1: \documentclass[pra,showpacs,twocolumn]{revtex4}
2: \usepackage{dcolumn}
3: \usepackage{graphicx}\usepackage{amsmath}
4: \newcommand{\opr}[1]{\mathaccent 94 #1}
5: \newcommand{\ket}[1]{\left|#1\right\rangle}
6: \newcommand{\bra}[1]{\left\langle#1\right|}
7: \newcommand{\avr}[1]{\left\langle #1 \right\rangle}
8: \begin{document}
9: \sloppy
10: \title{Simulation of tunneling in the quantum tomography approach}
11: \author{Yu. E. Lozovik}
12: \email{lozovik@isan.troitsk.ru}
13: \author{V. A. Sharapov}\email{vladsh@e-mail.ru}
14: \author{A. S. Arkhipov}\email{antoncom@id.ru}
15: \affiliation{Institute of Spectroscopy RAS, Moscow region, Troitsk, Russia, 142190}
16: \date{\today}
17: \begin{abstract}
18: The new method for the simulation of nonstationary quantum processes is proposed. The method is based on
19: the tomography representation of quantum mechanics, {\it i.e.}, the state of the system is described by the
20: {\it nonnegative} function (quantum tomogram). In the framework of the method one uses the
21: ensemble of trajectories in the tomographic space to represent evolution of the system (therefore
22: direct calculation of the quantum tomogram is avoided). To illustrate the method we consider the
23: problem of nonstationary tunneling of a wave packet. Different characteristics of tunneling, such
24: as tunneling time, evolution of spatial and momentum distributions, tunneling probability are
25: calculated in the quantum tomography approach. Tunneling of a wave packet of composite particle,
26: exciton, is also considered; exciton ionization due to the scattering on the barrier is analyzed.
27: \end{abstract}
28: \pacs{03.65.Wj, 02.70.-Ns, 03.65.Xp}
29: \maketitle
30:
31: \section{Introduction}\label{Introduction}
32: Nowadays simulation of quantum systems is developed to a high extent (see, {\it e.g.}, reviews
33: \cite{Lee,Cep}). However, common simulation methods, for example, Path Integral Monte Carlo
34: or Wigner dynamics, use non-positively defined functions (wave function, the Wigner function,
35: {\it etc.}) to describe a quantum state. This leads to the difficulties with convergence of
36: corresponding integrals, especially harmful for the simulation of Fermi systems (sign problem).
37: There is a hope that using a real nonnegative function, describing the quantum state,
38: one can avoid these difficulties.
39:
40: The real nonnegative function in the phase space, completely describing the quantum state, was
41: proposed 60 years ago (\cite{Husimi}, see also \cite{Lee,Glaub,Sudar}). During the last decade
42: another very interesting representation has been actively developed: the quantum tomography,
43: operating with the ensemble of {\it scaled and rotated reference frames}, instead of the
44: phase space \cite{Man'ko,ManMan'ko,DodMan'ko,Man'koMan'ko,KV,SM97}. In the framework of this
45: formalism the state-describing function (called marginal distribution or quantum tomogram) is
46: real and nonnegative. The advantage is that the quantum tomogram is a {\it probability
47: distribution} shown to completely describe the quantum state \cite{SM95,GMD}.
48: It is one of the reasons why the quantum tomography has become so popular.
49:
50: In this paper we propose a new method for computer simulation of nonstationary quantum processes
51: based on the tomography representation of quantum mechanics and illustrate it considering the
52: problem of nonstationary tunneling of a wave packet. Many simulation approaches in nonstationary
53: quantum mechanics are based on the numerical solution of the time-dependent Schr\"odinger
54: equation. There are also methods using the ensembles of classical trajectories to simulate
55: quantum evolution. For example, the method of "Wigner trajectories", based on the Wigner
56: representation \cite{Wigner}, is well known (see, {\it e.g.}, review \cite{Lee} for details) and
57: was recently successfully applied to investigate the tunneling of a wave packet \cite{DM,LF}.
58: Using such trajectories, one does not have to store the large arrays representing, say, the wave
59: function (contrary to grid methods), therefore there is a gain in computer memory.
60:
61: The quantum tomogram $w$ depends on the variables $\{X,\mu,\nu\}$, where $X = \mu q + \nu p$,
62: and $q,p$ are the coordinates and momenta of the system, respectively, $\mu, \nu$ are the
63: parameters of scaling and rotation of reference frame in the phase space. The quantum tomogram
64: is nonnegative and normalized in $X$ direction, therefore it can be interpreted as a distribution
65: function of the value $X$. In our method the ensemble of trajectories in space $\{X,\mu,\nu\}$
66: is introduced to describe the quantum evolution. The trajectories are governed by the dynamical
67: equations obtained from the evolution equation for the quantum tomogram.
68:
69: We demonstrate the method considering the nonstationary tunneling of a wave packet through the
70: potential barrier. For this problem we calculated tunneling times, which are of interest nowadays,
71: both for basic science ({\it e.g.}, what is the time spent by an atom to tunnel from the trap?)
72: and for applications (electronic tunneling time is connected with the operation rate of some
73: nanostructure-based devices). We also analyzed the evolution of the wave packet in coordinate and
74: momentum spaces in details. Another demonstration was designed to show that our method is not
75: restricted to one-particle simulations. Namely, we investigated the tunneling of a wave packet of
76: composite quasiparticle, an exciton (coupled electron and hole in semiconductor), in a
77: one-dimensional nanostructure (quantum wire). In this case we had two degrees of freedom. For
78: this problem, in addition to the probability density evolution, we determined the probability
79: of ionization due to electron and hole scattering on the barrier in different directions.
80:
81: In Sec.~\ref{Method} we present description of the method, in Sec.~\ref{Tunneling} describe the
82: model problem and main results for a wave packet tunneling. Exciton tunneling is considered in
83: Sec.~\ref{Exciton} and the work is summarized in Sec.~\ref{Conclusion}.
