1: \documentclass[12pt]{article}
2: \usepackage{graphicx}
3: \textwidth 6.5 true in
4: \textheight 8.5 true in
5: \oddsidemargin 0.0 true in
6: \evensidemargin 0.0 true in
7: \topmargin -0.5 true in
8:
9: \begin{document}
10:
11: \title{Time evolution of tunneling and decoherence: \\
12: soluble model.}
13:
14: \author{A. N. Salgueiro$^{(1)}$, A.F.R. de Toledo Piza$^{(1)}$ \\
15: J. G. Peixoto de Faria$^{(2)}$ and M.C. Nemes$^{(2)}$}
16:
17: \maketitle
18:
19: \begin{center}
20:
21: {$^{(1)}$ Departamento de F\'{\i}sica--Matem\'atica,
22: Instituto de F\'\i sica,
23: Universidade de S\~ao Paulo, \\ C.P. 66318,
24: CEP 05315-970 S\~ao Paulo, S.P., Brazil}
25:
26: {$^{(2)}$ Departamento de F\'\i sica, ICEX,
27: Universidade Federal de Minas Gerais,\\ C.P. 702, CEP
28: 30161-970 Belo Horizonte, M.G. Brazil}
29:
30: \end{center}
31:
32: \begin{abstract}
33:
34: Decoherence effects associated to the damping of a tunneling two-level
35: system are shown to dominate the tunneling probability at short times
36: in strong coupling regimes in the context of a soluble model. A
37: general decomposition of tunneling rates in dissipative and unitary
38: parts is implemented. Master equation treatments fail to describe the
39: model system correctly when more than a single relaxation time is
40: involved.
41:
42: \end{abstract}
43:
44: %\newpage
45:
46: \section{Introduction}
47:
48: Motion in tunneling phenomena can be seen quite generally as resulting
49: from dispersive effects in the time evolution of spatially localized
50: quantum states which are not energy eigenstates. Since, from this
51: point of view, quantum interference plays a central role in these
52: phenomena, one may ask how are they affected by quantum decoherence
53: processes which occur as a result of coupling to other degrees of
54: freedom. Effects of additional degrees of freedom in tunneling
55: processes have been studied now for a long time and in a variety of
56: contexts \cite{gen}, but emphasis falls usually on the question of the
57: observable changes on inclusive properties such as tunneling
58: probabilities. However, these questions are often addressed in
59: semi-classical terms, including generalizations of the WKB approach
60: \cite{BNW}, which tend to becloud the dispersive character of the
61: quantum dynamics.
62:
63: In order to analyze the dispersive dynamics of tunneling we consider in
64: this paper the simple but not unrealistic case of two-level
65: approximations to symmetric, bound, bi-stable systems whose energy
66: spectrum is characterized by a structure of doublets. Each member of
67: one doublet consists of symmetric/antisymmetric superpositions of
68: localized states which may be seen as coupled by effective tunneling
69: amplitudes. In the two-level approximation this is conveniently
70: expressed in terms of a spin-1/2 algebra. Dissipation and decoherence
71: phenomena can be introduced in this description by coupling the
72: two-level system to another system, possibly with a continuous energy
73: spectrum. A collection of harmonic oscillators is frequently used in
74: this connection, leading to the ``spin-boson'' model analyzed by
75: Leggett and collaborators \cite{legg}. There the effects of
76: dissipation on the tunneling rates are studied independently of
77: explicit control of properties of the tunneling state relating to
78: decoherence processes. A simpler, soluble model, which allows for a
79: control of this sort, results when the two-level system is suitably
80: coupled to a generic additional system with continuous spectrum, as
81: described in detail in section \ref{main} below. In this model the
82: coupled dynamics can be reduced to the problem of the spreading of a
83: product ``doorway state'' due to its coupling to a continuous
84: background also of product states, which allows for a treatment in
85: terms of the techniques developed long ago by Fano \cite{fano} in the
86: context of atomic scattering theory. While this schematic model is
87: less ``realistic'' from the point of view of energy-loss mechanisms,
88: it retains the overall dynamical features of the spin-boson model
89: relating to coherence-loss, which is our main concern here. The model
90: allows for the calculation of the reduced density associated with the
91: two-level system at all times. Model dependences are indicated by
92: differences in the results obtained in the two cases. We find, in
93: particular, that transition rates out of a localized initial state can
94: be substantially enhanced by decoherence in strong coupling regimes.
95:
96: The use of Master equations for the description of damped two-level
97: systems is well known \cite{livopt}. The usual derivation of these
98: equations \cite{meq} assume weak-coupling to an harmonic oscillator
99: reservoir and high temperature, so that their use in strong coupling
100: and zero temperature situations is not warranted. Comparison of the
101: exact model results with solutions of these Master equations agree for
102: particular forms of the coupling independently of the overall coupling
103: strength and at zero temperature. In general, however, the
104: perturbative ingredients of the Master equation derivations appear as
105: limiting factors.
106:
107: \section{Dissipative and unitary tunneling rates.}\label{Pt}
108:
109: The quantum state of a two-level subsystem can be generally described
110: in terms of a (time dependent) hermitian reduced density operator of
111: unit trace $\rho(t)$. The simple two-level effective Hamiltonian
112:
113: \[
114: H_\sigma=\frac{\epsilon}{2}\,(\sigma_3+\hat{1}_\sigma),
115: \]
116:
117: \noindent where $\hat{1}_\sigma$ stands for the unit $2\times 2$
118: matrix, can be understood as describing the tunneling between the
119: ``localized'' (``left'' and ``right'') states
120:
121: \[
122: |l\rangle\equiv\frac{|+\rangle+|-\rangle}{\sqrt{2}}\hspace{2cm}
123: {\rm and}\hspace{2cm}|r\rangle\equiv\frac{|+\rangle-|-\rangle}
124: {\sqrt{2}}
125: \]
126:
127: \noindent where $|\pm\rangle$ are the eigenvectors of the diagonal
128: Pauli matrix $\sigma_3$, and hence also energy eigenstates. The time
129: dependent density matrix corresponding to the initial condition
130:
131: \begin{equation}
132: \label{incond}
133: \rho(t=0)=|l\rangle\langle l|
134: \end{equation}
135:
136: \noindent is given in the $|\pm\rangle$ basis as (with $\hbar=1$)
137:
138: \begin{equation}
139: \label{pure}
140: \rho(t)=\frac{1}{2}\left(\begin{array}{cc} 1 & e^{i\epsilon t} \\
141: e^{-i\epsilon t} & 1 \end{array}\right)=\rho^2(t).
142: \end{equation}
143:
144: \noindent As is well known, the property of idempotency is equivalent
145: to the purity of the quantum state described by $\rho(t)$. A tunneling
146: probability $P(t)$ can be defined as the probability of finding this
147: system in the complementary localized state $|r\rangle$ at time $t$,
148: and is readily calculated as
149:
150: \begin{equation}
151: \label{ptun}
152: P(t)={\rm Tr}\left[|r\rangle\langle
153: r|\rho(t)\right]=\sin^2\frac{\epsilon t}{2}.
