copyright by alexandre s maltsev 2004tditmire/theses/maltsev.pdf · 2006. 7. 18. · alexandre s...

40
Copyright by Alexandre S Maltsev 2004

Upload: others

Post on 27-Jan-2021

2 views

Category:

Documents


0 download

TRANSCRIPT

  • Copyright

    by

    Alexandre S Maltsev

    2004

  • Above Threshold Ionization with Ultrahigh Intensity

    Laser Light

    by

    Alexandre S Maltsev, B.S.

    THESIS

    Presented to the Faculty of the Graduate School of

    The University of Texas at Austin

    in Partial Fulfillment

    of the Requirements

    for the Degree of

    MASTER OF ARTS

    THE UNIVERSITY OF TEXAS AT AUSTIN

    December 2004

  • Above Threshold Ionization with Ultrahigh Intensity

    Laser Light

    APPROVED BY

    SUPERVISING COMMITTEE:

    Todd Ditmire , Supervisor

  • To my special people...

  • Acknowledgments

    First of all I would like to thank my supervisor Prof. Todd Ditmire

    who made this work possible. Todd’s style of managing the group is abso-

    lutely great and I couldn’t wish for better. Watching him I learnt what a good

    team leader should be like.

    I am grateful to Anatoly Maksimchuk who was always helping me with

    the experiment. He also showed me how persistent a good scientist should be

    - as a matter of fact he was so persistent trying to make my experiment work

    that he got his family upset because of spending too much time in the lab.

    Special thanks go to Galina Kalintchenko and Vladimir Chvykov who spent

    long hours in the Hercules lab providing us with laser light.

    Finally I want to thank Prof. Michael Downer who was so kind to be

    the co-reader of this thesis.

    v

  • Above Threshold Ionization with Ultrahigh Intensity

    Laser Light

    Alexandre S Maltsev, M.A.

    The University of Texas at Austin, 2004

    Supervisor: Todd Ditmire

    This document has the form of a “fake” doctoral dissertation in order

    to provide an example of such, but it is actually a copy of Miguel Lerma’s doc-

    umentation for the Mathematics Department Computer Seminar of 25 March

    1998 updated in July 2001 and following by Craig McCluskey to meet the

    March 2001 requirements of the Graduate School.

    This document and its source file show to write a Doctoral Dissertation

    using LATEX and the utdiss2 package.

    vi

  • Table of Contents

    Acknowledgments v

    Abstract vi

    List of Figures viii

    Chapter 1. Theory 1

    1.1 Ionization by Laser Light . . . . . . . . . . . . . . . . . . . . . 1

    1.2 Calculating Ion Yields . . . . . . . . . . . . . . . . . . . . . . 6

    1.3 Exact fields in the focus . . . . . . . . . . . . . . . . . . . . . 10

    1.4 Non-Relativistic Above Threshold Ionization (ATI) . . . . . . 14

    1.5 Relativistic ATI . . . . . . . . . . . . . . . . . . . . . . . . . . 18

    1.6 Simulations of Relativistic ATI . . . . . . . . . . . . . . . . . . 27

    Bibliography 30

    Vita 32

    vii

  • List of Figures

    1.1 BSI: a) Coulomb potential and laser potential combine b) to setelectron free. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3

    1.2 Electron tunnels through the potential barrier. . . . . . . . . . 4

    1.3 Dependence of ion yields on intensity for neon. . . . . . . . . . 10

    1.4 Fields in a focused beam: a) paraxial approximation, b) com-plete description. . . . . . . . . . . . . . . . . . . . . . . . . . 11

    1.5 Electrons ”slide off the potential hill”. . . . . . . . . . . . . . 17

    1.6 Shape of the electron trajectory upon ionization at ultrahighintensity. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20

    1.7 Electron surfing a laser wave. . . . . . . . . . . . . . . . . . . 21

    1.8 Dependence of the electron energy on the position in the beam. 26

    1.9 Simulation results for 1×1018 W/cm2. Dependence of the ejec-tion angle on the electron energy for (a) paraxial approximation,(b) longitudinal fields included. . . . . . . . . . . . . . . . . . 28

    1.10 Simulation results for 5×1021 W/cm2. Dependence of the ejec-tion angle on the electron energy for (a) paraxial approximation,(b) longitudinal fields included. . . . . . . . . . . . . . . . . . 29

    viii

  • Chapter 1

    Theory

    Some theory goes next.

    1.1 Ionization by Laser Light

    Ionization by light can be described in different terms depending on the

    intensity of the laser and on the ionization potential of a given atom or ion.

    At lower intensities the process is most adequately described as a multiphoton

    ionization:

    An+ + N~ ω → A(n+1)+ + e−,

    where N~ ω > Ip (see also section Non-Relativistic Above Threshold Ionization

    (ATI)). However, we are interested in higher intensities, in which case the

    process of ionization is more appropriately described as a tunnelling process.

    These two regimes can be more quantitatively distinguished using the Keldysh

    tunnelling parameter [1]:

    γ = (Ip/2Φpond)1/2, (1.1)

    where Φpond is the ponderomotive potential of the laser and Ip is the ionization

    potential of the atom or ion. The ponderomotive potential is given by

    Φpond [eV] = e2E2/4meω

    2 = 9.33× 10−14Iλ2, (1.2)

    1

  • where E is the electric field of the laser, I is the corresponding intensity in

    W/cm2, and λ is the wavelength in µm.