84:
85: \section{The method of simulation}\label{Method}
86: The quantum tomogram $w(X,\mu,\nu)$ is connected with the density matrix $\rho(q,q')$ as \cite{SMJ,VIM}:
87: \begin{eqnarray}
88: && \rho(q,q') = \int w(X,\mu,q-q')e^{i(X - \mu (q+q')/2)}\frac{d\mu dX}{2\pi},\\ \label{WtoDM}
89: && w(X,\mu,\nu) = \nonumber\\
90: && \int e^{-i(k(X-\mu q-\nu p)+pu)}\rho(q+\frac u 2,q-\frac u 2)\frac{dp dk dq du}{2\pi^2}\label{DMtoW}
91: \end{eqnarray}
92:
93: Consider the case of the particle with mass $m$ in one-dimensional space. If the Hamiltonian of the system is
94: \begin{equation}
95: H = \frac{p^2}{2m}+V(q),
96: \end{equation}
97: then the integral transformation (\ref{DMtoW}) applied to the time-dependent
98: evolution equation for the density matrix gives \cite{Man'ko}
99: \begin{eqnarray}
100: \dot w - \frac \mu m \frac{\partial w}{\partial \nu} - 2\frac{\partial V(\tilde q)}{\partial
101: q}\left(\frac \nu 2 \frac{\partial}{\partial X}\right)w + \nonumber\\
102: 2\sum_{n=1}^\infty \frac{(-1)^{n+1}}{(2n+1)!}\frac{\partial^{2n+1}V(\tilde q)}{\partial
103: q^{2n+1}}\left(\frac \nu 2 \frac{\partial}{\partial
104: X}\right)^{2n+1}w = 0,
105: \label{EvT}
106: \end{eqnarray}
107: where we use $\hbar = 1$ and $\tilde q$ is given by
108: \begin{equation}
109: \tilde q = - \left(\frac{\partial}{\partial X}\right)^{-1} \frac{\partial}
110: {\partial \mu}
111: \label{OprQ}
112: \end{equation}
113:
114: Eq.(\ref{EvT}) can be rewritten as
115: \begin{eqnarray}
116: \frac{\partial w}{\partial t} + \frac{\partial w}{\partial X}
117: G_X(X,\mu,\nu) + \frac{\partial w}{\partial \mu}G_{\mu}(X,\mu,\nu) +\nonumber\\
118: \frac{\partial w}{\partial \nu}G_{\nu}(X,\mu,\nu) = 0,
119: \label{ContEqXMuNu}
120: \end{eqnarray}
121: where functions $G$ depend on quantum tomogram, its derivatives and antiderivatives (the latter
122: corresponding to terms with $(\partial/\partial X)^{-1}$ in Eq.(\ref{EvT})). Generalization for
123: the case of more variables is straightforward because the form of the equations does not change.
124: Functions $G$ for the problem under investigation are given in Sec.~\ref{Tunneling}. The
125: evolution equation rewritten as (\ref{ContEqXMuNu}) has the form of continuity equation for the
126: quantum tomogram
127: \begin{equation}
128: \frac{dw}{dt} = \frac{\partial w}{\partial t} + \frac{\partial w}{\partial X}
129: \dot X + \frac{\partial w}{\partial \mu}\dot \mu + \frac{\partial w}{\partial \nu}\dot\nu = 0
130: \label{ContEq}
131: \end{equation}
132: This equation is analogous to the continuity equation for classical distribution function
133: and Liouville equation. As known, the characteristics of Liouville equation are the classical
134: trajectories in phase space and they obey Hamilton equations of motion. The
135: quantum tomogram is nonnegative and we use it as a distribution function for trajectories
136: in the space $\{X, \mu, \nu\}$, obeying the equations analogous to Hamilton equations for
137: the classical trajectories. From the comparison of Eq.(\ref{ContEqXMuNu}) with Eq.(\ref{ContEq})
138: it is obvious that the trajectories are governed by the equations
139: \begin{equation}
140: \dot X = G_X(X,\mu,\nu),
141: \dot \mu = G_{\mu}(X,\mu,\nu),
142: \dot \nu = G_{\nu}(X,\mu,\nu)
143: \label{EqMotionG}
144: \end{equation}
145:
146: The trajectories are used to avoid the direct calculations of the distribution function (contrary
147: to grid methods where the wave function is calculated at every point to solve numerically
148: Schr\"odinger equation). Hence it is necessary to use some approximation for the quantum tomogram
149: and we use local exponential approximation (as in Ref.~\cite{DM} for the Wigner function):
150: \begin{equation}
151: w(X,\mu,\nu) = w_0e^{-[(y-y_a(t))A_a(t)
152: (y-y_a(t))+b_a(t)(y-y_a(t))]},
153: \label{TLocA}
154: \end{equation}
155: where $y = \{X, \mu, \nu\}$, and $y_a$ is the point under consideration. Parameters of this
156: approximation are matrix $A_a$ and vector $b_a$, and some combinations of these parameters enter
157: the evolution equation (\ref{EvT}), instead of the derivatives and antiderivatives of the quantum
158: tomogram. Calculation of average $X, \mu, \nu$ and their average products allows to obtain $A_a$
159: and $b_a$. After that, functions $G$ are known and dynamical equations (\ref{EqMotionG}) can be
160: solved numerically.
161:
162: We would like to emphasize that we use the local approximation (\ref{TLocA}) only for the
163: calculation of r.h.s. in the equations of motion (\ref{EqMotionG}). The use of ensemble of trajectories to
164: represent the quantum tomogram means the approximation of quantum tomogram as a set of
165: delta-functions, each delta-function corresponds to one trajectory. If the number of trajectories
166: approaches infinity, the quantum tomogram can be approximated by the set of delta-functions with
167: arbitrary precision. This is analogous to what is conventionally done in classical statistical
168: mechanics for the distribution function in phase space. But unlike the classical statistical
169: mechanics now the trajectories are not independent: the approximation (\ref{TLocA}) is used to
170: take the non-local character of quantum-mechanical evolution into account.
171:
172: The validity of this approximation holds if the quantum tomogram is smooth and the trajectories
173: are close to each other. For example, this approximation can fail if one tries to consider a
174: plain wave with wave vector $k$: in this case $w(X,\mu=0,\nu=1) = \delta(X-k)$. Approximation
175: (\ref{TLocA}) also works not well for unbounded motion, because the trajectories scatter with
176: time. If there are few trajectories in the region around the given point, then the approximation
177: (\ref{TLocA}) will not reconstruct the quantum tomogram well, due to lack of statistics.
178:
179: We consider the tunneling of wave packets through the barrier, and comparing our results with
180: exact quantum computation we see that approximation (\ref{TLocA}) holds for this problem (see
181: Secs.~\ref{Tunneling},\ref{Exciton}). For example, considering the tunneling through the
182: potential barrier from the well, we deal with both the region of bounded motion (in the well) and
183: unbounded motion (beyond the barrier), still the approximation works not bad (Sec.~\ref{Tunneling}).
184: For higher initial energy the penetration through the barrier is higher, and the evolution of most
185: trajectories corresponds to unbounded motion, then the validity of Eq.(\ref{TLocA}) becomes poorer
186: (see the end of Sec.~\ref{Tunneling}), in agreement with the discussion above. But in general local
187: approximation (\ref{TLocA}) works satisfactory even for quite long time intervals (see Fig.~\ref{Fig1}).
