resolve them directly by considering steady state; this solution will be
presented at the end of this section. However, to acquire the spectrum of Stoke field, we will
firstly address the Fourier transform of annihilation operator and collective atomic operator.
68
The fourier transform of annihilation operator and collective atomic operator
∫ ∞
â(z, t) = ãs (z, ω)e−iωt dω (4.17)
−∞
∫ ∞
F̃ij (z, t) = F̃ij (z, ω)e−iωt dω (4.18)
∫ −∞
∞
σ̃ij (z, t) = σ̃ij (z, ω)e−iωt dω (4.19)
−∞
It should be noted that we cannot apply original Fourier transform on creation operator with
this expression. Instead, it will be modified as follows
∫ ∞
â†s (z, t) = â†s (z, ω)eiωt dω (4.20)
−∞
∫ −∞
=− â†s (z, −ω)e−iωt dω (4.21)
∫ ∞∞
= â†s (z, −ω)e−iωt dω (4.22)
−∞
Besides, Langevin noise and collective atomic operator will follow this distinctive mathe-
matical structure.
By rewriting the first order Heisenberg-langevin equation in frequency domain
Γ (1) i ( (1) )
σ̃33 (z, ω) + σ̃13 (z, ω)Ω∗p − σ̃31 (z, ω)Ωp
(1) (1)
− iωσ̃11 (z, ω) = F̃11 (z, ω) + (4.23)
2 2
γ21 σ̃21 (z, ω) âs (z, ω)gs Ω∗p iσ̃23 (z, ω)Ω∗p
(1) (1)
(1)
− iωσ̃21 (z, ω) = F̃21 (z, ω) − − + (4.24)
2 Γ + 2i∆p 2
Γ + 2i∆p (1) i ( )
σ̃31 (z, ω) − Ω∗p σ̃11 (z, ω) − σ̃33 (z, ω)
(1) (1) (1)
− iωσ̃31 (z, ω) = F̃31 (z, ω) − (4.25)
2 2
(1) Γ (1)
− iωσ̃22 (z, ω) = F̃22 (z, ω) + σ̃33 (z, ω) (4.26)
2
Γ (1) i( )
− iωσ̃32 (z, ω) = F̃32 (z, ω) − σ̃32 (z, ω) − 2∆p σ̃32 (z, ω) + σ̃12 (z, ω)Ω∗p
(1) (1) (1)
(4.27)
2 2
i ( (1) )
− iωσ̃33 (z, ω) = F̃33 (z, ω) − Γσ̃33 (z, ω) + σ̃31 (z, ω)Ωp − σ̃13 (z, ω)Ω∗p
(1) (1) (1)
(4.28)
2
After resolving these simultaneous equations, one will receive that
(1)
σ̃23 (z, ω) = A(ω)âs (z, ω) + B(ω)F̃21 (z, ω) + U (ω)F̃23 (z, ω) (4.29)
Here we only present the solution of σ̃23 (z, ω) because what we emphasize on is the genera-
tion rate of stoke field; we define A(ω), B(ω), and U (ω) just for the simplicity of coefficient.
69
The structure of A(ω), B(ω), and U (ω) are
2gs |Ωp |2
A(ω) = ( )( ) (4.30)
Γ + 2i∆p (γ21 − 2iω)(iΓ + 2(ω + ∆p )) + i|Ωp |2
2iΩp
B(ω) = (4.31)
(γ21 − 2iω)(Γ − 2iω − 2i∆p ) + |Ωp |2
2(iγ21 + 2ω)
U (ω) = (4.32)
(γ21 − 2iω)(2ω + iΓ + 2∆p ) − i|Ωp |2
Remind that it is necessary to employ Maxwell-Schrödinger-equation to describe the time

and spatial evolution of Stoke field.
[ ]
∂ 1∂ iN gs∗
+ âs (z, t) = σ̃23 (z, t) (4.33)
∂z c ∂t c
by transforming operator from time domain into frequency domain. Maxwell-Schrödinger-

equation can be expressed as
[ ]
∂ iω iN gs∗
− âs (z, ω) = σ̃23 (z, ω) (4.34)
∂z c c
∂âs [ iω iN gs∗ ] iN gs∗ [ ]
+ − − A(ω) âs = B(ω)F̃21 + U (ω)F̃23 (4.35)
∂z c c c
∗
In further computation, I define p(ω) = − iωc − A(ω) iNcgs to simplify equation.
4.3 Computation of Stoke Field
Utilizing the method of integrating factor, the solution of âs (z, ω) is as below
∫
−p(ω)z −p(ω)z
z
iN gs∗ ( )
âs (z, ω) = âs (0, ω)e +e B(ω)F̃21 (x, ω) + U (ω)F̃23 (x, ω) ep(ω)x dx
0 c
(4.36)
Similarly,
∫ z′
∗ ′ ′ −p∗ (ω ′ )z ′ −iN gs ( ∗ ′ † ′ ′
â†s (z ′ , ω ′ ) =â†s (0, ω ′ )e−p (ω )z
+e B (ω )F̃21 (x , ω ) (4.37)
0 c
∗ ′ † ′ ′
) p∗ (ω′ )x′ ′
+ U (ω )F̃23 (x , ω ) e dx (4.38)
After that, we can write down ⟨â†s (L, ω ′ )âs (L, ω)⟩
70
Remember that we don’t inject stoke photon beforehand; stoke field are produced by Raman
process entirely. Thus, it is valid to suppose that there are no stoke photon in the beginning
of atomic ensemble ⟨â†s (z = 0)âs (z = 0)⟩ = 0
∗ ′ N 2 |gs |2
⟨â†s (L, ω ′ )âs (L, ω)⟩ = e−(p(ω)+p (ω ))L
[ c2
∫ L∫ L
∗ ′ ′ †
B ∗ (ω ′ )B(ω) e(p(ω)x+p (ω )x ) ⟨F̃21 (x′ , ω ′ )F̃21 (x, ω)⟩dxdx′ +
0 0
∫ L∫ L
∗ ′ ′ †
B ∗ (ω ′ )U (ω) e(p(ω)x+p (ω )x ) ⟨F̃21 (x′ , ω ′ )F̃23 (x, ω)⟩dxdx′ +
0 0
∫ L∫ L
∗ ′ ′ †
U ∗ (ω ′ )B(ω) e(p(ω)x+p (ω )x ) ⟨F̃23 (x′ , ω ′ )F̃21 (x, ω)⟩dxdx′ +
0 0
∫ L∫ L ]
∗ ′ (p(ω)x+p∗ (ω ′ )x′ ) † ′ ′ ′
U (ω )U (ω) e ⟨F̃23 (x , ω )F̃23 (x, ω)⟩dxdx (4.39)
0 0
As the annihilation operator at z = 0 is independent to Langevin noise, and the Langevin

noise possess haphazard fluctuation, one could regard it as
⟨âs (0, ω)F̃ij (x, ω)⟩ = ⟨âs )(0, ω)⟩⟨F̃ij (x, ω)⟩ = 0 (4.40)
Afterwards, we demonstrate the relation of ⟨F̃ij† F̃ij ⟩ between time domain and frequency
domain. Recall that
∫ ∞
1
F̃ij (z, ω) = F̃ij (z, t)e−iωt dt (4.41)
2π −∞
∫ ∞
† 1 ′
′
F̃ij (z, ω ) = F̃ij† (z ′ , t)eiω t dt (4.42)
2π −∞
∫ ∞∫ ∞
† 1 ′ ′
′ ′
⟨F̃ij (z , ω )F̃ij (z, ω)⟩ = 2 ⟨F̃ij† (z ′ , t′ )F̃ij (z, t)⟩eiω t e−iωt dtdt′ (4.43)
4π −∞ −∞
Associating with the definition of delta function, we can reduce it as

