ABSTRACT Title of dissertation: NOVEL APPLICATIONS OF HIGH INTENSITY FEMTOSECOND LASERS TO PARTICLE ACCELERATION AND TERAHERTZ GENERATION Andrew G. York, Doctor of Philosophy, 2008 Dissertation directed by: Professor Howard Milchberg Institute for Research in Electronics and Applied Physics We have investigated new applications for high intensity femtosecond lasers theoretically and experimentally, including a novel method to accelerate electrons to relativistic energy and a new type of coherent lasing medium for amplification of few-cycle, high energy pulses of terahertz radiation. We report the development of corrugated ?slow wave? plasma guiding struc- tures with application to quasi-phase-matched direct laser acceleration of charged particles. These structures support guided propagation at intensities up to 2 ? 1017 W/cm2, limited by our current laser energy and side leakage. Hydrogen, ni- trogen, and argon plasma waveguides up to 1.5 cm in length with corrugation pe- riod as short as 35 ?m are generated in extended cryogenic cluster jet flows, with corrugation depth approaching 100%. These structures remove the limitations of diffraction, phase matching, and material damage thresholds and promise to allow high-field acceleration of electrons over many centimeters using relatively small fem- tosecond lasers. We present simulations that show a laser pulse power of 1.9 TW should allow an acceleration gradient larger than 80 MV/cm. A modest power of only 30 GW would still allow acceleration gradients in excess of 10 MV/cm. Broadband chirped-pulse amplification (CPA) in Ti:sapphire revolutionized nonlinear optics in the 90?s, bringing intense optical pulses out of large govern- ment facilities and into the hands of graduate students in small university labs. Intense terahertz pulses (? 10 ?J, < 5 cycles), however, are still only produced at large accelerator facilities like Brookhaven National Labs. CPA is theoretically possible for terahertz frequencies, but no broadband lasing medium like Ti:sapphire has been demonstrated for terahertz. Dipolar molecular gases such as hydrogen cyanide (HCN) or nitrous oxide (N2O), ?aligned? or rotationally excited by intense optical pulses, are a novel and promising medium for amplification of broadband few-cycle terahertz pulses. We present calculations that show rotationally excited molecules can amplify a few-cycle seed pulse of terahertz radiation: a short, intense optical pulse (or sequence of pulses) aligns a dipolar gas (such as HCN), driving the molecules into a broad superposition of excited rotational states. A broadband seed terahertz pulse following the optical pulses can then be amplified on many pure rotational transitions simultaneously. We also discuss plans and progress towards experimental realization of a few-cycle terahertz amplifier. NOVEL APPLICATIONS OF HIGH INTENSITY FEMTOSECOND LASERS TO PARTICLE ACCELERATION AND TERAHERTZ GENERATION by Andrew G. York Dissertation submitted to the Faculty of the Graduate School of the University of Maryland, College Park in partial fulfillment of the requirements for the degree of Doctor of Philosophy 2008 Advisory Committee: Professor Howard M. Milchberg, Chair/Advisor Professor William Dorland Professor Julius Goldhar Professor Wendell T. Hill III Professor Steven Rolston c? Copyright by Andrew G. York 2008 Dedication To Kelli, naturally. Hey Kelli, let me know if you ever notice that I dedicated my thesis to you. ii Acknowledgments First and foremost, this thesis obviously wouldn?t be possible without Howard Milchberg. I?ve learned more in Howard?s lab than any other time in my life. Of course, this learning wouldn?t have been nearly as fun without Sanjay Varma to share the ups and the downs. My wife Kelli has listened to far more physics than she deserves, and that goes double for my old roommate Ben Cooper, but I think he likes it more. John Palastro, Brian Layer, and Yu-Hsin Chen were amazing colleagues and made so much possible that I could never have done alone, and Matt Aubuchon gives me hope for our lab?s future creativity and intelligence. Evan Merkel, Mike Harrington, Chris Pesto, and Darryl Waddy all showed me how much more I could have done with my undergrad years. Thomas Antonsen never locked me and my incessant questions out of his office for some reason, and helped guide my questions in more productive directions more than once. Tom Murphy let me ?borrow? an out-of-print textbook for more than a year, and helped correct several of my grosser misconceptions about fiber optics. Yongzhang Leng?s fabrication work made the corrugated waveguide possible, and his kindness and patience made it easier to keep bothering him. Edd Cole taught me everything I know about machining, and tells some pretty cool stories too. Jane Hessing saved me from my own mistakes more times than I can count, retroactively. Finally, I think this thesis owes a lot to my dad flipping over restaurant placemats to show me sine waves, my mom?s love of a good puzzle, and my brother being so much better at art that I had to pursue science. iii Table of Contents List of Tables vi List of Figures vii 0 Introduction 1 0.1 High intensity lasers . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 0.2 Some of applications of high intensity femtosecond lasers . . . . . . . 2 0.3 Outline of thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1 Review of laser-driven particle acceleration 6 1.1 Direct acceleration by electromagnetic waves . . . . . . . . . . . . . . 6 1.2 Wakefield acceleration . . . . . . . . . . . . . . . . . . . . . . . . . . 8 1.3 Proposed acceleration scheme . . . . . . . . . . . . . . . . . . . . . . 10 2 Direct laser acceleration in plasma 14 2.1 Toy model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 2.2 Finite difference time domain model: leakage, group velocity, dispersion 19 2.3 Analytic model, scaling law . . . . . . . . . . . . . . . . . . . . . . . 22 2.4 Transverse dynamics . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 2.5 Test particles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29 2.6 Remaining challenges: radial polarization, electron injection . . . . . 30 3 The corrugated plasma waveguide 32 3.1 Generation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 3.2 Plasma density measurement: phase extraction, Abel inversion . . . . 42 4 Review of pulse-driven molecular alignment 47 4.1 Laser-driven alignment, multishot measurement techniques . . . . . . 47 4.2 THz-driven alignment . . . . . . . . . . . . . . . . . . . . . . . . . . 49 4.3 Thepossibility ofterahertz emission from laser-driven molecular align- ment . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 5 Molecular alignment simulation 53 5.1 The rigid rotor model of molecular rotation . . . . . . . . . . . . . . . 53 5.2 Statistical description using a density matrix . . . . . . . . . . . . . . 56 5.3 Details of the simulation algorithm . . . . . . . . . . . . . . . . . . . 60 6 Using optical pulses to drive population inversion in aligned molecules 65 6.1 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 6.2 Low pressure, room temperature . . . . . . . . . . . . . . . . . . . . . 67 6.3 Moderate pressure, room temperature: the effects of dissipation . . . 68 6.4 Low temperature gas jet targets . . . . . . . . . . . . . . . . . . . . . 70 6.5 Overpumping . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72 iv 7 Using inverted rotational populations to amplify THz pulses 74 7.1 Amplification bandwidth, THz pulse duration . . . . . . . . . . . . . 74 7.2 Gain at lower temperature . . . . . . . . . . . . . . . . . . . . . . . . 80 8 Coherent terahertz absorption in laser-aligned molecules 82 8.1 Suppression or enhancement of absorption . . . . . . . . . . . . . . . 82 8.2 Extra terahertz revival caused by optical prealignment . . . . . . . . 84 9 Feasibility of experimental study of THz properties of aligned molecules 88 9.1 Discussion of gases- HCN, CH3Cl, OCS, N2O . . . . . . . . . . . . . . 88 9.2 Free space geometry . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 9.3 Wire guiding geometry . . . . . . . . . . . . . . . . . . . . . . . . . . 95 9.3.1 Coupling THz onto and off of the wire . . . . . . . . . . . . . 97 9.3.2 Skimming the hollow optical beam along the wire: ionization, polarization . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 9.4 Terahertz detection . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 Bibliography 104 v List of Tables 9.1 Overview of dipolar gases for optical/terahertz experiments. Data collected from [63, 86, 97, 98] . . . . . . . . . . . . . . . . . . . . . . 89 vi List of Figures 1.1 The Stanford Linear Accelerator (SLAC) accelerating structure is (a) a long copper waveguide filled with discs to control the phase veloc- ity of guided microwaves. (b) Breakdown of the copper limits the maximum microwave intensity, so to accelerate electrons to 40 GeV energy, (c) an extremely long structure is neccesary. . . . . . . . . . . 7 1.2 (a) A 40 TW, 40 fs laser pulse enters (b) an electric discharge plasma waveguide, where it drives (c) a nonlinear wakefield, which accel- erates (d) a nearly monoenergetic ? 30 pC bunch of electrons to ?1 GeV energy [26]. Figure adapted and modified from Leemans et al., Nature Physics 2, 696-699 (2006). . . . . . . . . . . . . . . . . . . 9 1.3 Our idea is to replace the SLAC structure, (a) a centimeter-scale, kilometers-long microwave-frequency waveguide made of copper, with (b) a micron-scale, centimeters-long, optical frequency waveguide made of plasma. Instead of (c) megawatt klystrons, we drive the structure with (d) gigawatt or terawatt laser pulses. . . . . . . . . . 11 2.1 Radially polarized light entering a plasma waveguide to accelerate a copropagating bunch of relativistic electrons. . . . . . . . . . . . . . 15 2.2 Radially polarized light entering a corrugated plasma waveguide to accelerate a copropagating bunch of relativistic electrons. Instead of phase matching the interaction with neutral gas as in Figure 2.1, the corrugations quasi-phase match the acceleration. . . . . . . . . . . . . 16 2.3 Toy model of electron density to estimate quasi-phase matched elec- tron acceleration gradients in the corrugated plasma waveguide. . . . 17 2.4 High magnification view of an experimental modulated waveguide. The waveguide has alternating peaks and troughs of central electron density, acting as focusing and defocusing microlenses. This geometry could potentially eliminate out-of-phase decelerating intensity on axis and further enhance in-phase acceleration intensity. . . . . . . . . . . 18 vii 2.5 We approximate the plasma waveguide shown in Fig. 2.2 with (a) a simple electron density profile to model laser pulse propagation and electron acceleration. FDTD simulation results show (b) accelerating and (c) focusing forces felt by a properly phased relativistic (v = c) electron copropagating with a femtosecond laser pulse. Several acceleration (yellow, below axis) and deceleration (blue, above axis) half-dephasing cycles in (b) are labeled to show the work done on the electron in that region, with acceleration clearly dominating. The focusing force (c) is similarly quasi phase matched for this electron, to a lesser degree. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 2.6 (a) Energy gain ?? vs. time with the slow wave phase velocity matched to the initial electron velocity and (b) the slow wave velocity is set to c, for?0 = 30, 100, and 1000. Allowing the electrons to catch up to the slow wave velocity reduces the dephasing due to acceleration at higher energies. (c) Average final z momentum (color scale) as a function of initial position (z0,x0). (d) Average final z momentum (color scale) as a function of final position (zf,xf) relative to the leading electron. (e) Final electron density as a function of final position (zf,xf). The electron beam has become bunched and focused. 27 3.1 Experimental geometry for creating the corrugated plasma waveg- uide, showing waveguides generated in air, argon clusters, and nitro- gen clusters, and an image of a guided mode. . . . . . . . . . . . . . . 33 3.2 Different corrugation methods: (a) Wire mesh obstructing cluster jet flow, producing a waveguide with ? 1 mm corrugation period and near 100% density modulation. (b) Transmissive ?ring grating? used to produce the waveguide shown in Fig. 3.1(c). This corruga- tion method modulates ionization and heating rather than material density. High-magnification image of the ring grating (c) shows the ?10 ?m groove period required to produce periodic ?35 ?m corru- gations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 3.3 Corrugated channels produced in a hydrogen cluster jet for different channel creation energies, and with and without guiding an intense optical pulse. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38 3.4 Corrugated channels in anargon cluster jet with and without injection of an intense guided pulse . . . . . . . . . . . . . . . . . . . . . . . . 40 viii 3.5 Abel-inverted electron density (a) of a wire-generated corrugation in nitrogen clusters 1 ns after creation shows nearly 100% density modulation. Phase images of these corrugations at 0.5, 1.0, and 2.0 ns delay in (b) nitrogen and (c) argon clusters show that this sharp density structure persists for the entire useful life of the waveguide. . 42 3.6 Abel transform of the 50-point FFT basis. Each row is the Abel inverse of the corresponding basis function used in the fast Fourier transform. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 4.1 Measured alignment ?cos2(?)? vs. time and space, of nitrogen (N2), oxygen (O2), and air, driven by a single 100 fs, ?1013 W/cm2, 800 nm pulse. The measurement technique is described in [69]. . . . . . . 48 4.2 Initiallyfew-cycle terahertzpulse measured aftertransmission through 38.7 cm of 800 hPa N2O. Adapted from Harde and Grischkowsky, J. Opt. Soc. Am. B Vol. 8, No. 8, Aug. 1991 [70]. . . . . . . . . . . . . 49 6.1 Proposed broadband terahertz amplification . . . . . . . . . . . . . . 66 6.2 Rotational state populations ?mjj vs. j of HCN gas for selected m- values (a) before illumination (thermal population), (b) after illumi- nation by a 15 TW/cm2, 100 fs, 800 nm optical alignment pulse, and (c) after a train of four such pulses separated by 11.49 ps (the revival time of HCN). Below are intensity plots of ?mjj vs. j and m for (d) thermal, (e) one-pulse, and (f) four-pulse illumination. . . . . . . . . 69 6.3 Population ?mj,j for different dissipation rates ? and number of optical pump pulses for (a) m = 0 and (b) m = 1. . . . . . . . . . . . . . . . 70 6.4 Population ?mj,j for static gas and gas jet initial temperatures T for (a) m = 0 and (b) m = 1, neglecting collisions (1/? ? 1ns). . . . . . 71 6.5 Population ?mj,j driven by a variable number of 15 TW/cm2, 100 fs optical pump pulses, for a static HCN gas target at 310 K neglecting collisions (1/? ? 1ns). . . . . . . . . . . . . . . . . . . . . . . . . . . 73 7.1 Amplification of a 200 femtosecond, single-cycle terahertz pulse . . . 75 7.2 Amplification of an 800 femtosecond, single-cycle terahertz pulse . . . 77 7.3 Scaled gain vs. frequency for a range of terahertz pulse fluences . . . 79 7.4 Comparison of extracted energy for low- and room-temperature targets. 81 ix 8.1 A 1TW/cm2 optical pulse weakly alignsHCN and stronglymodulates (a) terahertz energy absorbed/emitted per molecule, per m = 0, j ?j?1 transition depending on the optical-terahertz relative delay ?t. (b) Averaging over delay shows the ?incoherent? absorption that a terahertz pulse would experience if it was not collinear with the optical pump pulses. (c) Summing over j gives total absorption vs. ?t. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83 8.2 Expectation value of orientation?cos(?)?vs. time, following the same optical/terahertz pulse pair from Figure 8.1, for different relative de- lays ?t. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 8.3 Expectation value of orientation?cos(?)?vs. time, following the same optical/terahertz pulse pair from Figure 8.1, for selected relative de- lays ?t. (offset vertically for clarity) . . . . . . . . . . . . . . . . . . 86 9.1 Free-space geometry for combining optical and terahertz beams in a molecular gas jet. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 9.2 Free-space geometry for combining optical and terahertz beams in a static gas-filled tube; the tube is transparent to terahertz radiation. . 94 9.3 (a) Steel wire for terahertz waveguiding, coated with a poled polymer [106] and surrounded by a dipolar gas. (b) Optical excitation pro- duces a terahertz pulse which is guided along the wire. (c) A hollow optical pulse precedes the terahertz pulse, exciting the surrounding dipolar gas to amplify the terahertz. . . . . . . . . . . . . . . . . . . 99 9.4 Terahertz diagnostic for measuring electric field vs. time and beam profile. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 x Chapter 0 Introduction 0.1 High intensity lasers Multi-millijoule, subpicosecond laser pulses have given scientists amazing abil- ity to observe and manipulate matter. University laboratories have produced laser intensities of more than 1022 W/cm2, corresponding to an electric field of more than 1012 V/cm [1], almost 1000 times the electric field that binds an electron in a hydro- gen atom. More modest intensities (1015 W/cm2) are routinely achieved by compact commercial laser systems [2]. These compact, intense lasers allow small laboratories to ?push? on electrons with extreme power and precision, driving relativistic electron velocities, deep ionization of atoms and molecules, megavolt plasma temperatures, and even initiating nuclear reactions [3]. At the same time, the ultrashort nature of these pulses allows observation of the resulting dynamics with few-femtosecond time resolution [4], essential for observing such violent motion. Such short, intense pulses are made possible by several recent advances in laser technology. An optical pulse with duration <100 fs neccesarily has a bandwidth of many nanometers. Titanium sapphire, studied by Moulton et al., [5], is an excellent amplifying medium for such pulses, with an amplification bandwidth extending from 660-986 nm. Combined with the technique of Kerr-lens modelocking, titanium- sapphire oscillators routinely produce trains of optical pulses with duration less 1 than 30 femtoseconds [6]. Chirped pulse amplification (CPA), invented by Strickland and Mourou [7, 8] in the 1980?s, allows these short pulses to be amplified to many joules of energy with- out damage to their amplifiying medium. CPA uses dispersive optics to reversibly stretch ultrashort pulses to multi-picosecond duration, reducing their intensity to avoid nonlinear effects in amplifier rods. After amplification, similar dispersive op- tics return the amplified pulses to nearly their original femtosecond duration, giving high energy in a short pulse. Focused to micron-scale spots, these pulses reach enormous energy density despite containing fewer calories of energy than a stick of gum. 0.2 Some of applications of high intensity femtosecond lasers Femtosecond laser pulses have a tremendous number of applications. The ?Virtual Journal of Ultrafast Science? tracks journal articles related to femtosecond lasers from a variety of journals, and has routinely indexed more than 100 articles each month since 2002 [9], including entries on chemistry, biology, and materials science. The 1999 Nobel prize in chemistry was awarded for the study of chemical bonds and dynamics on a femtosecond timescale using subpicosecond pulsed lasers [10]. Subpicosecond lasers were used both to initiate and control chemical reactions, and also to probe the resulting behavior. Intense femtosecond lasers have enabled new kinds of imaging and manip- 2 ulation in biotechnology. Multiphoton imaging exploits nonlinear absorption to confine fluorescence to the laser?s focal region while minimizing heat and photo- damage; a sample can be transparent to a converging femtosecond pulse until the pulse focuses tightly enough to exceed the intensity threshold for nonlinear absorp- tion. Nanosurgery exploits a similar technique to destroy targeted regions in a 3-D sample without damaging the surrounding material [11]. Even simple laser ablation can be improved by intense femtosecond pulses. Laser ablation by nanosecond pulses typically heats a target, causing thermal stress. Femtosecond laser ablation ionizes and ablates without depositing much heat, and has been used in dentistry to reduce cracking caused by thermal stresses [12]. The 2005Nobel prize in physics was awarded forprecision spectroscopy, includ- ing the ?optical frequency comb?. By stabilizing the round-trip time and dispersion of a femtosecond laser oscillator cavity, the megahertz repetition rate of the cavity can be directly connected to the femtosecond-scale oscillations of the laser pulse, allowing researchers to literally count the number of optical cycles emitted by a source. This gives some of the most precise measurements of time ever made [13]. 0.3 Outline of thesis The compact, inexpensive nature of femtosecond oscillators and chirped pulse amplification has been crucial to all these applications, driving rapid progress by enabling many small labs to work simultaneously on a problem. The lesson is clear: if you can reach higher power, higher energy, shorter pulses, or a new type of beam 3 with a smaller, cheaper setup, countless applications will quickly follow. Laser wakefield acceleration [14] uses the extreme intensity of multi-terawatt femtosecond laser pulses to accelerate electrons to relativistic energy. This tech- nology aims to eventually replace enormous microwave-based electron accelerators with laboratory-scale laser systems. However, these multi-terawatt laser systems are much larger and more expensive than gigawatt regenerative amplifiers, and still out of the reach of many small labs and hospitals. In Chapters 1, 2, and 3, we de- scribe our work towards a new kind of laser-based electron accelerator which could function with much smaller lasers but still reach high electron energy. Chapter 1 briefly surveys the field of electron acceleration. Chapter 2 describes our theoretical work to design our novel acceleration structure, and Chapter 3 details our experimental realization of this structure. Terahertz generation by amplified femtosecond lasers [15] can produce short pulses that contain only a few cycles of the electromagnetic field. This technolgy is hampered by low conversion efficiency, limiting terahertz pulse energies to micro- joule levels, even with terawatt lasers or large electron accelerators. Chapters 4-9 describe our work to create a new kind of terahertz source based on rotating dipolar molecules, which could amplify existing few-cycle terahertz pulses to much higher energies. Chapter 4 briefly surveys the interaction of intense optical and terahertz pulses with molecular rotation. Chapter 5 describes the details of how we simulate these interactions. Chapter 6 describes our simulation results, which show intense optical pulses can drive a gas of molecules into a rotational population inversion, and Chap- 4 ter 7 shows how the resulting excited gas can amplify terahertz radiation. Chapter 8 shows some interesting results that suggest optically-aligned molecules could am- plify terahertz pulses even without rotational population inversion. Finally, Chapter 9 discusses experimental implementation of these schemes. 5 Chapter 1 Review of laser-driven particle acceleration 1.1 Direct acceleration by electromagnetic waves The Stanford Linear Accelerator (SLAC), shown in Figure 1.1, is currently the longest linear accelerator in the world [16]. The accelerator itself is fairly simple, a long copper tube with annular discs partially blocking the tube every few centime- ters. Charged particles are accelerated by the Ez field at the center of a TM-mode microwave guided in this axially modulated copper waveguide. In an ordinary unmodulated metal waveguide, the guided wave phase velocity vp = ?/k is greater than c [17], so a charged particle would be accelerated and then decelerated as the wave oscillations pass the particle, giving zero net acceleration. However, each annular copper disc partially reflects microwaves, forming a long chain of coupled microwave resonators. Through careful design, the phase of each resonator is related to the next by just the right amount such that vp < c, and particle and wave speeds can be matched. Called a ?slow wave structure?, this type of guide finds wide application in both accelerators and microwave sources [18]. The accelerating gradient of the SLAC structure is limited by the microwave damage threshold of the copper tube. Microwave fields stronger than ?107 V/m breakdown the walls catastrophically. In order to reach high energy (?50 GeV), the SLAC structure is roughly two miles long. Amplified femtosecond laser pulses, 6 Figure 1.1: The Stanford Linear Accelerator (SLAC) accelerating struc- ture is (a) a long copper waveguide filled with discs to control the phase velocity of guided microwaves. (b) Breakdown of the copper limits the maximum microwave intensity, so to accelerate electrons to 40 GeV en- ergy, (c) an extremely long structure is neccesary. in comparison, can easily reach intensities of 1015 W/cm2, corresponding to electric field strengths >60 GV/m, and laser intensities >1019 W/cm2 are not uncommon. If some fraction of this field could be used to accelerate electrons, then electron energies comparable to SLAC could be made over meters instead of miles. A wide variety of structures have been used to allow the exchange of energy between laser radiation and relativistic electrons, from a hydrogen-filled gas cell fed by a conically-focused, radially-polarized nanosecond CO2 laser [19], to a sim- ple reflective metal tape positioned at the focus of a picosecond laser [20]. While the efficiency and expense of these schemes varies greatly, the electron acceleration gradients they can achieve are ultimately limited by the radiation intensity they can produce and control, and typically are significantly less than 1 MV/cm. Mod- 7 ern femtosecond lasers based on chirped-pulse amplification [8] can produce focused field strengths in excess of 10 GV/cm. An accelerating structure that could control this field strength would allow enormous acceleration gradients, but no material can survive this intensity un-ionized. Schemes to use a laser?s electric field to accelerate electrons without an acceler- ation structure have been proposed [21], but none have been demonstrated, possibly because they violate the Woodward-Lawson theorem [22, 23] and are therefore im- possible. Generally, the electric field of radiation is perpendicular to the propagation direction. A relativistic electron in vacuum can either be phase-matched with radi- ation, or it can can have a component of velocity parallel to the radiation?s electric field, but not both since radiation moves at c in vacuum. Radiation which appears to have a component of its electric field point along its propagation direction can always be described as a superposition of plane waves, which either do not actually move in the apparent propagation direction, or have no component of their electric field pointing in this direction. 1.2 Wakefield acceleration Laser wakefield acceleration was proposed in 1979 [24] to solve the problems of phase matching and material damage threshold. An intense laser beam cannot accelerate electrons in vacuum [22, 23], and will ionize anything it touches, so high- intensity laser acceleration of electrons must therefore happen in plasma. The principle of wakefield acceleration is fairly simple: a sufficiently intense 8 Figure 1.2: (a) A 40 TW, 40 fs laser pulse enters (b) an electric discharge plasma waveguide, where it drives (c) a nonlinear wakefield, which accel- erates (d) a nearly monoenergetic?30 pC bunch of electrons to?1 GeV energy [26]. Figure adapted and modified from Leemans et al., Nature Physics 2, 696-699 (2006). laser pulse propagating in a plasma will deflect electrons to the side as it passes, through the nonlinear ponderomotive force [25]. Heavier ions are nearly unaffected (> 106? acceleration) by this ponderomotive effect [25], creating a region of net positive charge in the wake of the laser pulse. After the laser pulse is gone, this positive region pulls the expelled electrons back, creating a region of excess negative charge. This transverse motion creates a ?wake? that travels at the group velocity of the driving laser pulse. Even though the laser pulse itself has negligible electric field pointing along the laser propagation direction, the wake produces significant axial electric fields that can point and propagate in the same direction. Laser wakefield acceleration [24, 27, 28] of electrons to relativistic velocities 9 is often described as a ?tabletop? experiment [29], promising to make high-energy electrons available outside of large, multi-user accelerator facilities. Recent advances in wakefield acceleration have produced monoenergetic beams [30, 31, 32] and im- pressive electron energies exceeding 1 GeV [26]. These experiments, however, all use expensive multi-terawatt laser systems that cannot reasonably be considered ?tabletop?. Figure 1.2 illustrates the setup used by Leemans et al. to produce ?1 GeV electrons in a few centimeters [26]. A 40 TW, 40 fs laser pulse is injected into an electric discharge plasma waveguide. The waveguide helps suppress diffraction of the laser pulse, enabling high-intensity interaction over many centimeters. The intense pulse nonlinearly drives a wakefield, which has sufficient field strength to accelerate and trap thermal background electrons in the plasma. This accelerator produces a nearly monoenergetic ?30 pC bunch of electrons with ?1 GeV energy. 1.3 Proposed acceleration scheme Direct laser acceleration would be an attractive alternative to wakefield accel- eration for accelerating relativistic electrons if it weren?t for the damage threshold of solid or gaseous accelerating structures [19, 33]. Wakefield acceleration is a nonlin- ear process and is inefficient at sub-terawatt powers, but direct laser acceleration (a linear process) has no threshold intensity. Small, inexpensive few-mJ chirped-pulse regenerative amplifiers with gigawatt peak power that could never reach the intensity threshold for wakefield acceleration could still be used for direct laser acceleration. 10 Figure 1.3: Our idea is to replace the SLAC structure, (a) a centimeter- scale, kilometers-long microwave-frequency waveguide made of copper, with (b) a micron-scale, centimeters-long, optical frequency waveguide made of plasma. Instead of (c) megawatt klystrons, we drive the struc- ture with (d) gigawatt or terawatt laser pulses. 11 We describe experimental [34] and theoretical [35] progress towards a practical scheme for direct electron acceleration at multi-MeV/cm gradients by sub-terawatt femtosecond pulsed lasers. The essence of our scheme is illustrated in Figure 1.3. Just as the segmented copper structure of the Stanford Linear Accelerator channels the megawatt power of microwave-frequency klystrons to accelerate electrons over several kilometers [16], we have created a corrugated plasma structure to guide the giga- or terawatt power of optical frequency femtosecond pulses over centimeters. Our simulations show that if this structure is correctly tuned and injected with a radially polarized laser beam and a simultaneously copropagating electron bunch, true ?tabletop? electron acceleration is possible: a laser pulse power of 1.9 TW should allow an acceleration gradient larger than 80 MV/cm, and a modest power of only 30 GW would still allow acceleration gradients in excess of 10 MV/cm. Just as the annular discs in the SLAC structure control the effective phase velocity of guided microwaves, the corrugations in our plasma guiding structure control the local phase velocity of a guided optical pulse. Unlike any other direct acceleration scheme, our structure is completely ionized and can withstand laser intensity at least six orders of magnitude more than any solid accelerating structure. Unlike wakefield acceleration, we can use smaller, cheaper gigawatt regenerative amplifier laser systems to drive acceleration instead of larger, more expensive multi- terawatt systems. We describe the details of our simulations in Chapter 2 We recently achieved the first ever high intensity optical guiding in an ex- tended corrugated plasma slow wave structure. The application of these structures to coherent electromagnetic wave generation was first discussed in reference [36]. 12 Spontaneous, but uncontrolled modulated channels were reported in reference [37]. We have produced exceptionally stable plasma waveguides with adjustable axial modulation periods as short as 35 ?m, where the period can be significantly smaller than the waveguide diameter. The axial modulations can also be extremely sharp and deep, with nearly 100% modulation in plasma density. We have measured guided propagation at intensities up to 2?1017 W/cm2, limited only by our current laser energy and waveguide leakage. We describe the creation and characterization of this structure in Chapter 3. 13 Chapter 2 Direct laser acceleration in plasma 2.1 Toy model Using uncorrugated plasma waveguides [38] for direct electromagnetic accel- eration was suggested by Serafim et al. [39], who proposed guiding a radially polarized laser pulse to accelerate a copropagating relativistic electron bunch, as shown in Figure 2.1. The laser?s dominant radial component Er guides as a hollow mode with peak intensity at r = wch/?2, where the mode radius wch is given by wch = (1/?re?Ne)1/2, re is the classical electron radius and ?Ne is the electron den- sity difference between r = 0 and r = wch. The accelerating field is the associated axial component Ez, which peaks at r = 0 and passes through zero at r = wch. Fol- lowing reference [39], the peak axial acceleration gradient from hollow beam guiding in a plasma channel is given by E[GV/cm] = 98?P1/2/w2ch (2.1) where ? (the laser wavelength) and wch are in ?m and P is the peak laser power in TW. This accelerating field is proportional to the square root of laser intensity, which scales with P/wch. The accelerating field is also proportional to k?/|k| ? ?/wch, the effective ?ray angle? of the guided mode. Small guided mode diameters are therefore especially important for high acceleration gradients. 14 Figure 2.1: Radially polarized light entering a plasma waveguide to ac- celerate a copropagating bunch of relativistic electrons. For a 1.9 TW laser pulse with ? = 800 nm in a channel supporting wch = 15?m, Ez is an impressive 0.49 GV/cm. The plasma waveguide prevents the tightly focused pulse from diffracting rapidly away, extending this high energy density over many centimeters of propagation [40]. If there were no slippage between the laser phase velocity and the electron velocity (essentially c), this would compare very favorably to laser wakefield accel- eration: Malka et al. used a 30 TW laser to produce an acceleration gradient of 0.66 GV/cm (200 MV over 3 mm) [28]. Wakefield acceleration (a nonlinear process) would not even work with a laser pulse power of only 1.9 TW. However, even a regenerative amplifier with ? 1 mJ output (much smaller and cheaper than a ter- awatt laser) can easily produce 20 GW peak power, giving a 49 MV/cm gradient for direct acceleration. Of course, a means must be found to slow the laser phase velocity tocor less to 15 1 mm Corrugated Plasma Waveguide (a) e- (b) Relativistic electron bunch Radially-polarized femtosecond laser pulse Figure 2.2: Radially polarized light entering a corrugated plasma waveg- uide to accelerate a copropagating bunch of relativistic electrons. Instead of phase matching the interaction with neutral gas as in Figure 2.1, the corrugations quasi-phase match the acceleration. match the relativistic electron velocity. Neutral gas as proposed in [39] will not sur- vive the laser intensities essential for high values of accelerating field Ez; even pulses well below the terawatt level will propagate in fully ionized waveguides. Without neutral gas, the laser phase velocity in an uncorrugated plasma waveguide is strictly superluminal [38]: a relativistic electron would slip 2? out of phase with the accel- erating pulse after propagating a dephasing length Ld = ?(N0/Ncr + 2?2/?2w2ch)?1 [41], where N0 is the on-axis plasma electron density of the channel and Ncr is the critical plasma electron density for wavelength ?