188:
189: To obtain any information about the system, we have to calculate some average values. Consider an
190: arbitrary operator $A(\opr q,\opr p)$. Average value $\langle A\rangle$ of corresponding physical
191: quantity is calculated in the tomographic representation of quantum mechanics as \cite{OM}
192: \begin{equation}
193: \langle A\rangle = \int A(\mu,\nu)e^{iX}w(X,\mu,\nu)dXd\mu d\nu,
194: \label{Aver}
195: \end{equation}
196: where $A(\mu,\nu)$ is the Fourier component of the Weyl symbol $A^W(q,p)$ of operator
197: $A(\opr q,\opr p)$ (see, {\it e.g.}, \cite{Lee}):
198: \begin{equation}
199: A(\mu,\nu) = \int A^W(q,p)exp(-i(\mu q + \nu p))\frac{dqdp}{4\pi^2}\label{FofW}
200: \end{equation}
201:
202: For the calculation of average values we use the following approximation of quantum tomogram:
203: \begin{equation}
204: w(X,\mu,\nu,t) = \sum_{j=1}^J\delta(X-X_j(t))\delta(\mu-\mu_j(t))
205: \delta(\nu-\nu_j(t)),\label{TomAp}
206: \end{equation}
207: where the summation is made over all $J$ trajectories; $X_j(t),\mu_j(t),\nu_j(t)$ are the
208: coordinates of the $j$-th trajectory in $\{X,\mu,\nu\}$ space at time $t$. Such approximation
209: corresponds to use of the ensemble of trajectories. In the regions, where $w(X,\mu,\nu)$ is
210: small, trajectories are rare, and where it is large, trajectories are accumulated. The more
211: trajectories are used, the better the approximation (\ref{TomAp}) works. If during the simulation
212: the wave function has the form of a compact wave packet, even consisting of several distinct parts,
213: approximation (\ref{TomAp}) holds, because in this case one has the compact sets of trajectories
214: providing good statistics. This is the case for problems considered, and therefore the use of
215: this approximation does not change results essentially, in comparison with the exact quantum
216: computation (Secs.~\ref{Tunneling},\ref{Exciton}).
217:
218: For the operators $A(\opr q)$, depending on $\opr q$ only, expression for $\langle A\rangle$
219: takes the form:
220: \begin{equation}
221: \avr A = \int A(X)w(X,\mu=1,\nu=0)dX,\label{AverAq}
222: \end{equation}
223: where $A(X)$ is the function corresponding to the operator $A(\opr q)$ in coordinate
224: representation, $A(X) = A(q = X)$. The method of calculation of an average $\avr{A(\opr q)}$ at
225: arbitrary time $t$, with the approximation (\ref{TomAp}), is quite simple. One just takes into
226: account the trajectories with any $X$ and with $\mu(t), \nu(t)$ from the small region near
227: $\mu = 1, \nu = 0$ only, and performs a summation of $A(X)$ over all such trajectories.
228:
229: The developed method is similar to well-known method of Wigner trajectories (see \cite{Lee} for
230: review), where the ensemble of trajectories is introduced in the phase space, with the Wigner
231: function used as a quasi-distribution function. The quantum tomogram is defined in the space
232: $\{X,\mu,\nu\}$, that is not as simple for understanding as the phase space used in the Wigner
233: approach. On the other hand, quantum tomogram is true distribution function, while the Wigner
234: function can be both positive and negative. In spite of these differences the two approaches are
235: quite close in general, and there are some difficulties common for both methods. The discussion
236: (see above) of the approximation for the tomogram (such as (\ref{TLocA})) concerns, in principle,
237: the method of Wigner trajectories either. Another important example is the discontinuity of the
238: Wigner trajectories discussed in Ref.~\cite{SBMJChemPhys1993}, that is due to the fact that the
239: trajectories are not independent as in classical statistical mechanics, their evolution depends on
240: their local distribution. The same is applicable for the quantum tomography approach, where the
241: trajectories are also not independent for the same reason. In Ref.~\cite{SBMJChemPhys1993} a possible
242: alternative to Wigner function was proposed. Namely, in that work the authors proposed to use the
243: Weyl transforms of some operators (the Wigner function, up to the constant, is the Weyl transform
244: of the density operator) instead of the Wigner function to generate the ensemble of trajectories.
245: The same can be introduced for the quantum tomography. Applying the transform (\ref{DMtoW}) to the
246: matrix elements of an operator $A$ we obtain the symbol $w_A(X,\mu,\nu)$. This function is not
247: nonnegative in general, but in analogy with the Wigner trajectories one can use $w_A(X,\mu,\nu)$
248: to develop some new ensemble of trajectories. Probably, as with the Weyl transforms trajectories
249: \cite{SBMJChemPhys1993}, it will be more convenient to use the trajectories corresponding to
250: $w_A$ in certain cases. Of course, this problem needs further investigation.
251:
252: \section{Simulation of tunneling of a wave packet}\label{Tunneling}
253: \subsection{The model and calculated average values}
254: We choose the external potential to coincide with the potential used in \cite{DM}, for comparison
255: of the results of simulation in quantum tomography approach with those obtained by other methods.
256: We consider behavior of the wave packet in one-dimensional space in external potential
257: \begin{equation}
258: V(q) = \frac{m\omega_0^2q^2}{2} - \frac{bq^3}{3}
259: \label{Vext}
260: \end{equation}
261: As the potential has only the second and third powers of coordinate, all its derivatives of
262: order more than the third vanish. Evolution equation in this case has the form ($\hbar = 1$):
263: \begin{eqnarray}
264: \frac{\partial w}{\partial t} - \frac \mu m \frac{\partial w}{\partial \nu} +
265: 2\left[-\frac{\partial V(\tilde q)}{\partial q}\left(\frac \nu 2 \frac{\partial}{\partial X}\right)+\right.\nonumber\\
266: \left.\frac 1 6 \frac{\partial^3V(\tilde q)}{\partial q^3}
267: \left(\frac \nu 2 \frac{\partial}{\partial X}\right)^3\right]w = 0
268: \label{EvTX23}
269: \end{eqnarray}
270: For the potential given by Eq.(\ref{Vext}) evolution equation reads as:
271: \begin{eqnarray}
272: \frac{\partial w}{\partial t} - \frac \mu m \frac{\partial w}{\partial \nu}
273: + m\omega_0^2\nu\frac{\partial w}{\partial \mu} -
274: \frac{b \nu^3}{12}\frac{\partial^3 w}{\partial X^3} +\nonumber\\
275: b\nu\left(\frac{\partial}{\partial X}\right)^{-1}\frac{\partial^2
276: w}{\partial \mu^2} = 0,
277: \label{EvTX23Cont}
278: \end{eqnarray}
279: and dynamical equations have the form
280: \begin{eqnarray}
281: \frac{\partial X}{\partial t} &=& \frac{b\nu^3}{12} \frac
282: 1 w \frac{\partial^2 w}{\partial X^2}\nonumber\\
283: \frac{\partial \mu}{\partial t} &=& m\omega_0^2\nu - \frac{b\nu}{w}
284: \left(\frac{\partial}{\partial X}\right)^{-1}\frac{\partial w}{\partial \mu}\nonumber\\
285: \frac{\partial \nu}{\partial t} &=& - \frac \mu m
286: \label{EqMotionM}
287: \end{eqnarray}
288:
289: We use atomic units throughout, $\hbar = m_e = |e| = 1$, where $m_e$ and $e$ are the mass and
290: charge of a free electron. The particle with mass $m = 2000$ is regarded. Parameters of
291: the potential are $\omega_0 = 0.01$ and $b = 0.2981$. This potential has the minimum at
292: $q = 0$ $(V(0) = 0)$ and maximum at $q = 0.6709$ $(V(0.6709) = 0.015)$, therefore here we
293: consider the motion of a particle in the potential well with infinite left wall and the
294: barrier of height $0.015$ at $q = 0.6709$. This model problem roughly describes
295: nonstationary tunneling of an atom from the trap.