154: \end{equation}
155:
156: When the two-level subsystem is coupled to additional degrees of
157: freedom its state at time $t$ is still described by a reduced density
158: matrix which is non-negative, hermitian and of unit trace, but in
159: general {\it not} idempotent. In fact, the quantity $\delta(t)=1-{\rm
160: Tr}[\rho(t)^2]$ (referred to as the {\it linear entropy} or the {\it
161: idempotency defect}) is often used as a measure of the decoherence
162: suffered by the two level subsystem through its coupling to the other
163: degrees of freedom.
164:
165: A most convenient way of representing such a reduced density matrix
166: makes use of its eigenvalues and eigenvectors
167:
168: \[
169: \rho(t)|t,k\rangle=p_k(t)|t,k\rangle,\hspace{.5cm}k=1,2;\hspace{1cm}
170: \langle t,k|t,k'\rangle=\delta_{kk'},\hspace{.5cm}0\leq p_k(t)\leq 1
171: \]
172:
173: \noindent in terms of which one can write
174:
175: \begin{equation}
176: \label{natorb}
177: \rho(t)=\sum_k |t,k\rangle p_k(t)\langle t,k|.
178: \end{equation}
179:
180: \noindent It should be noted that in general both the eigenvectors and
181: the eigenvalues in (\ref{natorb}) are time-dependent. While the
182: eigenvectors evolve unitarily (as they define a time-dependent
183: orthonormal basis in the quantum phase-space of the two-level
184: subsystem) the time-dependence of the eigenvalues $p_k(t)$ reveal the
185: change in time of the coherence properties of the state of the
186: two-level subsystem. The idempotency defect is in fact given in terms
187: of the eigenvalues $p_k(t)$ as
188:
189: \begin{equation}
190: \label{delta}
191: \delta=1-\sum_k p_k(t)^2=2p_1(t)(1-p_1(t))
192: \end{equation}
193:
194: \noindent where use has been made of the unit trace property of
195: $\rho(t)$. In the case of eq. (\ref{pure}), it can be easily checked
196: that $p_1=1$ and $p_2=0$ at all times, so that $\delta\equiv
197: 0$. Assuming that the reduced density is that which evolves from the
198: initial condition (\ref{incond}), the tunneling probability is now
199: given by
200:
201: \[
202: P(t)=\sum_k p_k(t)|\langle r|t,k\rangle|^2.
203: \]
204:
205: \noindent In view of what has been said concerning the time-dependence
206: of the ingredients of this expression, it is immediately clear that
207: the tunneling {\it rate}
208:
209: \[
210: R(t)\equiv\dot{P}(t)=\sum_k\dot{p}_k(t)|\langle r|t,k\rangle|^2+
211: \sum_k p_k(t)\frac{d}{dt}|\langle r|t,k\rangle|^2
212: \]
213:
214: \noindent splits into two contributions which can be unambiguously
215: ascribed to changes in the coherence properties of the state of the
216: subsystem and to its unitary evolution, respectively:
217:
218: \begin{equation}
219: \label{Rd}
220: R_d(t)\equiv\sum_k\dot{p}_k(t)|\langle r|t,k\rangle|^2=\left(2
221: |\langle r|t,1\rangle|^2-1\right)\dot{p}_1(t)
222: \end{equation}
223:
224: \noindent and
225:
226: \begin{equation}
227: \label{Ru}
228: R_u(t)\equiv\sum_k p_k(t)\frac{d}{dt}|\langle r|t,k\rangle|^2=
229: \left(2p_1(t)-1\right)\frac{d}{dt}|\langle r|t,1\rangle|^2.
230: \end{equation}
231:
232: \noindent The last forms of $R_u$ and $R_d$ make use of the unit trace
233: of $\rho$ and also of the normalization of $|r\rangle$. In the case of
234: the purely unitary time evolution under the simple two-level
235: Hamiltonian $H_\sigma$ one has of course $R_d(t)\equiv 0$ and
236: $R_u(t)=R(t)=(\epsilon/2)\sin\,\epsilon t$. This decomposition of
237: transition rates in dissipative and unitary parts can be simply
238: extended to cases involving reduced densities of larger
239: dimensionality, as indicated in Appendix A.
240:
241: \section{Two-level tunneling with coupling to a
242: continuum.}\label{main}
243:
244: A simple model allowing for decoherence effects in the two-level
245: tunneling processes is one in which the two-level subsystem described
246: by the Hamiltonian $H_\sigma$ is coupled to an additional
247: ``nondescript'' set of degrees of freedom (which will be identified by
248: means of a subscript $b$) to which is associated an energy spectrum
249: containing one discrete state represented by a normalized state vector
250: $|0_b\rangle$ in addition to a continuum of states $|\eta\rangle$,
251: normalized in energy as $\langle\eta|\eta'\rangle=\delta(\eta-\eta')$.
252: The model is characterized further by the Hamiltonian
253:
254: \begin{eqnarray}
255: \label{Hmod}
256: H&=&\frac{\epsilon}{2}(\sigma_3+\hat{1}_\sigma)\otimes{\bf{\hat{1}}}_b
257: +\left(|0_b\rangle e_0\langle 0_b|+\int_{\eta_0}^{\bar{\eta}}d\eta\,
258: |\eta\rangle\eta\langle\eta|\right)\otimes\hat{1}_\sigma \nonumber \\
259: &&+\frac{1}{2}\int_{\eta_0}^{\bar{\eta}}d\eta\,\left[g(\eta)|\eta
260: \rangle\sigma_-\langle 0_b|+g^*(\eta)|0_b\rangle\sigma_+\langle\eta|
261: \right]\\ &&+\frac{1}{2}\int_{\eta_0}^{\bar{\eta}}d\eta\,\left[g'(
262: \eta)|\eta\rangle\sigma_+\langle 0_b|+g'^*(\eta)|0_b\rangle\sigma_-
263: \langle\eta|\right]\nonumber
264: \end{eqnarray}
265:
266: \noindent where $g(\eta)$ and $g'(\eta)$ are coupling matrix elements
267: $\langle -\eta|\hat{g}|+0_b\rangle$ and $\langle+\eta|\hat{g}'|-0_b
268: \rangle$ respectively, possibly dependent on $\eta$. In the special
269: case $g(\eta)=g'(\eta)$ the coupling terms reduce to
270:
271: \[
272: \sigma_x\int_{\eta_0}^{\bar{\eta}}d\eta\;\left[|\eta\rangle g(\eta)
273: \langle 0_b|+|0_b\rangle g^*(\eta)\langle\eta|\right]
274: \]
275:
276: \noindent which resembles in its structure the coupling term of the
277: spin-boson model\cite{legg}.