    Typically, in laser ionization experiments the measured quantity is the number

    of ions of a particular ionic species per shot. A popular way of estimating the

    minimum intensity at which the ionic species is observed (the ”appearance

    intensity”) is using the barrier suppression ionization (BSI) mechanism. BSI

    was demonstrated to predict the appearance intensities quite well by Augst et

    al. in [2]. The idea is to consider a system of an electron and a charged core,

    and describe this system using the simple Coulomb potential. The field of

    the laser is considered as uniform and static, which is reasonable because the

    spatial variations in a laser beam happen at scales much larger than the size of

    an atom, and the laser oscillation period is much larger than the characteristic

    atomic time scale. As shown in Fig. 1.1 a sufficiently strong electric field of

    the laser can completely suppress the Coulomb barrier and let the electron

    escape. Calculating the appearance intensities through BSI is quite simple.

    The total potential can be written as

    U(r) = −Ze2/r − eEr. (1.3)

    To find the maximum of the potential we use dU/dr = Ze2/r2 − eE = 0,obtaining

    rmax =√

    Ze/E. (1.4)

    We then assume that the ionization potential is equal to the maximum of the

    combined potential, U(rmax) = −Ip. This gives us the electric field required

    2

  • Figure 1.1: BSI: a) Coulomb potential and laser potential combine b) to setelectron free.

    for the barrier suppression:

    Ethreshold =I2p

    4e3Z. (1.5)

    The corresponding threshold intensity is found from Ithreshold = (c/8π)E2threshold,

    and transforming to the practical units:

    Ithreshold [W/cm2] ≈ 4× 109 Ip [eV]

    Z2. (1.6)

    The Z in this equation should refer to an ”effective charge” of the core, which

    should take into account the deviation of the actual potential, experienced by

    the electron in the atom or ion, from the assumed simple Coulomb potential.

    Very often, however, Z is taken to be simply the charge of the ion created in

    the ionization process. Appendix .. lists ionization potentials and calculated

    BSI thresholds for some important species.

    Quantum mechanics, of course, permits the electron to get ionized even when

    3

  • the Coulomb barrier is not completely suppressed - through tunnelling (Fig.

    1.2). There are several theories calculating the probability rate for the elec-

    Figure 1.2: Electron tunnels through the potential barrier.

    tron to tunnel ionize at a given intensity. A detailed comparison of the rates

    predicted by different theories with experimental results at intensities from

    mid-1013 W/cm2 to mid-1016 W/cm2 was performed by Augst et al. in [3].

    They showed that the species dependence of the ionization rates is best de-

    scribed by the theory developed by Ammosov, Delone and Krainov [4]. In ADK

    model the initial atomic or ionic state is described by the effective principal

    quantum number n∗ (n∗ = Z/(2Ip)1/2, where Z is the charge of the created

    ion), the orbital angular momentum, and the magnetic quantum numbers l

    and m. The ionization rate is given by

    W = ωAC2n∗lf(l, m) Ip

    (3E

    π(2Ip)3/2

    )1/2 (2

    E(2Ip)

    3/2

    )2n∗−|m|−1(1.7)

    × exp(− 2

    3E(2Ip)

    3/2

    ),

    4

  • where ωA is the atomic unit of frequency (ωA = 4.134 × 1016 sec−1), E is theelectric field in atomic units (EA = 5.142 × 109 V/cm), Ip is the ionizationpotential in hartrees (1 hartree = 27.2 eV) and the factors f and C are given

    by

    f(l,m) =(2l + 1)(l + |m|)!

    2|m|(|m|)!(l − |m|)! , (1.8)

    Cn∗l =(

    2e

    n∗

    )n∗1

    (2πn∗)1/2, (1.9)

    where e is the base of natural logarithm. The rate shown in Eq. (1.7) is an

    approximation that is valid for n∗ À 1, E ¿ 1, and ω ¿ Ip. Thus, the valid-ity of ADK improves as n∗ increases, i.e. for heavier atoms and higher charge

    states.

    I would like to make a remark that normally we consider ionization of multi-

    electron atoms or ions as a sequential process, i.e. the electrons are torn away

    one by one. This description seems to be reasonable because the time scale

    for atomic processes is on the order of 10−17 sec, while the characteristic laser

    time is on the order of 10−15 sec. But this reasoning is only perfectly true

    when the ionized electron cannot affect the remaining electrons anymore. It

    was experimentally shown that the process of non-sequential (NS) ionization

    can also take place. Two mechanisms for NS ionization were proposed, that

    use a simple two-step picture. In the first step the electron passes over or tun-

    nels through the Coulomb barrier. In the second step the electron’s motion

    is dominated by the action of the laser fields. At non-relativistic intensities

    the electron gets accelerated away from the core and then accelerated back

    5

  • towards the core during the next half cycle. In the mechanism proposed by

    Corkum [5] the electron can collisionally ionize the second electron (e → 2e)when it revisits the core (rescattering). Fittinghoff et al. [6] suggested that

    the second electron can get ionized because of the sudden loss of screening due

    to the rapid removal of the first electron (shakeoff). Walker et al. performed

    measurements of NS ionization rates over 12 orders of magnitude in ion signal

    [7]. At relativistic intensities, however, the rescattering process is suppressed

    because the v×B forces do not allow the first ionized electron to get back tothe core. This phenomenon of NS ionization suppression was experimentally

    verified by Chowdhury and Walker in [8].

    1.2 Calculating Ion Yields

    It was mentioned in the previous chapter that in laser ionization ex-

    periments the measured quantity is the number of ions of a particular species.