296:
297: Initially the particle represented by the wave packet is situated to the left from $q = 0$, its
298: mean momentum is zero. The particle can oscillate in the potential well and can tunnel or pass
299: above the barrier. The probabilities of these processes depend on the initial energy of the wave
300: packet. We consider the problem, where all parameters, except the initial mean coordinate $q_0$
301: of the wave packet, are fixed (initial mean momentum equals zero, dispersions of the wave packet
302: in coordinate and momentum spaces are $\approx 0.3$ and $\approx 1.6$, respectively).
303:
304: We solve the equations (\ref{EqMotionM}) numerically. As in \cite{DM} we consider three values
305: of $q_0$: $-0.2, -0.3$ and $-0.4$. The most interesting quantities characterizing tunneling
306: are reaction probability and tunneling time. Reaction probability is defined as
307: \begin{equation}
308: \int\limits_{q_a}^{\infty}|\psi(x,t)|^2dx, \label{ReactProb}
309: \end{equation}
310: where $q_a = 0.6709$ (the point where potential has the maximum), the maximum value of reaction
311: probability is unity. Reaction probability shows what part of the wave packet is currently beyond
312: the barrier.
313:
314: Tunneling time of the wave packet is also an important feature of tunneling. There are a lot of
315: methods to determine tunneling time \cite{BL1982,LM,SokConnor,Baz',Ryb,But,HS,BL,AB,Wlod,BEMuga,
316: LLBR,DelgMuga,MLPhysRep2000}. We use the approach where tunneling time is calculated as the
317: difference of {\it presence times} (see \cite{MLPhysRep2000} for review) at point $x_a$ and $x_b$,
318: located on the opposite sides of the barrier:
319: \begin{equation}
320: t_T(x_a,x_b) = \langle t(x_b)\rangle-\langle t(x_a)\rangle\label{TunTime}
321: \end{equation}
322: The presence time at arbitrary point $x_0$ is
323: \begin{equation}
324: \langle t(x_0)\rangle = \frac{\int\limits_0^\infty t |\psi(x_0,t)|^2dt}
325: {\int\limits_0^\infty |\psi(x_0,t)|^2dt}\label{PresTime}
326: \end{equation}
327:
328: \subsection{Reaction probability}
329: Here we present the results obtained by means of our method and compare them with the exact
330: numerical solution of Schr\"odinger equation. In Fig.~\ref{Fig1} we present the dependence of
331: reaction probability (\ref{ReactProb}) on time for three values of initial mean coordinate of the
332: wave packet $q_0 = -0.2, -0.3$ and $-0.4$, corresponding mean energies of the wave packet are
333: $\approx 0.75V_0$, $\approx 1.25V_0$ and $\approx 2.0V_0$, respectively. Solid lines represent
334: the results of simulation in the quantum tomography approach (QT) and dashed lines correspond to
335: the numerical solution of Schr\"odinger equation (exact quantum computation). Due to the increase
336: of initial mean energy with the increase of $|q_0|$, the portion of high energy components in the
337: wave packet grows. This leads to the higher portion of components, which pass through the barrier,
338: either because their energy is greater than the height of the barrier, or due to tunneling.
339: Therefore, with the growth of $|q_0|$, reaction probability becomes larger and one can see that
340: the curves corresponding to different $q_0$ are above each other in Fig.~\ref{Fig1}. The time
341: evolution of reaction probability is qualitatively the same for every $q_0$. The components,
342: which have passed through the barrier, can not return, because for $q > 0.6709$ potential
343: diminishes with the growth of coordinate, and so reaction probability can not decrease with
344: time. At first it grows rapidly due to transmission of components with the energy higher than
345: the height of the barrier (comparison with the classical solution of corresponding problem,
346: for which only transmission above the barrier is possible, convinces us in it). Then
347: reaction probability continues to grow slowly, because of the tunneling. All these features
348: present both for QT simulation and for exact quantum computation.
349:
350: \begin{figure}
351: \includegraphics[height=7cm]{fig1.eps}
352: \caption{\label{Fig1}The dimensionless reaction probabilities (\ref{ReactProb}) for three
353: values of initial mean coordinate of the wave packet: $q_0 = -0.2$, $-0.3$, and $-0.4$ a.u.
354: Solid lines are for the simulation in quantum tomography approach, dashed lines are
355: for the exact numerical solution.}
356: \end{figure}
357:
358: In comparison with the exact computation, reaction probability for the QT simulation is slightly
359: higher. Note also some difference in the character of increase of the reaction probability for QT
360: simulation and exact solution: in the former case the curves are not so smooth. These
361: differences are due to the finite number of trajectories used in QT simulation: for smaller
362: number of trajectories (not shown) reaction probability curves resemble staircase more evidently
363: (this is connected with the overestimation of the role of wave packet oscillations in the well
364: for the finite number of trajectories), and quantitative deviation from exact result is stronger.
365: But in general, for quite large number of trajectories, as for the case shown in Fig.~\ref{Fig1},
366: QT simulation results on reaction probability are quite close to those obtained through the exact
367: quantum computation (compare also with the method of "Wigner trajectories" in the work by
368: Donoso and Martens \cite{DM}).
369:
370: \subsection{Evolution of the wave packet and tunneling times}
371: Besides the reaction probability we also obtained a number of new qualitative and quantitative
372: results, which described in details the behavior of the wave packet during tunneling. We also
373: calculated tunneling times using the concept of presence time (see below).