[17]. Dephasing occurs because the guided mode phase velocity is not c. The term N0/Ncr is due to plasma dispersion and would occur even in plane-wave geometry, and the term 2?2/?2w2ch is geometric dispersion due to waveguiding; both these effects increase the laser phase velocity above c. The electron receives no net acceleration: it would accelerate for a distance Ld/2, then decelerate an equal amount over the next Ld/2. Since Ld is typically a 16 Figure 2.3: Toy model of electron density to estimate quasi-phase matched electron acceleration gradients in the corrugated plasma waveg- uide. few tens or hundreds of microns, limiting the interaction length to Ld/2 gives very little acceleration [20]. Figure 2.2 shows how the corrugated plasma waveguide can quasi-phase match this interaction [42]. Laser phase velocity is locally faster in high plasma-density regions and slower in low plasma density. If Ld and the corrugation period are matched, the symmetry between acceleration and deceleration in a dephasing cycle is broken, and a properly phased electron will gain net energy; this process can be viewed as the inverse of transition radiation [43]. Figure 2.3 shows a toy model of the corrugated plasma waveguide that allows simple estimates of effective acceleration gradients with quasi-phase matching. The waveguide is assumed to consist of alternating segments which support the same guided mode radius wch but have either high or low on-axis electron density N0. The length of each segment is chosen to equal the dephasing length Ld for that segment?s radially polarized guided mode; the low density regions are therefore longer than the high density regions. Assuming a small density discontinuity at the interface between segments, we can neglect reflections at the boundaries, and a 17 Figure 2.4: High magnification view of an experimental modulated waveguide. The waveguide has alternating peaks and troughs of central electron density, acting as focusing and defocusing microlenses. This geometry could potentially eliminate out-of-phase decelerating intensity on axis and further enhance in-phase acceleration intensity. properly phased relativistic electron will gain energy ?U = integraldisplay Ld1 0 eEz1 sin(?z/Ld1)dz? integraldisplay Ld2 0 eEz2 sin(?z/Ld2)dz (2.2) = 2?e(Ez1Ld1 ?Ez2Ld2) giving an effective accelerating gradient Ez,eff = 2Ez? Ld1 ?Ld2L d1 +Ld2 (2.3) where the axial accelerating fieldEz can be calculated from equation 2.1. Supposing N0,1 = 3?1018 cm?3 andN0,2 = 6?1018 cm?3 for a channel supportingwch = 15?m, then Ld1 ? 340 ?m and Ld2 ? 195 ?m, giving an effective accelerating field Ez,eff ? 0.17Ez,max. This promising result deserves further study with more sophisticated models. A more realistic model would include the effects of diameter modulation, backscattering, changes in electron velocity, nonparabolic waveguide profiles, and plasma and optical nonlinearities. Note also that this simple model is far from opti- 18 mized. A waveguide with alternating focusing and defocusing segments (see Figure 2.4) could allow further enhancement of in-phase acceleration and greatly reduce or eliminate out-of-phase acceleration. Annular plasma waveguides have been also been demonstrated [44, 45] which guide annular modes with zero on-axis intensity. A corrugated waveguide consisting of alternating annular-mode and Gaussian-mode regions is another possible scheme for eliminating the decelerating regions and giving a higher effective gradient. 2.2 Finite difference time domain model: leakage, group velocity, dispersion We obtain physical insight into the quasi-phase matching process from finite- difference time-domain (FDTD) simulations of linear pulse propagation in the sim- plified plasma density shown in Figure 2.5(a). This simulation was performed using MEEP, a freely available software package with subpixel smoothing for increased accuracy [46]. These simulations assume cylindrical symmetry, and include plasma dispersion, finite pulse duration, and pulse leakage out of the channel. Finite com- puting resources force us to use an unrealistically long wavelength of 6.4 ?m, so the waveguide density was scaled to make the laser phase velocity comparable to experimental conditions. We launch a radially polarized pulse into a corrugated plasma waveguide, and calculate the propagating electric and magnetic fields. Figure 2.5(b) shows the relative longitudinal and 2.5(c) the transverse electric field (scaled for visbility) felt 19 x1018 cm-3 (b) (a) (c) 3 mm E z (a .u. ) E r (a. u.) -75 +186 -133 -158 -155 -122 -76 +135 +192 +213 +131+80 4.5 0 Figure 2.5: We approximate the plasma waveguide shown in Fig. 2.2 with (a) a simple electron density profile to model laser pulse propa- gation and electron acceleration. FDTD simulation results show (b) accelerating and (c) focusing forces felt by a properly phased relativistic (v = c) electron copropagating with a femtosecond laser pulse. Sev- eral acceleration (yellow, below axis) and deceleration (blue, above axis) half-dephasing cycles in (b) are labeled to show the work done on the electron in that region, with acceleration clearly dominating. The fo- cusing force (c) is similarly quasi phase matched for this electron, to a lesser degree. 20 by a relativistic (vz = c) electron copropagating with the laser pulse nearly on- axis. The transverse force depends strongly on the electron?s transverse position; a truly on-axis electron would experience no transverse force due to the cylindrical symmetry of a radially-polarized beam. The channel?s corrugation period is matched to Ld, and the initial phase between the electron and the laser field is chosen so that the electron is accelerated in the low-density section of each corrugation. Phase velocity is lower in these regions, so dephasing is slower and the electron spends more than half of each Ld in-phase with the accelerating field. If the initial phase were shifted by ?, deceleration would occur instead of acceleration. This simple picture is complicated somewhat by the modulation in waveguide diameter, which causes additional modulations in the guided pulse phase and in- tensity, but the electron clearly gains more energy during acceleration than it loses during deceleration. Each oscillation in 2.5(b) represents one dephasing cycle, and the number at each half-cycle is proportional to the energy gained or lost by the electron in that region. The transverse electric field shown in 2.5(c) is similarly quasi-phase matched, which contributes to a net radial focusing/defocusing force. The laser group velocity is subluminal; the electron starts behind the laser pulse and overtakes it. This ?pulse length dephasing? limits the interaction length. Leakage of the laser pulse out of the waveguide is minimal, and plasma dispersion does not interfere with acceleration. Simulations with more extreme diameter modulation [47] show the importance of the detailed shape of the corrugation. The effects of diameter modulation can also quasi-phase match acceleration, and under some conditions this effect can even 21 oppose the acceleration due to modulation ofN0. Luckily, good experimental control of the corrugated waveguide has been demonstrated [34]. 2.3 Analytic model, scaling law Starting from these encouraging results, we introduce an analytic model in or- der to study more realistic parameters. This model neglects leakage and dispersion, which were found to be unimportant in FDTD simulations, and which will also be less pronounced for 0.8 ?m light than it was at 6.4 ?m. We start with the radial component of the laser vector potential: Ar = ?Ar(r,z,t)exp[i(k0z??0t)] + c.c. (2.4) where k0 and ?0 are the central wave number and frequency of the laser pulse respectively and ?Ar(r,z,t) is a slowly varying envelope. We assume the pulse is azimuthally symmetric, and consider plasma channels with low electron densities such that the plasma frequency satisfies ?p ??0, where ?2p = 4?e2Ne(r,z)/me, ?e is the electron charge, me the electron mass, and Ne is the electron density. In this regime the envelope ?Ar evolves on a time scale much longer than the laser period. The slowly varying envelope equation then determines the evolution of the laser pulse: bracketleftbigg 2ik0 parenleftbigg ? ?z + 1 c ? ?t parenrightbigg + 1r ??rr ??r ? 1r2 bracketrightbigg ?Ar = ?2p(r,z) c2 ?Ar (2.5) where ?0 = k0c, and we have assumed that the electron plasma responds as a linear non-relativistic cold fluid. Because the laser-electron dephasing length Ld depends on the electron?s velocity [41], acceleration of sub-relativistic electrons or 22 protons requires a structure with a graded modulation period to ensure Ld remains matched to the modulations over the entire interaction length. For mathematical simplicity we consider a fixed modulation period, suited to acceleration of electrons with ? ? 1 for which the relativistic electron velocity depends only weakly on? and Ld is nearly constant. We limit our analysis to a periodic electron density profile that models Figure 2.2(a), Ne(r,z) = N0(1 +?sin(kmz)) +N??0r2/2, where ? is the relative amplitude of the density modulation, N??0 determines the radial dependence, and km is the wave number describing the axial periodicity of the channel. Exact solutions to Eq. (2.5) exist for this profile, which simplifies analysis of electron beam dynamics. Once Ar has been determined, Az and the axial electric field can be determined by vector?? vectorA? 0, which is consistent with the assumption ?p ??0. The slowly varying envelope approximation neglects second derivatives in z and t in the wave equation which are responsible for subluminal group velocity, but the group velocity can be explicitly restored by replacing c?1?/?t with v?1g ?/?t in Eq. 2.5. Here vg = 1 ??2p,0/2?2 ? 4/(k0wch)2, and we define ?2p,0 = ??2p(0,z)?z, where the brackets represent an average over z. The lowest eigenmode solution of Eq. (2.5) is: ?Ar(r,z,t) = A0 r wche ?r2/w2ch?(z?vgt)2/?2z summationdisplay n ?nJn(?)e???+?(?k+nkm)z (2.6) where Jn(?) is the nth order Bessel function of the first kind, ? = ??2p,0/2c2k0km, and ?k = ?k?10 (?2p,0/2c2 + 4/w2ch). The pulse has a Gaussian temporal shape with duration ?z/vg. In Eq. (2.6), we view the laser pulse as a sum of axial (z) spatial harmonics. Because our channel is axially periodic, each spatial harmonic moves with its own 23 constant phase velocity and amplitude. Matching the corrugation period to Ld is equivalent to matching the phase velocity of a spatial harmonic to the electron velocity. The harmonics have a relative amplitude given by Jn(?) and the effective phase velocity vp,n for the nth spatial harmonic is vp,n/c = 1?nkm/k0 +?2p,0/2?2 + 4/(k0wch)2. An appropriate choice of n and km gives a ?slow wave? spatial harmonic (vp 5 seconds per processed frame). The same algorithm implemented in MATLAB is much faster (several frames/second), but was not easily interfaced with our CCD cameras, so we compiled a MATLAB DLL to implement the phase extraction algorithm and called it from within LabView to combine speed with convenience. At the time, this was nontrivial and poorly documented, but provided an elegant solution to our optimization problem. More recent versions of LabView have since improved the speed and convenience of matrix manipulation however, making this solution somewhat obsolete. One important step of the phase extraction algorithm is choosing a Fourier window to isolate the spatial frequencies of the interferogram which contain plasma density information. This choice is somewhat subjective, and can introduce artifacts into the measured plasma density if chosen poorly. Displaying extracted phase in 43 real time had the unexpected advantage that it becomes easy to tune this Fourier window to minimize artifacts. Another troublesome step of the phase extraction algorithm is phase unwrap- ping; phase jumps of 2? do not represent discontinuities in plasma density, just artifacts in the measurement which must be removed. The phase unwrapping al- gorithm included in MATLAB is fast but naive, simply moving along each column and each row adjusting neighboring pixels by 2? to remove these discontinuities. Noise which disrupts the measured phase at any single pixel by ?? can cause errors which propagate through the rest of the image, requiring annoying fixes ?by hand? to recover the data. To minimize these problems, we instead used the phase un- wrapping method described in [61], which is more computationally expensive but more robust. Errors in this phase unwrapping algorithm due to noise produce local disturbances, but do not disrupt the entire image. Initially, Abel inversion was another bottleneck in our data processing. Con- verting a measured interferogram to radial density profiles like those shown in Fig- ures 3.3, 3.4, and 3.5 involves computating a numerical integral of each vertical column of the measured data, for each pixel of the processed data [62]: f(r,z) = ?1? integraldisplay R r dS(y,z) dy parenleftbigy2 ?r2parenrightbig?1/2dy (3.1) where S(y,z) is our measured plasma density integrated along the x-direction (transverse to the plasma waveguide), andf(r,z) is the plasma density in cylindrical coordinates. Even on a reasonably fast computer, processing a single image took 44 5 10 15 20 25 30 35 40 45 50 5 10 15 20 25 30 35 40 45 50 Figure 3.6: Abel transform of the 50-point FFT basis. Each row is the Abel inverse of the corresponding basis function used in the fast Fourier transform. upwards of ten minutes using straightforward numerical integration. However, the algorithm of Kalal and Nugent [62] speeds up this process immensely. Since Abel inversion is a linear process, you can compute the Abel inversion of each function in a complete set of basis functions. Then, to perform Abel inversion of an arbitrary S(y), you represent S(y) as a sum of these basis functions. Since you have stored the Abel inversion of each of these basis functions, you can simply sum up the transformed basis functions with appropriate weights to give f(r). Figure 3.6 shows the Abel inversions of each basis function in the 50-point FFT basis. An FFT of a 50-pixel signal decomposes that signal into a sum of 50 Fourier 45 basis functions of the form exp(ikx); each horizontal row of Figure 3.6 is the Abel inversion of the corresponding FFT basis function. An arbitrary S(y) represented with 50 pixels or less can be padded with zeros to a length of 50 pixels, and then the FFT of this signal gives the weight of each Fourier basis function in the signal. Each row of Figure 3.6 is weighted by this factor, and then the rows are added together to give a single 50-pixel row which is f(r), the Abel inverse of S(y). Notice that the far left column of Figure 3.6 is much more intense than the rest of the figure, especially for the lower (high spatial frequency) rows. This is because measurement of S(y) gives very little information about f(r = 0). Only one pixel of S(y) contains any information about f(r = 0), and that column represents a ray of light that traveled through the full width of the plasma to get to the center of the waveguide. Noise or error in the measurement of S(y) for any y will have an impact on the value of f(r = 0), so uncertainty in plasma density increases towards r = 0. In practice, this uncertainty often takes the form of negative plasma density, which while physically absurd serves as a good indication of the region of validity of f(r). Computing the Abel inversion of these 50 basis functions is time consuming (> 10 minutes), but only has to be done once. We also computed Abel inverison of larger FFT basis sets to process signals with more pixels, but 50 pixels was often enough. This algorithm is much faster than direct numerical integration of S(y), finishing in milliseconds rather than minutes. This robust, real-time phase extraction and Abel inverion is crucial for fully exploiting the control we have over the fine structure of the corrugated plasma waveguide. 46 Chapter 4 Review of pulse-driven molecular alignment 4.1 Laser-driven alignment, multishot measurement techniques Small linear molecules like nitrogen (N2), oxygen (O2), or nitrous oxide (N2O) typically rotate at terahertz frequencies in their room-temperature gas phase [63]. Rotational transitions in molecules resonantly absorb and emit terahertz radiation if the molecule has a dipole moment, while optical frequencies are completely non- resonant with molecular rotation. However, intense optical pulses can influence the rotation of small molecules by distorting the molecule?s cloud of bound electrons [64]. This distortion oscillates rapidly with each optical cycle, but produces a cycle- averaged torque that aligns the molecular axis with the laser polarization direction. Molecular alignment by intense lasers has been used to control high-harmonic generation [65], where the efficiency of the harmonic generation process depends on the direction of the molecular axis relative to the exciting laser field. Molecular alignment drives refractive index changes in gas, which has been used to compress short wavelength pulses [66]. These same refractive index modulations have signif- icant impact on the propagation of intense laser pulses over extended distances in the atmosphere [67]. When the aligning laser pulse is short compared to the terahertz timescale of molecular rotation, molecular alignment can persist long after the laser pulse 47 0 2 4 6 8 10 12 (ps) -1/3 2 c113 O2 air N2 120 mc109 -0.06 0.06 Figure4.1: Measured alignment?cos2(?)?vs. time and space, of nitrogen (N2), oxygen (O2), and air, driven by a single 100 fs, ?1013 W/cm2, 800 nm pulse. The measurement technique is described in [69]. has departed. This ?field-free? alignment was originally observed using multi-shot pump-probe techniques [68]. Figure 4.1 shows data from a single-shot measurement technique used recently in our lab to measure the modulation of the refractive index of air by an intense optical pulse [69]. The optical pulse driving alignment is intense (? 1013 W/cm2) and short (?100 fs) compared to the picosecond timescale typical of molecular rotation. After a prompt increase due to this short pulse, the refractive index of the gas returns to its equilibrium value with ?cos2(?)? = 1/3. However, several picoseconds after the optical pulse, the molecules realign, producing a fast refractive transient. This is a purely quantum mechanical phenomenon, showing that molecular rotation is coherent and nearly undisturbed by collisions for many picoseconds, at room tem- perature and atmospheric pressure. A detailed study of these transients in air and other gases is given in [69], including the first single-shot measurements of alignment of deuterium and hydrogen molecules. 