    Thus, in order to compare the predictions of a chosen model of ionization with

    the experiment one needs to calculate the ion yields for different species. In an

    actual optical system the intensity distribution in the focal region can be very

    complex. Even though that could probably be taken into account in numerical

    calculations, usually the situation is simplified by assuming an ideal Gaussian

    focusing. In this case it is possible to find exact equations of the ”isointensity

    surfaces”, i.e. the 3D-surfaces of constant intensity. An isointensity surface

    is characterized by the parameter I0/I, where I0 is the peak intensity in the

    focus, and I is an intensity of interest. The volume of space limited by an

    6

  • isointensity surface can be calculated analytically, and is given by

    V

    (I0I

    )=

    π2w40λ

    (2

    9Z3 +

    4

    3Z − 4

    3tan−1 Z

    ), (1.10)

    where Z =√

    I0/I − 1, w0 is the beam waist (1/e2 radius for intensity), and λ -the laser wavelength. Using the knowledge about the isointensity surfaces one

    can split the focal volume into ”shells” - volumes limited by two neighboring

    isointensity surfaces. Shells can be made thin enough such that in the process

    of computations one could claim that all the volume in one shell is experiencing

    approximately the same intensity. In this case the ionization probability is the

    same for all the atoms or ions in one shell. There can be different methods

    of choosing the thicknesses of shells such that they could be considered thin.

    The intensity of the beam exactly in the focal plane is given by

    I(r) = I0 exp(−r2/w20).

    And let’s say we consider a shell thin if the difference of intensities between

    the two limiting isointensity surfaces is ∆I = αI0, where α is a small num-

    ber. The difference in intensities can be approximated by ∆I(r) ≈ I ′(r)∆r.Differentiating intensity

    I ′(r) = I0 · (−2r/w20) exp(−r2/w20).

    There are two methods for splitting into shells that seem most natural. First,

    one can choose to have shells of different thickness ∆r and keep |I ′(r)∆r| =|∆I(r)| = αI0, which gives

    (2r/w20) exp(−r2/w20)∆r = α =⇒ ∆r1 = α (w20/2r) exp(r2/w20). (1.11)

    7

  • Second method is to have the shells of the same thickness ∆r = const,

    but thickness small enough so that (∆I)max = |I ′(r)|max ∆r = αI0. Wecan, of course, find |I ′(r)|max by finding the second derivative of the inten-sity and equating it to zero. Ultimately we get |I ′(r)|max = |I ′(w0/

    √2)| =

    I0(√

    2/w0) exp(−1/2), so that

    ∆r2 = α

    √e

    2w0 ≈ α · 1.166 w0. (1.12)

    The first method is a little more difficult to realize, so at the time of writing

    the code for calculating ion yields I chose the second method. In general,

    the evolution of the different charge states in time is described by a series of

    first-order coupled ordinary differential equations [9]

    dnjdt

    =

    j−1∑i=0

    Wijni −Zmax∑

    k=j+1

    Wjknj(r, t), (1.13)

    where Wij is the ionization rate from species i to j, n(r, t) is the number

    density of species i at a specific point in space r at a time t, and Zmax is the

    maximum charge state. We, however, consider only sequential ionization, so

    we get a much simpler expression

    dnjdt

    = Wj−1, j nj−1 −Wj, j+1 nj. (1.14)

    Notice that we can integrate these series of equations for every shell separately,

    as ions presumably don’t have time to move much within the focus during

    the laser pulse propagation. The shape of the laser pulse is significant as it

    determines the rates Wij at a given time. So the description of the algorithm

    8

  • for calculating ion yields is:

    1) Initialize each shell by setting the number of atoms proportional to the

    volume of the shell, and setting the numbers of all other ionization states to

    zero.

    2) Solve the equations (1.14) for each shell separately to get the numbers

    (nKi )final, where K is the shell number and i is the ionization state.

    3) Sum the ion populations for different species over all the shells to get the

    total ion yields

    ntotali =Kmax∑K=0

    (nKi )final. (1.15)

    In order to make a comparison with a theory one needs to calculate the ion

    yields dependence on the peak intensity of the laser, so the three steps of the

    above algorithm must be repeated for every peak intensity in the region of

    interest. The ultimate result could look like Fig. 1.3.

    It is easy to notice in the figure that above certain intensity the curve

    turns into a straight line. This is a saturation effect that takes place when

    I0/Im À 1, where Im is the appearance intensity for ionization state m. Itcan be seen from Eq. (1.10) that in this case Z ≈ (I0/Im)1/2 and the Z3 termdominates. Thus a shell volume ∆V = V (Ii) − V (Ij) ∝ I3/20 , for arbitraryvalues of Ii and Ij (as long as they satisfy I0/I À 1). The saturation resultsfrom the fact that the intensity is so high that all the atoms/ions within the

    shell get ionized and the yield becomes simply proportional to the volume of

    the shell.

    9

  • Figure 1.3: Dependence of ion yields on intensity for neon.

    1.3 Exact fields in the focus

    For most applications the electromagnetic fields in a focused laser beam

    can be described using the so-called ”paraxial approximation”. Paraxial ap-

    proximation assumes that the electric and magnetic fields are exactly perpen-

    dicular to the direction of propagation of the beam and are mutually perpen-

    dicular (Fig. 1.4(a)). However if one tries to imagine the picture of propagation

    of a focused beam, it becomes clear that the paraxial approximation can only

    hold well for ”soft focusing”, that is only for systems with large f-numbers.

