THE JOURNAL OF CHEMICAL PHYSICS 128, 154313 2008

Theory of three-dimensional alignment by intense laser pulses

Maxim Artamonov and Tamar Seidemana
Department of Chemistry, Northwestern University, 2145 Sheridan Road, Evanston, Illinois 60208-3113, USA

Received 29 January 2008; accepted 15 February 2008; published online 18 April 2008 We introduce a theoretical framework for study of three-dimensional alignment by moderately intense laser pulses and discuss it at an elementary level. Several features of formal interest are noted and claried. Our approach is nonperturbative, treating the laser eld within classical and the material system within quantum mechanics. The theory is implemented numerically using a basis set of rotational eigenstates, transforming the time-dependent Schrdinger equation to a set of coupled differential equations where all matrix elements are analytically soluble. The approach was applied over the past few years to explore different adiabatic and nonadiabatic three-dimensional alignment approaches in conjunction with experiments, but its formal details and numerical implementation were not reported in previous studies. Although we provide simple numerical examples to illustrate the content of the equations, our main goal is to complement previous reports through an introductory discussion of the underlying theory. 2008 American Institute of Physics. DOI: 10.1063/1.2894876
I. INTRODUCTION

In the gas phase, a molecule free of external elds undergoes unconstrained rotation in its center-of-mass frame. Typically, this means that the angular distribution of an ensemble of molecules in three-dimensional 3D space is isotropic under zero-eld conditions. In the presence of an external optical eld, however, molecules experience angledependent forces that can generate angular localization, with one or more axes of the molecule aligned preferentially with the coordinate system dened by the eld polarization.1,2 The exploration of molecular alignment using moderately intense laser elds has developed rapidly during the past few years into an exciting eld of experimental and theoretical research. This activity has been fueled by appreciation of the rich dynamics displayed by rotating molecules, by the development of experimental techniques to visualize and quantify alignment, and by a broad variety of demonstrated and anticipated applications. These range from study and manipulation of chemical reactions35 and elucidation of molecular structure,68 through generation of ultrashort light pulses9 and of high-order harmonics of light,1017 to fundamental studies in coherence18,19 and dissipation20 and new routes to quantum information processing.21 Several properties of the external laser eld determine the nature of the interaction: The central frequency, the pulse duration and shape, the intensity, and the polarization. The qualitative physics underlying intense laser alignment is conveniently discussed in terms of limiting cases of these parameters, corresponding, respectively, to near- versus nonresonant excitation, adiabatic versus sudden excitation, and linear versus circular polarization. Cases intermediate to these limits give rise to a spectrum of differently constrained rotational motions.
a

Electronic mail: t-seideman@northwestern.edu.

At near electronic resonance frequencies, a moderately intense laser eld induces Rabi-type cycles between the electronic manifolds, each transition being accompanied by the exchange of J = 0, 1 units of angular momentum between the eld and the material systems; this results in the sequential population of coherent rotational wavepackets in both states. At nonresonant frequencies, rotational excitation takes place via sequential Raman-type, J = 0 , 2 transitions, producing a rotational wavepacket in the initial vibronic state. Rotational wavepackets can be excited also at near rotational transition frequencies;22,23 although experimentally, this route is more demanding. The different modes of rotational excitation give rise to wavepackets of very similar properties, although their experimental realization relies on rather different techniques. In fact, it is possible to convert the equations of motion pertaining to one mode of excitation into those corresponding to the other with a simple transformation.24 In the long pulse limit, rot, with rot denoting the rotational period, the rotational dynamics is adiabatic. During the turn on, each eigenstate of the eld-free Hamiltonian evolves smoothly into the corresponding state of the complete Hamiltonian. The latter state, a coherent superposition of many free rotor eigenstates, can be analytically shown to align along the polarization direction. During the slow turn off, this state evolves adiabatically into the isotropic, eldfree state from which it originated; the alignment is lost upon the pulse turn off. In the opposite limit of an ultrashort pulse, rot, rotational excitation is impulsive. Here, the interaction serves as a kick to the rotational dynamics; it transfers a substantial amount of angular momentum to the system on a time scale short with respect to the rotational periods, creating a broad rotational wavepacket in angular momentum space. Alignment will ensue but only after the pulse is gone. After the initial postpulse alignment, dephasing and rephasing of the
2008 American Institute of Physics