278:
279: A basis in the quantum phase-space of the composite system described
280: by equation (\ref{Hmod}) can be constructed as $\{|s0_b\rangle,\,\{|
281: s\eta\rangle\}\},\;\,s=+,-,\;\,\eta_0\leq\eta \leq\bar{\eta}$, and the
282: adopted structure of the coupling terms shows that the subspaces
283: spanned by the two sets of basis vectors $\{|+0_b\rangle,\,\{|-\eta
284: \rangle\}\}$ and $\{|-0_b\rangle,\,\{|+ \eta\rangle\}\}$ are closed
285: under the action of $H$. Different choices for the model parameters
286: $\epsilon$, $e_0$, the span of the continuum band $\eta_0$ to
287: $\bar{\eta}$ and the $\eta$-dependence of the coupling functions $g$
288: and $g'$ allow for a considerable variety of dynamical
289: regimes. Closest to the situation described by the spin-boson model is
290: that in which $e_0=\eta_0=0$ with $g(\eta)=g^*(\eta)=g'(\eta)$. Some
291: aspects of the relationship of these two models are briefly discussed
292: in Appendix B. The particular version of the model which will be used
293: in the following sections differs from this, however. A manifest
294: feature of this version is the assumption that one has
295:
296: \[
297: \eta_0\ll e_0\hspace{1cm}{\rm and}\hspace{1cm}
298: e_0+\epsilon\ll\bar{\eta}
299: \]
300:
301: \noindent so that the two product states $|\pm0_b\rangle$, with
302: eigen-energies $e_0$ and $e_0+\epsilon$ fall well within a
303: sufficiently wide continuum band of product states $|\pm\eta\rangle$.
304: This feature will in fact allow for the continuum spreading of the
305: states $|+0_b\rangle$ and $|-0_b\rangle$ through their coupling with
306: states $|-\eta\rangle$ and $|+\eta\rangle$ respectively with spreading
307: widths appreciably larger than the doublet splitting $\epsilon$, a
308: situation which will be seen to characterize strong damping
309: regimes. Furthermore, since these two spreadings are dynamically
310: independent, as they take place in different invariant subspaces of
311: $H$, there will be in general two distinct relaxation times
312: associated with the damped two-level subsystem.
313:
314: General state vectors in each of the invariant subspaces of $H$ can be
315: written as
316:
317: \begin{eqnarray*}
318: |\psi_a\rangle&=&a_0|+0_b\rangle+\int_{\eta_0}^{\bar{\eta}}d\eta\,
319: A(\eta)|-\eta\rangle \\
320: |\psi_b\rangle&=&b_0|-0_b\rangle+\int_{\eta_0}^{\bar{\eta}}d\eta\,
321: B(\eta)|+\eta\rangle.
322: \end{eqnarray*}
323:
324: \noindent In terms of these expansions, the eigenvalue problem for $H$
325:
326: \[
327: (E-H)|\psi_{a,b}^{(E)}\rangle=0
328: \]
329:
330:
331: \noindent can be reduced to two independent pairs of coupled equations
332: with similar structure for the expansion coefficients
333:
334: \begin{eqnarray}
335: \label{forw}
336: (E-\epsilon-e_0)a_0^{(E)}&=&\int_{\eta_0}^{\bar{\eta}}d\eta\,
337: g^*(\eta)A^{(E)}(\eta), \nonumber \\
338: (E-\eta)A^{(E)}(\eta)&=&g(\eta)a_0^{(E)}
339: \end{eqnarray}
340:
341: \noindent and
342:
343: \begin{eqnarray}
344: \label{backw}
345: (E-e_0)b_0^{(E)}&=&\int_{\eta_0}^{\bar{\eta}}d\eta\,
346: g'^*(\eta)B^{(E)}(\eta), \nonumber \\
347: (E-\epsilon-\eta)B^{(E)}(\eta)&=&g'(\eta)b_0^{(E)}.
348: \end{eqnarray}
349:
350: \noindent The solutions of these coupled equations has been given long
351: ago by Fano\cite{fano} and van Kampen\cite{vankampen}. In the case of
352: eqs. (\ref{forw}) one has
353:
354: \begin{equation}
355: \label{a0ls}
356: |a_0^{(E)}|^2=\frac{|g(E)|^2}{[E-\epsilon-e_0-F(E)]^2+\pi^2|g(E)|^4}
357: \end{equation}
358:
359: \noindent with $F(E)$ given by the principal value integral
360:
361: \begin{equation}
362: \label{shift}
363: F(E)={\cal{P}}\int_{\eta_0}^{\bar{\eta}}d\eta\,\frac{|g(\eta)|^2}
364: {E-\eta}.
365: \end{equation}
366:
367: \noindent The continuum coefficients are
368:
369: \begin{equation}
370: \label{Afroma0}
371: A^{(E)}(\eta)=\left[\frac{{\cal{P}}}{E-\eta}+z(E)\delta(E-\eta)\right]
372: g(\eta)a_0^{(E)}
373: \end{equation}
374:
375: \noindent with
376:
377: \[
378: z(E)=\frac{E-\epsilon-e_0-F(E)}{|g(E)|^2}.
379: \]
380:
381: \noindent The value of the amplitude $a_0^{(E)}$ involves a phasing
382: convention which is subsequently carried over to the $A^{(E)}(\eta)$
383: through equation (\ref{Afroma0}). The solution of eqs. (\ref{backw})
384: is of course completely analogous and does not have to be given
385: explicitly here.
386:
387: Strictly speaking, when the integration limits of the continuum
388: variable $\eta$ are {\it finite}, the spectrum of $H$ may include
389: discrete states outside the continuum range, subject to the
390: $\eta$-dependence of the coupling parameters $g(\eta)$. Whenever
391: needed, the discrete eigenvalues of $H$ can be obtained (e.g. in the
392: case of eqs. (\ref{forw})) as solutions $E_d<\eta_0$ or
393: $E_d>\bar{\eta}$ of the dispersion equation
394:
395: \[
396: E-\epsilon-e_0=\int_{\eta_0}^{\bar{\eta}}d\eta\,\frac{|g(\eta)|^2}
397: {E-\eta}.
398: \]
399:
400: \noindent In this case eq. (\ref{Afroma0}) is replaced by
401:
402: \[
403: A^{(E_d)}(\eta)=\frac{g(\eta)}{E_d-\eta}
404: \]
405:
406: \noindent and $a_0^{(E)}$ is determined from the normalization
407: condition (up to an arbitrary phase factor) as
408:
409: \[
410: |a_0^{(E_d)}|^2=\left(1-\left(\frac{dF(E)}{dE}\right)_{E=E_d}
411: \right)^{-1}.
412: \]
413:
414: \noindent The low-energy discrete solutions are specially relevant for
415: comparison of this model with other models of the damping mechanism,
416: such as the many-oscillators model (see Appendix B).
417:
418: \subsection{Time evolution of a localized initial
419: condition.}\label{tev}
420:
421: Using the stationary states of $H$ as determined above one can write
422: the time evolution of the localized initial state
423:
424: \begin{equation}
425: \label{tunic}
426: |t=0\rangle=\frac{|+\rangle+|-\rangle}{\sqrt{2}}\otimes|0_b\rangle=
427: |l\rangle\otimes|0_b\rangle.