    This point can be made using some simple physical considerations. A flow of

    electromagnetic energy in a point in space far from focus can be characterized

    by the local Pointing vector S =1

    µE×B. Tighter focusing thus implies that

    10

  • Pointing vectors make greater angles with the beam axis. But from the defi-

    nition of the Pointing vector it follows that the local plane of vectors E and

    B makes a sharper angle with the beam axis (Fig. 1.4(b)).

    Figure 1.4: Fields in a focused beam: a) paraxial approximation, b) completedescription.

    The exact fields of a focused beam can be calculated. For my work I

    used results by Quesnel and Mora [10]. As the starting point of their derivation

    they assume a Gaussian profile with beam waist w0 for the transverse electric

    field exactly at the focal plane, i.e. Ex(x, y, z = 0) = E0 exp(−(x2 + y2)/w20).They then derive integral expressions for all the components of both electric

    and magnetic fields, which were not of particular interest for me because they

    would require too much computational time if used in simulations. However,

    11

  • the authors proceed to derive the expansion of the integral expressions in

    terms of the small parameter ε = 1/kw0, where k is the wave number of

    the laser light. Turns out that the transverse field components (along x and

    y) contain only even powers of ε, while the longitudinal ones (along z) only

    contain odd powers. Thus the second term in every expansion is of the order

    ε2 as compared to the first term, and is not of great significance. Even though

    the components Ey and Bx are second order in ε, it might not be reasonable to

    neglect them completely from the very beginning, as one wouldn’t want to lose

    any interesting effects when doing simulations. But after I got some insight

    into the physics of electron acceleration by laser light it became evident that

    the most interesting effects come from the longitudinal electric field, as will

    be demonstrated later in this text, and small transverse components are of no

    significance. Following are the first order formulas that I used in my work:

    Ex = E0w0w

    exp(− r2

    w2) sin ϕG (1.16a)

    Ez = 2E0εxw0w2

    exp(− r2

    w2) cos ϕ

    (1)G (1.16b)

    By = Ex/c (1.16c)

    Bz = 2E0c

    εyw0w2

    exp(− r2

    w2) cos ϕ

    (1)G (1.16d)

    Ey = Bx = 0 (1.16e)

    where ϕG = ωt− kz +tan−1(z/zR)− zr2/zRw2−ϕ0, ϕ(1)G = ϕG +tan−1(z/zR);w = w0

    √1 + z2/z2R is the beam waist at longitudinal position z, zR = kw

    20/2

    is the Rayleigh length, and ϕ0 is an arbitrary constant.

    It may be of some interest to see how the characteristic numbers in these formu-

    12

  • las (w0, zR, ε) depend on the optical system used for focusing, i.e. ultimately

    on the f-number of the system - f# = F/D, where F is the characteristic

    focal length, and D is the characteristic aperture. For example, when using a

    parabolic mirror as a focusing element, F is the effective focal length of the

    mirror and D is the beam diameter (if it’s a flat-top beam and it can reflect

    completely off the mirror). I’d like to point out that for a tightly focusing

    parabolic mirror the definition of f# becomes somewhat vague.

    It is known that for a perfectly focused Gaussian beam the beam waist is given

    by w0 =2

    πλf#. Consequently we have:

    zR =kw202

    =4

    πλ(f#)2, ² =

    1

    kw0=

    1

    4f#.

    For a system with f# = 2 and λ=800 nm, which is close to the experimental

    conditions described later, we have: w0 ≈ 1 µm, zR ≈ 4 µm, ² = 0.125. Thisfocusing is already very tight and ² is quite large, though ²2 ≈ 0.01 is stillsmall.

    Usually in the lab the quality of focusing is characterized by the full width at

    half-maximum (FWHM) of the intensity distribution in the focal plane. At

    the same time w0 is 1/e radius for the electric field distribution. Using I ∝ E2

    to relate the two distributions, we find ρFWHM = w0√

    2 ln 2 ≈ 1.177w0 for anideal Gaussian beam.

    13

  • 1.4 Non-Relativistic Above Threshold Ionization (ATI)

    Albert Einstein received the Nobel prize in physics in 1921 for explain-

    ing the photoelectric effect in 1905. The photoelectric effect refers to the

    emission, or ejection, of electrons from the surface of, generally, a metal in

    response to incident light. If using the classical Maxwell wave theory of light,

    one would expect ejected electrons to have higher kinetic energy for higher

    intensity of the incident light. It was experimentally found, however, that

    the energies of the emitted electrons were independent of the intensity of the

    incident radiation. Einstein resolved this paradox by using quantum descrip-

    tion of light. Photoelectric effect is very close in nature to ionization of single

    atoms by light (photoionization). Einstein’s theory suggests that an atom can

    absorb a photon whose energy is higher than the atom’s ionization potential

    and the electron emitted in this process will have a kinetic energy given by

    K = ~ω− Ip, where ω is the photon’s frequency and Ip - the ionization poten-tial.

    In 1931 Goeppert-Mayer first considered two-photon transitions. Experimen-

    tal verification of his theory required very intense sources of EM energy. Such

    sources in radiofrequency domain were developed in 1950s and allowed for

    observation of multiphoton transitions between bound states of atoms and

    molecules. The first lasers in the early 1960s provided an opportunity to ob-

    serve multiphoton ionization.

    Intensity of light characterizes the ”concentration of photons” in a beam. Mul-

    tiphoton ionization is a probabilistic process requiring interaction of N photons

    14

  • with one bound electron. The probability that there is a photon in a vicin-

    ity of a particular electron is proportional to the concentration of photons -

    thus proportional to intensity. Therefore the N -photon ionization rate should

    vary as σNIN , where σN is a generalized N -photon ionization cross-section.

