AN ABSTRACT OF THE DISSERTATION OF
Andrew Stickel for the degree of Doctor of Philosophy in Physics presented on August 10, 2016.
Title: Terahertz Induced Non-linear Electron Dynamics in Nanoantenna Coated Semiconductors at the Sub-picosecond Timescale
Abstract approved: Yun-Shik Lee
This dissertation is an exploration of the material response to Terahertz (THz) radiation. Specifically we will explore the ultrafast electron dynamics in the non- perturbative regime in semiconductors that have been patterned with nanoantenna arrays using broadband, high intensity, THz radiation. Three main semiconductor materials will be studied in this work. The first is VO2 which undergoes a phase transition from an insulator, when it is below 67◦ C, to a metal, when it is above 67◦ C. The second and third materials are Si and GaAs which are two of the most commonly used semiconductors.
We study the insulator to metal transition (IMT) of VO2 and its response to high field THz radiation. The near room temperature IMT for VO2 makes it a very promising material for electrical and photonic applications. We demonstrate that with high field THz the IMT transition can be triggered. This transition is induced on a sub-cycle timescale. We also demonstrate a THz field dependent reduction in the transition temperature for the IMT when transitioning from both below Tc to above as well as from above Tc to below. This transition is not equal for the above and below cases and leads to a narrowing of the hysteresis curve of the IMT. The thin film Fresnel coefficients, along with a phenomenological model developed for the nanoantenna patterned VO2, are also used to calculate the sheet conductivity of the VO2 sample. We show, using this sheet conductivity and its relation to the band gap, that the bang gap in the insulating phase has a strong dependence on the incident THz radiation with larger fields reducing the band gap from 1.2 eV at low incident THz fields to 0.32 eV at high incident THz fields.
The ultrafast, non-equilibrium, electron dynamics of GaAs and Si were also explored. GaAs and Si are the two most prevalent semiconductors in use today and with the decrease in size and increase in clock speeds of transistors a deep understanding of the ultrafast high field electron dynamics of these materials is of vital importance. Using THz time domain spectroscopy we investigate the transition rates of electrons excited by a 800 nm optical pump. We show that the optically induced transition for GaAs happens on a shorter timescale than that of Si. We also investigate the THz transmission dependence on the indecent THz field. We show that intense THz fields enhance the transmission through the sample. The increase in transmission is due to intervalley scattering which increases the effective mass of the electrons resulting in a decrease in the conductivity the sample. c Copyright by Andrew Stickel August 10, 2016 All Rights Reserved Terahertz Induced Non-linear Electron Dynamics in Nanoantenna Coated Semiconductors at the Sub-picosecond Timescale
by
Andrew Stickel
A DISSERTATION
submitted to
Oregon State University
in partial fulfillment of the requirements for the degree of
Doctor of Philosophy
Presented August 10, 2016 Commencement June 2017 Doctor of Philosophy dissertation of Andrew Stickel presented on August 10, 2016.
APPROVED:
Major Professor, representing Physics
Chair of the Department of Physics
Dean of the Graduate School
I understand that my dissertation will become part of the permanent collection of Oregon State University libraries. My signature below authorizes release of my dissertation to any reader upon request.
Andrew Stickel, Author ACKNOWLEDGEMENTS
First and foremost I would like to thank Professor Yun-Shik Lee. For Advising me throughout the my 6 years of working for him. I would also like to thank Dr. Zack Thompson for working with me, teaching how to tighten a bolt, and not punching me in the face as much as he had the right to. I would also like to thank Dr. Byounghwak Lee and Ali Mousavian for their input and support during my work. All of my friends and faculty in the physics department that have listened to me whine...I mean helped me out. Specifically I would like to thank Dr. KC Walsh for being a mentor, friend, and all around amazing help who gave me important insights into grad school and helped me out immensely. I would like to thank my teachers Mr. Day, Mr. Morales, and Dr. Deal for encouraging my passion for science and guiding me when I needed it. Also I need to thank Dr. Ritter and Ms. Wise without who this 137 page document would be a dribbling pile of non-sense. Finally, and most importantly, I would like to thank my parents Bryan Stickel and Carol Ernst. From the beginning of my life they have done nothing but support my dreams and put up with my 19,762 questions. They kept me from killing myself, even when those cool metal disks on the oven seemed REALLY fun to play with. Mom, Dad, I love you both. TABLE OF CONTENTS Page 1 Terahertz Radiation and Applications1
1.1 History...... 2
1.2 Sources...... 2
1.3 Detection...... 4
1.4 Applications...... 6
2 Theoretical Foundations for Terahertz Spectroscopy of Semiconductors8
2.1 Maxwell’s Equations and Electromagnetic Radiation...... 8 2.1.1 Electromagnetic Plane Waves...... 10 2.1.2 Electromagnetic Waves at a Boundary...... 16
2.2 Non-linear Optics...... 23 2.2.1 Second Order Non-linear Effects...... 24 2.2.2 The χ(2) tensor...... 26 2.2.3 Phase Matching...... 28
2.3 Ultrafast Optics...... 30 2.3.1 Mode Locking...... 31 2.3.2 Pulse Width and Dispersion...... 34
2.4 Band Theory of Semiconductors...... 36 2.4.1 The Simple 1-D Chain...... 37 2.4.2 Beyond the 1-D Chain...... 39
2.5 THz Field Enhancement by Nanoantennas...... 41 2.5.1 Nanoantenna Enhanced Electric Fields for Long Wavelength Radiation...... 42 TABLE OF CONTENTS (Continued) Page 2.5.2 Nanoantenna Arrays...... 44
3 Terahertz Generation and Detection 46
3.1 THz Generation...... 46 3.1.1 ZnTe and Optical Rectification...... 46
3.1.2 LiNbO3 THz Generation...... 48
3.2 THz Detection...... 51 3.2.1 Pyroelectric Detectors...... 52 3.2.2 The Bolometer...... 53
3.3 Experimental Setup and Procedures...... 55 3.3.1 The Laser System...... 55 3.3.2 Layout of Experimental Setup...... 56 3.3.3 Experimental Designs...... 58
4 Vanadium Dioxide Field Induced Transition 65
4.1 Mott Insulators...... 65 4.1.1 The Hubbard model...... 65 4.1.2 Conductivity and the band gap...... 68 4.1.3 Mott Criterion...... 71
4.2 Insulator to Metal transition in VO2 ...... 72
4.2.1 The structural phase transition of VO2 and its consequences on conductivity...... 72
4.3 The Sample...... 74
4.4 THz Dependent Transmission...... 76 TABLE OF CONTENTS (Continued) Page 4.5 Frequency Response...... 80
4.6 THz dependent Hysteresis Narrowing...... 84
4.7 THz Dependent Sheet Conductivity...... 89
4.8 Time Dependent Transmission of the THz Beam and Real Time Fluc- tuations of the Conductivity...... 94
4.9 Conclusion...... 98
5 High Field Electron Dynamics at Sub-picosecond Timescale in GaAs and Si 99
5.1 Intervalley Scattering Induced by Strong THz Fields...... 99 5.1.1 Intervalley Scattering in GaAs...... 101 5.1.2 Intervalley Scattering in Si...... 104
5.2 Nanoantenna Patterned Samples...... 104
5.3 Optical Free Carrier Driven Conductivity...... 106
5.4 Temporal Evolution of Optically Excited Free Carriers...... 112
5.5 THz Induced Transparency Via Intervalley Scattering...... 120 5.5.1 Time Domain Spectroscopy Investigation of THz Intensity Dependent Transmission...... 121 5.5.2 Power Transmission Detection of THz Induced Transparency 127
5.6 Induced Sheet Conductivity...... 133
5.7 Conclusion...... 135
6 Conclusion 137 TABLE OF CONTENTS (Continued) Page Appendix 139
A Derivation of Index of Refraction Vs Wavelength...... 140 LIST OF FIGURES Figure Page 1.1 The electromagnetic spectrum with the THz region highlighted in blue...... 1
2.2 A schematic of all of the internal reflections inside a bulk, lossy, dielectric...... 20
2.3 A schematic of all of the internal reflections inside a bulk, lossy, dielectric coated with a thin film...... 21
2.4 A plot a cavity where N=200 modes have a) random phase (inco- herent) b) uniform phase (coherent)...... 32
2.6 A cartoon plot of a nanoantenna array (left) with an SEM images of an actual nanoantenna (right)...... 44
3.1 Three time steps in the generation of the THz pulse via optical rectification...... 48
3.2 Cartoon setup of the LiNbO3 based THz generation. The dark lines within the optical pulse represent the the pulse front...... 50
3.3 The basic circuit diagram for the pyroelectric detector made by New England Photoconductor (NEP)...... 53
3.4 The schematic of a basic bolometer [36]...... 54
3.5 The laser setup used in the lab...... 56
3.6 A schematic of the experimental set up used in the lab...... 58
3.7 A schematic of the experimental for Power Dependent Transmission. 59
3.8 A schematic of the experimental set up for Michaelson interferometry. 60 LIST OF FIGURES (Continued) Figure Page 3.9 A schematic of the experimental set up for Time Domain Spectroscopy. 62
3.10 A cartoon figure of how the TDS detection works...... 63
4.1 A plot of density of states for different values of α [127]. T0 is the energy of the top of the valance band and I is the band gap..... 68
4.2 A) Energy plot for the unstrained (M1 state) and strained (R state) V 3d states. B) A representation of the V 3d|| states superimposed of a section of the VO2 crystal lattice...... 73