374:
375: \begin{figure}
376: \includegraphics[height=7cm]{fig2.eps}
377: \caption{\label{Fig2} Probability density in coordinate space for QT simulation (histograms)
378: and exact solution (smooth lines), at times $t = 0$ a.u. (left) and $t = 200$ a.u. (right). The
379: barrier is at the point $0.6709$ a.u., $q_0 = -0.2$ a.u.}
380: \end{figure}
381:
382: \begin{figure}
383: \includegraphics[height=7cm]{fig3.eps}
384: \caption{\label{Fig3} Probability density in coordinate space for QT simulation (histogram)
385: and exact solution (smooth line), at time $t = 300$ a.u. The barrier is at the point $0.6709$
386: a.u., $q_0 = -0.2$ a.u.}
387: \end{figure}
388:
389: \begin{figure}
390: \includegraphics[height=7cm]{fig4.eps}
391: \caption{\label{Fig4} Probability density in coordinate space for QT simulation (histogram)
392: and exact solution (smooth line), at time $t = 400$ a.u. The barrier is at the point $0.6709$
393: a.u., $q_0 = -0.2$ a.u.}
394: \end{figure}
395:
396: The following discussion concerns tunneling of the wave packet with initial mean coordinate
397: $q_0 = -0.2$. Note that the maximum of the potential is at the point $x = 0.6709$. We present
398: the normalized probability density $|\psi(x)|^2$ in coordinate space
399: (Figs.~\ref{Fig2}~-~\ref{Fig4}) and $|\psi(p)|^2$ in momentum space
400: (Figs.~\ref{Fig5}~,~\ref{Fig6}) for several successive time moments. In these figures smooth
401: lines show the shape of the wave packet obtained by means of exact quantum computation.
402: Histograms represent the result of {\it single QT run}. One can consider many runs with the same
403: number of trajectories and average the probability density over all these runs to obtain smoother
404: picture. But here we would like to show what QT simulation can give for one run, in comparison
405: with the exact quantum computation. Therefore the histograms (QT) fit the smooth solid lines in
406: Figs.~\ref{Fig2}~-~\ref{Fig4} (exact solution) not ideally, still the resemblance is obvious.
407:
408: First, consider the probability density $|\psi(x)|^2$ in
409: coordinate space (Figs.~\ref{Fig2}~-~\ref{Fig4}). One can see
410: (Fig.~\ref{Fig2}) that initially the wave packet has Gaussian
411: form. It begins to move as a whole towards the potential minimum
412: at $x = 0$ (initial mean momentum is zero, but the potential
413: inclines in that direction), passes that point, is accelerated,
414: and collides with the barrier. During the motion the wave packet
415: broadens (due to dispersion in momentum space, compare right and
416: left plots in Fig.~\ref{Fig2}) but the interaction with the
417: barrier changes its form more substantially ($t = 300$ and $t =
418: 400$, Figs.~\ref{Fig3},~\ref{Fig4}). The wave packet shrinks a
419: little, some components pass through the barrier and transmitted
420: part can be seen beyond the barrier ($x = 0.6709$). As the
421: transmitted part can not return and accelerates (potential
422: diminishes with the distance for $x > 0.6709$), the enriching of
423: the wave packet by high-energy components must present (see
424: below).
425:
426: All features described in the previous paragraph are common for
427: both the exact solution and QT simulation. The histograms in
428: Figs.~\ref{Fig2}~-~\ref{Fig4} fit the smooth lines representing
429: exact solution better for earlier times, but even after the
430: interaction with the barrier (Fig.~\ref{Fig4}), resemblance is
431: quite close. This shows that approximations (\ref{TLocA}) and
432: (\ref{TomAp}) work not bad for the problem under consideration.
433:
434: \begin{figure}
435: \includegraphics[height=7cm]{fig5.eps}
436: \caption{\label{Fig5} Initial probability density in momentum space for QT simulation
437: (histogram) and exact solution (smooth line). $t = 0$ a.u.}
438: \end{figure}
439:
440: \begin{figure}
441: \includegraphics[height=7cm]{fig6.eps}
442: \caption{\label{Fig6} Probability density in momentum space for QT simulation (histogram) and
443: exact solution (smooth line) at $t = 400$ a.u.}
444: \end{figure}
445:
446: Proceed now to the evolution of the wave packet in momentum space. To
447: confirm our analysis on the acceleration of the transmitted part
448: of the wave packet, we present the probability density
449: $|\psi(p)|^2$ in momentum space in Figs.~\ref{Fig5},~\ref{Fig6},
450: at times $t = 0$ and $t = 400$, respectively. As the wave
451: function is initially the Gaussian wave packet, the initial
452: distributions both in coordinate and momentum space are Gaussian
453: (compare Fig.~\ref{Fig2} and Fig.~\ref{Fig5}). But after the wave
454: packet has interacted with the barrier, the distribution in momentum
455: space changes substantially (Fig.~\ref{Fig6}). The higher the
456: energy of incident particle, the greater the tunneling
457: probability. Therefore the barrier transmits mainly the wave
458: packet components with relatively high energy, serving as an
459: energy selector. Components with high momentum arise, as the
460: transmitted part of the wave packet is accelerated in the region
461: of lowering potential beyond the barrier, and we
462: observe the enriching of the wave packet by high energy
463: components. The resemblance between the histograms (QT
464: simulation) and smooth solid lines (exact solution) is somewhat
465: poorer for momentum distribution at large times ($t = 400$,
466: Fig.~\ref{Fig6}) than for coordinate distribution
467: (Fig.~\ref{Fig4}). This is due to the fact that one deals with
468: finite number of trajectories and has to sample quite large
469: interval in momentum space, because the transmitted part is
470: permanently accelerated. Therefore, with time, momentum
471: distribution spreads and there are not many
472: trajectories with $\mu, \nu$ close enough to $\mu = 0, \nu = 1$
473: for given momentum $p$ (see Sec.~\ref{Tunneling}). As for the
474: considered value of initial mean coordinate $q_0 = -0.2$ the initial
475: energy is not very large ($\approx 0.75V_0$, where $V_0$ is the
476: height of the barrier), the wave packet mainly stays in the well
477: (for the time considered $t = 400$ only $\approx 20\%$ of the
478: wave packet is transmitted, see Fig.~\ref{Fig1}) and the
479: distribution in coordinate space is more compact.