0021-9606/2008/128 15 /154313/9/$23.00

128, 154313-1

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-2

M. Artamonov and T. Seideman

J. Chem. Phys. 128, 154313 2008

quantized angular momentum components of the wavepacket allow for the periodic reconstruction of alignment at regularly spaced revival times. The phase relation between the wavepacket components, and hence the ensuing alignment dynamics, is determined largely by the temporal characteristics of the laser pulse. It can be mathematically shown that in the impulse limit, the time integral of the interaction along with the rotational temperature fully determines the alignment dynamics, such that pulse shape effects are eliminated.25 A combination of the adiabatic and sudden limits, where a long turn on on rot is followed by a sudden turn off off rot is likewise formally interesting,26 experimentally feasible,27 and useful for certain applications.2833 More generally, optimally shaped laser pulses were numerically and experimentally shown to provide control over the alignment evolution by iteratively adjusting the phase relation among the light frequency components so as to provide a desired phase relation among the rotational components.3438 Optimal control approaches were shown useful also in generating insights regarding the dissipation of rotational coherences in media.23 Most of the research on intense laser alignment to date focused on linear molecules. In this case, as in the case of symmetric top molecules, a linearly polarized eld sufces to constrain the rotational motion. Currently, however, the interest in alignment of complex polyatomic molecules, in particular, asymmetric tops, is rapidly growing.3947 Here, the choice of the eld polarization plays an important role. A linearly or a circularly polarized laser eld denes a single direction in 3D space and, hence, can induce only onedimensional 1D alignment. It converts the free rotation of the body-xed axis with respect to the space-xed z axis into small amplitude librations but leaves unhindered the rotation about both the body-xed and the space-xed z axes. Quantum mechanically, a linearly polarized eld can populate a broad wavepacket of total angular momentum eigenstates but conserves the projection of the angular momentum vector J onto the space-xed z axis and, hence, cannot give rise to localization of the motion in the angle of rotation about the space-xed z axis. Similarly, the projection of J onto the body-xed Z axis is either conserved for linear or symmetric top molecules or weakly modied for asymmetric top molecules with linear polarization and, hence, the motion in the angle of rotation about the body-xed Z axis cannot be sharply dened. For most applications in polyatomic research, it is necessary to hinder also the rotation in the azimuthal angles, i.e., to three-dimensionally align the molecule. 3D alignment can be realized by means of elliptically polarized elds39 or by means of a combination of two linearly polarized elds with different polarization directions.44,47 Formally and conceptually, these routes are equivalent; although experimentally and numerically, they differ. Here, a unique direction in space is no longer dened and, hence, a broad wavepacket of M sublevels is excited; the elliptical polarization is able not only to modify the magnitude of the angular momentum vector as does the linearly polarized eld but also to reorient it with respect to the space-xed axes. For 3D alignment, ones interest is in asymmetric top molecules, where K is not a con-

served quantum number. By contrast to a linearly polarized eld, an intense elliptically polarized one populates a broad wavepacket of K sublevels in asymmetric tops, corresponding to reorientation of the angular momentum vector also with respect to the body-xed axes. In addition to a variety of potential applications of 3D alignment in elds ranging from molecular spectroscopy6,48,49 through quantum information50 to ultrafast imaging,10 3D aligned molecules promise to provide a route to understanding the rotational motions of complex molecules. Similar to vibrational and electronic wavepackets, a rotational wavepacket approaches, in the limit of a broad superposition in the relevant quantum number space, the motion of a single classical particle, in the present case the rotations of a single rigid body. These classical rotations are particularly interesting in the case of asymmetric top molecules, as such tops exhibit unstable dynamics in the classical limit and are inherently multidimensional. The study of rotationally broad, spatially aligned wavepackets in asymmetric tops introduces an opportunity to understand the motions of asymmetric tops near the classical limit. Whereas the problem of 1D alignment was studied in considerable detail, that of 3D alignment in only beginning to be explored.39,43,44,47 It is illustrated numerically and experimentally in Ref. 39 in the adiabatic domain and in Refs. 43, 44, and 47 in the nonadiabatic domain. The theory underlying 3D alignment was not discussed in our previous brief publications and is currently of growing interest. The main purpose of the present work, and the subject of Sec. II, is to provide an account of the theory and its numerical implementation. Section III briey illustrates the content of the formulas of Sec. II by means of two simple examples, and the nal section concludes.

II. THEORY

We consider a polyatomic molecule subject to a nonresonant, moderately intense laser eld. The formalism is derived for the most general case of an asymmetric top molecule subject to an elliptically polarized eld or to two linearly and orthogonally polarized elds . The previously studied problems of a linear molecule subject to linearly and circularly polarized elds are obtained as specic cases. Vibrations are neglected in the present study. The effects of vibrational coherences and vibration-rotation coupling on wavepacket alignment have been discussed elsewhere,5153 as was the case of a molecule exhibiting torsion.4 The effects of centrifugal distortion on alignment of symmetric tops have been likewise studied, numerically as well as experimentally.54 Distortion of the molecule is expected at high levels of rotational excitation; however, the rotational constants are typically orders of magnitude larger than distortion constants, and as a result, centrifugal distortion effects on the alignment revival pattern become non-negligible only at long but temperature- and intensity-dependent times. We invoke the electric dipole approximation and treat the laser eld as a classical entity. The complete Hamiltonian is, thus,

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-3

Molecular alignment by intense laser pulses

J. Chem. Phys. 128, 154313 2008

H = Hrot

t ,

Hind =

1 4

* =
,

ind

, 6

where Hrot denotes the rotational Hamiltonian, Hrot =

2 JX e 2IXX

J2 Y 2Ie YY

2 JZ e 2IZZ

ind

1 4

*,

2 where , = x , y , z are the space-xed coordinates and ind denes an induced electric dipole vector see Ref. 58 and the Appendix of Ref. 1 . In Eq. 6 , the effect of the excited vibronic states in which real population does not reside is included via the dynamic polarizability tensor,
v

where Jk, k = X , Y , Z, indicate the components of the matter angular momentum J along the principal axes of the mole ecule, Ikk are the corresponding equilibrium moments of inertia, is the electric dipole vector, and t is the laser eld, t =
1 2

t ei t + c.c. .