    Though σN decreases with increasing N , an N -photon process can be observed

    at sufficiently high intensity.

    Initially for multiphoton processes a simple extrapolation of Einstein’s picture

    was suggested, i.e. after absorbing N photons an electron was expected to

    have kinetic energy K = N~ω − Ip. However, in 1979 Agostini et al. [11] ob-served that the energy spectrum of electrons produced in 6-photon ionization

    of xenon atoms consisted of two peaks corresponding to absorption of 6 and

    7 photons. Results produced in other labs for various atomic species, intensi-

    ties and laser wavelengths showed spectra consisting of multiple peaks evenly

    spaced by the photon energy. This phenomenon was named above threshold

    ionization (ATI). As is well known, a free electron cannot absorb a photon

    because that would break the law of conservation of momentum. So it was

    suggested that during the process of ionization electron stays in the neighbor-

    hood of its parent ion for long enough to absorb additional photons, the ion

    serving the function of a momentum sink.

    An important result was obtained in 1983 in 11-photon ionization of xenon

    at 1064 nm with a laser intensity of about 1013 W/cm2 (Kruit et al. [12]).

    This experiment revealed that the first peak of the electron spectrum was

    smaller than the peaks corresponding to absorption of larger numbers of pho-

    15

  • tons. Moreover, the first peak was decreasing with increasing intensity. Later,

    in experiments conducted in helium with the intensity of about 1015 W/cm2,

    disappearance of 30 peaks was demonstrated. This effect can in principle be

    explained by considering multiple stimulated scattering events, with photons

    in a focused beam carrying momenta of different directions. It is much more

    reasonable for high intensities, however, to describe the laser light classically.

    In the non-relativistic case when the maximum velocity achieved by the elec-

    tron is small (vmax ¿ c) the motion of the electron is completely dominatedby the electric field of the laser light. This motion will be simply oscillation

    with laser frequency. In case of a plane wave and infinite pulse duration the

    electron would be oscillating around a single point. Now if we look at more

    realistic situations, the electron motion should be affected by both temporal

    effects (intensity changing with time) and spatial effects (intensity gradient in

    a focused beam). The most widely accepted description of spatial effects is

    that using the concept of ”ponderomotive potential”. Ponderomotive poten-

    tial is given by:

    Up(r) =e2E2(r)

    4meω2.

    Also can be introduced the concept of ”ponderomotive force” as the gradient

    of ponderomotive potential. Basically ponderomotive potential describes the

    averaged motion of the electron, i.e. the motion of the point, around which

    16

  • the electron oscillates, as

    d2 < r >

    dt2= − 1

    me∇Up(r).

    In the limit of a very long pulse the electron has time to make a lot of oscil-

    lations and ”slide off the potential hill” (Fig. 1.5). In this case the pondero-

    motive energy gets converted completely into kinetic energy and the electron

    is registered outside the interaction region with the energy Etot = Up + Etrans,

    where Etrans is the kinetic energy given to the electron in the ionization process

    itself (Freeman et al. [13]). Respectively, for very short pulses there is almost

    no conversion of ponderomotive energy into kinetic energy and electrons are

    registered with the energy they had upon ionization, which depends on the

    position of a parent atom (or ion) within the beam.

    Figure 1.5: Electrons ”slide off the potential hill”.

    There is an interesting point to make concerning ponderomotive forces.

    17

  • For an ideal beam they would be evidently axially symmetric and independent

    of polarization. At the same time it would seem natural to suggest that direc-

    tion of polarization should be somehow preferential. Indeed, this symmetry of

    ponderomotive forces cannot be explained if we consider the beam in paraxial

    approximation. This symmetry is a consequence of the fact that in a laser

    beam ∇·E = 0 and ∇·B = 0. And if we look at the components of the fieldsit turns out that the symmetry arises thanks to the longitudinal component of

    the magnetic field Bz, which produces electron motion perpendicular to both

    the propagation direction and the electric field polarization.

    1.5 Relativistic ATI

    First, let us consider ATI in the weakly relativistic case, which corre-

    sponds to intensities of ∼ 1018−1020 W/cm2. In this case the electron achievesa velocities that is comparable with the speed of light. Recall that the motion

    of the electron is determined by the Lorentz equation:

    ṗe = −e(E + 1c

    v ×B). (1.17)

    It means that as v becomes greater the effect of the magnetic fields becomes

    larger. Let’s think in terms of paraxial approximation for a moment and

    imagine the beginning of electron’s acceleration. If electron was produced

    via ionization initially at rest then it would start accelerating along E (the

    direction of polarization). It would continue pretty much along this line until

    18

  • its speed becomes significantly large and there appears a component of force

    perpendicular to E due to the magnetic field. This force would tend to bend

    the trajectory towards the direction of propagation k. Let θ be the angle

    between k and the momentum of the electron pe. It can be shown (e.g. Moore

    et al. [14]) that, assuming paraxial approximation and electrons born at rest,

    there is an exact relation between the electron energy and its θ:

    tan θ =p⊥pz

    =

    √p2 − p2zpz

    =

    √γ2 − 1− (γ − 1)2

    γ − 1 =√

    2

    γ − 1 . (1.18)

    This relation holds at any point of the electron’s trajectory. In experiments by

    Meyerhofer et al. [15] it was demonstrated that this relation holds quite well

    for ultimate momenta of the electrons at the intensity of 1018 W/cm2. They

    registered electrons with energies of up to 140 keV at the corresponding angle

    θ ≈ 68o.Ionization provides an opportunity to ”inject” electrons into laser pulse around

    its peak intensity. This is very different from simply propagating a high inten-

    sity pulse through a medium of free electrons, because free electrons would be

    expelled from the beam by lower intensities and would never see the peak of the

    pulse. Hu et al. [16] showed by simulations that a linearly polarized Gaussian

    beam focused to the peak intensity of 8× 1021 W/cm2 with the beam waist ofw0 = 10 µm can accelerate the electrons produced from ionization of V

    22+ to

    energies of up to 2 GeV. However in their simulations they used the paraxial

    approximation for the fields while later in this chapter it will be demonstrated

    that it is absolutely necessary to use more exact description of the fields at

    these ultrahigh intensities. Their results for energy-angle distribution of the

    19

  • electrons were, naturally, consistent with Eq. (1.18).