∫ ∞
′ 1 ′
δ(x − x ) = eip(x−x ) dp (4.44)
2π −∞
1
⟨F̃ij† (z ′ , ω ′ )F̃ij (z, ω)⟩ = Dij †,ij δ(z − z ′ )δ(ω ′ − ω) (4.45)
2πρ
The diffusion coefficients are

γ21
D21† ,21 = γ21 ⟨σ̃11 ⟩ D23† ,21 = ⟨σ̃31 ⟩ (4.46)
2
γ21
D21† ,23 = ⟨σ̃13 ⟩ D23† ,23 =0 (4.47)
2
71
Hence, the expectation value can be reduced as
∗ ′ N 2 |gs |2 γ21
⟨â†s (L, ω ′ )âs (L, ω)⟩ = e−(p(ω)+p (ω ))L (4.48)
2πρc2
∫ z
[ ∗ ′ B ∗ (ω ′ )U (ω)σ̃13 + U ∗ (ω ′ )B(ω)σ̃31 ] ′ ∗ ′ ′
B (ω )B(ω) + δ(ω − ω)e[p(ω)+p (ω )]x dx′
0 2
(4.49)
4.4 Spectrum of Stoke Photons
We have already illustrated how to calculate generation rate in last section. And generation
rate, in fact, can be deemed as the collective generation rate of all the modes. As a result, one
can deduce the form of spectrum as follows
∫ ∫
c † c ′
R = ⟨âs (L, t)âs (L, t)⟩ = ⟨â†s (L, ω ′ )âs (L, ω)⟩e−i(ω−ω )t dω ′ dω (4.50)
L L
∫
= Rs (ω)dω (4.51)
∫
c ′
=⇒ Rs (ω) = ⟨â†s (L, ω ′ )âs (L, ω)⟩e−i(ω−ω )t dω ′ (4.52)
L
by substituting equation 55 into equation 59, we have
γ21 N 2 |gs |2 −(p(ω)+p∗ (ω))L

Rs (ω) = e (4.53)
2πρcL
∫ z
( ∗ σ̃13 σ̃31 ) (p(ω)+p∗ (ω))x′ ′
B (ω)B(ω) + B ∗ (ω)U (ω) + U ∗ (ω)B(ω) e dx (4.54)
2 2
0
[ ∗
γ21 ODΓ32 [ ∗ ∗ σ̃13 ∗ σ̃31 ] 1 − e−(p(ω)+p (ω))L ]
= B (ω)B(ω) + B (ω)U (ω) + U (ω)B(ω)
8πL 2 2 p(ω) + p∗ (ω)
(4.55)
Notice that ρ is the inverse of linear density of atoms. Besides, according to Weisskopf-
Wigner theorem [35], we have
√
1 L |dij |2 ωij2 d32 }ω
ρ= = Γij = and recall that gs = (4.56)
Nz N 3πϵ0 }c3 } 2ϵ0 V
Thence, we have simplified formula above by
N |gs |2 |d32 |2 ω32 N 3nLλ2 ODΓ32

= = Γ32 = (4.57)
c 2}ϵ0 V c 8πL 4L
where OD denotes optical depth (OD = nσL), n is the density of atom, and σ indicates the
72
cross-section of absorption.
4.5 Generation Rate of Stoke Photons
The generation rate of Stoke photon can be carried out by integrating of frequency mode in
spectrum.
∫ ∞
Rs = Rs (ω)dω (4.58)
−∞
4.6 Second-Order Autocorrelation Function
As mentioned in last chapter, the normalized second order auto-correlation is defined as
⟨â†s (L, t)â†s (L, t + τ )âs (L, t + τ )âs (L, t)⟩

g (2) (τ ) = (4.59)
⟨â†s (L, t)âs (L, t)⟩⟨â†s (L, t + τ )âs (L, t + τ )⟩
To facilitate computation on numerator, we employ Wick’s theorem likewise. Also, most of

terms are pretty similar to generation rate, so we multiply c2 /L2 on both denominator and
numerator. Now, we just focus on ⟨â†s (L, t)âs (L, t + τ )⟩
∫ ∫
c † c ∞ ∞ † ′ ′
⟨âs (L, t)âs (L, t + τ )⟩ = ⟨âs (ω )âs (ω)⟩eiω t e−iω(t+τ ) dωdω ′ (4.60)
L L −∞ ∞
∫ ∞ ∫ ∞
−iωτ c † ′ ′
= e ⟨âs (ω )âs (ω)⟩e−it(ω−ω ) dω ′ dω (4.61)
−∞ L
∫−∞
∞
= Rs (ω)e−iωτ dω (4.62)
−∞
recall that Lc ⟨â†s (L, t)âs (L, t+τ )⟩ is H(τ ), which is mentioned in the end of chapter 3, and it is
the first order auto-correlation function with coefficient modification. In chapter 3, we exploit
Fourier transform on auto-correlation function to acquire spectrum, and in this chapter we
apply inverse Fourier transform on spectrum to retrieve first order auto-correlation function.
The normalized second order auto-correlation function will be
R2 + |F.T .Rs (ω)|2