48 20 40 60 80 100 120 ?1 ?0.5 0 0.5 1 1.5 2 2.5 3 3.5 4 Time (ps) Terahertz electric field (arb. units) Figure 4.2: Initially few-cycle terahertz pulse measured after trans- mission through 38.7 cm of 800 hPa N2O. Adapted from Harde and Grischkowsky, J. Opt. Soc. Am. B Vol. 8, No. 8, Aug. 1991 [70]. 4.2 THz-driven alignment Molecular alignment can also be driven resonantly at terahertz frequencies. Figure 4.2 shows the result of passing a few-cycle, ?3 picosecond terahertz pulse through 38.7 cm of 800 hPa N2O [70]. The similarity to Figure 4.1 is striking; a prompt response driven by the initial pulse is followed by several ?bursts? of signal each separated by a constant number of picoseconds. However, this signal is not refractive index, as in Figure 4.1; it is the electric field of the transmitted terahertz pulse. Because N2O has a permanent dipole moment, its rotation can be resonantly 49 driven by terahertz radiation. During each oscillation of the terahertz pulse?s electric field, the molecule has enough time to change orientation significantly, exchanging energy with the terahertz field. The dipoles which are rotating away from the electric field direction are slowed down, and the dipoles which are rotating towards the field direction are sped up, driving coherence between the orientations of each molecule. By orienting the molecular dipoles, this drives a net polarization of the N2O gas. After the initial terahertz pulse has passed, the molecules continue rotation as in Figure 4.1, but when the molecules realign many picoseconds later, their dipole moments produce a net polarization that re-radiates at terahertz frequencies. This re-radiation produces the few-cycle pulses of terahertz radiation visible at delays near 55, 95, and 135 picoseconds in Figure 4.2. 4.3 The possibility of terahertz emission from laser-driven molecular alignment Figure 4.1 shows that intense laser pulses can control molecular rotation, and Figure 4.2 shows that molecular rotation can emit few-cycle pulses of terahertz radation. This raises an intriguing question: Can optical pulses be used to drive molecular rotation to emit terahertz radiation? High-power terahertz radiation is notoriously difficult to produce. CO2 lasers have been used to pump vibrational transitions in dipolar molecules like D2O and methanol (CH3OH), which can drive rotational population inversion for lasing at teraherz frequencies [71]. Commercial cw CO2/CH3OH terahertz lasers are avail- 50 able [72], but produce only a few tens of milliwatts of peak power. More powerful terahertz pulses can be produced by using short CO2-laser pump pulses [73, 74], but these vibrationally resonant techniques only drive population inversion of one rotational transition, producing terahertz pulses with hundreds of cycles. Amplified femtosecond lasers can produce few-cycle pulses of terahertz radi- ation using optical rectification [75] or photoconductive switching [76]. However, the conversion efficiency of these schemes is typically fairly low; a review of these methods is presented in [15]. One should read [15] carefully; they appear to confuse micro-joules with nano-joules in several places (perhaps due to typographical error), overstating terahertz pulse energies by three orders of magnitude. Recent work has produced terahertz pulses with ?10 ?J energy, using sub- picosecond, 20 mJ optical pulses and a novel phase matching technique in lithium niobate [77]. However, even this record energy still gives conversion efficiency of only 5?10?4. Higher-power terahertz pulses have been produced using transition radi- ation from relativistic electrons [78, 79], but this requires large accelerator facilities or multi-terawatt lasers, in addition to the health hazards of relativistic electrons. Molecular rotation after terahertz orientation as shown in Figure 4.2 produces few-cycle terahertz pulses, but obviously the energy of these pulses must be less than the energy of the incident terahertz pulse which produces the orientation. It is much easier to produce high-energy optical pulses, and as shown in Figure 4.1 optical pulses drive similar rotational behavior in small linear molecules. N2 and O2 will not re-radiate terahertz upon alignment, of course, because they have no permanent dipole moment. Even laser-aligned molecules with a dipole 51 moment (like N2O) will not re-radiate terahertz. The laser alignment mechanism torques a molecule to point co-axially with the electric field, but since this mech- anism does not depend on the dipole moment of the molecule [80], there is no preference between parallel and anti-parallel molecular orientations. An initially thermal gas will exhibit no net polarization after laser alignment, and so it will not re-radiate terahertz. However, since both optical and terahertz radiation interact with the rotation of small linear molecules, a combination of optical and terahertz pulses incident on a molecular gas could result in a transfer of energy from the optical pulse to the terahertz pulse. We hope to combine the high average power of optically pumped terahertz lasers [72] with the few-cycle pulses typical of impulsively-excited molecu- lar rotation to produce extremely high power terahertz pulses. In the remainder of this work, we describe our theoretical and experimental progress towards terahertz generation from laser-aligned molecules. 52 Chapter 5 Molecular alignment simulation 5.1 The rigid rotor model of molecular rotation A small linear molecule in a low-pressure, room temperature gas is likely to be in its vibrational and electronic ground state [63]. The lowest excited vibrational energy of hydrogen cyanide (HCN), for example, is 88.3 milli-eV higher than the ground state [81], and therefore ? 30 times less populated at 300 K. Rotational excitation is much more likely; HCN?s lowest excited rotational energy is only 0.36 milli-eV above the ground state [63]. At 300 K, the ground and first excited rota- tional state population of HCN differ by <2%. A rotating molecule in the electronic and vibrational ground state can be modeled as a rigid rotor [63]. The rigid rotor Hamiltonian ?Hrr is similar to the hydrogen atom Hamiltonian but with no radial degree of freedom, given in spherical coordinates by: ?Hrr = ??h2 2I bracketleftbigg 1 sin? ? ?? parenleftbigg sin? ??? parenrightbigg + 1sin2? ? 2 ??2 bracketrightbigg (5.1) where I is the moment of inertia of the molecule. The spherical harmonics Ymj (?,?) are eigenfunctions of this Hamiltonian: ?HrrYmj (?,?) = ?h2 2Ij(j + 1)Y m j (?,?) (5.2) Each Ymj (?,?) represents a state of definite rotational angular momentum; j is the 53 angular momentum quantum number, and m (|m|?j) is the quantum number for the z-component of angular momentum. Since the rigid rotor eigenenergies differ by integer multiples of the j = 0 ? j = 1 transition energy ?h2/I, a rigid rotor wavepacket ?rr(?,?,t) = ?summationdisplay j=0 jsummationdisplay m=?j amj Ymj (?,?)exp parenleftbigg ?i ?h2Ij(j + 1)t parenrightbigg (5.3) is periodic with period Tr = 2?I/?h (neglecting collisions). Each amj is a complex number specifying the amplitude and phase at t = 0 of each spherical harmonic in the wavepacket. Tr is typically many picoseconds for small linear molecules. HCN, for example has Tr = 11.49 ps, whereas for the heavier carbonyl sulfide (OCS), Tr = 82.65 ps [63]. Applied electric fields modify molecular rotation. A rigid rotor with permanent dipole moment ? along the internuclear axis and polarizability components ?bardbl and ?? parallel and perpendicular to this axis, in an external electric field vectorE has the modified Hamiltonian [82]: ?Hrr+vectorE = ?Hrr +V?(?) +V?(?) (5.4) = ?Hrr ??Ecos?? 12E2parenleftbig?bardbl cos2?+?? sin2?parenrightbig where ? is the angle between vectorE and the molecular axis, and E ?|vectorE|. Instantaneous ?eigenfunctions? of this modified Hamiltonian [82] are less useful to us, because in our lab the disturbance vectorE is often in the form of a subpicosecond pulse with pulse duration Tpulse ? Tr. For short pulses of vectorE it is convenient to work in the rigid rotor basis and keep track numerically of how vectorE drives changes in the coefficients amj of Equation 5.3. 54 Note that a circularly polarized optical pump pulse has nearly the same in- teraction with a gas of linear molecules as a linearly polarized pump pulse with the same pulse duration and energy. The only change in the cycle-averaged Hamilto- nian is that the effective sign of ?? changes from positive to negative, and the polar (z) direction now points in the laser propagation direction instead of in the laser polarization direction [83]. Our simulations (not shown) show that this interaction drives the same changes in rotational populations ??mj,j with positive or negative ??. Circular polarization has been observed to drive alignment and anti-alignment like linear polarization [84], in contrast with the predictions of [85]. In moderate strength, low frequency electric fields, the permanent dipole inter- action V?(?), if it exists, dominates the induced polarization interaction V?(?). For example, HCN has a dipole moment of 2.98 debye and a differential polarizability ??(??bardbl???) of 1.91 ?A3 [63, 86]. In a 1 V/m electric field,?HCNE = 9.9?10?30 J, while ??HCNE2 = 2.13?10?40 J. Gigahertz or terahertz (109 ? 1012 Hz) radiation is near resonance for most rotational transitions [63], and affects the rotation of molecules primarly through the permanent dipole interaction, if it exists. The electric fields of optical pulses, however, oscillate at nearly petahertz (1015 Hz) rates (much faster thanTr for HCN) and the V?(?) interaction time-averages to zero, leaving only the V?(?) interaction, which cycle-averages to half the DC value (V?,AC = 12V?,DC) for linearly polarized radiation [82]. Despite the smallness of ?? for most molecules, amplified femtosec- ond optical pulses can reach high enough intensity to drive significant changes in the amj coefficients and strongly modify molecular rotation [85, 87]. 55 Such strong optical pulses might also drive vibrational or electronic transi- tions in a molecule, invalidating the rigid rotor approximation. This excitation could have serious impact on the design of an optically pumped broadband ter- ahertz laser. However, electronic transitions from the ground state are generally resonant with ultraviolet wavelengths, and vibrational transitions from the ground state are typically resonant in the infrared [63]; many molecular gases such as HCN are transparent (nonresonant) near the ?800 nm wavelength of most intense fem- tosecond lasers. Nonresonant excitation of vibrational and electronic transitions is still possi- ble, but since electronic and vibrational timescales are so much shorter than most femtosecond optical pulses (Telectronic ? Tvibrational ? 15 fs < Tpulse), electronic and vibrational excitation can adiabatically remain in the ?ground state? of an applied optical pulse, and return smoothly to the true ground state after the pulse is gone [88]. Experiments comparing measured molecular alignment in CO2 to the rigid rotor model [89] have found good agreement even for laser intensity exceeding the molecular ionization threshold, implying that neglect of electronic and vibrational excitation is an acceptable approximation for 90 fs, 800 nm alignment pulses. 5.2 Statistical description using a density matrix To study the feasibility of an optically pumped broadband terahertz laser, we need to perform full-scale nonperturbative numerical simulations of the rotational quantum state of a gas of dipolar molecules illuminated by femtosecond optical 56 and terahertz pulses. Direct simulation of ?(?,?,t) is straightforward for a single molecule, but for an ensemble of molecules (such as a small volume of gas), the thermal distribution of initial quantum states means a large number (? 103/?m3) of molecules must be simulated and averaged for accurate results. The density matrix formalism [90] describes an ensemble of identical nonin- teracting quantum mechanical systems with a statistical distribution of initial con- ditions. If each quantum system can be described by a superposition of Nst basis states, the density matrix contains at most N2st variables. Direct simulation of each of Nsys quantum systems would require keeping track of NstNsys variables, so the computational savings are considerable when the number of systems greatly exceeds the number of quantum basis states needed to describe each system (Nsys ? Nst), as in the case of a gas of rigid rotors. Using the Dirac notation and defining |j,m? by ??,?|j,m? = Ymj (?,?) (5.5) where |?,?? is an angular position eigenstate, we define a density matrix ?: ? = ?summationdisplay j=0 ?jsummationdisplay m=j ?summationdisplay k=0 ?ksummationdisplay n=k ?m,nj,k |j,m??k,n| (5.6) ? describes the quantum state of an ensemble of identical, noninteracting rigid rotor molecules subjected to the same electromagnetic field, such as a volume of gas much smaller than one cubic optical wavelength. The expectation value of any observable O of the ensemble can be computed from ? according to: ?O? = Tr(O?) (5.7) 57 where Tr() is the trace operation [90]. A density matrix modeling a system with Hamiltonian ?H evolves in time according to: i?h???t = [ ?H,?] (5.8) where [A,B] ?AB?BA is a commutator. Substituting the Hamiltonian from equation 5.4 for ?H and assuming vectorE is linearly polarized in the z-direction, ? time-evolves according to [69, 91]: d dt? m j,k(t) = ?i?j,k? m j,k(t) + (d? m j,k/dt)diss + (5.9) ?i?? 4?h E 2 optical(t) summationdisplay q [Omj,q?mq,k(t)?Omq,k?mj,q(t)] + ?i? ?h Eterahertz(t) summationdisplay q [Tmj,q?mq,k(t)?Tmq,k?mj,q(t)] Where: ? ?j,k = ?j??k, ?j = (?/Tr)j(j+1), ?? is the difference in the molecule?s par- allel and perpendicular polarizability and ? is its permanent dipole moment. For HCN Tr = 11.49 ps, ?? = 1.91 ?A3, and ? = 2.98 debye [63, 86]. ? E2optical(t) is the squared envelope of a nonresonant optical electric field which interacts through V?(?), and Eterahertz(t) is the full electric field (not the enve- lope) of a low frequency terahertz or gigahertz pulse that interacts primarily through V?(?) [70, 83]. ? Since we are only considering electric fields which point in the z-direction, quantum states with different values of m are not coupled to each other [69] (they?re also zero in thermal equilibrium, as we will discuss). For this reason, 58 we neglect elements of the density matrix for which m negationslash= n and substitute ?m,nj,k ??mj,k. ? Omj,k and Tmj,k describe how optical (O) and terahertz (T) radiation couple different elements of the density matrix. Omj,k =?j,m|cos2 (?)|k,m?, which is zero unless j = k or j = k ? 2, so optical radiation drives coherence only between states with the same parity [69]. Sufficiently intense optical radiation can be nonlinearly absorbed (or emitted) by j ?j?2 population transitions [85]. The terahertz coupling term Tmj,k =?j,m|cos(?)|k,m? is zero unless j = k?1, so terahertz radiation drives coherence between states of opposite parity, and is resonantly absorbed and emitted by j ?j?1 transitions at frequency ?j,j+1. ? Finally, (d?mj,k/dt)diss accounts for the fact that a gas of molecules is not re- ally a set of identical noninteracting quantum systems. Dissipative effects like collisions and emission of radiation drive the gas towards thermal equilibrium [91]. Interactions between the gas molecules are generally extremely com- plicated and prohibitively difficult to model precisely, but this term at least acknowledges the complication and leaves it to be specified later. So? describes a gas of rigid rotor molecules (equation 5.6), from which we can compute observables (equation 5.7), and we know how to time-evolve ? (equation 5.9). We only need the initial condition of ? to begin simulation, and most of the gas targets in our lab start in thermal equilibrium. In thermal equilibrium at temperature T, the population of rotational states ?mj,j is given by Boltzmann 59 statistics: ?mj,j ?00,0 = exp parenleftbigg??h? j kBT parenrightbigg (5.10) Where kB is Boltzmann?s constant. In thermal equilibrium there is no coherence between molecules and no preferred direction, so ?mj,knegationslash=j = 0 and we normalize? such that summationtextj,m?mj,j = 1. Our simulations apply a sequence of optical and terahertz pulses to an intially thermal HCN gas, and the resulting changes in population ??mj,j give the average energy absorbed/emitted per molecule ?E = summationtextj,m?h?j??mj,j from each of the pulses. Since we neglect back-action of the molecular emission on the ensemble, we can calculate transitions driven by the applied pulses but not self-stimulated transitions like those observed in free-induction decay [70]. A more accurate simulation would take into account the macroscopic polarization and radiation of the medium and allow self-generated electric fields to influence ?, but this would require including nonlocal effects such as propagation (solving the coupled Maxwell-Bloch equations) and increase the simulation burden beyond our resources. Since we are primarily concerned with prompt absorption or amplification of short optical and terahertz pulses, and self-stimulated emission generally occurs many picoseconds after applied electric fields, this is a reasonable approximation. 5.3 Details of the simulation algorithm ? as described in equation 5.6 is specified by an infinite number of terms ?mj,k. In practice, only a finite number of these terms are nonzero. The Boltzmann factor 60 in equation 5.10 decreases rapidly with increasing j, and the vast majority of terms (?mj,knegationslash=j) in the thermal density matrix are negligible. Selecting which elements of the density matrix to include is the primary challenge of our numerical simulations. Applied optical and terahertz fields drive these initially negligible elements of the density matrix to nonzero values. However, because of the j ?{1,2} coupling in equation 5.9, external fields must drive amplitude changes sequentially. ?23,7, for example, will have negligible amplitude until the amplitude of at least one of ?23?{1,2},7 or ?23,7?