4.3 A cartoon picture of the VO2 sample. The top layer of gold and sapphire substrate are pictured sandwiching the 100 nm thick VO2 thin film. An incoming THz electric field is pictured to depict the which side of the sample was exposed to THz...... 74
4.4 An image of the mounted VO2 with the two peltier heaters on either side, the thermistor attached near the center, and the thermal paste and kapton tape attaching it to the mount...... 75
4.5 THz dependent relative transmission though NS coated VO2 at tem- peratures approaching the transition temperature...... 77
4.6 THz dependent relative transmission though bare VO2 at tempera- tures above and below the transition temperature...... 79
4.7 The transmitted THz spectra from bare coated VO2 for varying temperatures at different incident THz powers...... 81
4.8 The transmitted THz spectra from NS coated VO2 for varying tem- peratures at different incident THz powers...... 82
◦ 4.9 The transmitted THz spectra from NS coated VO2 at 64 C for two different incident THz powers...... 83 LIST OF FIGURES (Continued) Figure Page
4.10 The temperature dependent transmission through bare VO2 demon- strating the hysteretic behavior of the phase transition...... 84
4.11 The temperature dependent transmission through NS VO2 demon- strating the hysteretic behavior of the phase transition...... 86
4.12 The temperature dependent transmission, for sample temperature going from 50 degrees up to 80 degrees C, through NS VO2 demon- strating the hysteretic behavior of the phase transition...... 87
4.13 Plot of the narrowing of the hysteresis width as it varies with in- creasing THz...... 88
4.14 The normalized temperature dependent theoretical sheet conductiv- ity for high incident THz on bare VO2 compared with low incident THz on NS coated VO2...... 90
4.15 The semilog plot of the relative transmission vs both the calculated sheet conductivity based off the phenomenological model and the regression done with the bare sheet conductivity and the fitted sheet conductivity based on the thin film equations for the NS coated VO2. 92
4.16 The resistivity of the samples as a function of temperature for the NS coated VO2...... 93
4.17 Plots of the THz waveform transmitted through the bare VO2 sam- ple. This demonstrates the effects of only varying the temperature on the transmitted waveform...... 95
4.18 The time dependent transmitted THz wave through NS coated VO2 at a) 45 C, b) 62 C, c) 65 C, and d) 67 C on the increasing temper- ature side of the hysteresis curve...... 96
4.19 The transmitted THz field through NS coated VO2 comparing low THz at high temperature to High THz at low temperature...... 97 LIST OF FIGURES (Continued) Figure Page 5.2 A diagram of the intervalley scattering for GaAs as designed by Su et. al. [152]...... 102
5.3 A cartoon of the Nanoslot coated samples...... 105
5.4 THz waveform transmitted through the nanoantenna coated GaAs with τpp = −0.5ps and incident THz field of 321 kV/cm...... 107
5.5 THz waveform transmitted through bare GaAs with τpp = −0.5ps and incident THz field of 321 kV/cm...... 108
5.6 A plot of THz waveform transmitted through the nanoantenna coated Si sample at different incident pump powers. τpp = -3 ps and ET Hz = 320 kV/cm...... 109
5.7 A plot of THz waveform transmitted through bare Si sample at different incident pump powers. τpp = -1 ps and ET Hz = 334 kV/cm 110
5.8 The Fourier transform of the transmitted THz waveform when τpp = -3 ps...... 111
5.9 The transmitted THz waveform through nanoantenna coated GaAs −3 at different values of τpp for ET Hz = 99 kV/cm and Ne = 9.6e15 cm 113
5.10 The transmitted THz waveform through nanoantenna coated GaAs at different values of τpp for ET Hz = 348 kV/cm and Ne = 9.6e15 −3 cm . Not the THz wavefrom for τpp = −2, −3 ps has been magni- fied by a factor of 10...... 114
5.11 The differential transmission of the THz field through nanoantenna coated GaAs vs τpp ...... 115
5.12 The differential transmission of the THz field through nanoantenna coated GaAs vs τpp ...... 116 LIST OF FIGURES (Continued) Figure Page 5.13 The transmitted electric through the nanoantenna coated Si at dif- −3 ferent values of τpp for ET Hz = 340 kV/cm and Ne = 8.9e14cm .. 117
5.14 The transmitted electric through bare Si at different values of τpp −3 for ET Hz = 340 kV/cm and Ne = 8.9e14cm ...... 118
5.15 The differential transmission of the THz field through nanoantenna coated Si vs τpp ...... 119
5.16 The differential transmission of the THz field through nanoantenna coated Si vs τpp ...... 120
5.17 THz Waveform transmitted through nanoantenna coated GaAs at −3 different incident THz powers with Ne = 1.58e16cm and τpp is a) 1.5 ps, b) 0 ps, and c) -1.5 ps...... 122
5.18 THz Waveform transmitted through bare GaAs at different incident −3 −3 THz powers when τpp = −0.5 ps and Ne = a) 0 cm , b) 1.9e15 cm , c) 6.4e16 cm−3, d) 1.3 e16 cm−3, e) 1.8e16 cm−3...... 124
5.19 THz Waveform transmitted through nanoantenna coated Si at ET Hz = 33 kV/cm and 619 kV/cm. There is no optical pump for this ex- periment...... 125
5.20 THz Waveform transmitted through nanoantenna coated Si at ET Hz = −3 33 kV/cm and 619 kV/cm. τpp = −2 ps and Ne = 1.38cm ..... 125
5.21 THz Waveform transmitted through bare Si at different incident THz powers...... 126
5.22 Total relative power transmission through nanoantenna coated GaAs −3 −3 −3 for Ne = 6.4e15 cm , 1.6e16 cm , 2.5e16 cm and τpp = a) 3 ps, b) 1.5 ps, c) 0.5 ps, d) 0 ps, e) -0.5 ps, f) -1.5 ps, g) -3 ps, and h) -10 ps...... 128 LIST OF FIGURES (Continued) Figure Page
5.23 Total relative power transmission through bare GaAs for τpp = −3 ps and -1 ps...... 129
5.24 Total relative power transmission through bare GaAs at low incident pump powers such that the percent change in transmission due to excited carries is less that 25 % for τpp = a) -10 ps, b) -3 ps, and c) -1 ps...... 130
5.25 Total relative power transmission through nanoantenna coated Si −3 −3 −3 for Ne = 3.6e14 cm , 8.9e14 cm , 1.4e15 cm and τpp = a) 3 ps, b) 1.5 ps, c) 0.5 ps, d) 0 ps, e) -0.5 ps, f) -1.5 ps, g) -3 ps, and h) -10 ps...... 131
5.26 Total relative power transmission through bare Si for τpp = a) -3 ps, b) -1 ps, c) 1 ps)...... 132
5.27 The conductivity of GaAs vs incident THz fields at τpp = −10 ps and free carrier densities of 12.9, 31.8, and 51.1 ·1015 cm−3 ..... 134
5.28 The conductivity of Si vs incident THz fields at τpp = −10 ps and free carrier densities of 7.2, 28.8, and 28.7 ·1015 cm−3 ...... 135 1 Terahertz Radiation and Applications
Terahertz(THz) radiation is a band of the electromagnetic spectrum that ranges in frequency from 0.1·1012 Hz to 10·1012 Hz. It fills the gap between the far infrared and microwaves. Radiation of 1 THz has a wavelength of 300 µm which corresponds to an energy of 4.2 meV or 49 Kelvin. Due to this low energy THz radiation is an excellent tool for exploring the electron dynamics in many material systems. Its energy is well bellow the band gap of virtually all semiconductors, which allows for extensive analysis of intra-band electron dynamics without worrying about making excitations between the valence and conduction band. Its long wavelength also allows for it to pass through dielectric materials without much interaction.
Figure 1.1: The electromagnetic spectrum with the THz region highlighted in blue 2
1.1 History
THz radiation has been known about for over one hundred year. Long wavelength sources have been produced as far back as the turn of the 20th century [1]. In 1911 Ribens and Bayer showed that a mercury arc lamps with a quarts shell were able to reproduce long wave radiation reliably [2]. This opened the door to many new experiments that were unavailable before. However, it is only since the 1980s that coherent THz radiation was able to be reliably produced [3]. Using a free electron laser Elias, Hu, and Ramian were able to produce radiation with wavelengths between 390 and 1000 µm; the first time a stable coherent source of THz had been produced. Over the years other sources of THz radiation have been produced such as synchrotron radiation [4], air plasmas [5], and optical rectification [6].
1.2 Sources
As stated in section 1.1 historically there have been very few sources of THz and before the 1980s there were no sources that were either high powered or coher- ent. Since the 1980s a number of useful sources have been developed such as the free electron laser, synchrotron radiation, air plasma, photocurrent sources, and materials with non-centrosymmetric symmetry.
The first modern source of THz radiation that was developed was the free electron laser. The basic principle of a free electron laser is that a beam of free 3 electrons is accelerated down a path. The path is lined with a strong magnetic field that alternates direction every few centimeters. This has the effect of causing the electrons to oscillate or “wiggle” up and down. This oscillation causes the electrons to radiate at
2 eBwλw 1 + 2πm c λ = λ 0 (1.1) l w 2γ2
where λl is the wavelength of the emitted radiation, λw is the distance for the applied magnetic field to make one complete cycle, Bw is the average magnetic induction, and γ is the energy of the electron divided by its rest mass energy [7]. The radiation produced from this setup is incoherent. However, if a laser with a wavelength equal to λl is co-propagate with the electron beam then the electrons begin to emit in phase with the laser. It turns out that an an external laser source is not needed. The electron beam can be placed in a resonant cavity, which will produce its own coherent beam, allowing for the free electron laser to be tuned to virtually any wavelength.