g (2) (τ ) = (4.63)
R2
73
As shown, the prediction of stoke photon generation rate utilizing Fourier manner almost
matches to the outcome employing Laplace scheme.
4 0 0 0
3 5 0 0 F o u r ie r a p p r o a c h
L a p la c e a p p r o a c h
3 0 0 0
2 5 0 0
2 0 0 0
1 5 0 0
1 0 0 0
5 0 0
0
0 .6 0 .8 1 .0 1 .2 1 .4 1 .6 1 .8 2 .0
Ωp ( Γ)
Figure 4.1: Theoretical diagram of Generation rate with respect to pumping power. The
parameter setting: OD=100, ∆p = 500Γ, Γ32 = Γ/2, γ21 = 0.0001Γ, The gain lengths is
0.125 when pumping power is 0.5Γ, and gain length is 2 when pumping power is 2Γ
While pumping power rises, vacuum fields and noise fluctuation will be more likely to disturb
atomic system. Hence, the atomic system will possess a higher chance to create a two-photon
process.
1 2 0 0 0
F o u r ie r a p p r o a c h
1 0 0 0 0 L a p la c e a p p r o a c h
8 0 0 0
6 0 0 0
4 0 0 0
2 0 0 0
0
0 1 0 0 2 0 0 3 0 0 4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0
O D
rameter setting: Ωp = 0.5Γ, ∆p = 300Γ, Γ32 = Γ/2, γ21 = 0.0001Γ, The gain lengths is
about 3.5 when OD=1000
As the prediction in the previous approach, generation rate grows linearly in low gain and
74
develops exponentially in the high gain regime.
1 4 0 0 0 0
1 2 0 0 0 0 F o u r ie r a p p r o a c h
( 1 / s ) ( 1 / Γ)
1 0 0 0 0 0
8 0 0 0 0
6 0 0 0 0 5 ×1 0 -4
Γ
S p e c tru m
4 0 0 0 0
2 0 0 0 0
0
-0 .0 0 6 -0 .0 0 4 -0 .0 0 2 0 .0 0 0 0 .0 0 2 0 .0 0 4 0 .0 0 6
Figure 4.3: Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30
∆p = 300Γ, Γ32 = Γ/2, γ21 = 0.0005Γ, The gain lengths is about 0.02
1 4 0 0 0 0
1 2 0 0 0 0 F o u r ie r a p p r o a c h
( 1 / s ) ( 1 / Γ)
1 0 0 0 0 0
8 0 0 0 0
5 ×1 0 -4
Γ
6 0 0 0 0
S p e c tru m
4 0 0 0 0
2 0 0 0 0
0
-0 .0 0 6 -0 .0 0 4 -0 .0 0 2 0 .0 0 0 0 .0 0 2 0 .0 0 4 0 .0 0 6
Figure 4.4: Theoretical diagram of spectrum. The parameter setting: Ωp = 0.5Γ, OD=30
∆p = −300Γ, Γ32 = Γ/2, γ21 = 0.0005Γ, The gain lengths is about 0.02
75
Dressed state picture
2
Ω𝑝
4Δ𝑝
Δ𝑝 Ω𝑝 Stoke
ȁ3ۧ
Stoke ȁ𝐸1 ۧ
Ω𝑝
ȁ2ۧ ȁ2ۧ
2
Ω𝑝
ȁ𝐸2 ۧ
ȁ1ۧ 4Δ𝑝
Figure 4.5: Schematic illustration to realize the shift in spectrum by dressed state picture
Because we have transformed our system by slowly varying amplitude, the zero frequency
denotes the central frequency of Stoke photon. However, there is a shift in the above diagram.
One can follow the dressed state picture in AC Stark shift effect. By regarding the pumping
field and atoms as a new atomic system, the new energy levels will be shifted by |Ωp |2 /(4∆p )
for far detune condition, which we have mentioned in chapter two. We can interpret this effect
in our system as follows: by dividing Stoke photons into several proportions. That is, we
can regard there are one thousand Stoke photons, and every single of them will experience
the new atomic system constructed by other Stoke photons and atoms. The |E1 ⟩ and |E2 ⟩
represent the new atomic levels. In addition, both of |E1 ⟩ and |E2 ⟩ are the superposition of
|1⟩ and |3⟩, so population exist in both of the new atomic new levels which results in two
shifts in spectrum. Nevertheless, according to Equation (2.69, 2.70), |E2 ⟩ possess most of
population for far-detuned assumption. Consequently, we only have one shift. However,
this phenomenon doesn’t appear in the previous approach because the Stark shift term was
discarded. Otherwise, it would lead to obstacles in the further calculation.
76
( τ)
2 .0 F o u r ie r a p p r o a c h
(2 )
A u to c o r r e a tio n fu n c tio n g 1 .5
5 5 µs
1 .0
0 .5
0 .0
0 5 0 1 0 0 1 5 0 2 0 0 2 5 0 3 0 0 3 5 0 4 0 0 4 5 0 5 0 0
τ ( µs )
Figure 4.6: Schematic illustration second-order autocorrelation function. The parameter