    Good way of thinking about the electron acceleration by the ultrahigh inten-

    sity laser is to imagine it ”surfing the pulse”. Indeed, upon ionization the

    electron acquires the energy very quickly so that its relativistic gamma-factor

    goes into hundreds within a small part of the wave period. It results in v ≈ cand the electric and magnetic fields affecting the electron’s trajectory equally

    strongly. Consequently the electron starts moving nearly parallel to the direc-

    tion of propagation of the beam (Fig. 1.6). It means that at the electron’s

    position the phase of the light would be changing much slower than it would

    for the electron standing still (Fig. 1.7). This slowing of the rate of change

    would be determined by the electron’s gamma-factor and the angle between

    its trajectory and the direction of propagation of light.

    Figure 1.6: Shape of the electron trajectory upon ionization at ultrahigh in-tensity.

    Let us try to describe the electron motion in a more formal way (the

    following part of this section is an extended version of the paper [17] by Todd

    20

  • Figure 1.7: Electron surfing a laser wave.

    Ditmire and myself). Upon ionization the electron energy changes according

    to

    dE

    dt=

    d

    dt(γmc2) = −eE · v. (1.19)

    Thus the electron would be gaining energy from the laser light as long as E ·vstays negative, say for a period of time δτ - the ”dephasing time”.

    Let us first consider an ultra-relativistic electron (γ À 1) moving in the fieldof an infinite monochromatic plane wave, such that the electron’s momentum

    makes a small angle θ with the wave vector. In this case the dephasing time

    is completely determined by the difference between the z component of the

    electron velocity and the light phase velocity c.

    γ =

    (1− v

    2

    c2

    )−1/2=⇒ v = c

    √1− 1

    γ2≈ c

    (1− 1

    2γ2

    )

    vz = v cos θ ≈ v(

    1− θ2

    2

    )≈ c

    (1− 1

    2γ2− θ

    2

    2

    )

    If we presume that the electron is born at the peak of the field then the

    dephasing time will correspond to the time it takes for the light phase at the

    21

  • electron’s position to change by π/2. For the sake of convenience we say that

    the electron was born at z = 0. Another assumption that is going to be used

    several times in this section is:

    θ ¿ 1γ, (1.20)

    which is not necessarily very meaningful for a real ionization experiment. It

    might, however, represent an artificially realizable situation with preacceler-

    ated electrons ejected into the beam (an experiment of this kind was conducted

    at mildly relativistic intensities by Malka et al. [18]). We use this assumption

    because it gives an opportunity to compare different physical models of laser-

    electron interaction with potentially different resulting values of θ, using only

    the energetic parameter γ. With this said, we have:

    π/2 = ωt− kz = k(ct− z) = kt(c− vz) ≈ kct(

    1

    2γ2− θ

    2

    2

    )≈ kct/2γ2

    =⇒ δτplane ≈ πγ2/kc. (1.21)

    Let us now consider the motion of an electron in a focused beam. In Eqs.

    (1.16) t = 0, z = 0 corresponds to the peak field strength if ϕ0 = π/2. For

    simplicity we consider an electron that is born via ionization at t = 0 and

    x, y, z = 0, which is realistic because it corresponds to the peak of the field

    where the ionization probability is highest. Up to the first order in θ and 1/γ

    we have:

    z ≈ vzt ≈ ct(

    1− 12γ2

    − θ2

    2

    )≈ ct (1.22)

    x ≈ vxt = (vz tan θ) t ≈ vzt θ ≈ zθ ≈ θct. (1.23)

    22

  • At t = 0 the right hand side of Eq. (1.19) is positive and the electron is

    gaining energy from the field. The electron will start losing energy right after

    the moment when E · v = 0. Using Eqs. (1.16a) and (1.16b) we have:

    E · v = Exvx + Ezvz ∝ vx sin ϕG + vz · 2² xw

    cos ϕ(1)G

    ≈ vx(

    sin ϕG + 2²ct

    wcos ϕ

    (1)G

    ). (1.24)

    Let us consider the paraxial approximation first, which corresponds to ² = 0 in

    Eq. (1.24). Thus, we need to solve the equation sin ϕG = 0, which is equivalent

    to ϕG = 0 because ϕG(t = 0) = −π/2 and ϕG increases with time. For theparaxial approximation and γ À 1 the dephasing distance is much greater thanthe Rayleigh length (z ¿ zR). We then have the following approximations:

    tan−1 z/zR ≈ π/2− zR/z

    w2 = w20 (1 + z2/z2R) ≈ w20 z2/z2R

    Using these and Eqs. (1.22), (1.23) up to the second order in θ and 1/γ we

    ultimately get

    δτparaxial ≈ γw0c

    . (1.25)

    This dephasing time scales linearly with electron energy and is clearly much

    smaller than in the plane wave case, where it scales as the square of the

    electron energy. This can be explained by the fact that the field undergoes

    a Guoy phase shift as the electron propagates out of the focus, leading to

    reversal of the field at the electron much sooner.