THz synchrotron radiation is produced in a similar fashion, at least in the sense that it is depends on the acceleration of electrons to produce radiation. Unlike free electron lasers a synchrotron use orbital electron motion rather than a linear path with a wiggler. A synchrotron uses an electron storage ring, where electrons are kept in a closed loop, and moved in a circle via an external magnetic field.
The third source of THz radiation is laser generated air plasma. A high intensity 4 laser can be used to produce an air plasma. The resulting freed electrons can be accelerated either via a pondermotive force, due to a secondary applied laser [8], or by taking advantage of the oscillations produced by the ionization process itself [9].
Another important source of THz radiation takes advantage of photoconductive switches. The essence of this method of THz generation is a high power pulsed laser incident on a a semiconductor wafer, such as GaAs or InP, which has electrodes attached [11]. A bias is applied to the semiconductor synced to the laser pulse which, in conjunction with the incident laser pulse, can produce sub-picosecond THz pulses.
The final method that will be discussed is THz generation via optical recti- fication. Optical rectification is a second order, non-linear optical effect which mixes an incoming photon with a static electric field. Two types of crystals that are employed in the use of THz generation via optical rectification are ZnTe and
LiNbO3 [12]. As this is the primary method of THz generation which this work employs, a more in-depth explanation can be found in section 2.2
1.3 Detection
Due to it’s low energy THz radiation is extremely difficult to detect using standard photo current detectors, which rely on interband transitions. There are a number of other THz detection devices. Devices such as Golay cells, pyroelectric detectors, and bolometers have been developed as possible ways to detect THz radiation. 5
These types of detectors are often refereed to as power detectors, or incoherent detectors, as they only register the total power of the THz beam. Techniques such as electro-optic sampling allow for detection of the THz waveform, not just total integrated power. THz Time domain spectroscopy will be discussed in section 3.3.3.
The most accurate integrated power detector is the bolometer. The primary components of a bolometer is a thermally reactive resistive element that is attached to a material with well known thermal characteristics such as GaAs/AlGaAs [10] or Si. The base material is coupled to a low temperature bath to help reduce noise. A bias is applied to the resistive element and the resulting current is measured. When photons enter the detector their energy is deposited onto the base material and is transfered into thermal energy. This increase in temperature causes the resistive element to change its resistance, and thus the current. This is an extremely accurate system [13], but it usually requires liquid helium temperatures to run efficiently.
One alternative is a Golay cell. This functions based on the principles of the ideal gas law. A chamber that is filled with gas is separated from an optical ele- ment by a thin, flexible membrane. The thermal energy of the incoming radiation increases the temperature of the gas which causes an increase in pressure. This causes the membrane to shift, an effect which can be measured with high accu- racy [14]. Thus, if the change in the membrane is known the incoming energy can be calculated based on the particular details of the gas cell. 6
The final incoherent detector is the pyroelectric detector. Like the other de- tectors it operates on the conversion of photon energy into thermal energy. The main operating element of a pyroelectric detector is a crystal that undergoes spon- taneous polarization. This spontaneous polarization and the dielectric constant are dependent on the temperature of the crystal. Electrical plates, and an open circuit setup are employed so that when the crystal is heated, and it’s properties altered, a measurable current is produced [15].
1.4 Applications
While THz radiation is difficult to produce and detect, its properties lend them- selves to many useful applications. As THz radiation is extremely sensitive to the conductivity of a material [16] it is a useful tool for exploring the electrical properties of a material without having to attach electrodes to make physical con- tact with the sample. Thus THz radiation allows for non-contact non-destructive analysis of many interesting and novel materials and devices.
THz also has applications in medical imagining [17–20]. The energy of one THz photon is on the order of 4 meV. The energy splitting in many biological systems and materials is also on the order of 4 meV.
THz radiation is also very useful for security purposes. Over the last 15-20 years there has been a large increase in the need for security in high traffic areas such as airports, train stations, and ports. This has required the development 7 of devices such as the infamous x-ray scanner which is used to take a full body image of a person to look for any weapons or hidden items. However, as X-rays are ionizing radiation they are very hazardous to one’s health. THz radiation offers an alternative to X-rays, and will result in little to no damage to those being scanned [22,23]. THz radiation can also be employed in the detection of chemicals such as explosives or drugs [24,25]. 8
2 Theoretical Foundations for Terahertz Spectroscopy of
Semiconductors
The usefulness of THz radiation was laid out in Sec 1.4. In order to properly utilize THz radiation a through understanding of a number of concepts, such as electromagnetic radiation and ultrafast optics, must be understood.
2.1 Maxwell’s Equations and Electromagnetic Radiation
All electromagnetic radiation consists of coupled oscillations of electric and mag- netic fields. Therefor, before we can discuss any radiation a complete understand- ing of Electric and Magnetic fields must be developed. All Electric and Magnetic fields are governed by six main equations [26] 9
~ ~ ∇·D = ρf (2.1) ∂B~ ∇~ × E~ = − (2.2) ∂t ~ ~ ∇·B = 0 (2.3) ∂D~ ∇~ × H~ = J~ + (2.4) f ∂t ~ ~ ~ ~ D = 0 (1 + χE) E ≡ 0rE = E (2.5) ~ ~ ~ ~ B = µ0 (1 + χM ) H ≡ µ0µrH = µH (2.6)
where 0 and µ0 are the vacuum permittivity and permeability respectively, and χE and χM , in general, are tensors that are defined as
~ ~ P = 0χEE (2.7) ~ ~ M = µ0χM H (2.8)
Here, P~ and M~ are the microscopic polarization and magnetization of a mate- rial. For this work we will treat χE and χM as scaler operators as the materials that are being dealt with behave as such. 10
2.1.1 Electromagnetic Plane Waves
Applying the curl operator to equations 2.2 and 2.4, distributing the curl on the ~ ~ ~ ~ ~ ~ 2 ~ RHS, and applying the vector equality ∇ × ∇ × Q = ∇· ∇·Q − ∇ Q
! ∂∇~ × B~ ∇~ ∇~ E~ − ∇2E~ = − (2.9) · · ∂t ∂∇~ × D~ ∇~ ∇~ H~ − ∇2H~ = ∇~ × J~ + (2.10) · · f ∂t
By applying equations 2.1 and 2.5 to 2.9 and equations 2.2 and 2.6 to 2.10 yields
! ρ ∂ ∂E~ ∇~ f − ∇2E~ = − µJ~ + µ (2.11) · ∂t f ∂t ∂ ∂B~ ∇~ (0) − ∇2B~ = µ∇~ × J~ − µ (2.12) · f ∂t ∂t
Rearranging we get the equations 11
∂2E~ 1 ∂J~ ∇2E~ − µ = ∇~ ρ + µ f (2.13) ∂t2 f ∂t ∂2B~ ∇2B~ − µ = −µ∇~ × J~ (2.14) ∂t2 f
At this point we have something that looks like the inhomogeneous wave equa- tion. However, equations 2.13 and 2.14 can be simplified further if we make the assumption that the charges are uniformly distributed in space, ie that ∇ρf = 0. Also, if we assume a linear relationship between the electric field and the current density, also known as Ohm’s law [27], then the equations can be simplified further to be
∂2E~ ∂E~ ∇2E~ − µ = µσ (2.15) ∂t2 ∂t ∂2B~ ∂B~ ∇2B~ − µ = −µσ∇~ × E~ = µσ (2.16) ∂t2 ∂t
∂E~ ∂2E~ For non-matalic materials it is fair to assume that σ ∂t ∂t2 as the conduc- tivity of non-metalic materials is extremely low compared to the how rapidly the electric field is changing in time. The same can be said for the magnetic field. This means the first temporal derivative of the fields can be set to zero which yeilds the final equations 12
∂2E~ ∇2E~ − µ = 0 (2.17) ∂t2 ∂2B~ ∇2B~ − µ = 0 (2.18) ∂t2
This is the exact form of the wave equation and has a number of important
2 ~ 1 ∂2Q~ results. One is that the standard wave equation is ∇ Q − v2 ∂t2 = 0 which implies 1 that µ = v2 . Since there can be no polarization or magnetization in a vacuum, c 1 √ and v = , then c = √ and n = µrr. n µ00
Also, the general solution to the wave equation is known so these equations can be easily solved. It is not common to solve for the electric and magnetic fields themselves. We can return to equations 2.2 and 2.3 and derive a solution directly ~ ~ ~ from these equations. Since ∇·∇ × Q = 0 equation 2.3 is automatically satisfied if we assume
B~ = ∇~ × A~ (2.19)
where A~ is a generic vector field called the vector potential. Similarly, as ∇~ × ∇θ = 0 equation 2.2 is automatically solved if we let
˙ E~ = −∇φ − A~ (2.20) 13
Here φ is generic scaler function called the scaler potential. Equations 2.19 and 2.20 along with equations 2.5 and 2.6 can be plugged into equation 2.4 to produce
! ∂ ∂A~ ∇~ × ∇~ × A~ = µJ~ + µ −∇φ − (2.21) f ∂t ∂t
simplifying and rearranging give the equation
∂2A~ ∂φ µ − ∇~ 2A~ = µσE~ + ∇ −µ + ∇~ · A~ (2.22) ∂t2 ∂t
Again, if we make the assumption that the material is non-conductive then, for a reasonably sized electric field, µσE~ ≈ 0. The main reason for switch from the actual fields found in equations 2.17 and 2.18 to the potentials is that the potentials are gauge invariant. This allows us to produce new potentials without altering the ∂φ ~ associated fields. The gauge we will employ is the Lorentz gauge, µ ∂t − ∇A = 0. Applying these to equation 2.22 gives
∂2A~ µ − ∇~ 2A~ = 0 (2.23) ∂t2
which is another form of the wave equation and all of the same insights still hold. Weather we use equations 2.17 and 2.18 or equation 2.23 it is clear the solutions are simple sinusoidal functions. Focusing on equations 2.17 and 2.18 gives the solution of polarized plane waves of the form 14
~ ~ i(~k·~x−ωt) E (~x,t) = E0e (2.24)
~ ~ i(~k·~x−ωt) B (~x,t) = B0e (2.25)
~ ~ ~ ~ plugging equations 2.24 and 2.25 into Maxwell’s equations yields k·E0 = k·B0 = ~ ~ ~ 0 as well as k × E0αB0. This proves that the electric field and the magnetic field are perpendicular to each other and both are perpendicular to the wave vector ~k. Also, plugging equation 2.24 into equation 2.17 gives k2 − µω2 = 0 which shows the dispersion relationship for electromagnetic radiation to be
c ω = ~k (2.26) n which leads to c E~ = B~ (2.27) n
So far a non-conductive dielectric media has been assumed. However, if the material is metallic, that is to say that it is highly conductive, then the approxima- tion that lead to equations 2.17 and 2.18 are no longer valid. For a good conductor it is fair to say that σ ω which implies that the second time derivative of the electric and magnetic fields can be ignored in equation 2.15 and 2.16 instead of the first order. This leads to the dispersion relation 15
k2 ≈ iµσω (2.28)
Any complex number can be expressed as Q = |Q|eiφ where |Q| is the magni-
−1 Im(Q) 2 π tude of the complex number and φ = tan Re(Q) . For k the value of φ is 2 . Thus
r √ i π µσω k = µσωe 4 = (1 + i) (2.29) 2
Thus the real and imaginary parts of k are equal. Applying this result to the plane wave solution gives an oscillatory portion related to the real part of k and an exponentially decaying portion related to the imaginary portion. For a lossy
nω dielectric media k is expressed as c . As both ω and c are real, n must be a complex number which means
r krc σ nr = ni = = (2.30) ω 2ω0 16
2.1.2 Electromagnetic Waves at a Boundary
2.1.2.1 Bulk Dielectric Boundary Conditions
Now that we have shown that plane waves are a solution to Maxwell’s equations we can begin to apply these solutions to different situations. One of the most important case studies is what happens at a boundary interface between different materials. In terms of optics, the major difference between materials is the index √ of refraction which we showed can be expressed, in general, as n = rµr. Recall that µr = 1 + χM . As the magnetic susceptibility is very small compared to 1 [28] √ µr ≈ 1 and thus n ≈ r.