setting: Ωp = 1Γ, OD=50, γ21 = 0.001Γ, Γ32 = Γ/2, ∆p = 500Γ and gain lengths=0.025
As demonstrate, these two methods reveal that the second order correlation function of Stoke
photons is 2 when τ = 0, and it will decay to 1 eventually. Namely, Stoke field is a thermal
field, and Stoke photons are bunched together. (For more detail, one can follow Appendix
E)
77
Chapter 5
Conclusion
5.1 Summary and Outlook
In this research, we utilize Laplace transform to carry out the time-dependent Stoke
photon generation rate, spectrum, and second-order autocorrelation function. Owing to the
stimulation effect of Stoke photon, the generation rate enhances rapidly for high Raman gain.
To realize the role of Raman gain, we provide the relation between generation rate and Raman
gain. Next, we analyze the spectrum of Stoke photon and realize that Stoke photon bandwidth
is mostly dependent on the dephasing rate of the two ground states for low gain regime. It
would be compressed by stimulated Stoke photon for high gain regime. According to our
prediction, the bandwidth of Stoke photons is about γ21 for low gain condition, and it is
similar to single-mode photons for high gain condition. Consequently, Stoke photons possess
long coherent time. For gain-length is 0.025, and γ21 is 0.001Γ, the coherent time is about
27 µs. In addition to the above computation, we are inspired by Dr.Kolchin and employ
Fourier transform to resolve the generation rate of Stoke photon, which is extremely close to
the previous theory. However, we notice that the spectrum obtained by Fourier approach is
shifted by |Ωp |2 /(4∆p ), which is difficult to be retrieved by Laplace method.
In fact, these two approaches possess their own advantage. The first method could assist
us in determining when Stoke photons attain steady state and tell us how Raman gain affects
the time to reach steady state. The second method, on the other hand, is much more easier to
be resolved, and it can reveal the Stark shift effect.
Up to the present, one can derive biphoton scheme by Fourier transform because the
Stoke photons and anti-Stoke photons are generated simultaneously. Nevertheless, in DLCZ
protocol, although the second step of DLCZ protocol, injecting reading beam to generate anti-
Stoke photon, resembles an EIT process as mentioned in chapter two, the result is based on the
ground states coherence built by the first step, which still remains an obstacle in mathematical
computation. Therefore, our next step would focus on working out the solution of anti-Stoke
photon.
In the experiment, we have already rehabilitated the experimental system in our new
laboratory. We will firstly follow the above description to verify whether the theoretical
prediction of generation rate and spectrum of Stoke photon is consistent with the experimental
78
result. If everything goes on wheels, we will reduce the generation rate (≪ 1), inject reading
beam to withdraw anti-Stoke photons, and then compare the generation rate, spectrum, and
cross-correlation function of Stoke and anti-Stoke photons with biphoton scheme.
For the ultimate goal, we desire to follow DLCZ protocol and create entangled pairs with
the atomic ensemble in the other optical table.
79
References
[1] C. H. Bennett and G. Brassard, “Quantum cryptography: Public key distribution and
coin tossing,” arXiv preprint arXiv:2003.06557, 2020.
[2] A. K. Ekert, “Quantum cryptography based on bell’s theorem,” Physical review letters,
vol. 67, no. 6, p. 661, 1991.
[3] N. Gisin, G. Ribordy, W. Tittel, and H. Zbinden, “Quantum cryptography,” Reviews of
modern physics, vol. 74, no. 1, p. 145, 2002.
[4] H. J. Kimble, “The quantum internet,” Nature, vol. 453, no. 7198, pp. 1023–1030, 2008.
[5] M. A. Nielsen and I. Chuang, “Quantum computation and quantum information,” 2002.
[6] N. Sangouard, C. Simon, H. De Riedmatten, and N. Gisin, “Quantum repeaters based
on atomic ensembles and linear optics,” Reviews of Modern Physics, vol. 83, no. 1,
p. 33, 2011.
[7] C. H. Bennett, G. Brassard, S. Popescu, B. Schumacher, J. A. Smolin, and W. K. Woot-
ters, “Purification of noisy entanglement and faithful teleportation via noisy channels,”
Physical review letters, vol. 76, no. 5, p. 722, 1996.
[8] D. Bouwmeester and A. Zeilinger, “The physics of quantum information: basic con-
cepts,” in The physics of quantum information, pp. 1–14, Springer, 2000.
[9] W. K. Wootters and W. H. Zurek, “A single quantum cannot be cloned,” Nature,
vol. 299, no. 5886, pp. 802–803, 1982.
[10] C. H. Bennett, G. Brassard, C. Crépeau, R. Jozsa, A. Peres, and W. K. Wootters, “Tele-
porting an unknown quantum state via dual classical and einstein-podolsky-rosen chan-
nels,” Physical review letters, vol. 70, no. 13, p. 1895, 1993.
[11] M. Zukowski, A. Zeilinger, M. A. Horne, and A. K. Ekert, “” event-ready-detectors”
bell experiment via entanglement swapping.,” Physical Review Letters, vol. 71, no. 26,
1993.
[12] H.-J. Briegel, W. Dür, J. I. Cirac, and P. Zoller, “Quantum repeaters: the role of im-
perfect local operations in quantum communication,” Physical Review Letters, vol. 81,
no. 26, p. 5932, 1998.
[13] L.-M. Duan, M. D. Lukin, J. I. Cirac, and P. Zoller, “Long-distance quantum commu-
nication with atomic ensembles and linear optics,” Nature, vol. 414, no. 6862, pp. 413–
418, 2001.
[14] A. Kuzmich, W. Bowen, A. Boozer, A. Boca, C. Chou, L.-M. Duan, and H. Kimble,
“Generation of nonclassical photon pairs for scalable quantum communication with
atomic ensembles,” Nature, vol. 423, no. 6941, pp. 731–734, 2003.
80
[15] C. H. van der Wal, M. D. Eisaman, A. André, R. L. Walsworth, D. F. Phillips, A. S.
Zibrov, and M. D. Lukin, “Atomic memory for correlated photon states,” Science,
vol. 301, no. 5630, pp. 196–200, 2003.
[16] A. Einstein, B. Podolsky, and N. Rosen, “Can quantum-mechanical description of phys-
ical reality be considered complete?,” Physical review, vol. 47, no. 10, p. 777, 1935.
[17] J. S. Bell, “On the Einstein Podolsky Rosen paradox,” Physics Physique Fizika, vol. 1,
no. 3, p. 195, 1964.
[18] M. Weber, Experimental Quantum Memory Applications and Demonstration of an El-
ementary Quantum Repeater Link with Entangled Light-Matter Interfaces. PhD thesis,
05 2012.
[19] J. Hald, J. Sørensen, C. Schori, and E. Polzik, “Spin squeezed atoms: a macroscopic
entangled ensemble created by light,” Physical review letters, vol. 83, no. 7, p. 1319,
1999.
[20] C. Liu, Z. Dutton, C. H. Behroozi, and L. V. Hau, “Observation of coherent optical
information storage in an atomic medium using halted light pulses,” Nature, vol. 409,
no. 6819, pp. 490–493, 2001.
[21] D. F. Phillips, A. Fleischhauer, A. Mair, R. L. Walsworth, and M. D. Lukin, “Storage
of light in atomic vapor,” Physical review letters, vol. 86, no. 5, p. 783, 2001.
[22] S. E. Harris, J. Field, and A. Imamoğlu, “Nonlinear optical processes using electro-
magnetically induced transparency,” Physical Review Letters, vol. 64, no. 10, p. 1107,
1990.
[23] M. O. Scully and M. S. Zubairy, Quantum Optics. Cambridge University Press, 1997.
[24] R. W. Boyd, Nonlinear Optics, Third Edition. USA: Academic Press, Inc., 3rd ed.,
2008.
[25] M. Bajcsy, A. S. Zibrov, and M. D. Lukin, “Stationary pulses of light in an atomic
medium,” Nature, vol. 426, no. 6967, pp. 638–641, 2003.
[26] M. Fleischhauer and M. D. Lukin, “Dark-state polaritons in electromagnetically induced
transparency,” Physical review letters, vol. 84, no. 22, p. 5094, 2000.
[27] T.-N. Wang, “Quasi-phase-matching slow light propagation in efficient four-wave mix-
ing media,” National Cheng Kung University, 2020.
[28] E. Arimondo, “V coherent population trapping in laser spectroscopy,” Progress in op-
tics, vol. 35, pp. 257–354, 1996.
[29] B. D. Agap’ev, M. Gorny, B. G. Matisov, and Y. V. Rozhdestvenski, “Coherent popu-
lation trapping in quantum systems,” Physics-Uspekhi, vol. 36, no. 9, p. 763, 1993.
[30] P. Kolchin, “Electromagnetically-induced-transparency-based paired photon genera-
tion,” Physical Review A, vol. 75, no. 3, p. 033814, 2007.
81
[31] M. Raymer and J. Mostowski, “Stimulated Raman scattering: unified treatment of spon-
taneous initiation and spatial propagation,” Physical Review A, vol. 24, no. 4, p. 1980,
1981.
[32] J.-C. Tseng, “Theoretical study on photon generation based on stimulated raman scat-
tering,” National Cheng Kung University, 2020.
[33] C.-Y. Cheng, J.-J. Lee, Z.-Y. Liu, J.-S. Shiu, and Y.-F. Chen, “Quantum frequency
conversion based on resonant four-wave mixing,” Physical Review A, vol. 103, no. 2,
p. 023711, 2021.
[34] R. Kubo, “The fluctuation-dissipation theorem,” Reports on progress in physics, vol. 29,
no. 1, p. 255, 1966.
[35] V. Weisskopf and E. Wigner, “Berechnung der natürlichen linienbreite auf grund der
diracschen lichttheorie,” in Part I: Particles and Fields. Part II: Foundations of Quan-
tum Mechanics, pp. 30–49, Springer, 1997.
[36] D. Rowell, “Chapter 2, continuous and discrete,” in Signal Processing, Massachusetts
Institute of Technology.
[37] F. Rana, “Chapter 6, random signals and noise,” in Quantum Optics, Cornell University.
[38] Y. Yamamoto, “Chapter 1 mathematical methods,” in Fundamentals of Noise processes,
National Institute of Informatics, Japan.
[39] M. J. Buckingham, “Noise in electronic devices and systems.,” JOHN WILEY & SONS,
INC., 605 THIRD AVE., NEW YORK, NY 10158, USA, 1983, 368, 1983.
[40] I. Kostanic, “Lecture 2,” in Digital Communications, Wireless Center Of Excellence,
Florida Institute of Technology.
[41] R. Zwanzig, Nonequilibrium statistical mechanics. Oxford university press, 2001.
[42] J. Garrison and R. Chiao, Quantum optics. Oxford University Press, 2008.
[43] R. J. Glauber, “The quantum theory of optical coherence,” Physical Review, vol. 130,
no. 6, p. 2529, 1963.
[44] R. Loudon, The quantum theory of light. OUP Oxford, 2000.
[45] M. Fox, Quantum optics: an introduction, vol. 15. OUP Oxford, 2006.
82
Appendix A
Heisenberg-Langevin Equation
In this section, we will firstly demonstrate classical Langevin equation and explain the im-
portance of noise fluctuation. Later, we exploit Heisenberg equation for an open system and
follow Langevin approach to introduce Heisenberg-Langevin equation.
Figure A.1: Schematic diagram of pollen’s particle and stochastic noise from environment
To start with, let’s consider the equation of motion of a pollen’s particle underwater. [41] In
general, it can be governed by
d⃗v
m = F⃗ = −γ⃗v (A.1)
dt
where −γ⃗v denotes the friction force, as known to all, the solution of velocity is
γt
⃗v = ⃗v0 e− m (A.2)
Hence, as time evolutes, the velocity of the particle will decay to zero owing to friction
force. Alternatively, one can contend that the kinetic energy of particle will dissipate to the
environment by heat. Nevertheless, Does this outcome make sense? One should remind that
while a particle is in thermal equilibrium, its average kinetic energy is about KB2 T for one
dimension case. Thus, the square of velocity is
1 KB T KB T
m⟨v 2 ⟩ = ⇒ ⟨v 2 ⟩ = (A.3)
2 2 m
which is inconsistent with our conclusion above, In consequence, it is compulsory to intro-
duce another source, fluctuation. The equation of motion can be modified as follows
dv
m = −γv + ξ(t) (A.4)
dt
83
where ξ(t) indicates fluctuation of noise. To simplify problem, we consider the fluctuation
is a Markov processes, which is a no-memory progress. Namely, the fluctuation in future are
entirely independent of the past. Also this kind of fluctuation will follow below equation
⟨ξ(t)⟩ = 0 ⟨ξ(t)ξ(t′ )⟩ = Dδ(t − t′ ) (A.5)
That is to say, the ensemble average of noise is zero since noise is a stochastic result. Be-
sides, noises are time-uncorrelated, thus, it is governed by a delta function. On account of
the assistance of fluctuation, the mean square of velocity doesn’t vanish. One may recall that
⟨ξ(t)ξ(t′ )⟩ resembles the auto-correlation we mentioned in chapter 3. The Fourier transform
of auto-correlation is the spectrum of noise. As correlation is a delta function, its Fourier
transform will be a flat line. That is to say, it contains all kinds of modes. Markovian noise is
named as white noise as well. For those memory noises, they are categorized as Non-Markov
noise or so-called coloured noise. Generally, Morkov noise is for larger time-scale incidents.
If two incidents occur closely in time-scale, the noises preserve memory.
After that, Let’s consider Heisenberg equation of motion for open system. The total Hamil-
tonian can be expressed as
Htotal = HS + HR + HSR (A.6)
HS is the Hamiltonian of system, HR is the Hamiltonian of reservoir, and HSR denotes the
interaction Hamiltonian between system and reservoir. [23, 42]
HS = }νa† a (A.7)
∑
HR = }νk b̂†k b̂k (A.8)
k
∑ [ ]
HSR = } gk b̂†k a + a† b̂k (A.9)
k
it should be noted that we consider our system is a single-mode field. The evolution of a and
b can be expressed as
1[ ] ∑ [ † ]
â˙ = a, hνâ† â + } gk â, b̂k â + â† b̂k (A.10)
i} k
[ † ] ∑ [ ]
†
= − ν ââ â − â ââ − i gk ââ† b̂k − â† âb̂k (A.11)
k
∑
= − νâ − i gk b̂k (A.12)
k
where we have assume that the commutation of system operator and environment operator
84
are commute to each other
[ †] [ ]
â, b̂ = â, b̂ = 0 (A.13)
The reason is that system operator and bath operator can be deemed as the two different
operators acting on different Hilbert space. âs ⊗ IˆR , Iˆs ⊗ b̂k , Ĥs ⊗ IˆR , Iˆs ⊗ ĤR
The evolution of reservoir operator is
˙ 1[ ] 1[ ∑ ] 1[ ∑ [ ]]
b̂k = b̂k , Ĥ = b̂k , }νm b̂†m b̂m + b̂k , } gm b̂†m â + â† b̂m (A.14)
i} i} m
i} m
[ †
] [ † †
]
= − i b̂k , b̂k b̂k − igk b̂k , b̂k â + â b̂k (A.15)
= − ib̂k − igk â (A.16)
where [b̂k , b̂†m ] = [b̂k , b̂m ] = 0, if k ̸= m
db̂k (t)
⇒ + iνk b̂k (t) = −igk â(t) (A.17)
dt
The solution of bath operator can solved as
∫ t
−inuk t ′
b̂k (t) = b̂k (0)e − igk â(t′ )e−inuk (t−t ) dt′ (A.18)
0
the first part including reservoir contribution denotes the free evolution of reservoir opera-
tor; the second term indicates the interaction with harmonic oscillator. By substituting the
solution into equation (A.12)
∑ ∫
[ t
′ ]
˙
â(t) = − iνâ(t) − i gk b̂k (0)e−iνk t − −igk â(t′ )e−iνk (t−t ) dt′ (A.19)
k 0
∑ ∫ t
′
= − iνâ(t) + fa (t) − gk2 â(t′ )e−iνk (t−t ) dt′ (A.20)
k 0
where
∑
fa (t) = −i gk b̂k (0)e−iνk t (A.21)
k
85
After that, we can multiply exp(iνt) on the whole equation
∑ ∫ t
′ ′ ′
iνt
â(t)e = iνâ(t)e iνt
+ Fa (t) − gk2 â(t′ )eiνt e−iνt eiνt eνk (t−t ) dt′ (A.22)
k 0
∑ ∫ t
′
ã˙ = Fa (t) − gk2 ã(t′ )ei(ν−νk )(t−t ) dt′ (A.23)
k 0
where
∑
Fa (t) = −i gk b̂k (0)e−i(νk −ν)t (A.24)
k
We firstly work out the second term, and transform the summation into the integral into k-
space.
∑ ∫ 2π ∫ π ∫ ∞
2
⇒ dϕ dθsinθ k 2 dk (A.25)
k
( 2π
L
)3 0 0 0
where 2 in numerator is the result of polarizations. (2π/L)3 represents the volume that a state
occupies.
∑ ∫ t
′
⇒ gk2 ã(t′ )ei(ν−νk )(t−t ) (A.26)
k 0
∫ ∞ ∫ t
V ′
⇒ 2 2 2
gk k dk ã(t′ )ei(ν−νk )(t−t ) dt′ (A.27)
π 0
∫ ∞ 0
∫ t
V νk 2 νk ′
= 2 2
g νk ( ) d( ) ã(t′ )ei(ν−νk )(t−t ) dt′ (A.28)
π 0 c c c
∫ ∞ ∫ t 0
V ′
= 3 2 g 2νk νk2 dνk ã(t′ )ei(ν−νk )(t−t ) dt′ (A.29)
cπ 0 c
0
To simplify further calculation, we make Born-Markov approximation (reservoir is tremen-