480:
481: \begin{figure}
482: \includegraphics[height=7cm]{fig7.eps}
483: \caption{\label{Fig7} Tunneling times with errors for several values of initial mean coordinate
484: of the wave packet $q_0$. Results of the QT simulation (squares) are compared with exact quantum
485: computation (circles).}
486: \end{figure}
487:
488: In Fig.~\ref{Fig7} we present the dependence of tunneling time on
489: initial mean position of the wave packet. Tunneling time is
490: determined as the difference of presence times (\ref{PresTime})
491: for points $x_a = 0.5\cdot0.6709$ and $x_b = 2.0\cdot0.6709$ (at
492: $x = 0.6709$ potential has the maximum). Tunneling is usually
493: stronger for the higher energy. Therefore, the increase of
494: $|q_0|$ (and corresponding increase of initial mean energy) leads
495: to the growth of the average speed both of the transmitted part
496: and of the wave packet as a whole. The transmitted part passes
497: the region of the barrier (the space between the points $x_a$ and
498: $x_b$) faster, and so one expects that the increase of $|q_0|$
499: causes the decrease of tunneling time. Indeed, the value of
500: tunneling time drops with increase of $|q_0|$. Results of QT
501: simulation (squares in Fig.~\ref{Fig7}) deviate from those of
502: exact computation (circles) within the range of errors. The
503: deviation is maximal for large $|q_0|$. Probably, this is because
504: for large $|q_0|$ the wave packet leaves the well almost entirely
505: (see Fig.~\ref{Fig1}), and the evolution of most trajectories
506: corresponds to the unbounded accelerated motion. In such
507: situation trajectories scatter and approximation
508: (\ref{TomAp}) does not represent the quantum tomogram as exactly
509: as for smaller $|q_0|$.
510:
511: \section{Simulation of the exciton tunneling}\label{Exciton}
512: The method described in Sec.~\ref{Method} can be used to simulate the evolution of systems with
513: more than one degree of freedom. In this section we demonstrate such a possibility, considering
514: nonstationary tunneling of the composite particle, exciton, through the potential barrier in
515: one-dimensional (1D) semiconductor structure (quantum wire). Exciton is a bound state of electron
516: and hole in semiconductor, therefore we deal with two degrees of freedom in contrast to
517: Sec.~\ref{Tunneling}.
518:
519: Possible experimental realization is as follows. Consider semiconductor quasi-one-dimensional
520: nanostructure where the motion is allowed only in one direction (quantum wire). Transverse motion is
521: restricted due to strong confining barriers. Potential barrier in the direction of allowed motion
522: can be located at some point of the quantum wire either using the semiconductor heterojunction or
523: by the gate. Using femtosecond laser pulses, we can form an excitonic wave packet, either by the
524: quasiresonance pumping, or exciting an electron from valence band with formation of a hole and
525: subsequent binding of two particles into exciton. Then the excitonic wave packet can move to the
526: barrier and with the help of some detectors one can investigate scattering of the exciton.
527:
528: Keeping this in mind let us construct the model for the simulation. We use the constants
529: corresponding to GaAs for reference (dielectric constant $\varepsilon=12.5$, effective masses of
530: electron and hole are $m_e=0.07m_e^{(0)}$ and $m_h=0.15m_e^{(0)}$, respectively, where $m_e^{(0)}$
531: is the electron mass in vacuum). 3D exciton in bulk GaAs is characterized by effective Bohr radius
532: $a^*\approx 10 nm$ and binding energy $E_C^*\approx 4 meV$. We use unit of length $a^*$, unit of
533: mass $m_e$, and $\hbar=1$. Corresponding units of energy and time are
534: $E_0=\hbar^2/(m_ea^2)\approx10meV$ and $t_0 = m_ea/\hbar\approx100fs$.
535:
536: The energy spectrum and wave functions of relative electron and hole motion in 3D exciton are
537: analogous to those of hydrogen atom. But this is not the case for 1D exciton. First, electron-hole
538: effective interaction potential in quasi-1D structure is not Coulomb. Indeed, if the exciton size in
539: the direction of allowed motion is much greater than the width of the quantum wire (in the transverse
540: direction), then the adiabatic approximation is applicable and 3D interaction potential must be
541: averaged over the transverse degrees of freedom. Resulting 1D effective potential substantially
542: differs from Coulomb (see \cite{LozFilA} for the discussion of similar model). Second, corresponding
543: energy spectrum and wave functions of electron and hole relative motion also change in comparison
544: with the hydrogen-like states. We choose the wave function of the exciton ground state in Gaussian form.
545:
546: Excitonic wave packet can be represented as a Gaussian wave packet in center-of-mass coordinates:
547: \begin{equation}
548: \Psi(x_e,x_h,t=0)=\frac{e^{-r^2/(2\sigma)}}{(\pi\sigma)^{1/4}}\frac{e^{-(R-x_0)^2/(2S)+iRp_0}}{(\pi
549: S)^{1/4}},\label{InExcn}
550: \end{equation}
551: where $R=(m_ex_e+m_hx_h)/(m_e+m_h)$, $r=|x_e-x_h|$, $x_e$ and $x_h$ are electron and hole
552: coordinates, $x_0, p_0$ and $S$ are parameters; for them we used the following values:
553: $x_0=-10$, $p_0=3$, $S=2$ and $\sigma = 1$.
554:
555: External potential is assumed to be zero everywhere except the region of barrier; we use the
556: barrier of thickness equal to $5nm$, or $0.5$ in accepted units. For simplicity we set the barriers
557: for electron and hole to be the same and use both external and interaction potentials in quadratic
558: form, cut at some distance. Then external potential is given by
559: \begin{equation}
560: V_{ext}(x)=
561: \begin{cases}
562: C-Dx^2, & \text{if $|x| < \sqrt{\frac{C}{D}}$}, \\
563: 0, & \text{if $|x| \ge \sqrt{\frac{C}{D}}$}\label{VExcnExt}
564: \end{cases}
565: \end{equation}
566: $C$ is the height of the barrier, its width is $\sqrt{C/D}=0.5$.
567:
568: Interaction potential $V_{int}$ is also assumed to be quadratic:
569: \begin{equation}
570: V_{int}(r)=
571: \begin{cases}
572: Br^2-A, & \text{if $r < \sqrt{\frac{A}{B}}$}, \\
573: 0, & \text{if $r \ge \sqrt{\frac{A}{B}}$},\label{VExcnInt}
574: \end{cases}
575: \end{equation}
576: where $r=|x_e-x_h|$. Potential (\ref{VExcnInt}) can describe, {\it e.g.}, e-h interaction in
577: spatially indirect exciton, for example in coupled quantum wires with large interwire separation
578: \cite{LozWilAPA2000}. Initial wave function of relative motion, chosen to be Gaussian with
579: unity dispersion, is negligible within one percent accuracy at $r=3$. Thus we choose the radius of
580: electron-hole interaction to be $\sqrt{A/B}=3$.
581:
582: We assume that we deal with a quasi-1D exciton with binding energy $E_C = 1/8$. In fact, for an
583: exciton in quantum wire, the wave function, binding energy, etc., are essentially influenced by the
584: properties of quantum wire. We also neglect the possibility of electron and hole recombination at
585: the time scales studied.
586:
587: Here we consider just an example, therefore use relatively simple model. Still this model contains
588: the main features of exciton tunneling, such as the possibility of ionization, barrier and
589: interaction with realistic strength and size, the fact that composite particle is the bounded state
590: of two particles.