To understand how Electric and magnetic fields behave at the interface between two materials the standard pill box and loop can be applied, as in figure 2.1a. Assuming a normal incidence wave and taking equations 2.2 and 2.4 and integrating over the area of the loop, as the edges approach the interface, yields
−Ein + Eref + Etrans = 0 (2.31)
µ1Bin + µ1Bref − µ2Btrans = Kf (2.32) 17
(a) A cartoon of an a) pill box and b) (b) The vector picture of the incoming wave, the loop applied at the interface. reflected wave, and transmitted wave.
Assuming µ1 ≈ µ2 and no free charge, i.e. a non-magnetic dielectric, and plugging Equation 2.27 into 2.32 yields
n1Ein + n1Eref − n2Etrans = 0 (2.33)
By combining equations 2.31 and 2.33 yields the transmission and reflection coefficients
Etrans 2n1 t12 ≡ = (2.34) Ein n1 + n2 Eref n2 − n1 r12 ≡ = (2.35) Ein n1 + n2 18
It is important to note that t+r = 1. This process can also be done for oblique incidence waves and leads to effects such as the Brewster angle. The transmission and reflection coefficients above are just for the electric fields. For most laboratory settings, it is the intensity of the electromagnetic wave that is measured. The intensity of the electromagnetic wave is the magnitude of the Poynting vector [29]. The Poynting vector is defined as
S~ = E~ × H~ (2.36)
using equations 2.6, 2.27, that the electric field is perpendicular to the magnetic field, c2 = 1 , and the time average of a sinusoidal function is 1 leads to 0µ0 2
D E 1 2 hIi = S~ = cn E~ (2.37) 2 0
which gives the power transmission and reflection coefficients as
S n T ≡ trans = 2 |t|2 (2.38) Sin n1 S R ≡ ref = |r|2 (2.39) Sin
Again, we note that T +R = 1. For the power transmission and reflection coef- ficients there is a physical significance. The transmission plus reflection coefficients 19 must equal 1 in order for there to be energy conservation.
2.1.2.2 Thin Film Transmission and Reflection
So far bulk material has been assumed when working with transmission and re- flection coefficients, but interesting things happen when you assumes a thin film. In general, when measuring transmission through any material, the transmission detected will be more complicated than just applying the transmission coefficient. Consider a material of thickness L. Each interaction with an interface will have a reflected and transmitted element. This implies that there will be a Fabry-Perot effect. The total transmission can be calculated based on the this. Each round trip the light makes inside the material requires two reflections and by the time the light leaves the material there will have been a total of two transmissions. Thus a general form for the transmission of the electric field, after n round trips in the material, can be calculated
(n) i2φn iφ t = t12 r12r21e t21e (2.40)
2πnL where φ = λ To get the total transmission through the material, one must add all of the individual transmitted rays 20
∞ X (n) X i2φn iφ iφ X i2φn ttot = t = t12 r12r21e t21e = t12t21e r12r21e (2.41) n=0
Figure 2.2: A schematic of all of the internal reflections inside a bulk, lossy, dielec- tric.
As t12, t21, r12, and r21 < 1 the infinite sum is merely a geometric series which converges to.
2 iφ |t12| e ttot = (2.42) 2 i2φ 1 − |r12| e
This reasoning can also be expanded to thin films. A film is said to be thin if the wavelength of the light is much less than the apparent size of the material, 21
λ that is if nf d 10 . As most of the films dealt with in this work are approximately 100 nm and the wavelength for 1 THz is approximately 300 µm the thin film approximation is valid.
Figure 2.3: A schematic of all of the internal reflections inside a bulk, lossy, dielec- tric coated with a thin film.
When a thin film is applied, the total transmission stays mostly the same, as that of the bulk sample. However, for the thin film the second material is now the thin film and the bulk is the third.
(n) i2φn iφ t = t1f r1f rf3e tf3e (2.43)
which changes the total transmission to 22
iφ t1f tf3e ttot = i2φ (2.44) 1 + r1f rf3e
substituting equations 2.34 and 2.35 gives
2n1 2nf iφ n +n n +n e 4n n t = 1 f 3 f = 1 f tot n1−nf n3−nf −iφ iφ 1 − ei2φ e (n1 + nf )(n3 + nf ) + (n1 − nf )(n3 − nf ) e n1+nf n3+nf (2.45)
iφ ikx Equating the plain wave solution gives e = e |x=d and recalling that λ d implies that kd 1 and that e±iφ can be approximated as 1 ± iφ gives eequation 2.45
t13 (n1 + n3) ttot = (2.46) n1n3 n1 + n3 − iφnf 1 + 2 nf
For our materials, the thin film can be considered a good conductor which
n1n3 nω means that nf n1, n3 [30] and 2 1. Also we can equate φ to kd = d = nf c n2πf n2π c d = λ d. Using the relationship between the complex index of refraction of a 1 metal and its conductivity found in equation 2.30 and defining Z0 = = 376.7Ω c0 we get
t13 (n1 + n3) t13 (n1 + n3) ttot = = (2.47) n1 + n3 + Z0σd n1 + n3 + Z0σs 23
This is the thin film transmission coefficient where we have defined the sheet conductivity σs as σd. This is an incredibly useful tool for THz spectroscopy and thin film analysis as it allows for the computation of the sheet conductivity of a thin film only knowing the index of atmosphere and substrate and measuring the total transmission through the sample. This means that the conductivity, a purely material property, can be measured without any contact or destruction of the sample. A similar calculation can be done for the total reflection of the of the thin film. However, as no reflection measurements were taken that exercise will be left to the reader.
2.2 Non-linear Optics
In section 2.1, equation 2.5 stated that the microscopic polarization vector was linearly proportional to the applied electric field. This is actually just an ap- proximation for low strength electric fields. It is a good approximation for most applications [31]. However, when the applied electric field gets large enough, such as with a high fluence laser, then the polarization needs to be changed to
~ X ~ (n) X (n) ~ n P = P = 0 χ E (2.48)
where χ(n) is a rank n tensor that translates the applied electric field vector into the polarization vector. As each higher order term requires an increasingly 24 large electric field to have an appreciable effect one would expect that the χ(2) term would be the next dominate term. However, if the material in question has a centrosymmetric symmetry, then it can only have a symmetric response to an electric field. Thus, by symmetry, this type of material can only have odd terms in the Taylor expansion [32]. Thus, for any material with centrosymmetric symmetry the next highest order term will be χ(3). The χ(3) term gives rise to effects such as four wave mixing and self focusing [33].
There are two common ways to make use of the second order term, the term that the rest of this work will focus on. One is to have a material that has a noncentrosymmetric symmetry such as ZnTe. The other is to take advantage of an interface. Even if a material has centrosymmetric symmetry, at an interface there is no way for the response of an electron to be symmetric [34,35].