    Finally, we consider the more complete description of the laser fields, i.e. ² 6= 0

    23

  • in Eq. (1.24). Under our assumptions about the time and position of the

    ionization event, right after the ionization we have sin ϕG ≈ −1 and cos ϕ(1)G ≈cos ϕG > 0. It means that Eq. (1.24) contains two competing terms of opposite

    signs. This holds true for an extended period of time until ϕ(1)G becomes

    positive. Thus we can say that the longitudinal component Ez of the electric

    field serves to decelerate the electron even before the time δτparax. It is not

    possible to solve the equation E · v = 0 analytically in this case. However,during some preliminary numerical trajectory calculations it was noticed that

    the electrons would start decelerating around z ≈ zR. Knowing this fact, weTaylor expanded the expression in parentheses with z near zR, and found that

    the dephasing time in the properly treated Gaussian focus is

    δτGaussian ≈ zRc

    (1− kzR

    2γ2

    ). (1.26)

    For γ À 1 we can ignore the second term in parentheses in Eq. (1.26) andmake the following comparison:

    δτGaussian/δτparaxial ≈ kw0/2γ,

    meaning that the effect from the inclusion of the longitudinal components is

    larger for higher intensities (larger γ) and for tighter focusing (smaller w0, or,

    equivalently, larger ²).

    Now we can roughly estimate the energy pickup by an electron during its

    first accelerating cycle using Eq. (1.19). It is not possible to analytically

    calculate the energy pickup for ² 6= 0, that’s why we choose a different tactics.We estimate the energy pickup for ² = 0 assuming the average field strength

    24

  • during acceleration Ex = Emax/2, and then compare the resulting energies

    using the different dephasing times.

    γmax = − emec2

    ∫E · v dt ∝

    ∫Exvx dt =

    ∫Emax/2√1 + z2/z2R

    θc dt

    ≈ Emaxθ2

    zmax∫

    0

    dz√1 + z2/z2R

    =EmaxθzR

    2

    zmax∫

    0

    d(z/zR)√1 + (z/zR)2

    .

    Making substitution z/zR = sinh s, we have√

    1 + (z/zR)2 =√

    1 + sinh2 s =

    cosh s and d(z/zr) = cosh s · ds. Thus:

    γmax ∝ EmaxθzR2

    ∫cosh s ds

    cosh s=

    (EmaxθzR s

    2

    )z=zmaxz=0

    ⇒ γmax ≈ emec2

    Emax2

    θzR sinh−1

    (cδτ

    zR

    ). (1.27)

    Using Eqs. (1.25), (1.26) and (1.27) we get

    γparaxialmax /γGaussianmax ≈ ln(4γ/kw0). (1.28)

    Eq. (1.27) suggests that for the peak intensity of 5 × 1021 W/cm2, w0 = 5µm and θ = 3o the electron would acquire a maximum γ of about 950 (cor-

    responding to the energy of about 500 MeV). While in case of using paraxial

    approximation the maximum energy should be about 5 times higher, accord-

    ing to Eq. (1.28).

    To verify in a more exact way the significance of proper treatment of longitu-

    dinal fields, suggested by our very approximate theory, we conducted a set of

    computer simulations described in detail in the next section.

    One final remark I would like to make concerns the possibility of applying

    25

  • the concept of ponderomotive force to describe electron’s motion at ultrahigh

    intensities. There have been attempts to derive a relativistic version of the

    ponderomotive force. For example, Quesnel and Mora [10] analyze some pre-

    vious publications and give the following expression:

    dpedt

    = − e2meγ

    ∇Ã2⊥, (1.29)

    where the overlines mean averaging over the fast time scale, and Ã⊥ is the

    quickly varying component of the vector potential perpendicular to the direc-

    tion of propagation of the beam. However, this expression is only valid under

    the condition 1−vz/c À ², which does not hold for the ultrarelativistic regime,especially when the focusing is tight.

    Figure 1.8: Dependence of the electron energy on the position in the beam.

    Also, Eq. (1.29) involves averaging over the fast time scale, i.e. it pre-

    sumes that the electron makes many oscillations on its way out of the beam.

    However, as can be seen from Fig. 1.8, which depicts a typical computed

    26

  • dependence of the electron energy on its position in the beam, the electron

    practically does not oscillate. It means that the idea of describing the elec-

    tron motion through any averaged quantity does not make much sense in the

    ultrarelativistic case.

    1.6 Simulations of Relativistic ATI

    Computer simulations were performed primarily to verify the effects

    of inclusion of the longitudinal fields [17]. Here is a brief description of the

    algorithm used:

    1) A position within the focal region is chosen randomly and ion of a certain

    charge state is put into that position.

    2) A laser pulse is propagated through the ion. Ionization can occur in two

    ways: a) when the intensity becomes higher than the BSI threshold intensity

    for the current charge state; b) when the intensity is below the BSI threshold,

    the ADK ionization probability is calculated as p = WADKδt, where δt is the

    simulation time step; the probabilistic nature is modelled by generating a ran-

    dom number b in the range (0, 1) and comparing it to p - if b < p ionization

    occurs.

    3) Once ionization occurred in step 2 an electron, initially at rest, is placed

    in the ion position. Then the relativistic equation of motion (1.17) is numer-

    ically integrated to find the vector of momentum ultimately acquired by the

    electron. The time of ionization serves as one of the initial conditions in the

    integration. The information about ultimate momentum is saved in a file.