591:
592: For stationary state the binding energy is $-E_C = \int\Psi^*_{int}(r)H_{int}(r)\Psi_{int}(r)dr$,
593: where $\Psi_{int}(r)$ is the wave function of relative motion. Then, from Eq.(\ref{VExcnInt}) and
594: condition $\sqrt{A/B}=3$ we have $A\approx 18E_C/17$.
595:
596: We do not use the variables of relative and center-of-mass motion, for QT simulation it is easier to
597: deal with the initial conditions and evolution equation in coordinates of electron and hole $x_e$ and
598: $x_h$. Potentials (\ref{VExcnExt}) and (\ref{VExcnInt}) are quadratic, that makes tomographic
599: consideration of the problem easier (see Sec.~\ref{Tunneling}), discontinuity is neglected. On the
600: other hand, in coordinates $x_e$ and $x_h$ the evolution equations depend on trajectory distribution
601: (see Sec.~\ref{Method}), therefore the problem considered allows to employ all techniques, developed
602: for one degree of freedom, in this case of two degrees of freedom.
603:
604: \begin{figure}
605: \includegraphics[height=7cm]{fig8.eps}
606: \caption{\label{Fig8} Probability density distributions in coordinate space for electron
607: ($\rho_e(x)$) and hole ($\rho_h(x)$) at times $t=0$ and $t=10$. QT simulation (solid lines) is
608: compared with exact numerical solution (dashed lines). All values are in units $\hbar=m_e^*=E_C
609: =1$, $m_e^*$ is electron effective mass and $E_C$ is binding energy of the exciton. The height of
610: the barrier $C = 1$, width $\sqrt{C/D} = 0.5$.}
611: \end{figure}
612:
613: \begin{figure}
614: \includegraphics[height=7cm]{fig9.eps}
615: \caption{\label{Fig9} Probability density distributions in coordinate space for electron
616: ($\rho_e(x)$) and hole ($\rho_h(x)$) at times $t=20$ and $t=30$. QT simulation (solid lines) is
617: compared with exact numerical solution (dashed lines). The same units and barrier parameters
618: as in Fig.~\ref{Fig8} are used.}
619: \end{figure}
620:
621: \begin{figure}
622: \includegraphics[height=7cm]{fig10.eps}
623: \caption{\label{Fig10} Probability of exciton ionization $P_{Ion}$ due to electron and hole scattering
624: on the barrier in opposite directions {\it versus} barrier height $C$. Circles and
625: squares represent QT simulation and exact solution, respectively. Considered is the barrier of
626: thickness $0.5$. The units are the same as in Fig.~\ref{Fig8}.}
627: \end{figure}
628:
629: Results of exciton tunneling simulation are presented in Figs.~\ref{Fig8}-\ref{Fig10}. All
630: parameters are fixed (see above), except the barrier height $C$ in Fig.~\ref{Fig10}. In
631: Figs.~\ref{Fig8} and \ref{Fig9} we depict the evolution of probability density for electron and
632: hole, barrier height $C = 1$. Solid lines in Figs.~\ref{Fig8}~-~\ref{Fig9} and circles in
633: Fig.~\ref{Fig10} correspond to the simulation in quantum tomography approach, while dashed lines and
634: squares represent the exact numerical computation. Unlike in Figs.~\ref{Fig2}~-\ref{Fig6}, here we
635: show the results of several combined QT simulation runs, therefore corresponding lines are
636: relatively smooth. Coincidence of QT and exact computation results, initially very good, becomes
637: poorer with time (Figs.~\ref{Fig8} and \ref{Fig9}), but QT simulation reproduces main properties of
638: exciton tunneling: wave packets broadening with time and due to interaction with the barrier,
639: shrinking near the barrier, dividing into two parts (reflected and transmitted). Note that QT
640: results are quite close to exact ones even for long times ($t=30$), despite the fact that the motion
641: is unbounded (see Sec.~\ref{Method}). Integral values (in Fig.~\ref{Fig10}) obtained in QT approach
642: also agree with the exact results. Larger discrepancies correspond to higher barriers, probably due
643: to larger inaccuracies, introduced by neglecting the potential discontinuity in the case of
644: stronger interaction with the external potential.
645:
646: The electron and hole wave packets begin the motion from the point $x=-10$ and, shrinking near the
647: barrier, are partially reflected and transmitted. For the case presented in Figs.~\ref{Fig8}~-~
648: \ref{Fig9} about the half of wave packets is transmitted. Interesting is the question about the
649: ionization probability of exciton, induced by interaction with the barrier. If electron and hole are
650: scattered in different directions on the barrier, the distance between them can become quite large,
651: but, in principle, there is a possibility that exciton is not ionized after such scattering, because
652: one of the particles can be 'pulled' beyond the barrier, to the other particle, due to electron-hole
653: attraction. On the other hand, the electron-hole interaction is cut at the distance $\sqrt{A/B}$ in
654: our model. After the interaction with the barrier the wave packet divides into reflected and
655: transmitted parts moving in opposite directions. For the time large enough, these two parts are well
656: separated, the separation between them grows and the leakage through the barrier in both directions
657: is negligible. Denote the probability of ionization due to electron and hole scattering in different
658: directions as $P_{Ion}$. Then, the probability to find electron and hole in different directions in
659: respect to the barrier, with e-h distance being larger than $\sqrt{A/B}$, approaches $P_{Ion}$ in
660: the limit $t\rightarrow\infty$.
661:
662: The probability of ionization due to electron and hole scattering in different directions on the
663: barrier $P_{Ion}$ is presented in Fig.~\ref{Fig10}, depending on the barrier height $C$. For very
664: high and very low barriers $P_{Ion}$ must approach zero, because in former case both particles are
665: reflected and in the latter they both are transmitted. This trend is seen in Fig.~\ref{Fig10}, and
666: $P_{Ion}$ depending on $C$ is maximal (other parameters are fixed, see above) at $C\approx 1$. Note
667: that these features are obvious for curves representing both QT simulation (circles) and exact
668: computation (squares), and in general two curves are quite close to each other.
669:
670: \section{Conclusion}\label{Conclusion}
671: We have developed the new method of numerical simulation of quantum nonstationary processes
672: and applied it to the problem of tunneling of the wave packet through the potential barrier. The method is
673: based on tomographic representation of quantum mechanics. The quantum tomogram is used in a sense as the
674: distribution function for the ensemble of trajectories in space $X,\mu,\nu$, where $X = \mu q + \nu p$ is the
675: coordinate measured in rotated and scaled reference frame, $q,p$ are coordinate and momentum of the system,
676: respectively. The trajectories are governed by the equations, resembling the Hamilton
677: equations of motion, therefore, some analogue of molecular dynamics can be used. The
678: Gaussian approximation allows to avoid the direct calculation of quantum tomogram. Instead
679: of quantum tomogram, the parameters of the approximation are used in the equations of
680: motion. Those parameters can be obtained if one calculates the local moments of the
681: ensemble of the trajectories.