{1,2} is significant. Sufficiently weak applied fields may only drive one ?generation? of amplitude away from thermal equilibrium. In ref. [69], for example, experiment agrees well with ?first order? theory that only considers the time evolution of density matrix elements ?mj,j?2, for laser intensity ?1013 W/cm2. However, in order to study an optically-pumped terahertz laser, we must be able to accurately calculate large frac- tional changes in ?mj,j which are neccesary for population inversion. Clearly this requires more than a ?first order? simulation, in which changes in ?mj,j are assumed zero. We initially tried a naive approach of numerical simulation including all ?mj,k with {j,k,m} ? jmax. This approach is very accurate and simple to code, and allows simulation of very high intensity applied fields. However, this approach ul- timately was too computationally intensive and limited us to simulation of low initial-temperature gases for which jmax can be small. By including elements with large differences between j and k which have rapid phase evolution, we force our- selves to use a small timestep, and we waste simulation effort by including elements 61 which never become significantly nonzero. A more intelligent approach is required; we want our simulation to include the ?mj,k which will be nonnegligible, but no others, and we don?t know before we run a simulation if our parameters will drive significant amplitude into a given ?mj,k. Perhaps the best way to proceed is to grow our density matrix adaptively, adding elements ?mj,k to our simulation as they become important. The following algorithm provided a nice balance between computational burden and accuracy: For a given set of molecular parameters and a given applied optical/terahertz field pattern, determine which ?mj,k are above an ?amplitude threshold?, and declare this to be the set of ?fully considered? elements {?mj,k}f. Then: ? Construct the set of all elements which are not ?fully considered?, but are coupled to ?fully considered? elements by the applied optical and terahertz fields. Declare these elements to be ?boundary? elements {?mj,k}b. For example, if ?14,6 is a ?fully considered? element, and terahertz fields are present (j ? 1 coupling), and ?15,6 (which is coupled to ?14,6 by terahertz radiation) is not an element of {?mj,k}f, then ?15,6 is an element of {?mj,k}b. ? If the set of {?mj,k}f and {?mj,k}b is large enough for accurate simulation, the boundary elements should not get very big. If any of {?mj,k}b were to reach large amplitude, they should couple to elements that are not included in the simulation, and inaccuracy in the time evolution of the boundary elements will result. Therefore, if any of {?mj,k}b exceeds the ?amplitude threshold?, stop the simulation before much inaccuracy results. If no {?mj,k}b exceeds the amplitude 62 threshold at t = tfinal, the simulation is finished. ? If the simulation stops due to a large boundary element, promote that element to a ?fully considered? element, go back a few time steps in the simulation, and return to the first step in this list. Accuracy of the simulation is sensitive to the ?amplitude threshold?. Halving the amplitude threshold and rerunning the simulation is a good way to check results for convergence; the second simulation will take much longer, but if it returns the same result, you can trust it. Running the simulation over again from the start using {?mj,k}f and {?mj,k}b from a prior run with the same parameters often results in slightly different results, which is another way to tell if your amplitude threshold is too big. Symmetry offers several important speedups. Note that there is no difference in the differential equation for positive or negative values of m, and that ?mj,k = ?m,?k,j , which gives a factor of four improvement in speed and memory. The fact that different m-values do not couple is also helpful. For some simulation parameters, the full simulation including all ?mj,k is too big to fit nicely in the memory of a single workstation, and running separate simulations for different m-values simultaneously gives a nice speed boost. We took advantage of the Maryland high-performance computing cluster deepthought [92] to run 20 or more m-values at a time. One might expect the final sets {?mj,k}f and {?mj,k}b to give enough information for an intelligent guess about {?m?1j,k }f and {?m?1j,k }b. In practice, this could often lead to more rapid convergence to a sufficiently small ?amplitude threshold?, but it 63 also lead to unneccesary inclusion of ?mj,k with large values of |j?k|, which rapidly oscillate and slow the simulation. For serial jobs (one m after another) this offered some advantage, but was nearly useless for parallel jobs. Forfutureoptimization we plan tomake thechange ofvariables?mj,k = ?mj,kexp(?i?j,kt), which leads to a modified differential equation. Initial tests show this gives approx- imately a factor of two improvement in simulation speed, but this may depend on simulation parameters and the exact mechanism of speedup is unclear. 64 Chapter 6 Using optical pulses to drive population inversion in aligned molecules 6.1 Motivation Broadband chirped-pulse amplification (CPA) in Ti:sapphire revolutionized nonlinear optics in the 90?s, bringing intense optical pulses out of large govern- ment facilities and into the hands of graduate students in small university labs. High peak-power terahertz pulses (? 10 ?J, < 5 cycles), however, are still only produced at large accelerator facilities like BNL [78]. CPA is theoretically possi- ble for terahertz frequencies, but no broadband lasing medium like Ti:sapphire has been demonstrated for terahertz. Dipolar molecular gases such as hydrogen cyanide (HCN) or nitrous oxide (N2O), ?aligned? or rotationally excited by intense optical pulses, are a novel and promising medium for amplification of broadband few-cycle terahertz pulses. Using the techniques described in the previous chapter, we perform calcula- tions to study if rotationally excited molecules can amplify a few-cycle seed pulse of terahertz radiation. Our proposed scheme is shown in Fig. 6.1: a short, intense optical pulse (or sequence of pulses) aligns a dipolar gas (such as HCN), driving the molecules into a broad superposition of excited rotational states. If the right rota- 65 After: THz Seed, 800 nm Pump Amplified THz,Attenuated 800 nm Before: H C N Gas Cell Figure 6.1: Proposed broadband terahertz amplification tional ensemble is prepared, a broadband seed terahertz pulse following the optical pulses can then be amplified on many pure rotational transitions simultaneously. Since HCN or N2O have a dipole moment, they can absorb and emit radiation through rotational transitions. The rotational spectrum of a gas of linear molecules is a series of regularly spaced lines in the low-terahertz frequency range [63]. Absorp- tion and emission depend on the state of the gas; there is no coherence in thermal equilibrium (?mj,knegationslash=j = 0), so the strength of the linear absorption in the terahertz region depends on the population differences ??mj,jparenleftbig??mj,j ??mj?1,j?1parenrightbig between adja- cent rotational states [70]. Since ??mj,j < 0 in thermal equilibrium, the rotational spectrum is purely absorbing. Following intense, femtosecond optical pumping, however, molecular gas is driven into an ?aligned? excited rotational state [69]; the terahertz spectrum of this excited state is strongly modified. If the gas is sufficiently excited, ??mj,j > 0 for many j, and its rotational spectrum contains a set of regularly spaced amplification lines. The duration of the ?aligned? excited state is determined by collisions. The pressure-dependent alignment lifetime of small linear molecules is tens or hundreds of picoseconds at or below atmospheric pressure (68.5 ps for N2O at 1 atm [69]), 66 and increases linearly with decreasing pressure. A few-cycle terahertz seed pulse could follow a few picoseconds behind an optical pump pulse train and be strongly amplified by many of these lines simultaneously. Of course, if the gas is in a coherent ensemble, absorption can also depend on the correlation between molecules ?mj,knegationslash=j, as discussed in Chapter 8. 6.2 Low pressure, room temperature As shown in Fig. 6.2, a series of short intense optical pulses can drive HCN gas into an excited, inverted rotational state suitable for broadband terahertz amplifi- cation. Figure 6.2(a) shows the initial rotational population of the gas in thermal equilibrium at 310 K: ?mj,j decreases with j and is independent of m. Illumination by a single 15 TW/cm2, 100 fs, 800 nm optical pulse drives coherence (|?mj,knegationslash=j|> 0) between the gas molecules, but as shown in Fig. 6.2(b), causes little change in the rotational populations ?mj,j. Arbitrarily increasing the intensity of the optical pulse would drive larger changes in the rotational populations, but would also cause undesired effects like vibrational excitation, self-focusing, or ionization. Luckily, a train of pulses can drive much larger changes in rotational popula- tions. At low gas pressures (? 1 atm) collisions are rare and (d?mj,k/dt)diss ? Tr. The field-free ?(t) of a rigid-rotor gas is then perfectly periodic with period Tr, and a sequence of small pulses separated by Tr build up resonantly to the same effect as a single pulse with the same total energy. In Fig. 6.2(c), four 15 TW/cm2, 100 fs, 800 nm optical pulses separated by Tr,HCN = 11.49 ps strongly invert the popu- 67 lation of the HCN gas; each molecule absorbs ? 4.6?10?21 J of rotational energy (for pressures <1atm, this is a small fraction of the optical energy). ?mj,j increases with j for nearly every transition from j = 5 to j = 19; a broadband terahertz pulse could be amplified simultaneously by all these transitions. The optical pulses drive population to higher j values by strongly aligning the molecules; a rigid rotor highly localized in ? must occupy a broad spectrum of j states, and tighter angular localization projects into higher j. The irregularity of ?mj,j at low j shown in Figure 6.2(c) is due to the strong dependence of Omj,k (from equation 5.9) on j for j ?m. The simulation results are identical (neglecting ionization) to a single pulse with intensity 60 TW/cm2. Impressively, a train of 600, 0.1 TW/cm2 pulses sepa- rated by Tr gives the same result. Splitting the pulse energy into a train of smaller pulses avoids undesired intensity-dependent effects, and would allow recycling the same optical pulse in a cavity for high efficiency. The number of times a pulse can be reused for pumping is limited by the pressure-dependent alignment lifetime; the results in Fig 6.2(c, f) are appropriate for pressures < 10?2 atmospheres. 6.3 Moderate pressure, room temperature: the effects of dissipation High pressures are desirable for high single-pass gain and high saturation en- ergy. Low pressures are desirable because they allow many optical pulses to con- tribute to the population inversion; a compromise is necessary. In general, the dissipation of the density matrix toward thermal equilibrium is complicated [91] and not well studied experimentally, so we adopt the simple model (d?mj,k/dt)diss = 68 Total angular momentum (j) 0 15 ?15 30 ?30 Total angular momentum (j) Total angular momentum (j) 10 20 30 0 2 4 6 x 10?3 Population ( ? j,jm ) 10 20 300 2 4 6 x 10 ?3 10 20 300 1 2 3 x 10?3m=0 m=3 m=8 m=15 2 4 6 x 10?3 0 2 4 6x 10 ?3 0 2 x 10?3 (a) (b) (c) (d) (e) (f)Population Population PopulationZ?angular momentum (m) Figure 6.2: Rotational state populations ?mjj vs. j of HCN gas for se- lected m-values (a) before illumination (thermal population), (b) after illumination by a 15 TW/cm2, 100 fs, 800 nm optical alignment pulse, and (c) after a train of four such pulses separated by 11.49 ps (the re- vival time of HCN). Below are intensity plots of ?mjj vs. j and m for (d) thermal, (e) one-pulse, and (f) four-pulse illumination. ?(?mj,k??mj,k,thermal). As shown in Fig. 6.3, for very low pressures (1/? > 1 ns), four- pulse optical pumping drives a strong population inversion, but dissipation in the range 1/? = 200?300 ps (corresponding to pressures of ? 0.01 atm) decreases the strength and the bandwidth of the population inversion for them = 0,1 cases (other m?s are similar). Increasing the number of pump pulses from four to five, however, returns some of the inversion strength and more than restores the bandwidth. Es- timating ??j?1,j ? ??0,1 and ? ? 2???, these decay times are appropriate for a pressure of ? 0.014 (1/? = 300 ps) and 0.02 (1/? = 200 ps) atmospheres [93]. Gain at these pressures would be fairly low. Further simulations at higher pressure (> 0.02 atm) (not shown) suggest that the number of pump pulses required to drive strong population inversion rises prohibitively. Asthe collision timebecomes comparable to Tr, the decay of rotational excitation between pump pulses balances the pumping excitation and a pumping equilibrium is reached. For a given pump 69 5 10 15 202.5 2.6 2.7 2.8 2.9 3 3.1 3.2 x 10 ?3 Rotational angular momentum (j) Population ( ? j,jm ) 5 10 15 20 4 pulses, 1/?>1ns 4 pulses, 0.014 atm 4 pulses, 0.02 atm 5 pulses, 0.014 atm 5 pulses, 0.02 atm (a) (b) m=0 m=1 Figure 6.3: Population ?mj,j for different dissipation rates ? and number of optical pump pulses for (a) m = 0 and (b) m = 1. intensity, there is some critical pressure above which no population inversion is possible, no matter how many pump pulses are used. Fortunately, increasing the pressure of a room-temperature gas target is not the only way to increase gain. 6.4 Low temperature gas jet targets Decreasing temperature to decrease collisions without sacrificing density is appealing, but HCN?s vapor pressure drops off rapidly for temperatures below 300 K. The use of a gas jet rather than a static gas target allows the use of higher density targets at lower rotational temperature and long collision times. When a gas expands into a vacuum through a nozzle, the pressure and tem- perature both drop abruptly. If the backing pressure is high enough, collisions in 70 5 10 15 202 4 6 8 10 12 14 x 10 ?3 Rotational angular momentum (j) Population ( ? j,jm ) 5 10 15 20 4 pulses, static target (T=310K) 3 pulses, gas jet target (T=50K) 4 pulses, gas jet target (T=50K)(a) (b) m=0 m=1 Figure 6.4: Population ?mj,j for static gas and gas jet initial temperatures T for (a) m = 0 and (b) m = 1, neglecting collisions (1/? ? 1ns). the nozzle bring the gas molecules to nearly the same direction and velocity. This process efficiently cools molecular rotation and to a lesser extent, vibration [94, 95]. Rotational cooling in a supersonic gas jet has been shown to increase the efficiency of optical alignment of molecules [96]. As shown in Fig. 6.4, a low temperature (T = 50K) HCN gas jet target gives a much stronger population inversion after four-pulse optical pumping than in uncooled (T = 310K) static HCN, resulting in much more optical absorption and terahertz gain per molecule. Additionally, low temperatures concentrate population in low m-states which have a larger effective dipole interaction with terahertz fields. Estimating ?jet = ?staticradicalbigTjet/Tstatic at comparable density, collisional decay would be 2.5 times slower in a gas jet. The lower temperature of the gas jet also makes 71 population inversion easier to achieve. At low pressures (1/? ? 1ns), only three optical pulses still drive a strong population in 50K HCN, although with a reduced inversion bandwidth. 6.5 Overpumping As shown in Fig. 6.2(c), four pump pulses drive a population inversion in low-pressure, room temperature HCN gas from j = 5 to j = 19, corresponding to resonant transitions at frequencies < 1.6 THz. To amplify higher frequencies, the population inversion must be pushed to higher j values. One way to accomplish this is to use more pump pulses: Figure 6.5 shows the m = 0 populations ?0j,j vs j after each of 12 pulses in a train of of 15 TW/cm2, 100 fs pulses spaced by the revival timeTr, for the same parameters used in Figure 6.2. Four pulses drives a population inversion, but each successive pulse drives this inversion to higher j, extending the inversion as far as j = 36 by the twelfth pulse, nearly doubling the highest inverted resonant frequency. This behavior suggests a simple intensity-based mechanism to tune the center frequency of terahertz gain. Unfortunately, our computational resources are insuf- ficient to explore the limits of this behavior, but the trend shown in Figure 6.5 is promising. 72 0 5 10 15 20 25 30 35 40 450 1 2 3 4 5 6 7 x 10 ?3 Rotational angular momentum (j) Population ( ? j,jm ) No pulses 1 pulse 2 pulses 3 pulses 4 pulses 5 pulses 6 pulses 7 pulses 8 pulses 9 pulses 10 pulses 11 pulses 12 pulses Figure 6.5: Population ?mj,j driven by a variable number of 15 TW/cm2, 100 fs optical pump pulses, for a static HCN gas target at 310 K neglect- ing collisions (1/? ? 1ns). 73 Chapter 7 Using inverted rotational populations to amplify THz pulses 7.1 Amplification bandwidth, THz pulse duration In Chapter 6, we showed that four optical pulses can drive a strong population inversion in low pressure, room temperature HCN (Fig. 6.2). Figure 7.1 shows the effect of this four-pulse optical excitation on HCN?s absorption of a 200 fs, single- cycle terahertz pulse that follows Tr/2 behind the last optical pulse. In Chapter 6 we were interested in driving population inversion, so we displayed rotational populations ?mj,j. Here we are primarily concerned with absorption or amplification of terahertz pulses, which may drive small fractional changes in ?mj,j, so we instead display energy absorbed or emitted due to the terahertz pulse, ?Emj = ?h?j??mj,j. Since transition frequency is m-independent, we also display ?Ej = summationtextm?Emj . At 310 K without optical excitation, the j ? j ? 1 transitions are purely absorptive (Fig. 7.1(a, c)) at resonant frequencies 2jf0 when illuminated with the broadband single-cycle pulse shown in Fig 7.1(e). Some of this energy is reemited as a train of short terahertz pulses at integer multiples ofTr [70]; however, as mentioned in Chapter 5, this free induction decay is beyond the scope of our simulation. The width of these absorption lines (alternatively, the number of pulses emitted by free induction decay) depends on pressure. The HCN j = 0 ? 1 linewidth 2?f ? 38 GHz/atm [93], corresponding to an alignment lifetime of ?200 ps at 0.02 atm 74 10 20 30 0.09 0.7 1.31 1.92 2.52 10 20 3030 15 0 ?15 ?30 0.09 0.7 1.31 1.92 2.52 ?0.2 ?0.1 0 ?0.2 0 0.2?5 0 5 ?10 ?5 0 x 10?31 Energy emitted per molecule, per transition (J) Electric field (kV/cm) Time (ps)j? j?1 resonant emission frequency Initial rotational angular momentum (j) of transition Energy emitted per molecule (J ? 10 28 ) Z?angular momentum (m) (a) (b) (c) (d) (e) Figure 7.1: Energy absorbed/emitted (?Emj ) by 200 fs-pulse THz-driven j ? j ? 