=
v

v
v

v
,v

is the central frequency, t = t , is a In Eq. 3 , complex unit polarization vector, and t is a smooth envelope. The dependence of t on the spatial coordinates is not indicated in Eqs. 1 and 3 . Its effect on the center-of-mass motion and its application to focus and guide molecular beams are explored elsewhere2833 and are not considered here. We note only that in the long pulse case, these effects are substantial. Rotational motion typically occurs on a much faster time scale than translational motion, even for long pulses, typically ensuring adiabatic separability of the rotational degrees of freedom. We follow the standard convention of choosing the space-xed z axis to lie along the eld polarization vector in the case of linearly polarized light and along the laser beam propagation direction in the case of circularly or elliptically polarized light. In the former case, = z, and in the latter, = exx + ey y , where ex = cos ey = cos cos i sin sin i sin sin , 4 cos , and x , y , z denote unit vectors along the space-xed x , y , z Cartesian coordinates. In Eq. 4 , is the ellipticity, determined by the ratio of the minor and major axes of the polarization ellipse, and is the azimuth, specifying the orientation of the ellipse in the space-xed xy plane. For = 0 and = / 4 / 4 , Eq. 4 reduces to the standard expression for right left circularly polarized light. Rewritten in the forms t = tan
x
1 2

v
v

v
,v

where the v are vibrational quantum numbers, denotes a is the component of the collective electronic index, transition dipole moment operator, and v ,v = Ev Ev are transition frequencies. The level widths have been omitted from Eq. 7 , as we consider far-off-resonance frequencies. For sufciently low frequencies, v ,v , the dynamic polarizability converges to its dc limit. In terms of the body-xed coordinates, k , k = X , Y , Z, the polarizability components are given through =
kk

kk

are elements of the transformation matrix bewhere k tween the space-xed and body-xed frames, often referred to as direction cosines. Molecular alignment arises from the angular dependence of the Hamiltonian introduced by the direction cosines. In the next subsection, we examine the form of the induced dipole interaction of Eqs. 6 8 as a function of the Euler angles and its dependence on the eld polarization. In Sec. II B, we proceed to solve the time-dependent Schrdinger equation subject to this interaction.
A. The induced dipole interaction and the role of the polarization

t x ex cos tan ,

t+ tan

+ y ey cos = tan

t+

, 5

Considering rst the case of a linearly polarized eld, = z, we nd, by substituting Eq. 8 in Eq. 6 , Hind =
1 2 4

ZX

cos2 +

YX

sin2 sin2

= tan

cot ,

Eqs. 3 and 4 show that a phase lag between the two orthogonal eld components, which differs from the / 2 phase of the circular polarization case, will translate into a 0. Equations 3 and 4 can be exnonzero azimuth pressed in a variety of other equivalent ways,55,56 commonly in terms of Stokes parameters,57 but the form above is convenient for our present application. At laser frequencies far detuned from vibronic transition frequencies, the eld-matter interaction in Eq. 1 is readily converted into the form of an induced Hamiltonian,

where we introduced generalized polarizability anisotropies as ZX = ZZ XX and YX = YY XX, and R = , , are the Euler angles of rotation, specifying the orientation of the body- with respect to the space-xed frame. In deriving Eq. 9 , we used Table I of Ref. 59 to express the transformation elements in Eq. 8 in terms of the Euler angles and omitted a term independent of the angles, which amounts to an overall shift of all eigenvalues and has no effect on the alignment dynamics. To simplify the connection with the previous work on alignment of linear molecules, it is instructive to choose the

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-4

M. Artamonov and T. Seideman

J. Chem. Phys. 128, 154313 2008

body-xed coordinate system such that ZZ YY XX. With this choice, Hind of Eq. 9 has equivalent minima at = 0 , , giving rise to alignment of the body-xed Z axis to the eld polarization direction. The rst term of Eq. 9 is predominantly responsible for the alignment. The second term destroys the invariance of the interaction to rotation about the body-xed Z axis and results in nonconservation of Kthe quantum number corresponding to the body-xed Z projection of the matter angular momentum vector J. The overall magnitude of the rotational excitation the absolute value of J is determined by both terms of Hind, giving it a fairly direct dependence on the product of the intensity and the two anisotropies cf. the dimensionless parameters introduced in Ref. 1 ; the portion of the excitation directed about the molecular Z axis the body-xed Z projection of J , however, is determined only by YX and vanishes for symmetric top and linear molecules, where YX = 0. Due to the cylindrical symmetry of the linearly polarized eld, the interaction of Eq. 9 is invariant under rotation about the space-xed z axis. This is reected in the absence of -dependent terms in Eq. 9 . Thus, the space-xed z projection of J with the associated magnetic quantum number M is conserved and the rotation of the molecule about the polarization vector is free. In the limit of a linear or symmetric top molecule, Eq. 9 reduces to the form, Hind =
1 2 4

Turning to the general polarization case, we nd, by substituting Eq. 8 in Eq. 6 and expressing the transformation elements k in terms of the Euler angles Ref. 59, Table I ,
2

Hind =

t
XX

ex sin

sin

+ cos cos cos

2

cos

+ ey cos +
YY

sin

+ cos sin cos

ex sin cos

cos cos sin

sin
2 2

+ ey cos +
ZZ

cos sin

ex sin cos

+ ey sin sin

14

It is convenient to introduce a rotated polarization vector by applying a unitary transformation to ex , ey as ea eb = cos sin sin cos ex ey . 15

Equation 15 denes a frame that rotates with the molecule about the space-xed z axis, here, the laser beam propagation direction. The nite rotation on the right hand side is readily recognized as the rst of the three rotations that specify the orientation of the body-with respect to the space-xed frame. In terms of ea , eb , the interaction assumes a conveniently compact form,
2

cos2 ,

10

Hind =

ZX

sin2

ea 2 +

YX

cos2 sin2 ,

ea 2 16

familiar from a vast number of articles about alignment of linear molecules in linearly polarized elds that appeared in the past decade. In the case of a circularly polarized eld,
R = L =