dous. Thus, system barely alter reservoir, on the other hand, bath will only influence system
if the bath mode is nearly resonant to the system νk ≃ ν)
∫ ∞ ∫ t
V ′
⇒ gν ν 3 2
2 2
dνk ã(t′ )ei(ν−νk )(t−t ) dt′ (A.30)
c cπ 0
∫ ∞ ∫0 t
V ′
⇒ g 2ν ν 2 3 2 dνk ã(t′ )ei(ν−νk )(t−t ) dt′ (A.31)
c c π −∞ 0
∫ t
V
⇒ 2πg 2ν ν 2 3 2 ã(t′ )δ(t − t′ )dt′ (A.32)
c cπ 0
V
⇒ g 2ν ν 2 3 ã(t) (A.33)
c cπ
86
by analogy to classical Langevin equation,
V D
g 2ν ν 2 ã(t) = ã(t) (A.34)
c 2c3 π 2 2
2
2gν/c ν 2V 2
D= (A.35)
c3 π
Finally we have
dã D
= − ã(t) + Fa (t) (A.36)
dt 2
The first term on the right-hand side indicates the relaxation term in OBE, which is known as
damping or dissipative term. The second term denotes the noise fluctuation. The fluctuation
term can be identified as the quantity to maintain uncertainty principle. To be more specific,
let’s consider a simple case without noise
dã D
= − ã(t) (A.37)
dt 2
the solution of ã(t)
D
ã(t) = ã(0)e− 2 t (A.38)
And the commutation is

[ ] [ Dt D ]
ã(t), ã† = ã(0)e− 2 , ã† (0)e− 2 t = e−Dt (A.39)
which is inconsistent to [â(t), â† (t)] = 1. [23] To maintain the commutation relation (un-
certainty), fluctuation must be introduced. In other words, fluctuation always accompanies
dissipation. These two terms are inseparable, which coincides with the classical deduction.
87
Appendix B
Fluctuation-Dissipation Theorem
In this section, we follow previous calculation to derive the diffusion coefficient and Einstein
relation. To start with, we suppose that reservoir achieve thermal equilibrium
⟨b̂k (0)⟩R = ⟨b̂†k (0)⟩R = 0 (B.1)