The second order Polarization term is written as
~ (2) (2) ~ 2 P = 20χ E (2.49)
2.2.1 Second Order Non-linear Effects
Despite its simple appearance equation 2.49 has some very important consequences. If we assume that the applied electric field is made of two co-propagating waves then we can write the electric field as 25
~ ~ iω1t ~ iω2t E = E1e + E2e + c.c (2.50)
2 ~ (2) (2) ~ iω1t ~ iω2t and applying that to equation 2.49 gives P = 0χ E1e + E2e + c.c . Expanding the electric field term gives
2 i2ω1t 2 i2ω2t i(ω1+ω2)t ∗ i(ω1−ω2)t 2 2 E1 e + E2 e + 2E1E2e + 2E1E2 e + c.c. + 2 |E1| + |E2| (2.51)
These terms can be group according to the frequency of the resulting wave.
2 i2ω1t 2 i2ω2t E1 e ,E2 e , c.c. are the second harmonic generation (SHG)terms.
i(ω1+ω2)t 2E1E2e , c.c. are the sum frequency generation(SFG) terms.
∗ i(ω1−ω2)t 2E1E2 e , c.c. are the difference frequency generation (DFG) terms.
2 2 |E1| , |E2| are the optical rectification(OR) terms.
The second order Polarization term can be written as a sum of the different frequency terms
~ (2) X ~ (2) (2) ~ ~ P = P (ωn) = 20χ (ω1, ω2; ωn) E1(ω1)E2(ω2) (2.52)
The polarization is often discussed in terms of this frequency dependent form, 26 not the full polarization, as often only one particular type of effect is being used, such as OR or DFG.
2.2.2 The χ(2) tensor
A compact way of writing equation 2.49 is
(2) X (2) Pi = 2 χijkE1jE2k (2.53) j,k
This makes it clear that χ(2) is a rank 3 tensor. In general it can be written as a matrix. It can also be simplified into what is called the ”d-matrix“. This is done by taking advantage of the symmetries in equation 2.53. The d matrix is defined as dil = χijk.
l= 1 2 3 4 5 6
j,k= x,x y,y z,z y,z;z,y x,z;z,x x,y;y,x
This means that the 18 elements in the d-matrix can represent the 27 elements that are in χ(2). Using the d-matrix, equation 2.53 can be written as the matrix equation 27
E2 x E2 (2) y Px d11 d12 d13 d14 d15 d16 2 Ez P (2) = d d d d d d (2.54) y 21 22 23 24 25 26 2EyEz (2) Pz d31 d32 d33 d34 d35 d36 2E E x z 2ExEy
While this looks complicated, the d-matrix can often be simplified for specific materials. By making symmetry arguments based on the crystal lattice structure many of the elements in the d-matrix can be set to 0 and others can be shown to be equal. For example, in LiNbO3, which has the space group 3m, the d-matrix is [36]
0 0 0 0 d15 −d22 dLiNbO3 = −d d 0 d 0 0 (2.55) 22 22 15 d15 d15 d33 0 0 0
This d-matrix for LiNbO3 is actually very important for THz generation and we will return to it later. 28
2.2.3 Phase Matching
So far, it has been assumed that the radiation produced from these non-linear processes is in phase with the radiation that generates it and no interference be- tween the two occurs. While it is possible for the incident beam and the resulting beam to maintain constructive interference over the distance of the crystal, the fact that the index of refractions depends on the wavelength of the light means that the assumption is not universally true. If we return to equation 2.17 and use the relation that D~ = E~ + 1 P~ then we get the equation 0
2 ~ 2 ~ 2 ~ 1 ∂ E 1 ∂ P ∇ E − 2 2 = 2 2 (2.56) c ∂t 0c ∂t
Assume a 1-D plane wave for the generated wave and define it as E3(x, t) =
i(k3x−ω3t) i(k1x−ω1t) A3e with the two incident waves being defined as E1(x, t) = A1e
(1) (2) (2) ±i(k2x−ω2t) ∼ and E2(x, t) = A2e . Using the fact that P3 = P3 + P3 where P3 =
(2) χ E1E2 gives
2 2 2 2 (2) ∂ E3 n3 ∂ E3 1 ∂ χ E1E2 2 − 2 2 = 2 2 (2.57) ∂x c ∂t 0c ∂t
Substituting the functional form for Ei into the above and assuming that Ai is only spatially dependent yields the equation 29
2 2 2 ∂ A3 ∂A3 n ω + i2k − k2A + 3 3 A ei(k3x−ω3t) = ∂x2 3 ∂x 3 3 c2 3 (ω ± ω )2 χ(2)A A 1 2 1 2 i[(k1±k2)x−(ω1±ω2)t] − 2 e (2.58) 0c
2 2 2 n3ω3 As k3 = c2 and ω3 = ω1 ± ω2, depending if the phenomenon is sum or difference frequency generation, and assuming a slowly varying envelope for A3, ie
2 that ∂ A3 i2k ∂A3 , equation 2.58 simplifies to ∂x2 3 ∂x
∂A i (ω )2 χ(2)A A 3 3 1 2 i(k1±k2−k3)x = 2 e (2.59) ∂x 2k30c
It is useful to define a new quantity ∆k ≡ k1 ± k2 − k3. If ∆k = 0 then
A3 monotonically increases as a function of x, but if ∆k 6= 0 then A3 must have some spacial oscillations which means there will be points of constructive and destructive interference. Thus, for the best results for harmonic generation of
ωini 1 radiation ∆k = 0. Using ki = c ∆k becomes c (ω1n1 ± ω2n2 − ω3n3). As we know that the index of refraction of a material, in principle, is wavelength dependent (see appendixA) the only way to satisfy ∆ k = 0 and have ω3 = ω1 ±ω2 is if
n1 = n2 = n3 (2.60) 30
This is the phase matching criteria for non-linear wave mixing with electromag- netic radiation. In order to produce any useful radiation from harmonic mixing the index of refraction at the wavelength of the generated light must be equal to the wavelength of the source(s). This condition is difficult to meet for many materials. However there are a number of techniques used to obtain phase matching. One is take advantage of a material that is birefringent. Depending on the difference in the indexes it is possible to excite a material with a given polarization that will allow for the the index of the incoming wave to match the index of the wave being produced.
2.3 Ultrafast Optics
Ultrafast optics and spectroscopy have opened the door to a huge number of excit- ing and important fields [37–40]. Many chemical and electrical interactions happen in less than one nanosecond. Thus, in order to learn about the temporal dynamics of these systems we need to use tools that shorter characteristic times. Out of this necessity ultrafast optics was born. Ultrafast optics is the technique which use a series of coherent pulses that are incredibly short, instead of a contentious wave (CW) laser.
While there are many methods for producing ultrafast laser pulses two will be discussed here. The first is Q-switching. Q-switching works by placing some gain medium inside a resonant chamber that, some how, has a loss mechanism that can 31 be turn on and off. This can be some sort of saturable absorber, a rotating mirror, or laser modulator such as a pockel cell. Initially the loss mechanism is turned on and the gain medium is pump externally to produce a a population inversion. Once the population inversion N(t), is sufficiently high the loss mechanism is turned off. This creates a cavity for the laser to resonate in, causing a massive build up of photons n(t). However, as the photons are produced by the excitation of the population inversion, the number of photons produced per second will drop dramatically due to the lose of the population inversion N(t). This interaction between the population inversion of the gain medium and the photons is what creates the ultra short pulse.
2.3.1 Mode Locking
Mode locking is the primary method that is used to generate short pulses in this work. ML works based on cavity modes and broad band laser generation. The most common gain medium employed for ML is Ti:sapphire, which is titanium doped with sapphire(Al2O3). The primary benefit of this gain medium is its range of wavelengths it can generate [41]. It has been shown that it can produce light at wavelengths between 660 and 986 nm. For a laser cavity that is 1 m there are over 106 different modes that can be supported in this range of wavelengths. The total electric field can be represented as 32
N X i(ωnt+φn) Ene (2.61) n=1
Normally, having this number of modes would be of little use. This is due to the random phase that each mode has, φn. However, if one aligns these phases so they are no longer random then the different modes would interfere constructively. A simulation of a cavity with only 200 modes is displayed in figure 2.4.
Figure 2.4: A plot a cavity where N=200 modes have a) random phase (incoherent) b) uniform phase (coherent).
There are number of ways to induce mode locking. One is called active mode 33 locking. This is where an element modulates the loss of the beam. The element is made to periodically switch back and forth form disrupting the beam to not. This produces an electric field of
E(t) = E0 cos(iωnt + φn) (1 − α(1 − cos (Ωt)) ⇒
E(t) = E0 (cos(iωnt)(1 − α) + α cos ((ω ± Ω)t + φn))) (2.62)
2c Thus, if Ω = ωn+1 − ωn = L , where L is the length of the cavity, then this would force all modes to have one uniform phase.