+ cos2

* eb 2 cos sin 2 Re ea eb

1 2

iy ,

11

where we omitted an angle-independent shift and rearranged the terms in a form convenient for comparison with Eq. 13 . Equation 16 can be simplied by using Eqs. 4 and 15 to express the eld parameters as ea 2 = cos2 cos2 sin2 + + + sin2 + sin2 + sin2 cos2 . + + , , 17

we have
2

Hind =

ZX

8
2

cos2 +

YX

sin2 sin2

12

eb 2 = cos2

* 2 Re eaeb = cos 2 sin 2

ZX

sin2 +

YX

cos2 sin2

+ cos2

13

up to an angle-independent shift , where Eq. 12 is included to note the formal analogy of alignment in linearly and circularly polarized elds and Eq. 13 is given to facilitate comparison with the elliptically polarized case discussed below. The sign difference between Eqs. 9 and 12 arises from our choice of the polarization plane as the space-xed x , y plane in the case of circular polarization and the polarization vector as the space-xed z axis in the case of linear polarization. In the former case, the eld aligns the most polarizable axis of the molecule perpendicular to the spacexed z axis and in the latter parallel to this axis. The factor of 2 difference in the prefactors of Eqs. 9 and 12 arises from the fact that the eld oscillates along two orthogonal axes in the circular polarization case and along one axis in the linear polarization case. In both cases, the cylindrical symmetry of gives rise to a -independent potential and, hence, the space-xed projection of J is conserved.

From Eq. 17 , we have a nonzero azimuth equivalently, a phase lag between the x and y eld components that differs / 2 that can be readily eliminated from the interacfrom tion through a redenition of the coordinate system. Within the context of Eq. 6 , the azimuth can be made to vanish by a rotation of the laboratory coordinate system and, hence, for our specic application, can be set to 0 with no loss of generality. With this choice, in the strong eld limit, where 0 vicinity, we have ea the system is conned to the ex and eb ey. With set to 0, Eq. 6 can be written equivalently as
2

Hind = +

ZX

4
YX

ex 2 ey 2 x Z 2 + ey 2 sin2

ex 2 ey 2 x Y 2 + ey 2 cos2 . 18

+ cos2 sin2

Comparing Eq. 18 with the circular polarization case, we see that the elliptical polarization adds a term proportional to

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-5

Molecular alignment by intense laser pulses

J. Chem. Phys. 128, 154313 2008

ex 2 ey 2 to each of the terms of Eq. 12 . With Eq. 12 aligning the body-xed Y , Z plane to the polarization plane, the new terms break the cylindrical symmetry and give rise to alignment of the most polarizable molecular axis to the major polarization axis. The broken symmetry translates into alignment of the azimuthal Euler angle . We note also that the alignment terms depend on the ellipticity in Eq. 18 , whereas for the circular polarization case, the eld only appears as an overall intensity prefactor that affects the and alignment equally. set to / 4, Eq. 16 reduces to the circular With polarization case Eq. 13 . In this limit, the control over the relative magnitudes of the different terms in Eq. 16 offered by the elliptical polarization is lost. An interesting feature of Hind is explicit in Eq. 16 , namely, a term that is antisymmetric with respect to . We remark that the antisymmetric term vanishes in the absence of interference between the eld components in the rotated frame and in the limit of a symmetric top or linear molecule. That is, it requires the eld to be elliptically polarized with ex 2 ey 2 if is set to 0 and the polarizability tensor to be asymmetric with X Y . The former condition implies that the space-xed z projection of the material angular momentum J is not dened. The latter implies that the body-xed Z projection of J is not dened. It is worth noting that the presence of the antisymmetric term does not imply that the interaction of an elliptically polarized eld with an asymmetric top molecule can induce orientation. Inor , the interference term vanishes tegrated over either and the parity of the total angular momentum is conserved, as in a linearly polarized eld.
B. Alignment dynamics

tion of Zare.55 For linear molecules with zero electronic an gular momentum, R n reduce to spherical harmonics. For asymmetric top molecules, n = J M , where = J , J + 1 , . . . , J is the asymmetric top quantum number, n = J M =
K

aJ

JKM ,

22

and the aJ K are obtained through the solution of the eigenvalue problem, Hrot n = En n . 23

In the case of adiabatic or nearly adiabatic alignment, where most of propagation is carried out in the presence of the laser eld, it is convenient to use a symmetric top basis regardless of the molecular symmetry save for the linear case . The asymmetric top couplings are then accounted for along with the radiative coupling via the temporal propagation. With the choice of a symmetric top basis, we nd the following set of coupled differential equations for the expansion coefcients: iCJKM t =
J K

CJ
1 4

K M

t JKM Hrot J K M
K M

CJ
J K M

JKM

JK M

*,

24 where we substituted Eqs. 2 , 6 , and 20 in Eq. 19 and used the orthonormality of the JKM . In Eq. 24 , the diagonal J = J , K = K part of the rst term is the rotational energy, its off-diagonal part accounts for asymmetric top coupling, and the second term describes the eld-matter interaction. The nonradiative matrix elements are given as JK Hrot JK = Ae Be JK1 JK J+K+1 J+K+2 , 4 K =K2 25 Ae Be JK+1 JK+2 J+K1 J+K , 4 K = K + 2, with all other elements vanishing identically. To derive expressions for the radiative coupling elements in Eq. 24 , we rst transform the polarizability tensor to the space-xed frame Eq. 8 by expressing the transformation k in the symmetric top basis59 and using the elements product reduction expression in Eq. 4.3.2 of Ref. 60. The diagonal elements of the polarizability tensor in the spacexed frame are up to an angle-independent shift J + 1 J K2 A e + B e + K 2C e, 2 K =K

We proceed to solve the time-dependent Schrdinger equation, i t =H , 19 t in a suitable time20

with H = Hrot + Hind, and rst expand independent basis, t =

n

Cn t n .