⟨b̂†k (0)b̂′k (0)⟩ = δkk′ n̄k (B.2)
⟨b̂k (0)b̂′†
k (0)⟩ = δkk′ [n̄k + 1] (B.3)
⟨b̂k (0)b̂′k (0)⟩R = ⟨b̂†k (0)b̂′†
k (0)⟩R = 0 (B.4)
Hence, the expectation value of Langevin noise is zero
⟨F̂a (t)⟩R = ⟨F̂a† (t)⟩R = 0 (B.5)
And the autocorrelation function of noise can be written as

∑∑ ′
⟨F̂a† (t)F̂a (t)⟩R = gk gk′ ⟨b̂†k b̂k′ ⟩R ei(νk −ν)t e−i(νk′ −ν)t (B.6)
k k′
∑ ′
= gk2 n̄k ei(νk −ν)(t−t ) (B.7)
∫ ∞
k
V ′
= 2 k 2 n̄k ei(νk −ν)(t−t ) gk2 dk (B.8)
π 0
∫ ∞
V ′
= 2 3 dνk νk2 n̄νk /c gν2k /c ei(νk −ν)(t−t ) (B.9)
π c 0
2V ν 2 2
= 3
gν/c n̄ν/c δ(t − t′ ) (B.10)
πc
=D n̄th δ(t − t′ ) (B.11)
By comparing to classical Langevin expression
⟨F̂a† (t)F̂a (t)⟩R = 2⟨Dâ† â ⟩δ(t − t′ ) (B.12)

⇒ 2⟨Dâ† â ⟩ = D n̄th (B.13)
88
Following the same approach, we can write down
⟨F̂a (t)F̂a† (t)⟩R = D(n̄th + 1)δ(t − t′ ) (B.14)

⟨F̂a (t)F̂a (t)⟩R = ⟨F̂a† (t)F̂a† (t)⟩R = 0 (B.15)
Before retrieving Einstein relation, we firstly work out several quantities which are required
in the next step. By solving equation (A.36)
∫ t
−D D ′
ã(t) = ã(0)e 2
t
+ dt′ e− 2 (t−t ) Fa (t′ )dt′ (B.16)
0
Next
∫ t
D D ′
⟨Fa† (t)ã(t)⟩ =⟨Fa† (t)⟩ã(0)e− 2 t + dt′ e− 2 (t−t ) ⟨Fa† (t)Fa (t′ )⟩dt′ (B.17)
∫ t
0
D ′
= dt′ 2⟨Dâ† â ⟩δ(t − t′ )e− 2 (t−t ) dt′ (B.18)
0
=⟨Dâ† â ⟩ (B.19)
The reason for the one-half result is that the integration is only integrated up to t, which is at
the edge of the delta function. Likewise,
⟨ã† (t)Fa (t)⟩ = ⟨Dâ† â ⟩ (B.20)
Recall that
dã D
= − ã(t) + Fa (t) (B.21)
dt 2
2 ˙
⇒ â(t) = − [â(t) − Fa (t)] (B.22)
D
89
Finally we can evaluate the diffusion coefficient
d⟨â† (t)â(t)⟩ dâ† dâ(t)

=⟨ â(t)⟩ + ⟨â† (t) ⟩ (B.23)
dt dt dt
[ D ] [ D ]
=⟨ − ã† (t) + Fa† (t) â(t)⟩ + ⟨â† (t) − ã(t) + Fa (t) ⟩ (B.24)
2 2
D D
= − ⟨â† (t)â⟩ − ⟨â† (t)â⟩ + ⟨Fa† (t)â(t)⟩ + ⟨â† (t)Fa (t)⟩ (B.25)
2 2
D D
= − ⟨â† (t)â⟩ − ⟨â† (t)â(t)⟩ + 2Daˆ† â (B.26)
2 2
d⟨â† (t)â(t)⟩ D † D
⇒2Daˆ† â = + ⟨â (t)â⟩ + ⟨â† (t)â⟩ (B.27)
dt 2 2
d⟨â† (t)â(t)⟩ [ ] [ ]
⇒2Daˆ† â = − ⟨ â˙ † (t) − Fa† (t) â(t)⟩ − ⟨â† (t) â(t)
˙ − Fa (t) ⟩
dt
(B.28)
For more detail of fluctuation-dissipation theorem, one can follow“The fluctuation-dissipation

theorem”, R. Kubo and “Quantum optics”, Scully. [34, 23, 42]
90
Appendix C
Quantization of Electromagnetic Field
The wave nature of light reveals the interference effect; the particle nature of light tells us that
atom will be excited as it absorbs one photon. To address a unified picture, it is compulsory
to introduce the quantization of electromagnetic fields. To begin with, we can start from the
Maxwell equations in vacuum.
∇·E
⃗ =0 (C.1)
⃗
∇×E ⃗ = − ∂B (C.2)
∂t
∇·B =0
⃗ (C.3)
⃗
⃗ = µ0 ϵ0 ∂ E
∇×B (C.4)
∂t
Considering a single mode radiation field confined in a conducting cavity (length=L) prop-
agates along k̂ direction. Owing to the boundary condition (There is no electric field at the
boundary), the electric field can be expressed as
⃗ = Aq̂(t)sin(kz)î
E (C.5)
where k = mπ/L We assume that the it is polarized along x̂-direction; q(t) and sin(kz) is
the result of separation of variables, and A is a coefficient we attempt to resolve. by equation
(C.4), we have
∇×B ˙
⃗ = µ0 ϵ0 Aq̂(t)sin(kz)î (C.6)
the curl of magnetic field is
î ĵ k̂ [ ]
∂Bz ∂By
∂
∂x
∂
∂y
∂
∂z
= − î (C.7)
∂y ∂z
Bx By Bz
91
note that field propagates though ẑ, thus Bz = 0
∂By ˙
− = µ0 ϵ0 Aq̂(t)sin(kz) (C.8)
∂z
˙
⃗ = µ0 ϵ0 A q̂(t)cos(kz) ĵ
B (C.9)
k
Next, let’s recall the energy of electromagnetic field in free space is
∫ [ ]
1 1 ⃗ 2
H= dV ϵ0 |E| + |B|
⃗ 2
(C.10)
2 µ0
∫ [ 2 ]
1 2 2 2 µ0 ϵ20 A2 q̂˙ 2
= dV ϵ0 A q̂ (t)sin (kz) + cos (kz) (C.11)
2 k2
L3
notice that the space integral of sin2 (kz) is equals to 2
, thus
[ 2]
V ϵ0 A2 â2 µ0 ϵ20 A2 q̂˙
H= + (C.12)
2 2 2k 2
[ ]
A2 V ϵ 0 2 2 2
= ω q (t) + q̇ (t) (C.13)
4ω 2
which is pretty similar to the result of the quantization of harmonic oscillator
1[ 2 2 ]
H= ω q (t) + p2 (t) (C.14)
2
by comparing the coefficient, we can obtain A
√
2ω 2
A= (C.15)
V ϵ0
additionally, it is convenient to replace the canonical variable q̂(t) with annihilation operator
and creation operator (non-hermitian). In harmonic oscillator model
√
}[ ]
q̂ = â + aˆ† (C.16)
2ω√
}ω [ ]
p̂ = − i â − aˆ† (C.17)
2
Therefore, radiation field can be written as