428: \end{equation}
429:
430: \noindent Note that in the absence of coupling, $g(\eta)=g'(\eta)=0$,
431: the time evolution reduces to that found in section \ref{Pt}. In
432: general, the component states $|+0_b\rangle$ and $|-0_b\rangle$ can be
433: written in terms of the stationary states determined from the solutions
434: of eqs. (\ref{forw}) and (\ref{backw}), so that
435:
436: \begin{eqnarray*}
437: |t\rangle=e^{-iHt}|t=0\rangle&=&\frac{e^{-iHt}}{\sqrt{2}}\left(
438: \sum_{E_d}a_0^{(E_d)}|\psi_a^{(E_d)}\rangle+\int_{\eta_0}^{\bar{\eta}
439: }dE\;a_0^{(E)*}|\psi_a^{(E)}\rangle \right. \\ &+& \left.
440: \sum_{E'_d}b_0^{(E'_d)}|\psi_b^{(E'_d)}\rangle+\int_{\eta_0}^{\bar
441: {\eta}}dE\;b_0^{(E)*}|\psi_a^{(E)}\rangle\right)
442: \end{eqnarray*}
443:
444: \noindent When the continuum range extends over a sufficiently broad
445: energy interval on both sides of $e_0$ the discrete state amplitudes
446: $a_0^{(E_d)}$ and $ b_0^{(E'_d)}$ become very small and can be
447: neglected. In what follows we assume this to be the case in order to
448: avoid unessential complications. By re-expressing the stationary
449: states in terms of the factorized bases, forming the total density
450: $|t\rangle\langle t|$ and taking its trace over the $b$-states one
451: obtains for the reduced density of the two level system at time $t$
452:
453: \[
454: \rho(t)=\left(\begin{array}{cc} \rho_{++}(t) & \rho_{+-}(t) \\ &
455: \\ \rho_{+-}^*(t) & 1-\rho_{++}(t) \end{array}\right)
456: \]
457:
458: \noindent with
459:
460: \begin{equation}
461: \label{rho++}
462: \rho_{++}(t)=\frac{1}{2}\left(1+\left|\int_{\eta_0}^{\bar{\eta}}dE\,
463: |a_0^{(E)}|^2e^{-iEt}\right|^2-\left|\int_{\eta_0}^{\bar{\eta}}dE\,
464: |b_0^{(E)}|^2e^{-iEt}\right|^2\right)
465: \end{equation}
466:
467: \noindent and
468:
469: \begin{eqnarray}
470: \label{rho+-}
471: \rho_{+-}(t)&=&\frac{1}{2}\left(\int_{\eta_0}^{\bar{\eta}}dE\,
472: |a_0^{(E)}|^2e^{-iEt}\int_{\eta_0}^{\bar{\eta}}dE'\,|b_0^{(E')}|^2
473: e^{-iE't} \right. \nonumber \\ &&+ \left.
474: \int_{\eta_0}^{\bar{\eta}}dE\int_{\eta_0}^{\bar{\eta}}dE'
475: b_0^{(E)*}a_0^{(E')}e^{-i(E-E')t}\int_{\eta_0}^{\bar{\eta}}d\eta\,
476: B^{(E)}(\eta)A^{(E')*}(\eta)\right).
477: \end{eqnarray}
478:
479: \noindent The last term of eq. (\ref{rho+-}) is the only one in which
480: the continuum amplitudes $A^{(E')}(\eta)$ and $B^{(E)}(\eta)$
481: intervene explicitly. This term describes a re-correlation of
482: components of the state of the two-level system in the {\it single}
483: continuum subspace of system $b$, to which they couple through
484: $g(\eta)$ and $g'(\eta)$. In fact, a simple variant of the model that
485: avoids this re-correlation process can be devised by adding a second
486: continuum $\{|\zeta\rangle\}$ orthogonal to the first, i.e. $\langle
487: \eta|\zeta\rangle\equiv 0$, to which the component $|-0_b\rangle$ is
488: coupled through $g'(\zeta)$. It is easy to see that the resulting
489: expressions for $\rho_{++}$ and $\rho_{+-}$ are in this case identical
490: to eqs. (\ref{rho++}) and (\ref{rho+-}) except for the absence of the
491: last term in the latter.
492:
493: The tunneling probability at time $t$ of the two-level subsystem, from
494: the initial state $|l\rangle$ to the complementarily localized
495: state $|r\rangle$, defined in terms of the reduced density matrix
496: $\rho(t)$ as in eq. (\ref{ptun}), is generally given as
497:
498: \begin{equation}
499: \label{ptgen}
500: P(t)={\rm Tr}\left[|r\rangle\langle r|\rho(t)\right]=\frac{1}{2}
501: \left(1-2\,{\rm Re}\,\rho_{+-}\right)
502: \end{equation}
503:
504: \noindent and the idempotency defect (\ref{delta}) of the reduced
505: density is
506:
507: \begin{equation}
508: \label{delgen}
509: \delta(t)=2\left[\rho_{++}(t)\left(1-\rho_{++}(t)\right)-
510: |\rho_{+-}(t)|^2\right]=2\det\rho(t).
511: \end{equation}
512:
513: \subsection{Special case: single relaxation time.}\label{onel}
514:
515: An interesting particular case is that in which $g'(\eta)\equiv 0$, so
516: that $|-0_b\rangle$ is a stationary state of $H$ while $|+0_b\rangle$
517: is spread through its coupling to the $|-\eta\rangle$ continuum. In
518: this case the reduced density matrix elements are simply given in
519: terms of the spectral distribution $|a_0^{(E)}|^2$, eq. (\ref{a0ls}),
520: of the state $|+0_b\rangle$ as
521:
522: \begin{equation}
523: \label{rho++gp0}
524: \rho_{++}(t)\rightarrow\frac{1}{2}\left|\int_{\eta_0}^{\bar{\eta}}dE\,
525: |a_0^{(E)}|^2e^{-iEt}\right|^2,\hspace{1cm}g'(\eta)\equiv 0
526: \end{equation}
527:
528: \noindent and
529:
530: \begin{equation}
531: \label{rho+-gp0}
532: \rho_{+-}(t)\rightarrow\frac{1}{2}\int_{\eta_0}^{\bar{\eta}}dE\,
533: |a_0^{(E)}|^2e^{-i(E-e_0)t},\hspace{1cm}g'(\eta)\equiv 0.
534: \end{equation}
535:
536: \noindent These expressions can be immediately evaluated in closed
537: form when the matrix elements $g(\eta)$ are independent of $\eta$, and
538: the energy shift $F(E)$, eq. (\ref{shift}) is slowly energy dependent
539: over an interval of the order of $\pi|g|^2$ in the neighborhood of
540: $E=\epsilon+e_0$. This will in fact be the case when the continuum
541: range extends over sufficiently broad intervals both above and below
542: this energy. Equation (\ref{a0ls}) is then well approximated by the
543: Breit-Wigner form
544:
545: \[
546: |a_0^{(E)}|^2\rightarrow\frac{|g|^2}{(E-e_R)^2+\pi^2|g|^4}\equiv
547: \frac{1}{2\pi}\;\frac{\Gamma}{(E-e_R)^2-\frac{\Gamma^2}{4}}
548: \]
549:
550: \noindent where $e_R-\epsilon-e_0-F(e_R)=0$ and $\Gamma=2\pi|g|^2$ is
551: the usual width parameter of the Breit-Wigner distribution, given in
552: Golden Rule form. Note that for large $\bar{\eta}-e_0\simeq
553: e_0-\eta_0$ one has $F(e_R)\simeq F(e_0)\simeq 0$ so that
554: $e_R-e_0\simeq\epsilon$. Extending the integration limits to
555: $\pm\infty$ one obtains by simple contour integrations
556:
557: \begin{equation}
558: \label{ols}
559: \rho_{++}(t)=\frac{1}{2}e^{-\Gamma t}\hspace{1.5cm}{\rm and}
560: \hspace{1.5cm}\rho_{+-}(t)=\frac{1}{2}e^{-i(e_R-e_0)t-
561: \frac{\Gamma t}{2}}
562: \end{equation}
563:
564: \noindent so that
565:
566: \begin{equation}
567: \label{pt+}
568: P(t)=\frac{1}{2}\left(1-e^{-\frac{\Gamma t}{2}}
569: \cos(e_R-e_0)t\right).