The choice of an appropriate basis n depends on both the eld parameters and the molecular symmetry. In the case of nonadiabatic short-pulse-induced alignment, where a major part of the propagation is carried out under eld-free conditions, it is numerically advantageous to expand t in terms of the eigenstates of Hrot. With this choice, the postpulse propagation can be carried out analytically. In the limit of a symmetric top, the eigenstates of Hrot are normalized Wigner matrices, n = JKM , R JKM = 2J + 1 J* D R , 8 2 MK 21

where J, K, and M are the quantum numbers corresponding to the matter angular momentum operator and its body- and space-xed z projections, R = , , , and we use the nota-

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-6

M. Artamonov and T. Seideman

J. Chem. Phys. 128, 154313 2008

xx yy

1 2 = 6 D00

1 2 6

2 2 D20 + D20

ZX

YX

M2 t =

t M2

t =
JKM

CJKM t 2M 2 .

31

+ + 1

1 4

2 2 2 2 D22 + D22 + D22 + D22

The consequent 3D alignment is quantied using the expectation values of the Euler angles in the time-evolving wavepacket, 26 cos2 t =
JKMJ

2 6

2 2 D02 + D02

YX

CJKM t CJ

KM *

J KM cos2

JKM ,

32

zz

1 2 3 D00

ZX

YX

1 6

2 D02 +

2 D02

YX

27

cos2

t =
JKMJ K

CJKM t CJ

K M*

J K M cos2

JKM , 33

where the spatial dependence of rotation matrices is implied. zz is readily identied in the second term of Eq. 24 of Ref. 1, which denes the Hamiltonian matrix elements in the case of a linearly polarized eld. The integrals of the form 2 JKM Dsq J K M in the matrix elements of Eqs. 26 are proportional to a product of two 3-j symbols e.g., Eq. 3.118 of Ref. 55 . By noting the selection rule J 2 J J + 2 and using Table II of Ref. 60 as well as the symmetry properties of the 3-j symbols in Eqs. 3.7.5 and 3.7.6 of the same reference, the radiative coupling integrals in Eq. 24 can be readily expressed in a closed form. The resultant expressions are quite compact and can be inlined, thus, eliminating a time overhead associated with a function call. Computational efciency in evaluating Eq. 24 is further enhanced by using the properties of the 3-j symbols to identify the nonzero matrix elements of the induced Hamiltonian, which can be precalculated and saved as a set of look-up tables to be used as needed during the time propagation. Another method to evaluate radiative coupling elements in Eq. 24 is more straightforward to implement while more computationally demanding. In this approach, we rst transform the polarizability tensor to the body-xed frame as JKM JK M =
kk kk J K M

cos2

t =
JKMJ M

CJKM t CJ

KM

* J KM cos2

JKM . 34

The matrix elements in Eq. 32 are readily evaluated analytically. The integrations over in Eqs. 33 and 34 need to be carried out numerically. The expectation values of Eqs. 32 34 do not fully characterize the 3D alignment, since they convey only second order interferences, whereas the wavepacket contains all high-order coherences. For our present discussion, however, these observables sufce. The modication of Eqs. 24 34 to expand in an asymmetric top basis an eigenbasis of Hrot is formally straightforward. The rst term in Eq. 24 reduces to CnEn, with En determined through Eq. 23 , and the second term is replaced by a superposition of the JKM J K M with the aJ K expansion coefcients of Eq. 22 . Thermally averaged observables are obtained by solving the system of equations for all thermally populated initial states i, giving state-specic versions of the quantities O i t in Eqs. 29 34 . The corresponding thermal averaged observables are then obtained as O
T

JKM k

k JK M 28

t =

1 Q

wi T O i t ,
i

35

JK M

JK M ,

k are the transformation elements of Eq. 8 . where the The matrix elements of the direction cosines in the symmetric top basis are readily expressed analytically using Table I of Ref. 59 and Eq. 3.118 of Ref. 55. In the case of a linearly polarized eld and a symmetric top or linear molecule, Eqs. 24 28 reduce to Eq. 6 of Ref. 32. With the expansion coefcients determined through the solution of the set 24 , all observables of interest can be extracted from Eq. 20 . The degree and sense of rotational excitation are quantied through the time evolution of the expectation values of the angular momentum and its bodyand space-xed z components as J2 t = t J2 t =
JKM

CJKM t 2J J + 1 ,

29

K2 t =

t K2

t =
JKM

CJKM t 2K2 ,

30

where Q is the rotational partition function, wi T is a Boltzmann statistical weight function, and T is the rotational temperature. The effects of temperature on molecular alignment have been explored in several experimental and numerical studies39,42 and are not studied here. The revival pattern of asymmetric top molecules changes dramatically with increasing rotational temperature, as Boltzmann averaging washes out and narrows down portions of the structure. In dissipative media, temperature plays a more complex role,20 since rotational relaxation rates decrease with increasing rotational excitation and, hence, also with the rotational temperature. In the limit of a long pulse long with respect to the rotational periods , where the alignment dynamics is adiabatic, the time-dependent problem is formally equivalent to the solution of the eigenvalue problem, H t = t0 t = t0 = E t = t0 , where t0 denotes the peak of the laser pulse. From a numerical viewpoint, the time-dependent solution and the eigenvalue problem are comparably costly at zero temperature. At nite temperatures, however, the stationary