Another technique is Kerr lens mode locking. This takes advantage of the third order non-linear effect where the index of refraction is intensity dependent n = n0 +In1. A nonlinear material is placed next to a hard aperture and when the laser pulse strikes the nonlinear material, any portion of the beam that is mode locked will have a higher intensity, and thus experience a higher index, causing the ML section to be focused down more than the CW wave. The aperture can be designed so that the focused light will make it through but the diffuse light is blocked. Thus the random phases are filtered out and a uniform phase is built up. 34
2.3.2 Pulse Width and Dispersion
While ultrashort pulses are extremely useful they can be difficult to work with, mainly it is very hard to keep a characteristic temporal profile with an ultrafast pulse. The most common type of ultrafast pulse is the Gaussian pulse. It’s electric field can be represented as
Γt2+i(kx−ωt) E(t) = E0e (2.63)
Here Γ is a complex number that describes both pulse duration and the chirp of the pulse. A chirped pulse is a pulsed laser effect where a laser’s frequency can change with time. Ignoring the real part of Γ for a second, the imaginary part
2 of the exponent is φ = ω0t + Γit . As angular frequency is defined as the rate of
dφ change of the angle, we find that ω(t) = dt = ω0 + 2Γit. Thus the frequency of the laser increases or decreases with time, depending on the sign of Γi. This is called a chirp.
Lets consider the complex nature of Γ now. The spectral response can be calculated using a Fourier transform.
Z ∞ r 2 2 π −(ω−ω0) −Γt +iω0t iωt E(x = 0, ω) = E0e e dt = E0 e 4Γ (2.64) −∞ Γ
Thus the spectra of a Gaussian pulse is also a Gaussian, centered on the fre- 35
1 quency of the light pulse. The bandwidth of the laser pulse is directly related to Γ .
1 Γr−iΓi Recalling that Γ is a complex number Γ = Γr + iΓi it is better to write as 2 2 . Γ Γr+Γi Thus the intensity spectra I(ω) = |E(ω)|2 has a spectral bandwidth proportional
Γr to 2 2 . Γr+iΓi
Knowing the electric field in the frequency regime will also shed light on how a Gaussian pulse propagates through a dispersive media. If we know what the spectrum is at x = 0 we can propagate the electric field forward by E(x, ω) =
ikx nω E(x = 0, ω)e . As k = c if n is dependent on the frequency of the light then so must k. If we expand k around the central frequency we get
X 1 d(n)k 1 k(ω) = | (ω − ω )n ≈ k(ω ) + k0(ω )(ω − ω ) + k00(ω )(ω − ω )2 n! dω(n) ω=ω0 0 0 0 0 2 0 0 (2.65)
Using this expansion with the Fourier transform from the frequency regime back to the time domain gives
Z ∞ r 2 π −(ω−ω0) 0 1 00 2 i(k(ω0)+k (ω0)(ω−ω0)+ k (ω0)(ω−ω0) )x E(x, t) = E0 e 4Γ e 2 dω (2.66) −∞ Γ
0 letting ω = ω −ω0, combining terms, and completing the square gives the final result 36
00 Γ −2iΓ k x 2 − 0 0 (t−k0x) k(ω0) 1+(2Γ k00x) iω0 t− E(x, t)αe 0 e ω0 (2.67)
Making an analogy to equation 2.63 we find that phase velocity of the pulse is
ω 1 k , the group velocity of the pulse is k0 , and the pulse width Γ is now dependent on the distance traveled in the dispersive media
Γ − 2iΓ k00x Γ(x) = 0 0 (2.68) 00 2 1 + (2Γ0k x)
This means that the further a pulse travels into a dispersive media then the smaller Γ will be. This means that the pulse will be stretched out temporally. This is can be a major problem if an experiment requires extremely short pulses.
Also note that even if Γ0 is a real number the pulse will still have an imaginary component to the quadratic portion of its exponent after it passes through the media. This means that the dispersive media will chirp any Gaussian pulse.
2.4 Band Theory of Semiconductors
Virtually no atoms is in total isolation. While it is possible to do work with single atoms [42–44] it is not common. The vast majority of systems studied can be classified as some sort of bulk material. A material is considered bulk if the dimensions of the object are much greater than many of the characteristic lengths, such as the coherence length. This means that one needs to find a way to 37 characterize systems that are more than just single atoms.
2.4.1 The Simple 1-D Chain
Lets first start out with the most simple case; a series of hydrogen atoms arranged in a 1-D ring with N atoms. The first thing that must be done is to form the ˆ Hamiltonian of this system. Let Hs(x) be the the single atom Hamiltonian. As the system is periodic we can write the continuous variable x as xn = a · n where a is the inter-atomic spacing and n is the nth atom. Thus the total Hamiltonian is
N ˆ X ˆ H(x)tot = Hs(x − xn) (2.69) n=0
The first guess at the solution would be the solutions to the single hydrogen atom. We know these won’t be exact anymore due to the interactions between atoms, but they can give us a good start. Instead of just trying one of the hydrogen wave functions we will try a sum of all of them. This method is called the Linear Combination of Atomic Orbitals (LCAO)
X X Ψ(x) = bm(xn)ψ(x − xn)m = φm(xn) (2.70) n,m n,m
where ψ(x − xn)m is the m-th solution to the hydrogen atom centered at the nth lattice site. bm(xn) is the coefficient for the m-th orbital at the nth lattice 38
site. φm(xn) is some theoretical m-th full solution to the Hamiltonian at location xn. This can be simplified by using Bloch’s theorem which states that the wave function from different lattice sites can only differ by a phase factor [45]. Thus, by normalizing Ψ(x) and applying Bloch’s therm to equation 2.70 turns into
r 1 X Ψ(x) = eik(xn)ψ(x − x ) (2.71) N n m n,m
First we can determine something about k by applying the periodic boundary conditions of the 1-D chain. Thus φm(0) = φm(xN ) which implies that bm(0) =
ikxN bm(o)e which is only true if
q · 2π k = (2.72) aN
Where q = 1, 2, 3,...N-1. An expression for the energy of the system can be calculated by taking the expectation value of the Hamiltonian. For a first order approximation we will only consider the interaction of a single atomic orbital
X ˆ ik(xn−xn0 ) ˆ hΨ| H |Ψi ⇒ e hψ(x − xn0 )| H |ψ(x − xn)i = n,n0
X ik(xn−x 0 ) e n E hψ(x − xn0 )| ψ(x − xn)i (2.73) n,n0 39
The inner product on the RHS turns into a Kronecker delta δn,n0 . The LHS is slightly more complicated. The nearest neighbor approximation is needed to ˆ simplify the problem. Be letting hψ(x − xn±1)| H |ψ(x − xn)i = β, hψ(x − xn)| ˆ H |ψ(x − xn)i = α, and grouping like terms in the sums we get the relation
q2π E = eikaβ + α + e−ikaβ = α + 2β sin (ka) = α + 2β sin (2.74) N
This is the 1-D dispersion relation for a chain of hydrogen atoms only consid- ering a single state.
2.4.2 Beyond the 1-D Chain
In reality 3-D solids are much more complex. Not only are they shifting from a 1-D to a 3-D case but all of the atomic levels must be taken into account as well. More so, as only hydrogen is hydrogen there are no analytical forms for the atomic orbitals for other atoms. Thus other tools are required to compute the dispersion relations of different systems. Before the advent of modern computers a number of tools and approximations were developed to deal with the calculations of band structures [46–49]. Once computers had developed into what we think of them as now, a whole new world of calculations opened up [50–54]. Theories that before that had just been difficult at best to produce results with, such as 40 the Hartree-Fock method, became algorithms a computer could execute with ease. This allowed for a far more detailed analysis of band structures than was available with pen and paper methods.
When all of the terms neglected in the 1-D chain approximation are taken into account, a complete picture of the dispersion relationship for a material can be formed. This is often refereed to as the band structure of the material and can be seen in figures 2.5a and 2.5b. This is significant as many important properties of a material can be understood using the band structure, such as how energetic a photon needs to be to excite free electrons in a system and how good a material is at absorbing said light, ie direct vs indirect band gaps.
(a) A plot of electron energy vs momen- (b) A plot of electron energy vs momen- tum for GaAs [147] tum for Si [147]
However, these calculations can be misleading. As discussed in section 4.1 in some of these materials the approximations discussed in section 2.4.1 are invalid. When things such as electron electron correlations are neglected, which is often a 41 valid approximation to make, the results are sometimes incomplete, as is the case with Mott Insulators [55,56].
2.5 THz Field Enhancement by Nanoantennas
Under standard conditions light cannot pass through an aperture that is less than the size of its wavelength [57]. It was shown that the transmission through a small aperture goes as
r 4 T α (2.75) λ
Where λ is the wavelength of the light and r is size of the aperture. From this,
r it is clear that as λ drops below 1, the transmission through the aperture drops abruptly. However, in 1998 it was discovered that subwavelength apertures allow high transmission and enhance the electric field in the near zone, if a resonate condition is satisfied [58]. In recent years sub-wavelength apertures have gained importance [59–67]. In this work the subwavelength aperture that will be employed is the nanoantenna array pictured in figure 2.6 42
2.5.1 Nanoantenna Enhanced Electric Fields for Long Wavelength
Radiation
Antenna theory and applications have a wide verity of uses [70–72]. A recent application is for field enhancement of free space, long wavelength radiation. The design implemented in this work is a 2D array of slots cut out of a 100 nm think layer of gold that is deposited on top of the sample. Via Babinet’s principle these slots act very smiler to linear antenna of the same dimensions and can be considered bound charge oscillators (BCOs) [73].