570: \end{equation}
571:
572: \noindent It should be kept in mind that the model assumptions allow
573: $\Gamma > \epsilon$, in which case the initial time-dependence of
574: $P(t)$ is dominated by the decaying exponential factor. The
575: decoherence process undergone by the reduced density as measured by
576: the idempotency defect (\ref{delta}) is given by
577:
578: \begin{equation}
579: \label{del+}
580: \delta(t)=\frac{1}{2}e^{-\Gamma t}\left(1-e^{-\Gamma t}\right)
581: \end{equation}
582:
583: \noindent so that the tunneling and the decoherence {\it rates}
584:
585: \[
586: \dot{P}(t)=\frac{e^{-\frac{\Gamma t}{2}}}{2}\left(\frac{\Gamma}{2}
587: \cos(e_R-e_0)t+(e_R-e_0)\sin(e_R-e_0)t\right)
588: \]
589:
590: \noindent and
591:
592: \[
593: \dot{\delta}(t)=\frac{\Gamma}{2}e^{-\Gamma t}\left(2e^{-\Gamma t}-1
594: \right)
595: \]
596:
597: \noindent are both proportional to the damping width $\Gamma$, related
598: to the single relaxation time $\tau_d$ defined as $\tau_d=1/\Gamma$.
599: Moreover, $\dot{P}(t)$ contains the period of unitary evolution
600: $\tau_u\equiv 2\pi/(e_R-e_0)\simeq 2\pi/\epsilon$ as an additional,
601: independent time scale. In the strong coupling regime
602: $\tau_d\ll\tau_u$ eq. (\ref{pt+}) shows dominance of coherence loss
603: effects in $P(t)$ for times $t<\tau_d$ leading eventually to
604: saturation at $P=1/2$. The sorting of dissipative and unitary
605: contributions to the tunneling rate $\dot{P}(t)$ is achieved through
606: the calculation of the rates $R_d(t)$ and $R_u(t)$ introduced in
607: section \ref{Pt}. As shown in eqs. (\ref{Rd}) and (\ref{Ru}), this
608: involves the time dependence of the eigenvalues and eigenvectors of
609: the reduced density. The relevant ingredients can be expressed in a
610: straightforward way in terms of the reduced density matrix elements
611: $\rho_{++}(t)$ and $\rho_{+-}(t)$. Since the resulting expressions
612: are lengthy and not particularly illuminating they will not be given
613: here. Numerical results illustrating these features, obtained using
614: the matrix elements given in eq. (\ref{ols}), are shown in figs. 1 and
615: 2 respectively for the weak and strong coupling regimes. The figures
616: show $P(t)$, $\delta(t)$ and the tunneling rates $\dot{P}(t)$,
617: $R_d(t)$ and $R_u(t)$. The correlation of $\delta(t)$ with $R_d(t)$
618: and the dominance of the latter over $R_u(t)$ in the strong coupling
619: regime are clearly seen.
620:
621: \subsection{General case: two relaxation times.}\label{twol}
622:
623: When both $g(\eta)$ and $g'(\eta)$ are different from zero the matrix
624: elements of the reduced densities are given by the complete
625: expressions (\ref{rho++}) and (\ref{rho+-}). Using the tunneling
626: initial condition of subsection \ref{tev} and under the assumptions
627: used in subsection \ref{onel} the first of these can be evaluated to
628: yield
629:
630: \begin{equation}
631: \label{2lrho++}
632: \rho_{++}(t)=\frac{1}{2}\left(1+e^{-\Gamma t}-e^{-\Gamma' t}\right)
633: \end{equation}
634:
635: \noindent where $\Gamma'$, defined as the width parameter $\Gamma'
636: =2\pi|g'|^2$ ($g'$ having been assumed to be independent of $\eta$) of
637: the distribution
638:
639: \[
640: |b_0^{(E)}|^2=\frac{1}{2\pi}\;\frac{\Gamma'}{(E-e'_R)^2+
641: \frac{\Gamma'^2}{4}},
642: \]
643:
644: \noindent introduces an additional time scale in the evolution of the
645: composite system. As for $\rho_{+-}$ one gets, from eq. (\ref{rho+-}),
646: using eq. (\ref{Afroma0}) and the methods of ref. \cite{fano}
647:
648: \begin{equation}
649: \label{2lrho+-}
650: \rho_{+-}(t)=\frac{1}{2}\left[e^{-i(e_R-e'_R)t-\frac{\Gamma+\Gamma'}
651: {2}t}+\frac{2\sqrt{\Gamma\Gamma'}}{\Gamma+\Gamma'}e^{i(e_R-e'_R)t}
652: \left(1-e^{-\frac{\Gamma+\Gamma'}{2}t}\right)\right]
653: \end{equation}
654:
655: \noindent where the last term is due to the re-correlation processes
656: mentioned in subsection \ref{tev}. Note that under conditions in which
657: the shift functions $F(e_R)$ and $F'(e'_R)$ are small one has
658: $e_R-e'_R\simeq\epsilon$. This will be assumed to be the case from
659: here on in order to save notation.
660:
661: The tunneling probability is given by eq. (\ref{ptgen}) as
662:
663: \begin{equation}
664: \label{pt2gamma}
665: P(t)=\frac{1}{2}-\frac{\cos\epsilon t}{2}\left[\frac{2\sqrt{\Gamma
666: \Gamma'}}{\Gamma+\Gamma'}+e^{-\frac{\Gamma+\Gamma'}{2}t}\left(1-
667: \frac{2\sqrt{\Gamma\Gamma'}}{\Gamma+\Gamma'}\right)\right]
668: \end{equation}
669:
670: \noindent which shows that the re-correlation processes give rise to
671: steady state oscillations of $P(t)$ after the transient associated
672: with the decaying exponential has died down. The amplitude of the
673: steady-state oscillations depends on the damping widths $\Gamma$ and
674: $\Gamma'$, however, and is maximum when $\Gamma=\Gamma'$, in which
675: case the transient actually cancels out and the undamped form of
676: $P(t)$, eq. (\ref{ptun}), is recovered.