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-7

Molecular alignment by intense laser pulses

J. Chem. Phys. 128, 154313 2008 TABLE I. Rotational constants and polarizabilities of the model in reduced units. BX 6/8 BY 3/8 BZ 9/8
XX YY ZZ

1/6

2/6

3/6

FIG. 1. Averaged measures of the rotational excitation right column and associated 3D alignment left column , here represented by the expectation values of the squares of the rotational quantum numbers and the corresponding lowest moments of the rotational probability distribution. The pulse is taken sufciently long to affect adiabatic dynamics. The solid curves show the results for an ellipticity = / 5, with the circular dashed, = / 4 and linear dotted, = 0 polarization cases provided for comparison. The time is e given in reduced units Ref. 1 , = / t, where = 1 / 2 Ie + IYY . t I I XX

solution has a considerable advantage, as all eigenpairs of relevance for thermal averaging at different temperatures are determined with one diagonalization.
III. NUMERICAL RESULTS

In this section, we briey discuss two simple examples that serve to illustrate the content of the equations of the previous section. For generality, we reexpress the complete Hamiltonian in Eqs. 1 , 6 , and 2 in reduced units,1 i.e., H H I
2

=
k=X,Y,Z

2 B kJ k R
, =x,y,z

36

where we dene dimensionless angular momentum operators, rotational constants, inertia components, and interaction parameters as = Jk , Jk 2 = I
1 2

k = X,Y,Z,

37 38

IXX + IYY , kk =
kk

= I , Bk 2Ikk = I R 4
2 2

39

40

In Eqs. 39 and 40 , is the trace of the polarizability tensor, = XX + YY + ZZ / 3. Finally, we introduce reduced time and temperature variables as = t / and =kT / 2, ret I T I spectively. Figure 1 explores the 3D alignment of an asymmetric top in the adiabatic limit. The parameters of the model stud-

ied are given in Table I. We have chosen the most polarizable molecular axis to dene the body-xed Z axis, with the least polarizable molecular axis dening the X axis. The ellipticity is / 5 and the azimuth is 0 so that the major axis of the polarization ellipse is along the space-xed x axis. The reduced Rabi coupling coefcient is R = 169 and all observables are thermally averaged over a Boltzmann distribution at the reduced temperature of 0.625. The observed decrease of cos2 below the isotropic value of 1 / 3 corresponds to connement of the molecular Z axis to the space-xed xy with respect to the isotroplane, and an increase of cos2 2 pic cos = 1 / 2 corresponds to preferential alignment of Z to the x axis. For comparison, we provide in Fig. 1 illustrations of the effects of circularly dashed curves and linearly dotted curves polarized elds. With circular polarization, the system undergoes planar connement but displays isotropic distribution of the probability density within the polarization plane. The motion in the coordinate, by contrast, is not isotropic, since the model is an asymmetric top, for which there is no cylindrical symmetry with respect to the bodyxed Z axis. With linear polarization, the molecule displays strong alignment along the z axis, as reected by the positive cos2 1 / 3. The antialignment is very weak for the same reason for which the linearly polarized eld does not induce alignment in , i.e., the cylindrical symmetry of the eld polarization vector. As the alignment increases, the bodyxed XY plane becomes increasingly aligned with the spacexed xy plane resulting in diminished interaction between the eld vector and the polarizability anisotropy in the XY plane. This reduction in the interaction strength can also be seen in the expression for the induced Hamiltonian in Eq 9 . The magnitude of the -dependent term decreases with an increase in the alignment. In this regime, the degree of K excitation caused by the asymmetric top coupling is small with respect to the intense J excitation by the laser eld. In the limit of strong adiabatic 1D alignment, corresponding to the excitation of a broad wavepacket of J states with J2 K2 , the rotation about the body-xed Z axis converges to that about the space-xed z axis and approaches an isotropic distribution. The latter effect is general and useful: The strong eld renders weak molecular interactions in this example, the asymmetric top coupling a perturbation and, hence, simplies the rotational dynamics. The solid curves of Fig. 1, corresponding to elliptical polarization, clearly illustrate the anticipated localization in all three Euler angles and the associated excitation of a broad wavepacket in J, K, and M spaces. Also seen is the expected trade off of the alignment quality in one motion that accompanies enhanced alignment of the other. At a given eld in-

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-8

M. Artamonov and T. Seideman

J. Chem. Phys. 128, 154313 2008

FIG. 2. As in Fig. 1 for nonadiabatic excitation.