682:
683: The problem of nonstationary tunneling of the wave packet was considered. Our method gave the
684: results in agreement with those obtained by the method of "Wigner trajectories" and by exact
685: quantum computation.
686:
687: Of course, we made only the first step in the development of this simulation method, having
688: considered one-dimensional problem. But it is obvious that the generalization for the case of
689: several dimensions is straightforward. In this case the method possesses the additional advantage
690: of more rapid convergence due to the fact the quantum tomogram is nonnegative. This may help to
691: overcome the problem of sign for fermionic systems. In the next work we intend to
692: consider the many-body problem for fermionic and bosonic systems by means of the quantum
693: tomography method. The similar approach (in the framework of "Wigner trajectories") has been
694: already applied for the investigation of the tunneling of two identical particles \cite{LozFilA}.
695:
696:
697: \begin{acknowledgments}
698: Authors are grateful to INTAS, RFBR and Ministry of Science for
699: financial support. A. A. also acknowledge the financial help of
700: the Dynasty foundation and ICFPM. We thank V. I. Man'ko for
701: fruitful discussions.
702: \end{acknowledgments}
703: \appendix
704: \begin{references}
705: \bibitem{Lee} H.W. Lee, Phys.Rep. {\bf 259}, 147 (1995).
706: \bibitem{Cep} D.M. Ceperley, Rev.Mod.Phys. {\bf 67}, 279, (1995).
707: \bibitem{Husimi} K. Husimi, Proc.Phys.Math.Soc. Japan {\bf 22}, 264 (1940).
708: \bibitem{Glaub} R.J. Glauber, Phys.Rev.Lett. {\bf 10}, 84 (1963).
709: \bibitem{Sudar} E.C.G. Sudarshan, Phys.Rev.Lett. {\bf 10}, 277 (1963).
710: \bibitem{Man'ko} V.I. Man'ko, arXiv:quant-ph/9902079.
711: \bibitem{ManMan'ko} S. Mancini, V.I. Man'ko and P. Tombesi, Phys.Lett.A {\bf 213}, 1 (1996).
712: \bibitem{DodMan'ko} V.V. Dodonov and V.I. Man'ko, Phys.Lett.A {\bf 229}, 335 (1997).
713: \bibitem{Man'koMan'ko} V.I. Man'ko and O.V. Man'ko, JETP {\bf 85}, 430 (1997).
714: \bibitem{KV} K. Vogel and H. Risken, Phys.Rev.A {\bf 40}, 2847 (1989).
715: \bibitem{SM97} S. Mancini, V.I. Man'ko and P. Tombesi, Found.Phys. {\bf 27}, 801 (1997).
716: \bibitem{SM95} S. Mancini, V.I. Man'ko and P. Tombesi, J.Opt.B {\bf 7}, 615 (1995).
717: \bibitem{GMD} G.M. D'Ariano, S. Mancini, V.I. Man'ko and P. Tombesi, J.Opt.B {\bf 8}, 1017
718: (1996); S. Mancini, V.I. Man'ko and P. Tombesi, Europhys.Lett. {\bf 37}, 79 (1997).
719: \bibitem{Wigner} E. Wigner, Phys.Rev. {\bf 40}, 749 (1932).
720: \bibitem{DM} A. Donoso and C.C. Martens, Phys.Rev.Lett., {\bf 87}, 223202 (2001).
721: \bibitem{LF} Yu.E. Lozovik and A.V. Filinov, Zh.Eksp.Teor.Fiz. {\bf 115}, 5, 1872 (1999) [JETP, {\bf 88}, 5, 1026 (1999)].
722: \bibitem{SMJ} S. Mancini, V.I. Man'ko and P. Tombesi, J.Mod.Opt. {\bf 44}, 2281 (1997).
723: \bibitem{VIM} V.I. Man'ko, L. Rosa and P. Vitale, Phys.Rev.A {\bf 57}, 3291 (1998).
724: \bibitem{OM} O. Man'ko and V.I. Man'ko, J.Russ.Laser Res. {\bf 18}, 407 (1997).
725: \bibitem{SBMJChemPhys1993} R. Sala, S. Brouard and J.G. Muga, J. Chem. Phys. {\bf 99}, 2708 (1993).
726: \bibitem{BL1982} M. B\"uttiker and R. Landauer, Phys.Rev.Lett. {\bf 49}, 1739 (1982).
727: \bibitem{LM} R. Landauer and Th. Martin, Sol.St.Com., {\bf 84}, 115 (1992).
728: \bibitem{SokConnor} D. Sokolovski and J.N.L. Connor, Phys.Rev.A {\bf 47}, 4677 (1993).
729: \bibitem{Baz'} A.J. Baz', Sov.J.Nucl.Phys. {\bf 5}, 161 (1967).
730: \bibitem{Ryb} V.F. Rybachenko, Sov.J.Nucl.Phys. {\bf 5}, 635 (1967).
731: \bibitem{But} M. B\"uttiker, Phys.Rev.B {\bf 27}, 6178 (1983).
732: \bibitem{HS} E.H. Hauge and J.A. Stovneng, Rev.Mod.Phys. {\bf 61}, 917 (1989).
733: \bibitem{BL} M. B\"uttiker and R. Landauer, IBM J.Res.Dev. {\bf 30}, 451 (1986).
734: \bibitem{AB} Y. Aharonov and D. Bohm, Phys.Rev. {\bf 122}, 1649 (1961).
735: \bibitem{Wlod} J.J. W\l odarz, Phys.Rev.A {\bf 65}, 044103 (2002).
736: \bibitem{BEMuga} A.D. Baute, I.L. Egusquiza and J.G. Muga, Phys.Rev.A {\bf 65}, 032114 (2002).
737: \bibitem{LLBR} S. Longhi, P. Laporta, M. Belmonte and E. Recami, Phys.Rev.E {\bf 65}, 046610 (2002).
738: \bibitem{DelgMuga} V. Delgado and J.G. Muga, Phys.Rev.A, {\bf 56}, 3425 (1997).
739: \bibitem{MLPhysRep2000} J.G. Muga and C.R. Leavens, Phys. Rep. {\bf 338}, 353 (2000).
740: \bibitem{LozFilA} Yu.E. Lozovik, A.V. Filinov and A.S. Arkhipov, Phys.Rev.E., {\bf 67}, 026707 (2003).
741: \bibitem{LozWilAPA2000} Yu.E. Lozovik and M. Willander, Appl. Phys. A, {\bf 71}, 379 (2000).
742: \end{references}
743:
744: \end{document}
745: