1 Numerical Simulations of an Atomic Beam Focused by an Optical Potential T. Kidan, Joan Adler and A. Ron Physics Department, Technion-Israel Institute of Technology, Haifa 32000, Israel. e-mail Address: kidan@tx.technion.ac.il May 6, 1998 Abstract: We present a model to describe the behavior of an atomic beam in a focusing experiment and discuss its implementation. Such modeling is relevant to the development of atom lithography. For simplicity, the light-atom interaction may be treated as a conserving potential (neglecting spontaneous emission) so that the Schrodinger equation is sufficient for the description of the problem. We start from a two dimensional time dependent Schrodinger equation that is reduced to a one dimensional equation by a paraxial approximation, and then solved iteratively via Pade approximation. The solution is implemented in an interactive routine with Matlab(5.x) to compute and visualize the atomic wave packet development as it propagates through the laser light optical potential. We present examples of our implementation, starting with cases which have known analytic solutions such as propagation of the atomic wave packet through free space and through a parabolic potential. We also present a case which does not have a simple analytic solution, namely, the propagation of the atomic wave packet through a focusing optical potential, which has a Gaussian 1 2 profile in the propagation direction and a parabolic profile in the perpendicular direction. 2 3 I. Introduction Optical manipulations of atoms have been widely studied in recent years [1,2]. One possible application of this growing field is atom lithography, which has many advantages over other lithography techniques. For example, electron beam lithography is limited in speed and resolution by space charge effects, which are absent in atom lithography because the atom is neutral. X-ray and electron beam lithography induce damage resulting in limitation to the lithography resolution, while this is probably absent for low energy atoms. As a consequence of the small de Broglie wavelength, atom lithography has the potential for high resolution [3]. The general configuration of the atom lithography process is illustrated in Figure 1. The atomic beam is initially mechanically collimated, and then further collimated by passing through a region in which there is a laser light radiation field which collimates the atomic beam transversely, reducing the transverse velocity spread and hence the divergence of the beam. The highly collimated beam of atoms then passes through a standing wave of the radiation field, which is formed by a retro-reflected near resonant laser beam grazing across the surface of the substrate on to which the atoms are to be deposited. While passing through the standing wave the atoms experience a dipole force, at the node, which acts as a lens, causing them to focus and to form patterns as they deposit onto the surface. We will describe the experiment in detail in section II, where the atomic beam will be represented by a wave packet, interacting with the conserving 3 4 Atom source Vacuum chamber Slit Laser light for optical collimating Standing wave laser light Mirror Substrate Sample holder Figure 1: A typical apparatus for atom lithography utilizing laser focusing of atoms in a standing wave 4 5 light potential (neglecting the spontaneous emission) [1,4]. The form of Schrodinger equation used by these computer simulations will be obtained in section III from a two dimensional time dependent equation into an effectively one dimensional equation, by utilizing the paraxial approximation [5,6]. In section IV we will present the numerical solution. This solution is found by reducing the parabolic equation to a system of differential equations (via Pade approximants), whose solutions are standard finite differences equations [7]. The computer simulations and visualizations of the various cases will be described in section V, starting with the examination of the numerical algorithm for cases with known analytic solutions and proceeding with the case of the focusing experiment. In section VI we summarize and discuss applications of our routines. II. Description of the focusing experiment The beam of atoms in a typical focusing experiment are predominantly traveling in the z–direction (downwards in Figure 1), which traverses a focusing light potential, aligned both in the x (left to right in Figure 1) and z directions. Close enough [8] to the anti-node of the standing wave light field the potential has an approximately parabolic shape along the x–direction and acquires the Gaussian profile of the laser beam light field along the propagation direction, which is the z–direction. The focusing light potential at the anti-node region of the large period [9] optical standing wave is produced by the reflection of the red detuned laser beam from a glass surface. Assuming both 5 6 that the kinetic energy of the incoming atoms in the z–direction is much larger than the interaction energy and that the focusing region is uniform in the ydirection, the focusing is then described by an effectively one dimensional model of the quantum mechanical motion along the x–direction. Utilizing a paraxial approximation, the spatial evolution of the atomic wave function ( x, z ) along the z–direction is then described by the Schrodinger equation: 2 2 iv 0 ( x, z ) V ( x, z ) ( x, z ) 2 z 2m x (1) where v0 is the center of mass velocity of the incoming atoms in the zdirection, m is the mass of the atoms and V ( x, z ) is the interaction operator between the atoms and the focusing optical potential. (The derivation of this equation is presented in detail in section III). The presence of this potential is but a small perturbation (see section III) to the propagation of the wave packet along the z–direction. Equation (1) gives an effectively one-dimensional model for this focusing experiment. III. Construction of the effectively one dimensional model The two dimensional time dependent Schrodinger equation is: H ( x, z , t ) i ( x, z , t ) t (2) where the hamiltonian is H H0 V 2 2 V ( x, z ) 2m 6 (3) 7 and thus the evolution of the wave function in time is: ( x, z, t ) e i Ht ( x, z,0) . (4) The initial wave packet describing the atom in the focusing experiment is chosen to be a Gaussian in both z and x directions, propagating as a plane wave with momentum k z along the z-direction, 0 ( x, z,0) 1 ( x,0) 2 ( z,0) 1 4 2 x2 e x2 1 4 x2 2 z2 4 e ikz 0 z e z2 4 z2 (5) where the widths of these Gaussians are chosen to be z x and thus ( x, z , t ) e i E ( k z0 ) t ( x, z,0) (6) Substituting (6) into (2) yields the the stationary Schrodinger equation in two dimensions 2 x, z 2m E k V x, z x, z 0 2 z0 (7) For the selected conditions of the typical experiment, described above (equations (5), (6)), it is convenient to express the wave function ( x, z) as a product of ( x, z ) and a carrier wave moving in the positive z-direction: ( x, z) ( x, z)e ikz0 z (8) where kz 0 2mEk z 0 (9) Substitution of the wave function (8) into the two dimensional stationary Schrodinger equation (7) gives: 7 8 2 2 2 ( x, z ) 2ik z ( x, z ) 2 ( x , z ) z z x 0 2mE k 2 z0 2mV 2 ( x, z ) ( x , z ) k z ( x, z ) 2 (10) 0 The last equation can be simplified to become equation (1) (which we will solve numerically in the next section), in a few steps: First, utilization of a paraxial approximation will lead to the negligence of the second order derivative (the first term on the left-hand side of equation (10)). This is made possible by the fact that the function ( x, z ) changes slowly (see equations (5) to (6)), in the z-direction compared to the x-direction. Second, the use of relation (9) leads to the cancellation of the first term on the right-hand side of equation (10) with the third term from the right hand side. Note that the slow change of ( x, z ) comes about because the motion of the center of the wavepacket along the z-direction, for which most of the momentum is concentrated in only one spatial mode k z , is on the trajectory z v0t . In this case where 0 most of the momentum of the wave packet is concentrated in one spatial mode k z there is practically no reflection of the wave packet. 0 IV. The algorithm for the solution of the effectively one dimensional model Equation (1) is a parabolic partial differential equation that can be solved by different techniques. One convenient method is to use the finite-differences approximation via the ordinary differential equations. Consider equation (1) rewritten as 8 9 2 ( x, z ) a 2 bV ( x, z ) ( x, z ) z x where a (11) i i and b . 2mv 0 v0 The function ( x, z ) satisfies the initial condition (x,0) , which is a Gaussian form of the wave packet. The second derivative with respect to x at (x,z), on the right hand side of equation (11), is replaced by 1 ( x dx, z ) 2( x, z ) ( x dx, z ) so the equation becomes in its (dx) 2 matrix form: : b 1 2 (dx ) 2V1 1 1 a b 2 2 1 2 (dx ) V2 1 2 a . d a 2 dz (dx ) b N 2 1 2 (dx ) 2VN 2 1 N 2 a b 2 1 2 (dx ) VN 1 N 1 N 1 a (12) where i and Vi are the wave packet and the potential vectors respectively. The solution of this ordinary scalar differential equation, can be written in a more concise form as d A dz (13) where the matrix A is in general dependent on z and (z) satisfies the initial condition (0) , is simply: 9 10 ( z dz) exp(dzA)( z) (14) For the restriction on dz see [7]. The exponential function of the matrix A appearing on the right-hand side can be approximated by a [1,1] Pade’ approximant to give: 1 1 A 2 O[ A 3 ] exp( A) 1 1 A 2 (15) (Pade approximants are ratios of polynomials, for example, the [1,1] Pade approximant has polynomials of order A in both numerator and denominator.) This approximation is second order accurate in z, having an error term of the 1 third order in A with a coefficient of 12 . Although the [1,1] Pade approximant produces unwanted finite oscillations near points of discontinuity it is adequate for this simulation since the wave function is continuous and smooth over the grid and zero at the edges. The evolution of the wave packet through the two dimensional potential region can now be calculated iteratively from (14) and (15). The output is a matrix with the values of the propagating wave packet and its distortion caused by the potential. V. Computer simulations and visualization of the wave packets In order to demonstrate this solution, an interactive visualization using Matlab(5.x) has been prepared. Free propagation of the atomic wave packet and propagation through a parabolic potential in the x-direction for all z were 10 11 calculated, visualized and compared to the analytical known solutions of the evolution for a Gaussian wave packet. Both the simulation and the visualization were done using the convenient tools of Matlab. The program m-files are available on the world wide web [10]. The input parameters are the velocity of the atoms along the z - direction which is of the order of 1m/sec, the laser intensity, the width of the atomic wave packet, the length of the z grid, the number of lattice points and the location of the laser beam (the optical potential), again along the z – direction. All these parameters can take a wide range of values depending on the details of the experimental implementation. Some specific parameters are given in the example m-files. The optical potential for the focusing experiment has the form of ( zz 2 1 2 2 V ( x, z ) 2 x e 2 L 0) 2 L 2 where L is the laser's wavelength, L stands for the width of the laser's beam profile and z 0 is the location of the laser beam along the z-direction. The output is the square of the amplitude of the wave function, indicated by P 2 on the graph. Each simulation is visualized with two subplots, one is three dimensional and the other two dimensional. The color bar at the side of the lower plot indicates the P 2 value. The dual visualization technique reinforces visual perception. 11 12 On a PC with a Pentium II 233MHz processor and 64Mb SDRAM both the simulation and the visualization for 100 100 lattice took approximately 3 seconds for the free propagation simulation up to 40 seconds for the propagation through the parabolic - Gaussian potential. (On a 200Mhz with 32Mb RAM these times were 10 and 200 seconds respectively) We now present a series of figures illustrating the use of the program to visualize both free and constrained propagation of Rb atoms through a region one wavelength wide with the antinode in the middle (where the laser's wavelength is L = 0.760 m ). The laser's Gaussian profile (along the zdirection) is L =500 m wide. In Figure 2 the main menu of the program is pictured. In Figure 3 the free propagation of the atomic wave packet is shown. Here V(x,z)=0 and it is easily seen that the Gaussian wave packet spreads as it propagates, as expected. Figure 4 illustrates the propagation of the atomic wave packet through the parabolic (along the x – direction) potential. The Gaussian wave packet’s width oscillates as it propagates, as expected. In Figure 5 the propagation of the atomic wave packet through the parabolic (along the x – direction) potential is again shown but with the atomic beam off center. Again the Gaussian wave packet oscillates as it propagates. Figure 6 illustrates the propagation of the atomic wave packet through a potential that is parabolic along the x – direction, and Gaussian along the z – direction. The Gaussian wave packet is focused as it propagates through the potential and then spreads as it leaves the region of the potential. In Figure 7 we give a closer look at the propagation of the atomic wave packet through the same potential as in Figure 12 13 6, parabolic along the x – direction and Gaussian along the z – direction. Again the Gaussian wave packet is focused as it propagates through the potential and then spreads as it leaves the region of the potential. For those readers who are interested in the details of the routines, we conclude with a brief outline of the different files that we have placed on the web [10]. The README file tells the user to enter matlab and then write main at the matlab prompt. This displays the menu pictured in Figure 2. The intro.m, intro1.m etc files provide a short introduction to the physics of the problem; the help1.m file gives advice on running the interactive routine and the about.m gives contact information for Tal Kidan. The program contains five sets of demonstrations for the three different potentials, which are accessed thru the file demon.m, and the outputs of these are presented in figures 3-7 above. These have parameters already encoded for typical cases in the corresponding files evop.m (free), evop1.m (parabolic), evop1off.m (off-center parabolic), evop2.m (parabolic-Gaussian), evop3.m (parabolic-Gaussian), ready to reproduce the demonstration figures. The user can set her/his own values for each of the three preprogrammed potential cases (free, parabolic, parabolicGaussian) thru the interactive simulation option with sliders that move within the allowable parameter ranges. For all of these no programming skills are needed, just download the files into a single directory and enter matlab! The programs work both under windows and under UNIX (tested on AIX, HP/UX and LINUX). For other potentials it is necessary to edit a file of the evop.m 13 14 type, this can be done by simply replacing evop3.m if you do not want to bother with editing the interface. VI. Applications and summary An effectively one dimensional model was introduced in order to simulate the propagation of an atomic wave packet through a two dimensional potential. A simple algorithm was used to calculate the evolution of the wave packet and the visualization tools of Matlab allowed convenient visualization. This routine could be used to teach students about wave packet propagation. It has already found research application in a laboratory of Atom Optics where it is used to estimate the spot size of the beam of Rb atoms deposited on the surface of the substrate. The user interface which allows simultaneously changing parameters, such as the intensity of the laser light potential or the velocity of the atoms, has been very helpful to this experimental Atom Optics group. VII. Acknowledgements The motivation for this study came from Dr. N. Davidson, head of the experimental Atom Optics group at the Physics Department, Weitzman Institute of Science - Israel. The numerical solution and visualization were carried out as a project by T. Kidan in a computational physics class at the Physics Department, Technion - Israel Institute of Technology. 14 15 VIII. References [1] “Atom Optics”, C. S. Adams, M. Sigel and J. Mlynek, Phys. Rep. 240, 143 (1994). [2] Reference [1] covers the subject in depth including discussions on the validity of all the approximations needed to model various experiments. There are extensive references to previos calculations and experiments. [3] S. Nowak, T. Pfau and J. Mlynek, Appl. Phys. B 63, 203 (1996). [4] “Atomic motion in Laser light”, C. Cohen-Tannoudji, Fundamental systems in QO, Elsevier Science Publications B. V. (1991), p. 4-166. [5] “Computation of mode properties in optical fiber waveguides by a propagating beam method”, M.D. Feit and J. A. Fleck, Applied Optics, 19, 7, 1154 (1989). [6] “Atomic Fresnel images and possible applications in atom lithography”, U. Janicke, J. Phys. II France, 4, 1975 (1994). [7] “Numerical Solution of PDE: Finite Difference Method”, G.D. Smith, Oxford Press, 3rd ed. (1993), p. 111-126. [8] Ref [1] - Section III B1. [9] T. Sleator, T. Pfau, V. Balykin and J. Mlynek, Appl. Phys. B 54, 375, (1992). [10] http://comphy.technion.ac.il/~comphy/talkdn.html 15 16 Figure captions Figure 1: A typical apparatus for atom lithography utilizing laser focusing of atoms in a standing wave. Figure 2: The program’s main menu. Figure 3: Free propagation of the atomic wave packet. It is easily seen that the Gaussian wave packet is spreading as it propagates, as expected. Figure 4: Propagation of the atomic wave packet through the parabolic (along the x – direction) potential. The Gaussian wave packet’s width is oscillating as it propagates, as expected. Figure 5: Propagation of the atomic wave packet through the parabolic (along the x – direction) potential is again shown but with the atomic beam off center. Again the Gaussian wave packet is oscillating as it propagates. Figure 6: Propagation of the atomic wave packet through a potential that is parabolic, along the x – direction, and Gaussian along the z – direction. The Gaussian wave packet is focused as it propagates through the potential and then spreads as it leaves the region of the potential. Figure 7: A closer look at the propagation of the atomic wave packet through the same potential as in Figure 6, parabolic along the x – direction and Gaussian along the z – direction.. Again the Gaussian wave packet is focused as it propagates through the potential and then spreads as it leaves the region of the potential. 16