√
1 }ω [ ]
⃗
E= â + â† sin(kz)î (C.18)
2 V ϵ0
√
1 }ω [ ]
⃗ =
B â − â† cos(kz)ĵ (C.19)
2ic V ϵ0
92
By analogy to quantum theory in harmonic oscillator, n̂ = â† â denotes the number operator
√
â|n⟩ = n|n − 1⟩ (C.20)
√
â† |n⟩ = n + 1|n + 1⟩ (C.21)
n̂|n⟩ = n|n⟩ (C.22)
|n⟩ indicates Fock state which represents the number of photon at certain mode; the Hamil-
tonian can also be modified as
[ 1]
H = }ω â† â + (C.23)
2
As shown, the eigen energy is }ω/2 for vacuum state (|n = 0⟩), which implies vacuum fluc-
tuation will contribute to zero-point energy. Vacuum fields, in general, spread in the entire
space; they are composed of infinite number of modes, with zero-point energy }ω/2 indi-
vidually. If we take all the modes into account, an abnormal outcome will appear. The total
zero-point energy of the universe tends to infinity. Nevertheless, there are several phenomena
that can be explained sufficiently by zero-point energy. For instance, owing to the ”stimula-
tion” of vacuum field, atoms will ”spontaneously” emit photons stochastically, which results
in most of visible light around as. Besides, Casimir effect, which denotes that two neutral
metal plates will attract to each other if the distance is small enough, can also be realized as
the outcome of zero-point energy.
Utilizing Heisenberg equation of motion, we acquire
dâ 1[ ] 1 [ ]
= â, Ĥ = }ω ââ† â − â† ââ = −iωâ (C.24)
dt i} i}
dâ† 1[ † ] 1 [ ]
= â , Ĥ = }ω â† â† â − â† ââ† = iωâ† (C.25)
dt i} i}
the solutions are
â(t) = â(0)e−ωt (C.26)

â† (t) = â(0)eiωt (C.27)
Finally, the radiation field can be quantized as

√ [ ]
⃗ = }ω i(kz−ωt) † −i(kz−ωt)
E â(0)e + â (0)e (C.28)
2ϵ0 V
93
Appendix D
Maxwell-Schrödinger Equation
The derivation is pretty similar to chapter 2. The difference is that density matrix is replaced
with collective atomic operator, thus, we just start from
⃗
1 ∂ 2E ∂ 2 P⃗
− ∇ 2⃗
E = −µ 0 (D.1)
c2 ∂t2 ∂t2
The left hand side is
√ [ ][ ]
⃗
1 ∂ 2E }ωs 1 ∂ 2 ∂2
− ∇2 E
⃗ =x̂ − âs (z, t)e i(kz−ωs t)
+ h.c. (D.2)
c2 ∂t2 2ϵ0 V c2 ∂t2 ∂z 2
√ [ ][ ][ ]
}ωs 1 ∂ ∂ 1∂ ∂
=x̂ + − âs (z, t)e i(kz−ωs t)
+ h.c. (D.3)
2ϵ0 V c ∂t ∂z c ∂t ∂z
√ [ ][
}ωs 1 ∂ ∂ 1 ∂âs i(kz−ωs t) 1
=x̂ + e + âs (−iωs )ei(kz−ωs t) (D.4)
2ϵ0 V c ∂t ∂z c ∂t c
]
∂âs i(kz−ωs t)
− e − ikâs ei(kz−ωs t) + h.c. (D.5)
∂z
√ [ ][ ]
}ωs 1 ∂ ∂ −iωs i(kz−ωs t)
=x̂ + âs e − ikâs e i(kz−ωs t)
+ h.c. (D.6)
2ϵ0 V c ∂t ∂z c
√ [ ]
}ωs 1 ∂ ∂
= − 2ike i(kz−ωs t)
x̂ + âs (z, t) + h.c. (D.7)
2ϵ0 V c ∂t ∂z
where I have assumed that Stoke field varies slowly in both space and time domain | ∂â
∂t
s
|≪
ωs , and | ∂z | ≪ k
∂âs
Right hand side

[ ]
∂ 2 P⃗ ∂2
−µ0 2 = − x̂nd23 µ0 2 σ̃23 e i(kz−ωs t)
+ h.c. (D.8)
∂t ∂t
[ ]
∂ ∂ σ̃23 i(kz−ωs t)
= − x̂nd23 µ0 e − iωs σ̃23 ei(kz−ωs t)
+ h.c. (D.9)
∂t ∂t
=x̂ωs2 nd23 µ0 ei(kz−ωs t) σ̃23 + h.c. (D.10)
94
Next, let’s reorganize them
√ [ ]
}ωs 1 ∂ ∂
⇒ −2ike i(kz−ωs t)
x̂ + âs (z, t) = x̂ωs2 nd23 µ0 ei(kz−ωs t) σ̃23 (D.11)
2ϵ0 V c ∂t ∂z
[ ] √
1∂ ∂ ωs2 nd23 µ0 2ϵ0 V
⇒ + âs (z, t) = σ̃23 (D.12)
c ∂t ∂z −2ik }ωs
[ ] √
1∂ ∂ 2ϵ0 V ωs2 nd23 µ0 }ωs
⇒ + âs (z, t) = σ̃23 (D.13)
c ∂t ∂z }ωs (−2ik) 2ϵ0 V
[ ]
1∂ ∂ 2V n ∗
⇒ + âs (z, t) = g σ̃23 (D.14)
c ∂t ∂z −2ic s
[ ]
1∂ ∂ iN gs∗
⇒ + âs (z, t) = σ̃23 (D.15)
c ∂t ∂z c
95
Appendix E
Correlation Function
In this section, we just briefly illustrate the concept and properties of correlation function.
The measurement of correlation function is not only one of the technique to rebuild the in-
formation of signal but also a realization to distinguish whether signal is a quantum result.
In general, we can interpret the measurement of correlation as an approach to quantify the
similarity of signals. For an ergodic process, the cross-correlation function is defined as
∫ ∞
Cf,g (τ ) = f (t)g ∗ (t + τ )dt (E.1)
−∞
which resembles the convolution of f (t) and g(t). If g(t) is replaced with f (t), it denotes
the autocorrelation function. Next, let’s consider a simple case to manipulate autocorrelation
function.
𝜏
signal
Time
−𝜏
−𝜏/2
0
𝜏/2
𝜏
Autocorrelation function Delay time=0
−𝜏 −𝜏/2 0 𝜏/2 𝜏 Time
Figure E.1: Schematic illustration of autocorrelation function.
When we calculate the correlation function, only the red section of signal may contribute to
96
the outcome. The integration of f (t) and f (t ± τ ) is zero; the integration of f (t) and f (t)
attain maximum. On the last page, the signal is completely clean, which may not exist in re-
alistic channel. In practical, experiments are not faultless, and equipments exist imperfection
somehow. Therefore real signal may be immersed in noise. However, we still can retrieve
the information of real signal by autocorrelation function.
𝜏
signal
Time
White noise
Time
Time
f(t)
ACF of f(t)= ACF of signal