tensity, the alignment in induced by the elliptically polarized eld is slightly degraded with respect to that induced by the eld with circular polarization. In Fig. 2, we show the analogous calculation for the nonadiabatic case. In this case, the reduced Rabi coupling coefcient R is 677 and the reduced temperature is 6.25. The sense of alignment of all three motions is common to the adiabatic and nonadiabatic cases, whereas the extent of both the angular connement and the rotational excitation differ. Before proceeding to discuss the gure, we remark that our results are not meant to compare the efciency of adiabatic versus nonadiabatic 3D alignment but included merely to provide an illustration of the general features of the observables described in Sec. II. In the short pulse limit, the alignment can be analytically shown to be determined by the uence alone;25 hence, the intensity is an irrelevant criterion. More generally, neither constant uence nor constant peak intensity calculations can serve to compare the adiabatic and nonadiabatic alignment cases. The most useful comparison is between calculations or measurements carried out in both pulse-length regimes at intensities that are just short of ionizing the molecules in question. Several features are noteworthy in Fig. 2. K2 is not constant after the pulse turn off by contrast to J2 and M 2 , since in the asymmetric top, K is not a conserved quantum number. The transient alignment revivals in all three Euler angles are clearly visible in the case of the elliptically polarized laser eld. Similar to the previous example of the adiabatic limit, the dynamics in and produced by the circularly polarized eld are very close to those produced by the eld with elliptical polarization. The increase in the degree of rotational excitation in the case of linear polarization compared to the other two cases is due to the fact, as mentioned in Sec. II A, that the eld oscillates along one axis in the linear polarization case as opposed to two orthogonal axes in the case of circular or elliptical polarizations.
IV. CONCLUSIONS

ment by moderately intense laser elds. We have been motivated by the recent realization of 3D alignment in three different types of experiments.39,44,47 We expect that the interest in 3D alignment will continue to grow in the coming years, as the range of applications of alignment continues to shift to the domain of polyatomic systems. Although the different approaches applied so far to achieve 3D alignment rely on different experimental techniques, they are formally equivalent and, hence, can be and have been39,47 equally well studied within the theoretical framework described above. Although exposition of the theory and numerical methods has been our main objective, we concluded for completeness with two numerical examples that bring across the relation of the formulas developed to experimental observables and illustrate a number of simple and general features of 3D alignment of asymmetric tops. Future research will make use of the above theory to explore new applications of 3D alignment, which we expect to be of formal and experimental interest. These range from high harmonic generation from aligned polyatomic molecules1016 through structural determination of complex molecules8 to eld-guided molecular assembly.61
T. Seideman and E. Hamilton, Adv. At., Mol., Opt. Phys. 52, 289 2006 . H. Stapelfeldt and T. Seideman, Rev. Mod. Phys. 75, 543 2003 . 3 J. J. Larsen, H. Sakai, C. P. Safvan, I. Wendt-Larsen, and H. Stapelfeldt, J. Chem. Phys. 111, 7774 1999 . 4 S. Ramakrishna and T. Seideman, Phys. Rev. Lett. 99, 103001 2007 . 5 S. Fleischer, I. S. Averbukh, and Y. Prior, Phys. Rev. Lett. 99, 093002 2007 . 6 C. Riehn, Chem. Phys. 283, 297 2002 . 7 V. V. Matylitsky, C. Riehn, M. F. Gelin, and B. Brutschy, J. Chem. Phys. 119, 10553 2003 . 8 J. C. H. Spence, J. Electron Microsc. 54, 163 2005 . 9 R. A. Bartels, T. C. Weinact, N. Wagner, M. Baertschy, C. H. Greene, M. M. Murnane, and H. C. Kapteyn, Phys. Rev. Lett. 88, 013903 2002 . 10 J. Itatani, J. Levesque, D. Zeidler, H. Niikura, H. Pepin, J. C. Kieffer, P. B. Corkum, and D. M. Villeneuve, Nature London 432, 867 2004 . 11 M. Kaku, K. Masuda, and A. Miyazaki, Jpn. J. Appl. Phys., Part 2 43, L591 2004 . 12 R. de Nalda, E. Heesel, M. Lein, N. Hay, R. Velotta, E. Springate, M. Castillejo, and J. P. Marangos, Phys. Rev. A 69, 031804 2004 . 13 T. Kanai, S. Minemoto, and H. Sakai, Nature London 435, 470 2005 . 14 X. X. Zhou, X. M. Tong, Z. X. Zhao, and C. D. Lin, Phys. Rev. A 72, 033412 2005 . 15 C. B. Madsen and L. B. Madsen, Phys. Rev. A 74, 023403 2006 . 16 S. Ramakrishna and T. Seideman, Phys. Rev. Lett. 99, 113901 2007 . 17 P. Lan, P. Lu, W. Cao, Y. Li, and X. Wang, Phys. Rev. A 76, 021801 2007 . 18 M. Leibscher, I. S. Averbukh, and H. Rabitz, Phys. Rev. Lett. 90, 213001 2003 . 19 M. Leibscher, I. S. Averbukh, and H. Rabitz, Phys. Rev. A 69, 013402 2004 . 20 S. Ramakrishna and T. Seideman, Phys. Rev. Lett. 95, 113001 2005 . 21 K. F. Lee, D. M. Villeneuve, P. B. Corkum, and E. A. Shapiro, Phys. Rev. Lett. 93, 233601 2004 . 22 J. Salomon, C. Dion, and G. Turinici, J. Chem. Phys. 123, 144310 2005 . 23 A. Pelzer, S. Ramakrishna, and T. Seideman, J. Chem. Phys. 126, 034503 2007 . 24 T. Seideman, J. Chem. Phys. 115, 5965 2001 . 25 T. Seideman, Phys. Rev. Lett. 83, 4971 1999 . 26 Z.-C. Yan and T. Seideman, J. Chem. Phys. 111, 4113 1999 . 27 J. G. Underwood, M. Spanner, M. Y. Ivanov, J. Mottershead, B. J. Sussman, and A. Stolow, Phys. Rev. Lett. 90, 223001 2003 . 28 T. Seideman, J. Chem. Phys. 106, 2881 1997 . 29 T. Seideman, J. Chem. Phys. 107, 10420 1997 . 30 H. Stapelfeldt, H. Sakai, E. Constant, and P. B. Corkum, Phys. Rev. Lett.
2 1