Choe at al. show that for a slot, cut out of a film of metal, of dimensions a by b, where a b, when light polarized perpendicular to side b is incident in the slot the component of the normalized pointing vector pointing into the slot is
32Re [W ] SN = s (2.76) z 2 2 π [Wa + Ws] where Ws/a corresponds to εsample/air and is equal to
ab Z ∞ Z ∞ ε k2 − k2 W = s/a 0 x s/a 2 q 8π −∞ −∞ 2 2 2 k0 εs/ak0 − kx − ky bk π + ak π − ak 2 sinc2 y sinc x + sinc x dk dk (2.77) 2 2 2 x y 43
If ba< λ this can be simplified to
3/2 32ε ab 2 2 2 2 s/a ib λ π b λ Ws/a ≈ + εs/a − ln εs/a − + 2γ − 3 (2.78) 3πλ2 λ 4a2 λ2 4a2 where γ ≈ .5777 is the Euler-Gamma constant. it can be shown that resonance will occur when Im[Wa + Ws] = 0 [74]. This leads to the resonance condition for the nanoslot which is
p 2 λres = 2(n + 1)a (2.79)
For Si, which has an index of nSi = 3.418, and a = 60 µm, λres ≈ 300µm which is right at 1 THz. When radiation of the proper wavelength is applied to these systems it has been shown to generate large field enhancements [74]. 44
2.5.2 Nanoantenna Arrays
Figure 2.6: A cartoon plot of a nanoantenna array (left) with an SEM images of an actual nanoantenna (right)
A nanoantenna array is a periodic array of slits etched out of a metal film that has been layered on top of a substrate. The primary characteristics are the hight (h) and width (w) of the nanoslot(NS), as well as its the slit spacing in the x and y
(dx and dy) directions. Each slot acts as an antenna.
A number of important quantities have been derived for the nanoantena array such as the charge distribution between slots. The surface charge distribution, σ, for a setup like figure 2.6 is
r ε0E0 λ iωt − iπ σ(x, t) = √ e e 4 (2.80) 2π x
Where E0 is the applied electric field, ω is the angular frequency of the light, and x is the distance from the slit edge. It is interesting to note the phase factor at the 45 end of equation 2.80. This implies that the emitted radiation will be out of phase with the incident light.
Possibly the most important quantity that can be calculated is the field en- hancement γ of the incident radiation [76]
α(ω) Etrans γ = = Einc (2.81) β hw dxdy
Using this formula, it is possible to calculate the strength of the electric field inside the gap of the antenna. 46
3 Terahertz Generation and Detection
With the essential theory laid out in Chapter2 it is important to discuss THz radiation itself, how its is generated, how it is detected, and how THz spectroscopy is implemented in a lab setting.
3.1 THz Generation
A wide range of setups and techniques have been used to generate THz radiation such as organic molecules [77], organic crystals such as 4-Dimethylamino-N-methyl- 4-stilbazolium Tosylate (DAST) [84–86], semiconductors such as GaAs [78–80], photocurrent and antenna systems [11, 81], and photoionization of gases [82, 83]. The main method that will be discussed in this section is THz generation via optical rectification in non-centrosymmetric systems, mainly ZnTe in section 3.1.1 and LiNbO3 in section 3.1.2.
3.1.1 ZnTe and Optical Rectification
ZnTe is an inorganic crystal structure with a cubic lattice space group of 43m¯ [28]. It has a lattice spacing of 0.61020 nm. One of its most important properties it’s 47 index of refraction as it allows for simple a means of THz generation via optical rectification [87–90].
Optical rectification, as discussed in section 2.2.2, is a second order non-linear effect. As such, it is based on the d-matrix shown in equation 2.54. Due to the high symmetry of the 43m¯ spacegroup the d-matrix for ZnTe can be simplified to [91]
0 0 0 d14 0 0 0 0 0 0 d 0 (3.1) 14 0 0 0 0 0 d14
pm The d14 element for ZnTe is 4.2 V [99]. ZnTe, with the laser system described kV in section 3.3.1, will produce a maximum field of approximately 100 cm when focused to a beam waist of 400 µm.
THz generation via optical rectification may seem odd at first glance, using a static field to create a time dependent field. However, it is important to remember that the optical excitation laser is not a continuous wave. As it has a finite pulse duration there is a formation of the static electric field, and then a decay. Thus, even though the produced field itself has no frequency component, the temporal limits of the laser pulse drive a time dependent polarization in the ZnTe crystal and this creates the THz field. This is depicted in figure 3.1 48
Figure 3.1: Three time steps in the generation of the THz pulse via optical recti- fication
3.1.2 LiNbO3 THz Generation
ZnTe is a very useful tool for generating THz as it is very easy to phase match with due to the closeness of the index of refraction for each wavelength [92, 93]. However, the intensity of the THz radiation emitted from ZnTe is fairly low. In order to produce high field THz radiation a different non-linear crystal is needed. 49
LiNbO3 is an excellent crystal for generating high field THz radiation and has a number of useful properties [94–98]. LiNbO3 is a 3m crystal configuration [91] which gives it a d-matrix of [36]
0 0 0 0 d15 −d22 −d d 0 d 0 0 (3.2) 22 22 15 d15 d15 d33 0 0 0
pm It has a d33 value of 27 V [98]. This makes it highly efferent at generating
THz radiation [100–103]. However, a major downside for LiNbO3 is the fact that the index of refraction for 800 nm is vastly different than for 300 µm. This makes standard phase matching extremely difficult. A solution to this issue is to tilt the pulse front of the optical beam and take advantage of the cherenkov radiation created by the velocity mismatch caused by the 800 nm pulse, which can be thought of as a particle, travailing faster than the local speed of light for the THz [104,105].
In order to take advantage of this, the Pulse front of the optical excitation beam must be tilted to the same angle as the cherenkov cone. The set up for THz generation via optical rectification in a LiNbO3 crystal uses a blazed grating to tilt the pulse front. A pair of lens (either spherical or cylindrical will work, each having its advantage and disadvantage) are employed to focus the tilted pulse. The first
first lens, with focal lengths f1, is set f1 distance away from the diffraction grating.
The second lens, with focal length f2, is set f1 + f2 away from the first lens, creating a telescope. The optical pulse, which now has been magnified by f1 , irradiates the f2 50
LiNbO3 crystal and produces THz parallel to the tilted pulse front. This setup is picture in figure 3.2
Figure 3.2: Cartoon setup of the LiNbO3 based THz generation. The dark lines within the optical pulse represent the the pulse front.
The angle of the pulse front tilt can can be mathematically expressed as vph,opt cos (γ) = vgr,T Hz. The Pulse front tilt angle, γ is equal to [106]
mλ p tan (γ) = 0 (3.3) ngrβ1cos (θd)
Here m is the diffraction order, λ0 is the central wavelength of the optical
(excitation) pulse, p is density of lines in the grating, ngr is the group index of refraction, β1 is the horizontal magnification of the optical pulse, and θd is the diffraction angle. The last variable, the diffraction angle, is the most important for the setup as it is the actual angle we will need to use when aligning the LiNbO3.
In order for the highest efficiencies of the LiNbO3 the pulse tilt γ needs to be equal to the tilt of the grating image inside LiNbO3 crystal. To ensure this, the tilt angle θ is used and is defined as 51
tan (θ) = nphβ2 tan (θd) (3.4)
here n is the refractive index of LiNbO3 for the phase velocity of the optical pulse and β2 is equal to the horizontal magnification of the lens inside the LiNbO3. Using these equations the set up can be optimized to generate THz radiation in
MV excess of 1 cm . One byproduct of this process is even though the generating pulse is on the order of 100 fs, the THz pulse is on the order of 1 ps.
3.2 THz Detection
As section 1.3 discusses, THz radiation is particularly difficult to detect, mainly due to the low energy of their photons; approximately 4.1 meV. However, over the last century and a half there are a large number of different types of directors that have been developed. All of these detectors are based around converting the photon energy into thermal energy.
A number of different detectors were developed to detect the ”new” long wave radiation. Long wave radiation ranges from the far infrared to the THz range, ie about 15 µ m to about 1 mm [107]. One detector developed for this type of range of wave lengths is the Pyrgeometer. A Pygeometer works using a hemispherical filter which blocks all radiation below 3.5 µm. A thermophile is placed inside and coated with a black paint which is responsive between 3 to 50 µm. The system 52 is connected to a circuit system such that when the thermophile is irradiated it will alter the voltage in the system [108,109]. Another type of detector is called a Golay detector. A Golay detector works based off of the thermal expansion of a gas against a flexible membrane with an attached mirror [110–112]. As the THz radiation heats the gas it expands, causing the flexible film to flex and altering the aliment of the attached mirror.
These detectors, and others, are good ways to detect THz radiation. However, the detectors that will be focused on in this work will be the Pyroelectric detector and the Bolometer. Both are integrating power detectors.
3.2.1 Pyroelectric Detectors
Pyroelectric detectors work off the principle that certain types of materials will generate a voltage when heated or cooled [113–115]. Crystals can also exhibit this type of property [116]. By taking advantage of this it is possible to build a THz detector based off of this material. By connecting a pyroelectric crystal to a circuit it is possible to measure the change in voltage of the crystal due to the thermal heating caused by the incident THz radiation. 53
Figure 3.3: The basic circuit diagram for the pyroelectric detector made by New England Photoconductor (NEP).
3.2.2 The Bolometer
While the pyroelectric detector is convenient in that it functions at room temper- ature, the primary integrating power detector used in this work is the bolometer. The bolometer has a responsivity that is a factor of approximately 103 times as large as the pyroelectric, making it a much more sensitive device. Originally used as a long range thermal bovine detector [117] bolometers are used to detect THz radiation.