    27

  • 4) Steps 1 to 3 are repeated until the maximum ionization stage is reached, or

    until the laser pulse has propagated far enough so it clearly can not result in

    further ionization.

    5) Steps 1 to 4 are repeated for a desired number of ions, large enough to

    collect a representative statistics.

    The first set of simulations was performed for the weakly relativistic intensity

    of 1 × 1018 W/cm2 for ionization of atomic neon. Laser pulse width was 30fs, wavelength 800 nm. The results are shown in Fig. 1.9 with the green line

    corresponding to the dependence (1.18) tan2 θ = 2/(γ − 1). In (a) the longi-tudinal components were deliberately excluded and in (b) they were properly

    included.

    Figure 1.9: Simulation results for 1×1018 W/cm2. Dependence of the ejectionangle on the electron energy for (a) paraxial approximation, (b) longitudinalfields included.

    It can be seen that at this intensity the angle-energy dependence is

    28

  • close to (1.18) in agreement with the experiment by Meyerhofer et al. [15].

    The maximum achieved energy does not change with inclusion of longitudinal

    components.

    The other set of simulations was performed rather closely following the param-

    eters of simulations by Hu and Starace [16]. We calculated ionization of Ar17+

    ions by an 800 nm, 30 fs laser pulse focused to w0 = 5 µm and the intensity of

    5×1021 W/cm2. The results are shown in Fig. 1.10. Clearly, at this ultrarela-tivistic intensity introduction of the longitudinal fields has drastic effects: the

    dependence (1.18) doesn’t hold anymore, and the maximum achieved energy

    decreases approximately 2.5 times. According to the previous section, one

    could expect that these effects are even stronger for tighter focussing.

    Figure 1.10: Simulation results for 5×1021 W/cm2. Dependence of the ejectionangle on the electron energy for (a) paraxial approximation, (b) longitudinalfields included.

    29

  • Bibliography

    [1] L. V. Keldysh, Sov. Phys. JETP 20, 1307 (1965).

    [2] S. Augst et al., Phys. Rev. Lett. 63, 2212 (1989).

    [3] S. Augst, D. D. Meyerhofer, D. Strickland, and S. L. Chin, J. Opt. Soc.

    Am. B 8, 858 (1991).

    [4] M. V. Ammosov, N. B. Delone, and V. P. Krainov, Sov. Phys. JETP 64,

    1191 (1986).

    [5] P. B. Corkum, Phys. Rev. Lett. 71, 1994 (1993).

    [6] D. Fittinghoff et al., Phys. Rev. Lett. 69, 2642 (1992).

    [7] B. Walker et al., Phys. Rev. Lett. 73, 1227 (1994).

    [8] E. A. Chowdhury and B. C. Walker, J. Opt. Soc. Am. B 20, 109 (2003).

    [9] M. D. Perry et al., Phys. Rev. A 37(3), 747 (1988).

    [10] B. Quesnel and P. Mora, Phys. Rev. E 58, 3719 (1998).

    [11] P. Agostini et al., Phys. Rev. Lett. 42, 1127 (1979).

    [12] P. Kruit et al., Phys. Rev. A 28, 248 (1983).

    30

  • [13] R. R. Freeman, P. H. Bucksbaum, and T. J. McIlrath, IEEE Journal of

    Quantum Electronics 24(7), 1461 (1988).

    [14] C. I. Moore et al., Physics of Plasmas 8(5), 2481 (2001).

    [15] D. D. Meyerhofer, J. P. Knauer, S. J. McNaught, and C. Moore, J. Opt.

    Soc. Am. B 13(1), 113 (1996).

    [16] S. X. Hu and A. F. Starace, Phys. Rev. Lett. 88, 245003 (2002).

    [17] A. Maltsev and T. Ditmire, Phys. Rev. Lett. 90(5), 053002 (2003).

    [18] G. Malka, E. Lefebvre, and J. L. Miquel, Phys. Rev. Lett. 78, 3314

    (1997).

    [19] P. Agostini et al., Phys. Rev. A 36, 4111 (1987).

    [20] V. P. Krainov, J. Opt. Soc. Am. B 14, 425 (1997).

    [21] V. P. Krainov and B. Shokri, Laser Phys. 5, 793 (1995).

    [22] V. P. Krainov and S. P. Roshupkin, J. Opt. Soc. Am. B 9, 1231 (1992).

    [23] C. H. Keitel, J. Phys. B 29, L873 (1996).

    [24] J. Prager and C. H. Keitel, J. Phys. B 35, L167 (2002).

    [25] M. Protopapas et al., Rep. Prog. Phys. 60, 389 (1997).

    [26] U. W. Rathe et al., J. Phys. B 30, L531 (1997).

    31

  • Vita

    Alexander Sergeevich Maltsev was born in Cheboksary, Russia on 5

    March 1980, the son of Tatyana Nikolaevna Maltseva and Sergey Vsevolodovich

    Maltsev. Received Bachelor of Science degree in Applied Mathematics and

    Physics from Moscow Institute of Physics and Technology in June 2001. Started

    graduate studies in physics program of the University of Texas at Austin in Fall

    of 2001. In January 2002 started work as a Research Assistant for professor

    Todd Ditmire.

    Permanent address: Russia, Cheboksary 428027Egersky blvd, 43-153

    This thesis was typeset with LATEX† by the author.

    †LATEX is a document preparation system developed by Leslie Lamport as a specialversion of Donald Knuth’s TEX Program.

    32