Our main goal in the work presented above has been to detail the theory and numerical implementation of 3D align-

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp

154313-9

Molecular alignment by intense laser pulses

46

J. Chem. Phys. 128, 154313 2008 L. Holmegaard, S. S. Viftrup, V. Kumarappan, C. Z. Bisgaard, H. Stapelfeldt, E. Hamilton, and T. Seideman, Phys. Rev. A 75, 051403 2007 . 47 S. S. Viftrup, V. Kumarappan, S. Trippel, H. Stapelfeldt, E. Hamilton, and T. Seideman, Phys. Rev. Lett. 99, 143602 2007 . 48 J. C. H. Spence and R. B. Doak, Phys. Rev. Lett. 92, 198102 2004 . 49 R. Santra, C. Buth, E. R. Peterson, R. W. Dunford, E. P. Kanter, B. Krssig, S. H. Southworth, and L. Young, J. Phys.: Conf. Ser. 88, 012052 2007 . 50 E. A. Shapiro, I. Khavkine, M. Spanner, and M. Y. Ivanov, Phys. Rev. A 67, 013406 2003 . 51 C. M. Dion, A. Keller, O. Atabek, and A. D. Bandrauk, Phys. Rev. A 59, 1382 1999 . 52 S. C. Althorpe and T. Seideman, J. Chem. Phys. 113, 7901 2000 . 53 F. Lgar, S. Chelkowski, and A. D. Bandrauk, Chem. Phys. Lett. 329, 469 2000 . 54 E. Hamilton, T. Seideman, T. Ejdrup, M. D. Poulsen, C. Z. Bisgaard, S. S. Viftrup, and H. Stapelfeldt, Phys. Rev. A 72, 043402 2005 . 55 R. N. Zare, Angular Momentum Wiley, New York, 1988 . 56 L. D. Barron, Molecular Light Scattering and Optical Activity Cambridge University Press, Cambridge, 1982 . 57 M. Born and E. Wolf, Principles of Optics Pergamon, New York, 1968 . 58 N. Moiseyev and T. Seideman, J. Phys. B 39, L211 2006 . 59 T. Seideman, Chem. Phys. Lett. 253, 279 1996 . 60 A. R. Edmonds, Angular Momentum in Quantum Mechanics Princeton University Press, Princeton, New Jersey, 1957 . 61 I. Nevo, S. Kapishnikov, S. R. Cohen, K. Kjaer, T. Seideman, and L. Leizerovitch, Inducing Alignment in Self-Assembled Molecules on the Water Surface via Linearly Polarized Laser Pulses, Science submitted .

79, 2787 1997 . H. Sakai, A. Tarasevitch, J. Danilov, H. Stapelfeldt, R. W. Yip, C. Ellert, E. Constant, and P. B. Corkum, Phys. Rev. A 57, 2794 1998 . 32 T. Seideman, J. Chem. Phys. 111, 4397 1999 . 33 H. S. Chung, B. S. Zhao, S. H. Lee, S. Hwang, K. Cho, S.-H. Shim, S.-M. Lim, W. K. Kang, and D. S. Chung, J. Chem. Phys. 114, 8293 2001 . 34 O. Atabek, C. Dion, and A. Haj-Yedder, J. Phys. B 36, 4667 2003 . 35 A. B. Haj-Yedder, A. Auger, C. M. Dion, E. Cances, A. Keller, C. L. Bris, and O. Atabek, Phys. Rev. A 66, 063401 2002 . 36 R. de Nalda, C. Horn, M. Wollenhaupt, M. Krug, L. Baares, and T. Baumert, J. Raman Spectrosc. 38, 543 2007 . 37 D. Pinkham, K. E. Mooney, and R. R. Jones, Phys. Rev. A 75, 013422 2007 . 38 E. Hertz, A. Rouze, S. Gurin, B. Lavorel, and O. Faucher, Phys. Rev. A 75, 031403 2007 . 39 J. J. Larsen, K. Hald, N. Bjerre, H. Stapelfeldt, and T. Seideman, Phys. Rev. Lett. 85, 2470 2000 . 40 E. Pronne, M. D. Poulsen, C. Z. Bisgaard, H. Stapelfeldt, and T. Seideman, Phys. Rev. Lett. 91, 043003 2003 . 41 M. D. Poulsen, E. Pronne, H. Stapelfeldt, C. Z. Bisgaard, S. S. Viftrup, E. L. Hamilton, and T. Seideman, J. Chem. Phys. 121, 783 2004 . 42 E. Pronne, M. D. Poulsen, H. Stapelfeldt, C. Bisgaard, E. L. Hamilton, and T. Seideman, Phys. Rev. A 70, 063410 2004 . 43 J. G. Underwood, B. J. Sussman, and A. Stolow, Phys. Rev. Lett. 94, 143002 2005 . 44 K. F. Lee, D. M. Villeneuve, P. B. Corkum, A. Stolow, and J. G. Underwood, Phys. Rev. Lett. 97, 173001 2006 . 45 M. Poulsen, T. Ejdrup, H. Stapelfeldt, E. L. Hamilton, and T. Seideman, Phys. Rev. A 73, 033405 2006 .
31

Downloaded 18 Apr 2008 to 129.105.55.191. Redistribution subject to AIP license or copyright; see http://jcp.aip.org/jcp/copyright.jsp