677:
678: This last result does not imply, however, that there is no decoherence
679: when $\Gamma=\Gamma'$. The general expression for $\delta(t)$ obtained
680: from eq. (\ref{delgen}) can be written as
681:
682: \begin{eqnarray*}
683: \delta(t)&=&\frac{1}{2}\left(1+e^{-\Gamma t}-e^{-\Gamma' t}\right)
684: \left(1-e^{-\Gamma t}+e^{-\Gamma' t}\right) - \frac{1}{2}e^{-(\Gamma+
685: \Gamma')t}\\ &&-\frac{2\sqrt{\Gamma\Gamma'}}{\Gamma+\Gamma'}\left(1-
686: e^{-\frac{\Gamma+\Gamma'}{2}t}\right)\left[e^{-\frac{\Gamma+\Gamma'}
687: {2}t}\cos2\epsilon t+\frac{\sqrt{\Gamma\Gamma'}}{
688: \Gamma+\Gamma'}\left(1-e^{-\frac{\Gamma+\Gamma'}{2}t}\right)\right]
689: \end{eqnarray*}
690:
691: \noindent and reduces, when $\Gamma=\Gamma'$, to
692:
693: \[
694: \delta(t)\hspace{.5cm}\stackrel{\Gamma=\Gamma'}{\longrightarrow}
695: \hspace{.5cm} 2e^{-\Gamma t}\left(1-e^{-\Gamma t}\right)
696: \sin^2\epsilon t
697: \]
698:
699: \noindent which vanishes only asymptotically, as $t\rightarrow
700: \infty$. In the general case $\Gamma\neq\Gamma'$ with $\Gamma,
701: \Gamma'\neq 0$, one has the asymptotic limit
702:
703: \[
704: \lim_{t\rightarrow\infty}\delta(t)=\frac{1}{2}-\frac{2\Gamma\Gamma'
705: }{(\Gamma+\Gamma')^2}.
706: \]
707:
708: \noindent The restriction $\Gamma,\Gamma'\neq 0$ is necessary since
709: the limit $t\rightarrow\infty$ does not commute with the limit
710: $\Gamma'\rightarrow 0$. If the alternate version of the model with two
711: orthogonal continua is used, all terms involving factors
712: $\sqrt{\Gamma\Gamma'}/{(\Gamma+ \Gamma')}$ are absent, both in $P(t)$
713: and in $\delta(t)$. In this case the two-level system decoheres
714: maximally when $t\rightarrow\infty$ as two orthogonal components of
715: the two-level subsystem become correlated to orthogonal continuum
716: wave packets.
717:
718: The dissipative and unitary tunneling rates $R_d$ and $R_u$ can also be
719: obtained from the reduced density in a straightforward way. A typical
720: result for moderately strong coupling is shown in fig. 3.
721:
722: \subsection{Interpretation of model results.}
723:
724: In spite of the schematic character of the model, the features
725: emerging from its analysis underline relevant aspects which should be
726: kept in mind in the study of more realistic cases. As shown explicitly
727: e.g. in the general equation (\ref{ptgen}), the coherence properties
728: directly relevant for localization are, in the representation used
729: there, contained in the {\it real part} of the matrix element
730: $\rho_{+-}$. The harmonic time dependence of this quantity in the
731: absence of any external entanglements is modified when these effects
732: are turned on by making $g$ and $g'$ (or, equivalently, $\Gamma$ and
733: $\Gamma'$) different from zero, as can be read from the time dependent
734: term of equation (\ref{pt2gamma}). In the case considered in section
735: \ref{onel} ($\Gamma\neq 0$, $\Gamma'=0$) the harmonic oscillations are
736: damped by the effects of the external coupling, leading, in the strong
737: coupling limit, to tunneling rates dominated by the associated
738: relaxation times, rather that by the oscillation period. When also
739: $\Gamma'\neq 0$, on the other hand, a new undamped oscillatory term
740: arises due to contributions to ${\rm Re}\,\rho_{+-}$ coming from
741: continuum components, an effect which has been referred to as a
742: re-correlation process in section \ref{twol}. The amplitude of the
743: damped component is at the same time reduced, leading to its complete
744: cancellation in the limiting case $\Gamma=\Gamma'\neq 0$.
745:
746: It is important to stress that the coherence properties bearing on the
747: purity of the complete tunneling {\it state}, which can be measured in
748: a basis-independent way in terms of the determinant of the reduced
749: density matrix (see equation (\ref{delgen})), involve other dynamic
750: quantities besides the localization-specific correlation part ${\rm
751: Re}\,\rho_{+-}$, and therefore in general correlate poorly with the
752: tunneling probability $P(t)$. In particular, when $\Gamma=\Gamma'\neq
753: 0$ the transient decoherence undergone by the tunneling state has no
754: effect on the tunneling probability, since the localization-specific
755: coherence properties remain the same as in the absence of the external
756: couplings due to the re-correlation process. The decomposition of the
757: tunneling rate $\dot{P}(t)$ into dissipative and unitary parts
758: $R_d(t)$ and $R_u(t)$ can thus be seen as expressing the dependence
759: of the time rate of change of the localization-specific correlation
760: part ${\rm Re}\,\rho_{+-}$ in terms of contributions associated to the
761: (unitary) time evolution of the eigenvectors and from the
762: (non-unitary) time evolution of the eigenvalues of the reduced density
763: which describes the tunneling state. Only the latter contributions
764: relate to the overall decoherence of the tunneling state.
765:
766:
767: \section{Master equation results.}
768:
769: The dissipative dynamics of the reduced density matrix of a two-level
770: system is often described in terms of a master equation valid in a
771: weak-coupling regime, derived \cite{meq} under the assumption of
772: linear coupling to a reservoir of harmonic oscillators, which is
773: subsequently eliminated from the description under the assumption of
774: the validity of a Born-Markov approximation. In addition, one assumes
775: in the derivation that correlation functions involving reservoir
776: degrees of freedom can be meaningfully evaluated for all times using
777: the initial state of the reservoir, usually taken to be a thermal
778: equilibrium state. The usual form of the master equation \cite{livopt}
779: is obtained assuming that the coupling to the reservoir is given by
780:
781: \[
782: H_{\rm int}=\sum_k \left(g_k b_k^\dagger\sigma_-+g_k^* b_k\sigma_+
783: \right)
784: \]
785:
786: \noindent where $b_k$, $b_k^\dagger$ are the lowering and rising
787: operators for the $k$-th bath oscillator. This coupling term is
788: similar to the term involving $g(\eta)$ in eq. (\ref{Hmod}) if one
789: associates the ground state of the reservoir with the discrete state
790: $|0_b\rangle$ of the model. The resulting form of the equation is
791: then, in the limit of zero temperature,
792:
793: \[
794: \dot{\rho}(t)=-i\frac{\epsilon}{2}\left[\sigma_3,\rho(t)\right]+
795: \frac{\gamma}{2}\left(2\sigma_-\rho(t)\sigma_+-\sigma_+\sigma_-
796: \rho(t)-\rho(t)\sigma_+\sigma_-\right)
797: \]
798:
799: \noindent where $\gamma$ is a damping coefficient related to $H_{\rm
800: int}$ and to the distribution of reservoir frequencies.