1 transitions vs. j and m of the upper state for (a) an ini- tially thermal HCN gas and (b) the optically-excited rotational pop- ulations shown in Fig. 6.2(c, f). Summed over m, this gives energy absorbed/emitted (?Ej) near each resonant frequency 2jf0 for (c) ther- mal and (d) excited gas. The ? 200 fs terahertz pulse driving these transitions (e) follows at Tr/2 after the last optical excitation pulse. 75 pressure. However, if the same terahertz pulse follows Tr/2 behind the last optical exci- tation pulse from Fig. 6.2(c), the 19 lowest absorption lines become amplification lines. This comb of amplification frequencies extends from 0.08 THz to 1.6 THz, as shown in Fig. 7.1(b, d) and the degree of amplification is comparable to the degree of absorption without excitation. In the 0-1.6 THz band, each unexcited HCN molecule would absorb about ?6.4?10?29 J of energy; with excitation, each molecule instead emits ? 3.6 ? 10?29 J. The seed terahertz pulse has a fluence of about 5 nJ/cm2, which could be generated in femtosecond-illuminated ZnTe. Extracting 3.6 ? 10?29 J from each molecule means a small-signal gain of about 0.85 nJ/cm2 per centimeter of propagation, per atmosphere of HCN pressure. After propagating only a few centimeters, the high-frequency component (> 1.6 THz) of the seed pulse?s spectrum will be completely absorbed and the remaining ampli- fied spectrum will closely match the amplification band (0 ? 1.6 THz), indicating significant pulse reshaping. The absorption of the terahertz pulse depends on relative delay between the optical pump pulses and the terahertz pulse. ?t = Tr/2 was chosen after lim- ited exploration because it seemed to maximize gain, but full exploration of ?t for the parameters discussed here is beyond our computational resources. Chapter 8 discusses the effect of ?t on terahertz absorption in more detail. The ? 200 fs terahertz pulse used in Fig. 7.1 is too short to be efficiently amplified, since most of its spectrum lies above the 0-1.6 THz amplification band. Figure 7.2 shows the same calculation for a longer (?800 fs) terahertz pulse with the 76 10 20 30 0.09 0.7 1.31 1.92 2.52 10 20 3030 15 0 ?15 ?30 0.09 0.7 1.31 1.92 2.52 ?0.8 ?0.6 ?0.4 ?0.2 0 0.2 ?0.5 0 0.5 ?2 ?1 0 1 2 ?4 ?2 0 2 x 10?30 Energy emitted per molecule, per transition (J) Electric field (kV/cm) Time (ps)j? j?1 resonant emission frequency Initial rotational angular momentum (j) of transition Energy emitted per molecule (J ? 10 28 ) Z?angular momentum (m) (a) (b) (c) (d) (e) Figure 7.2: Energy absorbed/emitted (?Emj ) by 800 fs-pulse terahertz- driven j ? j ? 1 transitions vs. j and m of the upper state for (a) an initially thermal HCN gas and (b) the optically-excited rotational populations shown in Fig. 6.2(c, f). Summed over m, this gives energy absorbed/emitted near each resonant frequency 2jf0 for (c) thermal and (d) excited gas. The ?800 fs terahertz pulse driving these transitions (e) follows at Tr/2 after the last optical excitation pulse. 77 same fluence. The longer pulse better overlaps both the amplification and thermal absorption bands. In the 0-1.6 THz band, each unexcited HCN molecule would absorb about ?8?10?28 J of energy; with excitation, each molecule instead emits ? 3 ? 10?28 J, giving a more respectable gain of 7.1 nJ/cm2 per centimeter of propagation per atmosphere of HCN. Note also that there is very little absorption of this pulse by the excited molecules shown in Fig. 7.2(b, d), indicating little pulse reshaping. The temporal shape of this terahertz emission depends on collisions. The time- domain simulation used here only calculates the energy ?E absorbed or emitted while the seed pulse interacts with the molecules, so post-pulse terahertz emission is not considered. At low pressures (? 1 atm), in addition to absorbing or amplifying the original pulse, molecular rotations will remain coherent for many Tr and will also generate a following train of few-cycle terahertz pulses spaced by Tr [70]; this is beyond the scope of our simulation. However, use of a cavity to pass the terahertz pulse repeatedly through the excited low-pressure gas allows collection of this excess energy into the original seed pulse, further increasing gain. The extracted energy shown in Figures 7.1 and 7.2 is less than 10?6 of the absorbed optical pump energy. Maximum extractable energy can be estimated be- cause the first three optical pump pulses drive HCN to the cusp of inversion, and only the fourth optical pump pulse drives population inversion. A strong terahertz pulse which saturates extraction would drive the populations back to this uninverted state, extracting ?41 mJ/cm3 per atmosphere of pressure. For pressures of ?1% of an atmosphere, this is a small fraction of the input pulse. 78 5 10 15 20 25 30 35 ?1 ?0.5 0 0.5 1 1.5 2 2.5 3 x 10?29 Transition upper level (j)Scaled energy released (Joules ? I/I 0) 5 nJ/cm2 328 ?J/cm2 655 ?J/cm2 1.31 mJ/cm2 2.62 mJ/cm2 5.24 mJ/cm2 10.5 mJ/cm2 Figure 7.3: Scaled gain vs. frequency for a range of terahertz pulse fluences Saturated extraction will generally drive population transitions and nonlinear effects, but the 5 nJ/cm2 terahertz pulse fluence studied here is well below satura- tion; doubling the terahertz intensity precisely doubles the extracted energy from each transition. Figure 7.3 shows the effects of saturation on gain for high tera- hertz pulse fluences, comparing the extracted energy per transition shown in Figure 7.2(d) for a 5 nJ/cm2 terahertz pulse to a range of higher pulse fluences. For easy comparison, this extracted energy is scaled by pulse fluence, so if gain stays linear, the plots will fall on top of one another. In fact, it takes 16 doublings of pulse fluence to show any significant deviation from this linear behavior; extraction of less than 100 ?J/cm3 per atmosphere leaves the rotational populations changed by less than one part in 500. For terahertz fluence of ?328 ?J/cm2, the shape of the gain curve is barely distinguishable from 79 the 5 nJ/cm2 case. Not until ? 10.5 mJ/cm2 does saturation finally drastically change the shape of the gain curve. Since demonstration of a 100 ?J terahertz pulse at a large accelerator facility is a significant achievement [78], and terahertz wavelengths make it difficult to focus to spot areas much smaller than 1 mm2 (which would give a fluence of 100 mJ/cm2 for the 100?J pulse in [78]), this is an extremely promising result. 7.2 Gain at lower temperature All the data presented in this chapter so far is for low-pressure gas targets where collisions are negligible, corresponding to pressures of < 10?2 atmospheres. Gain at these pressures would be fairly low, requiring long interaction lengths for largeamplification. Decreasing temperature to decrease collisions without sacrificing density is appealing, but HCN?s vapor pressure drops off rapidly for temperatures below 300 K. The use of a gas jet rather than a static gas target allows the use of higher density targets with long collision times. Gas jet targets also have a much lower rotational temperature [96], which makes it easier to drive population inversions, as shown in Figure 6.4 for a low temperature (T = 50 K) HCN gas jet target. Figure 7.4 compares the terahertz gain in this gas jet target to the gain shown in Figure 7.2(d), using the same 800 fs single-cycle pulse. The energy extracted per molecule is much larger, about 2.8? more than a room temperature target. This is both because of the steeper population inversions driven in the cold target, 80 5 10 15 20 25 30 35 ?5 0 5 10 15 x 10?29 Energy emitted per molecule, per transition Initial j of transitionEnergy (J) per molecule, per transition 300K HCN target 50K HCN target About 2.8x more gain in 50K target Figure 7.4: Comparison of extracted energy for low- and room- temperature targets. and because low temperatures concentrate population in low m-states which have a larger effective dipole interaction with terahertz fields. Note that there is more loss at higher frequency in the cold target, indicating that the gain bandwidth of the cold target does not extend quite as high. The 800 fs terahertz pulse used here is not ideally matched to the cold target; a slightly longer pulse with the same fluence would experience even more gain. Estimating ?jet = ?staticradicalbigTjet/Tstatic and assuming rotational and translational temperatures are similar [94], collisional decay would be 2.5 times slower in 50 K HCN at comparable density. Gas jets can probably achieve even lower rotational and translational temperature [96], allowing higher density with fewer collisions for efficient pumping and long excitation lifetimes, and further increasing terahertz gain. 81 Chapter 8 Coherent terahertz absorption in laser-aligned molecules 8.1 Suppression or enhancement of absorption The terahertz amplification shown in Chapter 7 is calculated for a terahertz pulse copropagating with the optical excitation pulses, with constant relative de- lay ?t = Tr/2. Interestingly, varying the relative delay between the optical and terahertz pulses can strongly modulate the terahertz absorption or amplification; ?t = Tr/2 was chosen empirically to optimize amplification. Chapters 7 and 6 focused on driving population inversions for amplification, but optical pulses that are far too weak to drive changes in population (??mj,j ??mj,j) can still create coherence (?mj,j?2 negationslash= 0). Physically, these pulses are too weak to change the mean rotational energy of a gas of molecules (?E ?kT), but they can still make different molecules in the gas more likely to point in the same direction. This coherence can have significant impact on terahertz absorption in the excited gas, as shown in Figure 8.1. A 1 TW/cm2, 100 fs optical pulse weakly aligns a gas of HCN molecules (|?mj,j?2|> 0 and?mj,j?2/?mk,k?2 = |?mj,j?2/?mk,k?2| at the peak of the optical pulse) while driving negligible change in populations (?mj,j ??mj,j,thermal). The same ? 200 fs input terahertz pulse as Fig. 7.1 follows the optical alignment pulse with variable ?t; HCN?s rotational absorption/emission is wildly modulated by this alignment depending on ?t. For clarity and computational reasons we show 82 ?4 ?2 0 30 25 20 15 10 5 Emission (J?1031) per molecule Which j ? j?1 transition 1 1.25 1.5 1.75 2 ?60 ?50 ?70 Optical/terahertz relative delay (revival periods) Emission (J ?10 31 ) per molecule ?2 ?1.5 ?1 ?0.5 0 0.5 1 x 10?30(a) (b) (c) Emission per transition per molecule (J) Figure 8.1: A 1 TW/cm2 optical pulse weakly aligns HCN and strongly modulates (a) terahertz energy absorbed/emitted per molecule, perm = 0, j ?j?1 transition depending on the optical-terahertz relative delay ?t. (b) Averaging over delay shows the ?incoherent? absorption that a terahertz pulse would experience if it was not collinear with the optical pump pulses. (c) Summing over j gives total absorption vs. ?t. 83 only the m = 0 case, but the behavior is similar for all m. A terahertz pulse copropagating with the optical excitation pulse would have roughly constant ?t, and absorption could be enhanced (?t = nTr, where n is an integer), suppressed ( ?t = (n + 1/2)Tr ), or oscillatory in spectrum, as shown in Fig. 8.1(a). In Figure 8.1(c), absorption is summed over j to show absorbed energy ?E vs. relative delay ?t. Absorption is suppressed by more than 30% at ?t = (n+ 1/2)Tr, and enhanced by ? 25% at ?t = nTr. If this result for linear molecules could be extended to asymmetric-top molecules like H2O, then perhaps an optical prepulse could be used to suppress the absorption of terahertz pulses in the atmosphere by water vapor which hampers terahertz remote sensing. If the terahertz pulse is propagating at some angle to the optical beam path, then ?t will slip, and the terahertz absorption will average over delays as shown in Fig. 8.1(b). The absorption shown in Fig. 8.1(b) is very similar to the thermal absorption for the m=0 portion of Figure 7.1(a). HCN optically driven into a population inversion generally must be described with a much larger density matrix, making an exhaustive scan of ?tcomputationally infeasible. Our limited simulations (not shown) show that this ?incoherent? delay-averaged absorption depends mainly on populations ?mj,j rather than coherences ?mj,knegationslash=j, as in [70]. 8.2 Extra terahertz revival caused by optical prealignment As described in Chapter 5, we neglect back-action of the molecular emission on the ensemble. We can calculate transitions driven by the applied pulses but 84 Time (in revival periods) Relative optical/THz delay (in revival periods) 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 3 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 ?2 ?1 0 1 2 x 10?5 Figure 8.2: Expectation value of orientation ?cos(?)? vs. time, follow- ing the same optical/terahertz pulse pair from Figure 8.1, for different relative delays ?t. not self-stimulated transitions like those observed in free-induction decay [70]. The absorption shown in figure 8.1 considers only energy absorbed or emitted during the externally applied 200 fs terahertz pulse. Modeling the few-cycle terahertz pulses emitted picoseconds later by free-induction decay would require including nonlocal effects such as propagation (solving the coupled Maxwell-Bloch equations) and increase the simulation burden beyond our resources. However, we can compute the expectation value of orientation ?cos(?)? driven by the combination of optical and terahertz pulses. Since this orientation leads directly to macroscopic polarization (Pz = N??cos(?)?) and emission of short ter- 85 1 1.2 1.4 1.6 1.8 2 2.2 2.4 2.6 2.8 ?1.5 ?1 ?0.5 0 0.5 1 1.5 2 2.5 Time (in revival periods) Orientation ( ?Cos[ ?]? ?10 5 ) 0 TR 0.125 TR 0.25 TR 0.375 TR 0.5 TR Figure 8.3: Expectation value of orientation ?cos(?)? vs. time, follow- ing the same optical/terahertz pulse pair from Figure 8.1, for selected relative delays ?t. (offset vertically for clarity) ahertz pulses [70], it is useful for estimating the effects of optical prealignment on terahertz-driven free induction decay. Figure 8.2 shows ?cos(?)? vs. time, driven by the same combination of optical and terahertz pulses from Figure 8.1, for a range of relative optical/terahertz delays ?t; the optical pulse arrives at t = 0. In the absence of the optical pulse, the terahertz pulse would drive prompt polarization followed by subpicosecond bursts of polarization separated by Tr, as observed in [70]. The addition of the optical pulse causes extra revivals, that follow atnTr?2?t after the terahertz pulse, where n is an integer. When the ?extra? revivals caused by the optical pulse are well separated from the ?normal? revivals that would occur with the terahertz pulse alone, they are significantly larger. For certain delays (?t = nTr/2), these extra revivals merge into the ?normal? revivals, and become similar in size to ?normal? revivals. 86 Figure 8.3 shows the same ?cos(?)? vs. time for selected ?t. Each plot is offset vertically for clarity. The ?extra? revivals are much larger than the ?normal? revivals, as much as five times larger for the ?t = Tr/4 case. Without considering complicated propagation effects, it is hard to say if this large revival of ?cos(?)? would give rise to an intense terahertz pulse. Because of energy conservation, the short terahertz pulses produced in nor- mal free induction decay [70] must contain less energy than the terahertz pulse that drives them. However, the addition of an intense optical prepulse provides an additional reservoir of energy which this process might efficiently transfer to a trailing terahertz pulse. If so, this would represent a promising source of intense terahertz radiation that requires significantly less optical intensity than the ampli- fication scheme described in Chapter 7. We plan to pursue this question in future work. 87 Chapter 9 Feasibility of experimental study of THz properties of aligned molecules 9.1 Discussion of gases- HCN, CH3Cl, OCS, N2O Table 9.1 gives an overview of the properties of several linear, dipolar molecular gases that our lab might use to study the effects of molecular alignment on terahertz absorption and emission. The importance of these properties can be understood from Equation 5.9, which describes the time evolution of the rigid rotor density matrix. The effective strength of the terahertz electric field in Eq. 5.9 is multiplied by the dipole moment ?, so a large value of ? is essential for strong terahertz interaction. Similarly, optical field intensity is multipled by ??, which is important for strong laser-driven alignment. The rotational revival timeTr (related to moment of inertia I by Tr = 2?I/?h) influences the equation in two ways: it sets the field-free spacing between rotational energy levels throughEj = (?h?/Tr)j(j+1), which in turn influences the thermal equilibrium distribution of j-states through the Boltzmann factor exp parenleftBig?E j kBT parenrightBig . Effectively, ?, ??, and the temperature are all normalized by their product with the moment of inertiaI. Finally, the boiling point isimportant for static gas targets, which give better terahertz amplification when cooled as discussed in Chapter 7. A lower boiling point allows more cooling before condensation limits 88 gas pressure. Table 9.1: Overview of dipolar gases for optical/terahertz experiments. Data col- lected from [63, 86, 97, 98] Gas Diff. polarizability Revival time Perm. dipole Boiling point ?? (esu?10?25) Tr (ps) ? (Debye) TB (K) HCN 19.1 11.5 2.98 299 CH3Cl 12.8 37.6 1.9 249 OCS 40-60 83.4 0.709 223 N2O 27.9 40.5 0.166 185 All of the simulations discussed so far use hydrogen cyanide (HCN) gas. HCN is particularly appealing for optical alignment because of its large ratio of differen- tial polarizibility ?? to moment of inertia I, making it easier to drive population inversions; all terms in the rigid rotor equation of HCN?s low moment of inertia also means that at a given temperature, more population is concentrated at low j- values which makes it easier to drive population inversion (as shown in Figure 6.4). Low moment of inertia also simplifies multi-pulse optical pumping: a short revival time allows more optical pulses to fit into the same time window, allowing higher pressures before collisions interfere with population inversion. HCN also has a par- ticularly large permanent dipole moment ?, so it interacts strongly with terahertz radiation; the strength of terahertz absorption/amplification generally scales like ?2 [63]. However, HCN is both poisonous and explosive, which make it undesirable for lab use. Any gas jet target using HCN must solve the difficult engineering problem of how to safely pump the target chamber. HCN?s vapor pressure is a bit low, which limits cooling of a static gas target much below room temperature. 