Time
2𝜏
ACF of white noise
Time
Figure E.2: Schematic illustration of autocorrelation function for regular signals immersed
in noise and white noise.
Signal 2
𝜏
Time
ACF of signal 2 Time
2𝜏
Figure E.3: Schematic illustration of autocorrelation function for stochastic signals im-
mersed in noise
Even though it is difficult to discern the real signal from f (t), we can exploit ACF to extract
the information of signal out. Because noise is stochastic, it only correlates to itself when
τ = 0. Thus, the result of noise proportion in f (t) is cancel out by integration. Additionally,
97
the autocorrelation function of white noise is a delta function, which will be a flat line by
Fourier transformation. That’s the reason we call it white noise.
For those stochastic signals, the autocorrelation function is totally correlated to themselves
when τ = 0. For large τ , even though correlation is broken, the integration result can still
lead to an offset because these random signals are positive.
Besides the construction of real signal, correlation function can assist us in seizing anti-Stoke
photons. On account of the noise, which origins from electronic noise or fluorescence, anti-
Stoke photons are hidden. Hence, we utilize cross-correlation function to measure the simi-
larity of Stoke and anti-Stoke photons, and obtain delay time of the latter one.
In experiment, we inject reading pulse to readout the spin-wave coherence. These readout
anti-Stoke photons might not be readout immediately. As a result, there is a delay time be-
tween Stoke photon and the corresponding anti-Stoke photon.
Stoke
Delay time
Time
Anti-Stoke
Time
Delay time
Cross-correlation function
0 Time
Figure E.4: Schematic illustration of cross-correlation function for Stoke and anti-Stoke
photons.
In experiment, we choose the circumstance that we detect a Stoke photon as the trigger point
to start the measurement of anti-Stoke photon, which is so-called post-selection. By em-
ploying two detectors for Stoke and anti-Stoke photon and repeating rounds of experiments,
we can obtain the cross-correlation function. To have a deeper interpretation of correlation
function, one can download a software, Multi-Instrument, on Virtins to operate virtual oscil-
loscope and function generator.
In addition to these signal processing, the second-order correlation provides us the charac-
teristic of signal. (R J. Glauber explained Hanbury Brown and Twiss effect, 1963) [43]
98
Adjust Stop
Start
Figure E.5: Schematic illustration of Hanbury Brown and Twiss experiment
This experiment contains two detectors, which is referred to as second-order correlation or

intensity-intensity measurement. The left-hand side, they firstly inject a stream of photons
and then arrive start and stop channel randomly. For our experiment, the start signal denotes
the trigger. Stop detection records the time it takes to receive the next photon. Because what
we measure is the photon number, which is similar to the intensity, we can write down the
correlation function as follows
⟨nstart (t)nstop (t + τ )⟩
g (2) (τ ) = (E.2)
⟨nstart (t)⟩⟨nstop (t + τ )⟩
where n represents the number of counts. The physical interpretation of g (2) (τ ) is the proba-
bility that we detect the second photon on stop channel in those cases that we have detected
the first photon on start channel.
After that, we consider classical cases. For coherent light, the photons are emitted randomly,
and the intervals between successive photons are irregular. As a result, the probability to
detect the second photon on stop channel is independent of time. The value of g (2) (τ ) is a
constant. For thermal fields, photons are bunched together. Hence, the probability to detect
the second photon declines as τ increases, and that’s our theoretical prediction in chapter
3. So far, these cases possess their corresponding classical interpretation. However, if all
successive photons are regularly spaced by a large distance, the probability of detecting the
second photon is zero at τ = 0, which denotes simultaneous detection, and then it reaches
maximum at a specific τ . In 1977, Kimble illuminated powerful laser on a single atom and
observed this pure quantum phenomenon. For more detail of correlation function and the
deviation of g (2) (τ ) below, one can follow The Quantum Theory of Light, R.Loudon, and
Quantum Optics, Mark Fox. [44, 45]
Bunched light : g (2) (0) > 1, g (2) (0) > g (2) (τ ) (E.3)
Coherent light : g (2) (0) = 1 = g (2) (τ ) (E.4)
Antibunched light : g (2) (0) < 1, g (2) (τ ) > g (2) (0) (E.5)
99
2
𝑔 (𝜏)
Coherent 1
Time
Bunched 1
Time
Antibunched 1
Time
Figure E.6: Schematic illustration of antibunched, coherent, and bunched photons
100

Narrownamd Stoke Photons Based on DLCZ Protocol

Uploaded by

Copyright:

Available Formats

Narrownamd Stoke Photons Based on DLCZ Protocol

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Narrownamd Stoke Photons Based on DLCZ Protocol

Uploaded by

Copyright:

Available Formats

National Cheng Kung University

學生： 楊智閔 Student： Chih-Min Yang

關 : 拉曼散射, DLCZ 機制, 電磁波引發透明, 量子中繼站

Keyword: Raman Scattering, DLCZ protocol, Electromagnetically induced transparency,

Chapter 2. Theoretical Model of Electromagnetically Induced Transparency 11

Chapter 3. Stoke Photons Generation Based on DLCZ Protocol 33

Chapter 4. An Alternative Approach to Resolve Stoke Photons 67

Appendix A. Heisenberg-Langevin Equation 83

Appendix B. Fluctuation-Dissipation Theorem 88

Appendix C. Quantization of Electromagnetic Field 91

Appendix D. Maxwell-Schrödinger Equation 94

Appendix E. Correlation Function 96

1.1 Schematic illustration of entanglement swapping by employing Bell-state

3.1 Schematic diagram of energy level of stimulated Raman scattering, where

E.1 Schematic illustration of autocorrelation function. . . . . . . . . . . . . . . 96

With the development of quantum technology in recent years, quantum communica-

Fortunately, in 1993, Charles H. Bennett proposed a novel scheme called entanglement

1.1 The Swapping of Quantum Entanglement

As known, quantum entanglement has always been a fundamental and counter-intuitive

|Ψ⟩ =|Φ, +⟩AB ⊗ |Φ, +⟩CD (1.3)

Hence, we can modify Equation (1.5) as follows

by extracting bell-state individually and rearranging them, we obtain

Alice Bob Charlie David

Alice Bob Charlie David

Figure 1.1: Schematic illustration of entanglement swapping by employing Bell-state mea-

So far, we have reviewed several major elements to achieve quantum communication.

L Filter Channel Detector 1

R Filter Channel Detector 2

𝑙1 Filter Channel Channel Filter 𝑙2

Figure 1.7: Schematic illustration of entanglement swapping in DLCZ scheme

Theoretical Model of Electromagnetically Induced

2.1 Electromagnetically Induced Transparency

As is known to all, if we project frequency-tunable light on atoms, light will be absorbed

ȁ1ۧ ȁ1ۧ ȁ1ۧ

2.2 Optical-Bloch Equation

2.2.1 Density Matrix Expression

where pi is the probability to find an individual subsystem of ensemble at state |ψ⟩

The element of the density matrix is expressed as:

ρmn = ⟨ϕm |ρ|ϕn ⟩ (2.3)

|ϕm ⟩ and |ϕn ⟩ are one of the eigenstate.

2.2.2 The Equation of Motion

Considering the Hamiltonian of our system:

Ĥ = Ĥ0 + ĤI (2.9)

Take the ĤDirac,13 and ĤDirac,23 term as an example,

where |ψ̃⟩ = Ŝ † |ψ⟩

Finally, we can obtain the formulation of transformation Hamiltonian:

We next write down the final Hamiltonian as below

2.3 Steady-State Solution

2.3.1 Brief Review of Perturbation Theory

In practice, dynamic problems in quantum mechanics are too complex to be resolved

Similarly, density matrix and relaxation term can be viewed as:

 (0) (0) (0)

(1) (1) (1)

The Gauss law for electric field : ∇ · D = ρfree (2.39)

In our experimental arrangement, there are no magnetic material M = 0, current Jfree = 0,

notice that d⃗ij = d⃗12 = d⃗21 = 0 as i = j

∂Ep,c (z, t) ∂Ep,c (z, t)

L.H.S of MSE can be simplified as:

In an isotropy and linear dielectric medium, the electrical polarization is :

ndij ρ̃ji Γij 3nλ3 ODΓij ρ̃ji

學生：楊智閔 Student： Chih-Min Yang