801:
802: The solution of this equation satisfying the initial condition
803:
804:
805: \begin{equation}
806: \label{icm}
807: \rho_{++}(0)=\rho_{+-}(0)=\frac{1}{2}
808: \end{equation}
809:
810: \noindent is
811:
812: \[
813: \rho_{++}(t)=\frac{1}{2}e^{-\gamma t}\hspace{1cm}{\rm and}
814: \hspace{1cm}\rho_{+-}(t)=\frac{1}{2}e^{i\epsilon t-\frac{\gamma}{2} t}
815: \]
816:
817: \noindent which reproduces the result (\ref{ols}), with the
818: correspondence $\gamma\leftrightarrow\Gamma$, up to details related to
819: the shift function $F(E)$ in that case. It should be kept in mind,
820: however, that the derivation of the master equation is limited to weak
821: coupling regimes due to the Born-Markov approximation.
822:
823: One may next try and include effects analogous to those produced by
824: the coupling term involving $g'(\eta)$ in the model Hamiltonian
825: (\ref{Hmod}) by adding a new term to $H_{\rm int}$ in the derivation
826: of the master equation. Thus, using now
827:
828: \[
829: H_{\rm int}\rightarrow\sum_k \left(g_k b_k^\dagger\sigma_-+g_k^* b_k
830: \sigma_+\right)+\sum_k \left(g'_k b_k^\dagger\sigma_+ +{g'}_k^* b_k
831: \sigma_-\right)
832: \]
833:
834: \noindent we obtain, following the standard derivation procedure
835: \cite{meq}, the modified master equation
836:
837: \begin{eqnarray}
838: \label{cme}
839: \dot{\rho}(t)&=&-i\frac{\epsilon}{2}\left[\sigma_3,\rho(t)\right]+
840: \frac{\gamma}{2}\left(2\sigma_-\rho(t)\sigma_+-\sigma_+\sigma_-
841: \rho(t)-\rho(t)\sigma_+\sigma_-\right) \\ &&+\frac{\gamma'}{2}\left(2
842: \sigma_+\rho(t)\sigma_--\sigma_-\sigma_+\rho(t)
843: -\rho(t)\sigma_-\sigma_+\right)+\frac{\sqrt{\gamma\gamma'}}{2}
844: \left(2\sigma_+\rho(t)\sigma_++2\sigma_-\rho(t)\sigma_-\right)
845: \nonumber
846: \end{eqnarray}
847:
848: \noindent where $\gamma'$ a the constant analogous to $\gamma$ but
849: related to the new terms added to $H_{\rm int}$. The solution of this
850: equation satisfying the same initial condition (\ref{icm}) is
851:
852: \[
853: \rho_{++}(t)=\frac{1}{2}e^{-(\gamma+\gamma')t}+\frac{\gamma'}
854: {\gamma+\gamma'}\left(1-e^{-(\gamma+\gamma')t}\right)
855: \]
856:
857: \noindent and
858:
859: \[
860: \rho_{+-}(t)=\frac{e^{-(\gamma+\gamma')t}}{2\sqrt{\epsilon^2-\gamma
861: \gamma'}}\left[\sqrt{\gamma\gamma'}-i\epsilon\,{\rm sen}\left(2
862: \sqrt{\epsilon^2-\gamma\gamma'}\;t\right)+\sqrt{\epsilon^2-\gamma
863: \gamma'}\cos\left(2\sqrt{\epsilon^2-\gamma\gamma'}\;t\right)\right]
864: \]
865:
866: \noindent which is different from the corresponding results obtained
867: in subsection \ref{twol}. In particular, the model result for
868: $\rho_{++}(t)$, eq. (\ref{2lrho++}), involves two relaxation times
869: associated to the damping widths $\Gamma$ and $\Gamma'$ appearing in
870: the difference of separate exponentials, while the master equation
871: result involves just one ``effective'' relaxation time appearing in a
872: single exponential involving the sum $\gamma+\gamma'$. The relation
873: between these two results can be revealed by expanding both solutions
874: to {\it first order} in the damping parameters. One obtains, from the
875: master equation result,
876:
877: \[
878: \rho_{++}(t)\rightarrow\frac{1}{2}-\frac{\gamma}{2}t+
879: \frac{\gamma'}{2}t
880: \]
881:
882: \noindent which is identical to what one obtains from
883: eq. (\ref{2lrho++}) with the replacements $\gamma\leftrightarrow
884: \Gamma$ and $\gamma'\leftrightarrow\Gamma'$. The expansion of the
885: master equation result for $\rho_{+-}(t)$ coincides likewise with the
886: expansion of eq. (\ref{2lrho+-}). These facts can be understood by
887: recalling that a Born-Markov approximation is involved in the
888: derivation of the master equation. The Born approximation implies a
889: truncation of an expansion in the couplings, producing a result which
890: is subsequently ``exponentiated'' to all orders through the Markov
891: assumption. As a result of this there is no general guarantee for
892: results of the master equation which go beyond the order of the
893: truncation involved in the Born approximation. In particular, the
894: dynamics leading to two different relaxation times is not properly
895: accounted for. Note, in this connection, that when $\gamma=\gamma'$
896: (i.e. when the two relaxation times coincide) the master equation
897: result reproduces the model result up to a small shift of the bare
898: frequency $\epsilon$, if one restricts oneself to weak coupling
899: regimes.
900:
901: It may be argued that the master equation (\ref{cme}) is derived on
902: the assumption of a reservoir of harmonic oscillators, which differs
903: from the dynamical assumptions of the model defined by the Hamiltonian
904: (\ref{Hmod}). However, the same master equation also follows from this
905: Hamiltonian, with the usual (in fact, even somewhat weaker) derivation
906: assumptions, establishing, in particular, the correspondence
907: $\gamma\leftrightarrow\Gamma$ and $\gamma'\leftrightarrow\Gamma'$.
908: The Born approximation has to be performed in the same way in both
909: cases, indicating again that it is at the root of the discussed
910: discrepancies.
911:
912: \section{Concluding remarks.}
913:
914: The main general conclusion to be drawn from the analysis of the model
915: described by the Hamiltonian (\ref{Hmod}) with localized initial
916: condition (\ref{tunic}) is that a clear distinction should be made
917: between coherence properties relating a) to localization in the
918: tunneling degree of freedom and b) to the degree of quantum purity of
919: the reduced density (\ref{natorb}) describing the quantum state of the
920: tunneling system, as measured e.g. by the linear entropy
921: (\ref{delta}). Coherence properties of class a) are carried by the
922: real part of off-diagonal matrix element $\rho_{+-}(t)$ in the basis
923: which diagonalizes $H_\sigma$ or, equivalently, by the diagonal
924: matrix elements in the localized basis $\{|r\rangle,|l\rangle\}$,
925: e.g. $\rho_{rr}(t)=1/2-{\rm Re}\, \rho_{+-}=P(t)$. Coherence
926: properties of class b), on the other hand, are carried by the
927: eigenvalues of the reduced density, and involve therefore other
928: dynamic quantities in addition to that which is relevant for class
929: a).