89 Methyl chloride, (CH3Cl) isinferior in many ways to HCN for optical alignment and terahertz amplification. It has a lower dipole moment, higher moment of inertia, and a smaller differential polarizability. However, CH3Cl is less dangerous than HCN, so it would be easier to use in a gas jet target, and it has a slightly higher vapor pressure, so it could be cooled more than HCN as a static gas target. CH3Cl is not a linear molecule, so technically the analysis presented so far does not apply to CH3Cl. However, since hydrogen is so light, the energy levels corresponding to rotation about the C-Cl axis of CH3Cl are very widely spaced, and are therefore neglectable for the same reason as vibrational excitation. Finally, the chlorine in CH3Cl is commonly an isotopic mixture, so unless isotopically pure CH3Cl is used, multi-pulse optical pumping will discriminate between the two isotopes [99]. Carbonyl sulfide (OCS) is also safer than HCN, and has a very high differen- tial polarizability, which is important for efficiently absorbing energy from optical pulses. The high moment of inertia of OCS makes population inversion difficult to drive, especially at high pressures. Our simulations (not shown) also show ini- tial population inversion in OCS will amplify lower frequencies than HCN, in the 150-450 GHz range; harder optical pumping could drive this amplification to higher frequencies as demonstrated in Figure 6.5. However, once population inversion is achieved in OCS, the high moment of inertia would mean terahertz amplification saturates at much higher energies than in HCN. OCS could be used in a gas jet like CH3Cl, and can be cooled even more as a static target before condensing. However, the single biggest drawback of OCS is its low dipole moment, leading to a weaker interaction with terahertz than HCN or CH3Cl. Note that ?? of OCS is reported 90 as ?60?10?25 esu in [86], but only ?40?10?25 esu in [100], which would lead to significantly different optical pumping requirements. Nitrous oxide (N2O) is even safer than CH3Cl or OCS, has a low boiling point, and a decent differential polarizability. However, its low dipole moment makes terahertz interactions extremely weak compared to the other gases considered here. The optical ionization threshold of these gases is also an important parameter, but because the ionization threshold is less clearly defined than the parameters in Table 9.1, it is slightly harder to compare among the different gases. For our purposes, ?ionization threshold? should mean the laser intensity a gas can withstand before the destruction of molecules interferes with amplification and/or the density of free electrons becomes high enough to interfere with the propagation of a terahertz pulse. Reference [80] estimates an ionization threshold of 140 TW/cm2 for HCN. If HCN really could withstand this intensity without ionization, a single optical pulse with 60 TW/cm2 intensity could drive strong population inversion in HCN at extremely high pressure, and a terahertz pulse following close behind could be amplified hundreds of times more strongly than in the low pressure environment required for multi-pulse optical pumping. Experimental investigation is neccesary, but could give a tremendous boost to terahertz amplification. The ionization threshold of CH3Cl is well studied in [101], where a laser in- tensity of 46 TW/cm2 in a 40 fs pulse produces noticable ionization. Since CH3Cl requires significantly more laser intensity than HCN to drive population inversion, this probably cannot be done with a single pulse without excessive ionization. 91 Laser ionization of OCS is studied in [102], and is estimated to occur at 30 TW/cm2. A single optical pulse with this intensity drives OCS to the cusp of population inversion, so this is certainly worth further study. 9.2 Free space geometry The suggested terahertz amplification scheme in Figure 6.1 assumes a plane- wave geometry. In practice, construction of a terahertz laser requires getting the propagation of an intense optical pulse and a terahertz pulse to overlap significantly in a dipolar gas. This presents several difficulties: ? A terahertz pulse following an intense optical pulse in a dipolar gas can be amplified, but is strongly absorbed by the same gas in the absence of optical pumping, as shown in Figure 7.1. The terahertz beam must therefore overlap the optical beam nearly everywhere in the amplifying medium. ? Because terahertz radiation has much longer wavelength than optical radia- tion, a terahertz beam will diffract much more rapidly than an optical beam with the same diameter. This means the optical pump beam must be large to overlap the terahertz beam (at least several mm2), which requires a high-power optical beam to reach high intensity. ? A beam of optical pulses intense enough to cause rotational population inver- sion in a gas is typically too intense to touch any solid object (like a lens or mirror) without damaging it; diffraction of the optical pulse can reduce the intensity, but this constrains experimental geometry. 92 Figure 9.1: Free-space geometry for combining optical and terahertz beams in a molecular gas jet. ? The intense optical beam must not cause excessive ionization of the dipolar gas, which could disrupt terahertz propagation. Controlling diffraction of the optical and terahertz pulses is essential so that the optical and terahertz beams interact for a long enough distance to give strong am- plification. Figure 9.1 shows one possible way to combine an intense optical beam with a terahertz beam in a gas jet. The optical beam is nearly collimated, with a few-mm diameter beam waist in the center of the figure. The focusing optic for the optical beam is many centimers away from the interaction region, where the beam is large enough to avoid damaging the optic. The terahertz beam has a similarly sized beam waist, but diffracts much more rapidly because of its longer wavelength; it is focused and collimated with a pair of reflective off-axis paraboloids (OAP?s) which 93 Figure 9.2: Free-space geometry for combining optical and terahertz beams in a static gas-filled tube; the tube is transparent to terahertz radiation. have a focal length of only a few centimeters. The optical beam passes through a narrow hole drilled in each OAP, and the terahertz and optical beams intersect in a molecular gas jet. The entire assembly is enclosed in a vacuum chamber which must be continuously pumped to keep the backgroud gas pressure low. The terahertz beam only encounters gas directly below the gas jet, which avoids absorption of the terahertz beam before it overlaps the optical beam. Figure 9.2 shows a similar experimental setup for using a static gas target. The target dipolar gas is confined to the interaction region by a tube made of a material that is transparent to terahertz, such as teflon or polyethylene. The terahertz only passes through gas which is pumped by the intense optical beam, and then diffracts out of the tube to be recollimated by the second OAP. This setup does not require a vacuum chamber or pumping, but the optical beam must enter the gas-filled tube many centimeters from the interaction region to avoid damaging the input window. 94 In both these cases, spherical aberration of the optical beam is actually de- sirable; a diffraction-limited spot would be either too intense (ionizing the target gas) and too small (underfilling the terahertz spot), or would diffract very slowly, requiring its focusing optics to be very far from the interaction region. Diffraction of the terahertz beam limits the interaction length in both these scenarios. A 1 THz beam has a wavelength of ? 300 ?m, so if it is focused to a 1 mm2 spot, it will have a Rayleigh length of only a few millimeters. Weaker focusing to larger spots will give longer interaction lengths, but as the terahertz beam waist grows, the optical beam diameter must also grow. Since optical powers of > 1 TW/cm2 are required to see strong optical-terahertz interaction, only multi- terawatt lasers could have very long interaction lengths in free-space geometry. Since the absorption of the optical beam by rotational transitions is very low (as estimated in Chapter 6, in section 6.2), the same optical pump pulse could be recycled and passed through the same gas target many times. Similarly, the terahertz pulse could be passed through the gas for higher gain. Both beams could be placed in a cavity for efficient multi-pass operation. 9.3 Wire guiding geometry If terahertzdiffraction could besuppressed, optical-terahertzinteraction lengths could be greatly extended. Of course, the waveguide structure would have to with- stand a high-intensity optical beam and allow terahertz radiation to propagate in a dipolar gas. 95 Wang and Mittleman demonstrated that a bare steel wire can guide terahertz radiation as a surface wave, extending into the gas around the wire [103]. This wire waveguiding method has been used to perform terahertz spectroscopy of lactose powder spread on the surface of the wire [104]. Normally, terahertz spectroscopy requires a high-volume sample [70] for a large-diameter terahertz beam to have a long interaction length. In [104], the wire suppresses diffraction, allowing a very small volume of powder coating the surface of the wire to interact with the terahertz pulse over a long distance. Since the terahertz surface wave only extends ?1 mm from the wire, a significant fraction of the beam interacts with the powder over many centimeters of propagation. If this same wire terahertz waveguide were in a gas chamber surrounded by HCN or CH3Cl, the guided terahertz pulse would be absorbed by a thin layer of gas surrounding the wire. However, if this thin layer of gas was optically pumped into a rotational population inversion, then the guided terahertz pulse would be amplified. A high-intensity optical pulse could propagate around the wire (as described in Section 9.3.2), driving this thin layer into population inversion. Since the hollow optical beam could have a cross sectional area of only a few mm2, this interaction could be very efficient. As demonstrated in [103], the 1/eattenuation length for wire waveguiding is> 50 cm for frequencies >0.25 THz. This is near the break-even point for amplification in low-pressure HCN; 1% of an atmosphere of HCN has a gain length of 70 cm for the amplification parameters used in Figure 7.2. Gain can be increased past this break-even by several methods. Using higher 96 pressure and an extra pump pulse, as studied in Figure 6.3, can cut the gain length by a factor of two or three. Using a gas jet rather than static gas to surround the wire could increase gain by ?10 or more as discussed at the end of Chapter 7, giving gain lengths less than 10 cm. If a single 60 TW/cm2 optical pulse could drive population inversion in HCN without causing enough ionization to disrupt terahertz guiding, then gas pressures of 1-10 atm could be used, giving centimeter or millimeter gain lengths. The primary advantage of the wire waveguide is the possibility of extended interaction length. Limited only by diffraction of the optical beam, it could extend interaction for tens or hundreds of centimeters. Since the optical beam needs a cross-section of only a few mm2, then a regenerative amplifier-based system with a peak power of only a few tens of gigawatts could be used as the optical pump for the coherent effects discussed in Chapter 8. A 2 TW, 100 fs laser system could be used as a pump for the four-pulse population inversion described in Figure 6.2. 9.3.1 Coupling THz onto and off of the wire Wang and Mittleman use a second steel wire to couple terahertz pulses onto their waveguide [103]. This coupling method is not suitable for the geometry con- sidered here, because it would block the path of the intense optical beam. Cao and Nahata demonstrated an alternative terahertz input coupler by milling a groove di- rectly into the guide wire [105]. This input coupler is well suited for optical/teraherz interaction and would not shadow the optical beam. However, this method suffers 97 from low coupling efficiency. The same group later improved on this scheme, coating their milled groove with a poled polymer and producing a terahertz pulse directly on the wire [106]. Figure 9.3 shows how this type of terahertz coupling could be used for ex- tended optical/terahertz interaction in a dipolar gas. The input coupler is excited by an optical pulse producing a guided terahertz wave (a pulse from a femtosecond oscillator was used in [106]). The entire assembly is surrounded by a dipolar gas such as HCN. An intense, hollow optical beam precedes the terahertz pulse, driving the surrounding gas into a rotational population inversion. The guided terahertz pulse follows behind and is amplified by the excited gas. Wire-guided terahertz is most simply outcoupled by terminating the wire; the terahertz rapidly diffracts away [103]. If the wire is not rigid enough and needs sup- port, a deep, wide milled groove like that used in [105] could serve the same purpose. Diffraction will rapidly separate the optical pump beam from the terahertz beam, which could then be divided using a drilled OAP like the ones shown in Figures 9.1 and 9.2. Alternatively, a teflon or polyethylene window in the terahertz path would transmit the terahertz while blocking the optical pulse, preventing absorption of the terahertz radiation by the target gas. 98 Figure 9.3: (a) Steel wire for terahertz waveguiding, coated with a poled polymer [106] and surrounded by a dipolar gas. (b) Optical excitation produces a terahertz pulse which is guided along the wire. (c) A hol- low optical pulse precedes the terahertz pulse, exciting the surrounding dipolar gas to amplify the terahertz. 99 9.3.2 Skimming the hollow optical beam along the wire: ionization, polarization It is crucial that the intense optical beam not damage the wire terahertz waveg- uide. As a preliminary investigation, we suspended a steel wire and surrounded it with an intense optical beam as illustrated in Figure 9.3(c). We used a slightly coverging, 500 GW, 100 fs, 800 nm pulse, and it ranged from ?1 cm diameter to ?2 mm diameter. No sign of damage could be detected. The wire surface appeared unchanged under a microscope, no ?snapping? noise was audible, and there was no noticable smell. It is possible, of course, that some ionization occured at the surface of the wire, but a more sensitive experiment would be neccesary to measure this effect. The polarization direction of the optical pulse used in Figure 9.3(c) is also important. All analysis performed in previous chapters considers linearly polarized optical and terahertz polarization, and assumes these polarization vectors point in the same direction. However, the terahertz mode guided by wire waveguides is radially polarized. The optical pump pulse could also be radially polarized for ease of analysis, but this presents experimental complications. As noted in Chapter 5, circularly polarized light can also drive rotational population inversion. Using a circularly polarized optical pump pulse will generate a cylindrically symmetric gain profile to match the cylindrically symmetric mode of the wire-guided terahertz pulse. Our simulation code currently can only model terahertz polarization in the z-direction, but future work will explore the case of 100 terahertz linearly polarized in the x or y direction. We expect the optical pump polarization to affect coherent dynamics like those in Chapter 8, but since they drive the same population dynamics as linearly polarized pulses, terahertz gain should be similar. 9.4 Terahertz detection The detection schemes used in most wire-based terahertz waveguide experi- ments are multi-shot, scanning techniques [103, 105, 106]. These experiments typi- cally use unamplified femtosecond oscillators with multi-megahertz repetition rates. Molecular alignment typically requires amplified optical pulses with kilohertz or 10- hertz repetion rates, and such systems are not as stable shot-to-shot as unamplified oscillators. Single-shot terahertz detection as used in [107] allows accurate mea- surement despite shot-to-shot fluctuations. We have constructed a similar terahertz pulse diagnostic based on electro-optic sampling to characterize our terahertz pulses, shown in Figure 9.4. The terahertz beam is combined with an optical beam using a pellicle. Both beams then pass through a zinc telluride (ZnTe) crystal where the ter- ahertz beam underfills the optical beam. The terahertz pulse induces birefringence in the crystal, which distorts the polarization of the optical pulse. The optical pulse is then imaged through a polarizer onto a CCD camera or imaging spectrometer. The polarizer is oriented to minimize transmission of the optical pulse in the absence of terahertz, so any change in transmission indicates the presence of ter- ahertz radiation. If the optical pulse is transform-limited, the CCD displays a 101 Figure 9.4: Terahertz diagnostic for measuring electric field vs. time and beam profile. ?snapshot? of the terahertz beam profile with time resolution equal to the optical pulse duration. We can measure the electric field vs. time of the terahertz beam by scanning the relative delay between the optical and terahertz beams. If the optical pulse is chirped, the different frequency components of the optical pulse sample different delays relative to the terahertz pulse. The imaging spectrom- eter displays the transmission of each frequency component, which can be processed into the electric field vs. time of the terahertz pulse as described in [107]. The pellicle we use has a thickness comparable to our optical wavelength, and exhibits strong variation in reflectivity vs. frequency over the spectrum of the optical pulse. In addition, it is very susceptible to vibration; even a nearbye conversation causes noticable deflection of the optical beam. The ZnTe crystal [108] shows residual birefringence comparable to the bire- 102 fringence generated by our terahertz beam. Combined with the spectral modulation of the pellicle and the attenuation due to the grating in our imaging spectrometer, this has prevented us from operating in single-shot mode so far. We plan to replace the pellicle with a small metal mirror. The mirror will block part of the terahertz beam, but will have better stability and higher, more spectrally uniform reflectivity. The brighter probe beam, combined with a more uniform spectrum, should allow single-shot measurement of our terahertz pulse electric field vs. time, with 1-D spatial resolution. 103 Bibliography [1] V. Yanovsky, V. Chvykov, G. Kalinchenko, P. Rousseau, T. Planchon, T. Mat- suoka, A. Maksimchuk, J. Nees, G. Cheriaux, G. Mourou, and K. Krushel- nick. Ultra-high intensity- 300-tw laser at 0.1 hz repetition rate. Opt. Express, 16(3):2109?2114, 2008. [2] http://lasers.coherent.com/lasers/regenerative [3] T. Ditmire. Nuclear fusion from explosions of femtosecond-laser heated deu- terium clusters. APS Meeting Abstracts, pages 105?+, November 1999. [4] K. Kim, I. Alexeev, and H. Milchberg. Single-shot measurement of laser- induced double step ionization of helium. Opt. Express, 10(26):1563?1572, 2002. [5] P. F. Moulton. Spectroscopic and laser characteristics of Ti:Al2O3. Journal of the Optical Society of America B Optical Physics, 3:125?133, January 1986. [6] C. Spielmann, P. F. Curley, T. Brabec, and F. Krausz. Ultrabroadband fem- tosecond lasers. IEEE Journal of Quantum Electronics, 30:1100?1114, April 1994. [7] D. Strickland and G. Mourou. Compression of amplified chirped optical pulses. Optics Communications, 56:219?221, December 1985. [8] Michael D. Perry and Gerard Mourou. Terawatt to petawatt subpicosecond lasers. Science, 264(5161):917?924, 1994. [9] www.vjultrafast.org/. [10] A.H. Zewail. Femtochemistry: Atomic-scale dynamics of the chemical bond. Journal of Physical Chemistry A, 104(24):5660?5694, 2000. [11] K. Koenig. Femtosecond laser application in biotechnology and medicine. In I. Miyamoto, H. Helvajian, K. Itoh, K. F. Kobayashi, A. Ostendorf, and K. Su- gioka, editors, Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, volume 5662 of Society of Photo-Optical Instrumentation Engineers (SPIE) Conference Series, pages 255?267, October 2004. [12] J. Kr?uger, W. Kautek, and H. Newesely. Femtosecond-pulse laser ablation of dental hydroxyapatite and single-crystalline fluoroapatite. Applied Physics A: Materials Science & Processing, 69:403?407, 1999. [13] John L. Hall. Nobel lecture: Defining and measuring optical frequencies. Reviews of Modern Physics, 78(4):1279, 2006. 104 [14] C. Joshi. The development of laser- and beam-driven plasma accelerators as an experimental field. Physics of Plasmas, 14(5):055501, 2007. [15] T Lo?er, M Kress, M Thomson, T Hahn, N Hasegawa, and HG Roskos. Comparative performance of terahertz emitters in amplifier-laser-based sys- tems. Semiconductor Science And Technology, 20(7):S134?S141, JUL 2005. One should read this reference carefully; they appear to confuse micro-joules with nano-joules in several places (perhaps due to typographical error), over- stating terahertz pulse energies by three orders of magnitude. [16] http://www2.slac.stanford.edu/vvc/accelerators/structure.html. [17] J. D. Jackson. Classical electrodynamics. 92/12/31, New York: Wiley, 1975, 2nd ed., 1975. [18] Levi Schachter. Beam-wave interaction in perodic and quasi-periodic struc- tures. Berlin :Springer-Verlag, 1997. [19] W. D. Kimura, G. H. Kim, R. D. Romea, L. C. Steinhauer, I. V. Pogorel- sky, K. P. Kusche, R. C. Fernow, X. Wang, and Y. Liu. Laser acceleration of relativistic electrons using the inverse cherenkov effect. Phys. Rev. Lett., 74(4):546?549, Jan 1995. [20] T. Plettner, R. L. Byer, E. Colby, B. Cowan, C. M. S. Sears, J. E. Spencer, and R. H. Siemann. Visible-laser acceleration of relativistic electrons in a semi-infinite vacuum. Physical Review Letters, 95(13):134801, 2005. [21] D. Li and K. Imasaki. Vacuum laser-driven acceleration by a slits-truncated Bessel beam. Applied Physics Letters, 86(3):031110?+, January 2005. [22] J D Lawson. Lasers and accelerators. Nuclear Science, IEEE Transactions on, 26(3):4217?4219, June 1979. [23] Eric Esarey, Phillip Sprangle, and Jonathan Krall. Laser acceleration of elec- trons in vacuum. Phys. Rev. E, 52(5):5443?5453, Nov 1995. [24] T. Tajima and J. M. Dawson. Laser electron accelerator. Phys. Rev. Lett., 43(4):267?270, Jul 1979. [25] H. A. H. Boot and R. B. R.-S.-Harvie. Charged Particles in a Non-uniform Radio-frequency Field. Nature, 180:1187?+, November 1957. [26] W. P. Leemans, B. Nagler, A. J. Gonsalves, C. T?oth, K. Nakamura, C. G. R. Geddes, E. Esarey, C. B. Schroeder, and S. M. Hooker. GeV electron beams from a centimetre-scale accelerator. Nature Physics, 2:696?699, October 2006. [27] E. Esarey, P. Sprangle, J. Krall, and A. Ting. Overview of plasma-based accelerator concepts. Plasma Science, IEEE Transactions on, 24(2):252?288, Apr 1996. 105 [28] V. Malka, S. Fritzler, E. Lefebvre, M.-M. Aleonard, F. Burgy, J.-P. Chambaret, J.-F. Chemin, K. Krushelnick, G. Malka, S. P. D. Mangles, Z. Najmudin, M. Pittman, J.-P. Rousseau, J.-N. Scheurer, B. Walton, and A. E. Dangor. Electron acceleration by a wake field forced by an intense ultrashort laser pulse. Science, 298(5598):1596?1600, 2002. [29] ?The Light Fantastic?, The Economist, Sep 28th 2006. [30] S. P. D. Mangles, C. D. Murphy, Z. Najmudin, A. G. R. Thomas, J. L. Collier, A. E. Dangor, E. J. Divall, P. S. Foster, J. G. Gallacher, C. J. Hooker, D. A. Jaroszynski, A. J. Langley, W. B. Mori, P. A. Norreys, F. S. Tsung, R. Viskup, B. R. Walton, and K. Krushelnick. Monoenergetic beams of relativistic elec- trons from intense laser-plasma interactions. Nature, 431:535?538, September 2004. [31] C. G. R. Geddes, C. Toth, J. van Tilborg, E. Esarey, C. B. Schroeder, D. Bruhwiler, C. Nieter, J. Cary, and W. P. Leemans. High-quality elec- tron beams from a laser wakefield accelerator using plasma-channel guiding. Nature, 431:538?541, September 2004. [32] J. Faure, Y. Glinec, A. Pukhov, S. Kiselev, S. Gordienko, E. Lefebvre, J.-P. Rousseau, F. Burgy, and V. Malka. A laser-plasma accelerator producing monoenergetic electron beams. Nature, 431:541?544, September 2004. [33] T. Plettner, P. P. Lu, and R. L. Byer. Proposed few-optical cycle laser-driven particle accelerator structure. Phys. Rev. ST Accel. Beams, 9(11):111301, Nov 2006. [34] B. D. Layer, A. York, T. M. Antonsen, S. Varma, Y.-H. Chen, Y. Leng, and H. M. Milchberg. Ultrahigh-intensity optical slow-wave structure. Physical Review Letters, 99(3):035001, 2007. [35] A. G. York, H. M. Milchberg, J. P. Palastro, and T. M. Antonsen. Direct acceleration of electrons in a corrugated plasma waveguide. Physical Review Letters, 100(19):195001, 2008. [36] H. M. Milchberg, T. R. Clark, C. G. Durfee, III, T. M. Antonsen, and P. Mora. Development and applications of a plasma waveguide for intense laser pulses. Physics of Plasmas, 3:2149?2155, May 1996. [37] J. H. Cooley, T. M. Antonsen, H. M. Milchberg, J. Fan, L. Margolin, and L. Py- atnitskii. Parametric instability in the formation of plasma waveguides. Phys- ical Review E (Statistical, Nonlinear, and Soft Matter Physics), 73(3):036404, 2006. [38] T. R. Clark and H. M. Milchberg. Optical mode structure of the plasma waveguide. Phys. Rev. E, 61(2):1954?1965, Feb 2000. 106 [39] P Serafim, P Sprangle, and B Hafizi. Optical guiding of a radially polarized laser beam for inverse Cherenkov acceleration in a plasma channel. IEEE Transactions on Plasma Science, 28(4):1190?1193, AUG 2000. [40] CG Durfee and HM Milchberg. Light pipe for high-intensity laser-pulses. Physical Review Letters, 71(15):2409?2412, OCT 11 1993. [41] A. York, B. D. Layer, and H. M. Milchberg. Application of the corrugated plasma waveguide to direct laser acceleration. Advanced Accelerator Concepts: 12th Advanced Accelerator Concepts Workshop, 877(1):807?811, 2006. [42] A. Yariv. Quantum Electronics. Wiley, 1989. [43] Ming Xie. Plasma inverse transition acceleration. Particle Accelerator Con- ference, 2001. PAC 2001. Proceedings of the 2001, 5:3876?3878 vol.5, 2001. [44] CG Durfee, TR Clark, and HM Milchberg. Mode control in a two-pulse- excited plasma waveguide. Journal of the Optical Society of America B-Optical Physics, 13(1):59?67, JAN 1996. [45] TR Clark and HM Milchberg. Laser-driven implosion of a cylindrical plasma. Physical Review E, 57(3, Part B):3417?3422, MAR 1998. [46] A. Farjadpour, David Roundy, Alejandro Rodriguez, M. Ibanescu, Peter Bermel, J. D. Joannopoulos, Steven G. Johnson, and G. W. Burr. Improving accuracy by subpixel smoothing in the finite-difference time domain. Opt. Lett., 31(20):2972?2974, 2006. [47] A. York, B. D. Layer, T. M. Antonsen, S. Varma, Y.-H. Chen, and H. M. Milchberg. Ultra-high intensity optical slow wave structure and applications. AIP Conference Proceedings, 926(1):152?161, 2007. [48] R.F. Hubbard, P. Sprangle, and B. Hafizi. Scaling of accelerating gradients and dephasing effects in channel-guided laser wakefield accelerators. Plasma Science, IEEE Transactions on, 28(4):1159?1169, Aug 2000. [49] Andrew G. York, B. D. Layer, J. P. Palastro, T. M. Antonsen, and H. M. Milchberg. Ultrahigh-intensity optical slow-wave structure for direct laser electron acceleration. J. Opt. Soc. Am. B, 25(7):B137?B146, 2008. [50] glue.umd.edu/?york/scat.mpg. [51] G. Machavariani, Y. Lumer, I. Moshe, A. Meir, and S. Jackel. Efficient ex- tracavity generation of radially and azimuthally polarized beams. Opt. Lett., 32(11):1468?1470, 2007. [52] T. van Oudheusden, E. F. de Jong, S. B. van der Geer, W. P. E. M. Op ?t Root, O. J. Luiten, and B. J. Siwick. Electron source concept for single-shot sub- 100 fs electron diffraction in the 100 kev range. Journal of Applied Physics, 102(9):093501, 2007. 107 [53] T Palchan, S Eisenmann, A Zigler, D Kaganovich, RF Hubbard, M Fraenkel, D Fisher, and Z Henis. All optical electron injector using an intense ultrashort pulse laser and a solid wire target. Applied Physics B-Lasers and Optics, 83(2):219?223, MAY 2006. [54] H. Sheng, K. Y. Kim, V. Kumarappan, B. D. Layer, and H. M. Milchberg. Plasma waveguides efficiently generated by bessel beams in elongated cluster gas jets. Physical Review E (Statistical, Nonlinear, and Soft Matter Physics), 72(3):036411, 2005. [55] S. P. Nikitin, I. Alexeev, J. Fan, and H. M. Milchberg. High efficiency coupling and guiding of intense femtosecond laser pulses in preformed plasma channels in an elongated gas jet. Phys. Rev. E, 59(4):R3839?R3842, Apr 1999. [56] http://www.iceditors.com/. [57] http://www.microtronicsinc.com/. [58] http://www.lps.umd.edu/. [59] Mitsuo Takeda, Hideki Ina, and Seiji Kobayashi. Fourier-transform method of fringe-pattern analysis for computer-based topography and interferometry. J. Opt. Soc. Am., 72(1):156?160, 1982. [60] http://www.ni.com/labview/. [61] Marvin A. Schofield and Yimei Zhu. Fast phase unwrapping algorithm for interferometric applications. Opt. Lett., 28(14):1194?1196, 2003. [62] Milan Kalal and Keith A. Nugent. Abel inversion using fast fourier transforms. Appl. Opt., 27(10):1956?1959, 1988. [63] Charles Townes and Arthur Schawlow. Microwave spectroscopy. McGraw-Hill, 1955. [64] H Stapelfeldt. Colloquium: Aligning molecules with strong laser pulses. Re- views of Modern Physics, 75(2):543?557, APR 2003. [65] R Velotta, N Hay, MB Mason, M Castillejo, and JP Marangos. High-order harmonic generation in aligned molecules. Physical Review Letters, 8718(18), OCT 29 2001. [66] RA Bartels, TC Weinacht, N Wagner, M Baertschy, CH Greene, MM Mur- nane, and HC Kapteyn. Phase modulation of ultrashort light pulses using molecular rotational wave packets. Phys. Rev. Lett., 88(1), JAN 7 2002. [67] S. Varma, Y.-H. Chen, and H.M. Milchberg. Effect of aligned nitrogen molecules on atmospheric propagation of ultrashort laser pulses. Lasers and Electro-Optics, 2008 and 2008 Conference on Quantum Electronics and Laser Science. CLEO/QELS 2008. Conference on, pages 1?2, May 2008. 108 [68] J. P. Heritage, T. K. Gustafson, and C. H. Lin. Observation of coherent transient birefringence in cs2 vapor. Phys. Rev. Lett., 34(21):1299?1302, May 1975. [69] Y.-H. Chen, S. Varma, A. York, and H. M. Milchberg. Single-shot, space- and time-resolved measurement of rotational wavepacket revivals in H-2, D-2, N-2, O-2, and N2O. Optics Express, 15(18):11341?11357, SEP 3 2007. [70] H Harde and D Grischkowsky. Coherent transients excited by subpicosecond pulses of terahertz radiation. Journal of the Optical Society of America B- Optical Physics, 8(8):1642?1651, AUG 1991. [71] LH Xu, RM Lees, ECC Vasconcellos, SC Zerbetto, LR Zink, and KM Even- son. Methanol and the optically pumped far-infrared laser. IEEE Journal of Quantum Electronics, 32(3):392?399, MAR 1996. [72] http://lasers.coherent.com/lasers/sifir-50. [73] SH Lee, SJ Petuchowski, AT Rosenberger, and TA DeTemple. Synchronous, mode-locked pumping of gas-lasers. Optics Letters, 4(1):6?8, 1979. [74] AT Rosenberger, HK Chung, and TA DeTemple. Sub-T2 optical pulse gener- ation - application to optically pumped far-infrared lasers. IEEE Journal of Quantum Electronics, 20(5):523?532, 1984. [75] BB Hu, XC Zhang, DH Auston, and PR Smith. Free-space radiation from electrooptic crystals. Applied Physics Letters, 56(6):506?508, FEB 5 1990. [76] DH Auston, KP Cheung, and PR Smith. Picosecond photoconducting Hertzian dipoles. Applied Physics Letters, 45(3):284?286, 1984. [77] K.-L. Yeh, M. C. Hoffmann, J. Hebling, and Keith A. Nelson. Generation of 10 mu J ultrashort terahertz pulses by optical rectification. Applied Physics Letters, 90(17), APR 23 2007. [78] Y. Shen, T. Watanabe, D. A. Arena, C.-C. Kao, J. B. Murphy, T. Y. Tsang, X. J. Wang, and G. L. Carr. Nonlinear cross-phase modulation with intense single-cycle terahertz pulses. Physical Review Letters, 99(4), JUL 27 2007. [79] WP Leemans, CGR Geddes, J Faure, C Toth, J van Tilborg, CB Schroeder, E Esarey, G Fubiani, D Auerbach, B Marcelis, MA Carnahan, RA Kaindl, J Byrd, and MC Martin. Observation of terahertz emission from a laser-plasma accelerated electron bunch crossing a plasma-vacuum boundary. Physical Re- view Letters, 91(7), AUG 15 2003. [80] CM Dion, A Keller, O Atabek, and AD Bandrauk. Laser-induced alignment dynamics of HCN: Roles of the permanent dipole moment and the polarizabil- ity. Physical Review A, 59(2):1382?1391, FEB 1999. 109 [81] Arthur Adel and E. F. Barker. The vibrational energy level system of the linear molecule hcn. Phys. Rev., 45(4):277?279, Feb 1934. [82] Bretislav Friedrich and Dudley Herschbach. Alignment and trapping of molecules in intense laser fields. Phys. Rev. Lett., 74(23):4623?4626, Jun 1995. [83] C. H. Lin, J. P. Heritage, T. K. Gustafson, R. Y. Chiao, and J. P. McTague. Birefringence arising from the reorientation of the polarizability anisotropy of molecules in collisionless gases. Phys. Rev. A, 13(2):813?829, Feb 1976. [84] Klaus Hartinger and Randy A. Bartels. Single-shot measurement of ultra- fast time-varying phase modulation induced by femtosecond laser pulses with arbitrary polarization. Applied Physics Letters, 92(2):021126, 2008. [85] D. V. Kartashov, A. V. Kirsanov, A. M. Kiselev, A. N. Stepanov, N. N. Bochkarev, Yu. N. Ponomarev, and B. A. Tikhomirov. Nonlinear absorption of intense femtosecond laser radiation in air. Optics Express, 14(17):7552? 7558, AUG 21 2006. [86] Joseph Oakland Hirschfelder, Charles F. Curtiss, and Byron R. Bird. Molec- ular theory of gases and liquids. Wiley, 1954. [87] Andrew G. York and H. M. Milchberg. Broadband terahertz lasing in aligned molecules. Optics Express, 16(14):10557?10564, JUL 7 2008. [88] David Griffiths. Introduction to Quantum Mechanics. 2nd edition edition, 2005. [89] V. Renard, M. Renard, A. Rouz?ee, S. Gu?erin, H. R. Jauslin, B. Lavorel, and O. Faucher. Nonintrusive monitoring and quantitative analysis of strong laser- field-induced impulsive alignment. Phys. Rev. A, 70(3):033420, Sep 2004. [90] Michael A. Nielsen and Isaac L. Chuang. Quantum computation and quantum information. Cambridge University Press, 2000. [91] S Ramakrishna and T Seideman. Dissipative dynamics of laser induced nona- diabatic molecular alignment. Journal Of Chemical Physics, 124(3), JAN 21 2006. [92] deepthought.umd.edu. [93] Alex. G. Smith, Walter Gordy, James W. Simmons, and William V. Smith. Microwave spectroscopy in the region of three to five millimeters. Phys. Rev., 75(2):260?263, Jan 1949. [94] Roger Campargue. Atomic and Molecular Beams. Springer, 2000. [95] Sebastiaan Y. T. Van De Meerakker, Hendrick L. Bethlem, and Gerard Meijer. Taming molecular beams. Nature Physics, 4(8):595?602, AUG 2008. 110 [96] Vinod Kumarappan, Christer Z. Bisgaard, Simon S. Viftrup, Lotte Holmegaard, and Henrik Stapelfeldt. Role of rotational temperature in adia- batic molecular alignment. The Journal of Chemical Physics, 125(19):194309, 2006. [97] DS Elliott and JF Ward. Polarizability anisotropies of CO2, N2O, and OCS from measurements of the intensity-dependent refractive-index in gases. Phys- ical Review Letters, 46(5):317?320, 1981. [98] http://webbook.nist.gov/chemistry/. [99] Sharly Fleischer, I. Sh. Averbukh, and Yehiam Prior. Isotope-selective laser molecular alignment. Physical Review A (Atomic, Molecular, and Optical Physics), 74(4):041403, 2006. [100] D. S. Elliott and J. F. Ward. Polarizability anisotropies of co2, n2o, and ocs from measurements of the intensity-dependent refractive index in gases. Phys. Rev. Lett., 46(5):317?320, Feb 1981. [101] Michinori Tanaka, Masanao Murakami, Tomoyuki Yatsuhashi, and Nobuaki Nakashima. Atomiclike ionization and fragmentation of a series of CH3-X (X : H, F, Cl, Br, I, and CN) by an intense femtosecond laser. Journal Of Chemical Physics, 127(10), SEP 14 2007. [102] V. Loriot, P. Tzallas, E. P. Benis, E. Hertz, B. Lavorel, D. Charalambidis, and O. Faucher. Laser-induced field-free alignment of the OCS molecule. Journal of Physics B-Atomic Molecular and Optical Physics, 40(12):2503?2510, JUN 28 2007. [103] KL Wang and DM Mittleman. Metal wires for terahertz wave guiding. Nature, 432(7015):376?379, NOV 18 2004. [104] Markus Walther, Mark R. Freeman, and Frank A. Hegmann. Metal-wire terahertz time-domain spectroscopy. Applied Physics Letters, 87(26):261107, 2005. [105] H Cao and A Nahata. Coupling of terahertz pulses onto a single metal wire waveguide using milled grooves. Optics Express, 13(18):7028?7034, SEP 5 2005. [106] Wenqi Zhu, Amit Agrawal, Hua Cao, and Ajay Nahata. Generation of broad- band radially polarized terahertz radiation directly on a cylindrical metal wire. Optics Express, 16(12):8433?8439, JUN 9 2008. [107] K. Y. Kim, B. Yellampalle, G. Rodriguez, R. D. Averitt, A. J. Taylor, and J. H. Glownia. Single-shot, interferometric, high-resolution, terahertz field diagnostic. Applied Physics Letters, 88(4):041123, 2006. [108] http://www.evproducts.com/. 111