A bolometer uses the conversion of electromagnetic energy into thermal energy to detect incoming THz radiation. However, it uses change in resistance to detect THz radiation. A bolometer uses a thermal absorber, such as a large piece of Si, which is coupled to a cold bath such as a reservoir of liq nitrogen or helium. A ther- 54 mistor is attached to the thermal absorber to measure the changes in temperature caused by the incoming THz radiation. The reason for the cold bath is to reduce thermal noise and fluctuations. As THz radiation is on the order of 4 meV and the thermal energy of a room temperature object, 300 K, is approximately 40 meV any signal that would be generated by the incoming THz radiation would almost certainly be lost within the noise of the thermal fluctuations. Thus by lowering the temperature of the system to 4.1 k with a liquid helium bath, the thermal energy can be lowered to around 0.35 meV. Even liquid nitrogen temperatures would be insufficient as the thermal energy would be approximately 6.5 meV. This is on the order of the energy of the THz photons and is thus too large.
Figure 3.4: The schematic of a basic bolometer [36]. 55
3.3 Experimental Setup and Procedures
With the methods for producing and detecting THz radiation established it is important now to discuss the physical setup of the system. The laser system is a four stage set-up ending with a 800 nm laser pulse with a temporal width of 130 fs and a pulse energy of 1 mJ. The beam is split into a pump beam and a THz beam. The THz beam is split again into a TDS probe beam and a final THz beam. The
THz beam is sent through the LiNbO3 setup described in section 3.1.2. The THz beam is focused onto the sample and the transmitted beam is collected by various detection rigs, depending on the experiment.
3.3.1 The Laser System
The laser system consists of four main parts. The first is a 10 W diode laser centered at 532 nm called the Verdi. The laser from the Verdi is sent into the second stage; the Mira. The Mira uses a Ti:Sapphire crystal as a gain medium and produces the 100 fs pulses via mode locking; as discussed in section 2.3.1. The output of the Mira is centered at 800 nm and has a bandwidth of 10-12 nm. This is sent to the final stage; the Legend. The Legend takes the seed pulse from the Mira and, by using another external pump, amplifies the pulses to 1 mJ of energy per pulse. The output pulsed laser has a repetition rate of 1 kHz and each pulse is 130 fs in duration giving a duty cycle of approximately 10−12. This extremely small duty cycle is vitally important to the experiments that will be discussed in 56 section 3.3.3.
Figure 3.5: The laser setup used in the lab.
3.3.2 Layout of Experimental Setup
The output of the laser system in figure 3.5 passes through a 95/5 beam splitter. The 5 % is used as an optical excitation beam with a variable attenuator to control the fluence of the pump beam. The other 95% passes through a second 95/5 beam splitter. The 5 % from the second beam splitter is taken off to used as the THz Time Domain Spectroscopy (THz-TDS) probe beam, which will be discussed in more depth in chapter 3.3.3.3. What is now roughly 90 % of the original laser passes into the setup described in section 3.1.2 to generate the THz radiation. The 57
THz is then focused down to minimum beam waist of approximately 450 µm. The sample is inserted into a 3 axis control stage and moved to the location of the beam waist minimum.
The optical pump line is then aligned to the center of the THz beam so the two are spatially overlapped and co-propagating. Then a wafer of GaAs is used, along with a delay stage set into the THz path, to find the temporal overlap of the optical pump beam and the THz beam. After the sample the THz pulse is once again focused down onto a THz detector. For the experiments in this work three main detection arrangements were implemented; total integrated power transmission, michaelson interferometry, and time domain spectroscopy. 58
Figure 3.6: A schematic of the experimental set up used in the lab.
3.3.3 Experimental Designs
The setup illustrated in figure 3.6 can be used to implement a number of different experiments; everything from simple transmitted power detection to wave form capture using THz time domain spectroscopy. 59
3.3.3.1 Power Dependent Transmission
Power dependent transmission is the most straight forward experiment to imple- ment. Either a pyroelectric detector or a Bolometer can be used for these experi- ments. However, the Bolometer is much more accurate and reliable so it was used for all main detection schemes. To do power dependent transmission measure- ments the Bolometer is simply placed at the second focus of the THz beam, after the sample, and is optimized to ensure that all light is being captured.
Figure 3.7: A schematic of the experimental for Power Dependent Transmission.
3.3.3.2 Michaelson Interferometry
The second type of experiment is Michaelson interferometry. Smiler to power dependent transmission, Michaelson interferometry use the Bolometer. However, in this set up a Si wafer is used as a 50/50 beam splitter for the THz beam transmitted through the sample. One of of the beam travels along a fixed path while the other is reflected off a mirror that is on a translational stage. The two 60
THz split THz beams are recombined and sent into the bolometer. Using the translational stage the relative path length between the two legs can be altered, this creating an interferogram.
The benefit of this set up is it allows the analysis of the spectral response of the sample; something straight power measurements won’t. However, as THz is readily absorbed by ambient water vapor the area around the THz generation and sample must be purged of all air. This is done by placing an acrylic purge box around the area and filling it with nitrogen gas.
Figure 3.8: A schematic of the experimental set up for Michaelson interferometry.
3.3.3.3 THz Time Domain Spectroscopy
The final experiment is THz time domain spectroscopy (THz-TDS). Unlike Michael- son interferometry or power dependent transmission THz-TDS does not use the bolometer. THz-TDS takes advantage of the pockels effect, a second order non- linear effect [118]. As discussed in 2.2 the second order polarization of a material 61 can be described as.
~ (2) X (2) ~ ~ Pi = 20 χijk (ω1, ω2; ωn) E1j(ω1)E2k(ω2) (3.5) jk
As the temporal width of the THz pulse is much longer than that of the 800 nm pulse and the frequency is much lower, a factor of 103, the THz pulse can be treated as a static electric field. This can be simplified to
~ (2) X (2) ~ Pi = 20 χij (ω1, ω2; ωn) E1j(ω1) (3.6) jk where (2) X (2) ~ χij = χijk (ω1, 0; ωn) E2k(0) (3.7) k This shows that the polarization of the material is dependent on applied static field. Thus our THz field, which is quasi static, is able to induce a birefringence in a non-linear crystal, such as GaAs. The set up for TDS can be seen in figure 3.9 62
Figure 3.9: A schematic of the experimental set up for Time Domain Spectroscopy.
Focusing on the detection part of the TDS set up there are three main optical elements; the detection crystal where the birefringence is induced, a quarter wave plate, and a wollaston prism. The wollaston prism separates the two polarizations of the TDS probe beam which are able to be picked up by the final element, a balanced photo diode. The system is aligned such that when there is no THz beam the intensity on each photo diode, Ix and Iy, are equal and the output signal is zero. However, as the THz induces a birefringence the circularly polarized optical pump beam becomes elliptically polarized and Ix is no longer equal to Iy and a signal is produced. 63
Figure 3.10: A cartoon figure of how the TDS detection works.
The time delay between the TDS probe and the THz can be varied allowing for the entire TDS wave form to be mapped out. However, in order to truly understand the output it is necessary to connect how the THz radiation alters the birefringence and thus how the output of the photo diode relates to the THz waveform. As the GaAs can be considered a lossless medium the pockels effect uses the same d-matrix element as optical rectification which for ZnTe is the d14 element. It can be shown that the change in difference in the phases for the x and y polarizations is [36]
Lω Lω ∆φ = (n − n ) = n3 r E (3.8) x y c c O 14 T Hz 64
where nO is the index of refraction of the optical (800 nm) beam. The important thing to note is that the change in angle is linearly dependent on the applied THz
field. With the change in phase it is possible to determine the change in Ix/y
I I = I (1 − (+)sin (∆φ)) = 0 (1 − (+)sin (∆φ)) (3.9) x(y) 0,x(y) 2
if the change in the phase is small sin(∆φ) ≈ ∆φ and the difference in in the intensities is
2Lω I − I ≈ 2∆φ = n3 r E (3.10) y x c O 14 T Hz thus the response from the balanced photo diode will be linearly proportional to the applied THz field. This lets us easily detect and map out the transmitted THz waveform. This is a very powerful tool as it allows for not only the characterization of the frequency response but, unlike Michaelson interferometer, TDS preserves the phase information of the THz pulse, as well as the frequency response, leading to a wealth of information lost in other experiments. 65
4 Vanadium Dioxide Field Induced Transition
Vanadium dioxide, VO2, is an incredibly interesting material and has been widely studied for many years [119–122]. VO2 is an insulator at room temperature. How- ever, above 340 K, or 67◦ C, the crystal structure changes from a monoclinic to a rutile structure [123]. This transition has the effect of changing the the VO2 from being an insulator to a conductor [124]. This insulator metal transition(IMT) is caused by changes in the crystal structure which alters the electron correlations which, in turn, changes whether it is energetically favorable to have a free electron gas [125]. This type of transition is a called a Mott transition and forms a class of materials called Mott insulators.
4.1 Mott Insulators
4.1.1 The Hubbard model
The theoretical foundation of Mott Insulators is based on the work of the Hubbard model [126, 127].One of his main contributions was a new approximation for the Hamilton of a many particle systems; the Hubbard model. This model, which accounts for electron-electron(e-e) interactions, is expressed as 66
ˆ X † X † † ˆ ˆ Hhub = −t ai,σaj,σ + U ai,↓ai,↓ai,↑ai,↑ = H0 + Vee (4.1) ,σ i
(†) < i, j > denotes the nearest neighbor approximation, aq,σ is the lowering(raising) operator for the electron at the qth lattice site with spin σ. U and t are defined as
1 X k2 t ≡ t = eik·a (4.2) N 2m k ZZ U = ψ∗ (r)ψ (r)V (r − r‘)ψ∗ (r‘)ψ (r‘)drdr‘ (4.3) Ri Ri ee Ri Ri
N is the number of lattice sites, a is the lattice vector from lattice point Ri to