930:
931: The contribution of decoherence processes undergone by the reduced
932: density to the coherence properties of class a) can be isolated
933: through the general decomposition of the tunneling rate $\dot{P}(t)$
934: into dissipative and unitary parts, eqs. (\ref{Ru}) and (\ref{Rd}).
935: Results for the damped two-level model show that the dissipative
936: component contributes most at times short on the scale of the bare
937: frequency $\epsilon$. The enhanced initial tunneling rate in the
938: strong coupling case with no re-correlation effects (see
939: fig. \ref{fig2}) is dominated by the dissipative component, and can
940: thus be interpreted in terms of a large effect of the coherence loss
941: of the tunneling state also on the localization-specific correlation
942: properties of the initial state of the two-level subsystem.
943:
944: Comparison of the exact solution of the model with master equation
945: results derived in the usual way from the model Hamiltonian indicates
946: that master equations appear to be inadequate to deal with situations
947: involving more than a single relaxation time. This difficulty can be
948: ascribed to the Born approximation involved in the derivation of these
949: equations.
950:
951: \vspace{.3cm}
952:
953: \noindent {\bf Acknowledgements.} JGPF and MCN are partially supported
954: by Conselho Nacional de Desenvolimento Cient\'{\i}fico e Tecnol\'ogico
955: (CNPq), Brazil. ANS is supported by Funda\c{c}\~ao de Amparo à
956: Pesquisa do Estado de S\~ao Paulo (FAPESP). MCN has been partly
957: supported by FAPESP during the later stages of this work.
958:
959: \section*{Appendix A.}
960:
961: A general reduced density matrix can always be written in terms of its
962: time dependent natural orbitals $|n(t)\rangle$ and associated
963: eigenvalues $p_n(t)$ as
964:
965: \[
966: \rho(t)=\sum_n|n(t)\rangle p_n(t)\langle n(t)|.
967: \]
968:
969: \noindent The probability that the state thus described is found in a
970: given (time independent) subspace defined in terms of a projector
971:
972: \[
973: {\cal{P}}\equiv\sum_k |b_k\rangle\langle b_k|,
974: \]
975:
976: \noindent where the $|b_k\rangle$ constitute an orthonormal set of
977: state vectors, is given by
978:
979: \[
980: P(t)={\rm Tr}\left[\rho(t){\cal{P}}\right].
981: \]
982:
983: \noindent The vectors $|b_k\rangle$ can be expanded in the natural
984: orbitals of $\rho(t)$ as $|b_k\rangle=\sum_n\beta_n^{(k)}(t)|n(t)
985: \rangle$ so that $P(t)$ is expressed as
986:
987: \[
988: P(t)=\sum_n p_n(t)\;\sum_k|\beta_n^{(k)}|^2.
989: \]
990:
991: \noindent Derivation with respect to time then gives dissipative and
992: unitary contributions proportional to $\dot{p}_n$ and
993: $\dot{\beta}_n^{(k)}$ respectively.
994:
995: \section*{Appendix B.}
996:
997: In view of the widespread use of coordinate coupling to a collection
998: of harmonic oscillators in order to introduce damping effects on the
999: dynamics of simple quantum-mechanical systems (including two-level
1000: systems), and since the results obtained in this way seem to be at
1001: variance with those of the continuum model for the damping as used in
1002: section \ref{main}, this Appendix deals briefly with a comparison of
1003: features of the two models. Instead of the implementation which
1004: minimizes ``threshold'' effects adopted in that section, the case
1005: $e_0\simeq\eta_0\simeq 0$ with $g(\eta)=g^*(\eta)=g'(\eta)$ of the
1006: continuum model will be considered, together with with the ``zero
1007: bias'' spin-boson model \cite{legg}.
1008:
1009: In both cases the Hamiltonian can be split as $H_\sigma+H_b+H_{int}$
1010: with $H_\sigma=\frac{\epsilon}{2} \sigma_3$ and
1011:
1012: \[
1013: H_b=|0_b\rangle e_0\langle 0_b|+\int_{\eta_0}^{\bar{\eta}}d\eta\,
1014: |\eta\rangle\eta\langle\eta|,\hspace{1cm}H_{int}=\sigma_1\int_{
1015: \eta_0}^{\bar{\eta}}d\eta\,g(\eta)\left(|\eta\rangle
1016: \langle0_b|+|0_b\rangle\langle\eta|\right)\equiv\sigma_1 G
1017: \]
1018:
1019: \noindent for the continuum model of equation (\ref{Hmod}) and
1020:
1021: \[
1022: H_{b}=\sum_\alpha \hbar\omega_\alpha a_\alpha^\dagger a_\alpha,
1023: \hspace{1cm}H_{int}=\sigma_1\sum_\alpha g_\alpha\sqrt{\frac{\hbar}
1024: {2m_\alpha\omega_\alpha}}\left(a_\alpha^\dagger+a_\alpha\right)
1025: \equiv\sigma_1 G.
1026: \]
1027:
1028: \noindent for the spin-boson model. A useful way of representing these
1029: operators uses the spin basis of localized states in which $\sigma_1$
1030: is diagonal. The Hamiltonian is then represented in both cases as
1031:
1032: \[
1033: H=\left(\begin{array}{cc}H_b+G & -\frac{i}{2}\epsilon \\ & \\
1034: \frac{i}{2}\epsilon & H_b-G \end{array}\right).
1035: \]
1036:
1037: \noindent The operators $H_b\pm G$ can be separately diagonalized
1038: exactly in the two cases. This solves the problem completely in the
1039: extreme adiabatic limit $\epsilon\rightarrow 0$. In general, matrix
1040: elements in the off-diagonal blocks involving $H_\sigma$ are modified
1041: by overlap factors involving eigenstates of the two operators $H_b\pm
1042: G$. Rather than dealing with the full problem, we restrict ourselves
1043: to the perturbative effects of $H_b$ for the lowest adiabatic
1044: states. These are, in both cases, a degenerate doublet which we denote
1045: by $|\phi_0^{\pm}\rangle$, with energy $E_0$. The secular matrix to be
1046: considered is then
1047:
1048: \[
1049: \left(\begin{array}{cc} E_0 & -\frac{i}{2}\epsilon\langle\phi_0^{(+)}|
1050: \phi_0^{(-)}\rangle \\ & \\ \frac{i}{2}\epsilon\langle\phi_0^{(-)}|
1051: \phi_0^{(+)}\rangle & E_0 \end{array}\right).
1052: \]
1053:
1054: In the case of the spin-boson model the overlap $\langle\phi_0^{(+)}|
1055: \phi_0^{(-)}\rangle$ involves oscillator coherent states with opposite
1056: displacements which depend, in particular, on the coupling constants
1057: $g_\alpha$. The ensuing reduction of the overlap implies the reduction
1058: of the splitting of the perturbed states, favoring the slowing down of
1059: transition rates out of localized states. In the case of the continuum
1060: model, on the other hand, $E_0$ is the root which lies below $\eta_0$
1061: of the dispersion equation
1062:
1063: \[
1064: E_0-e_0=\int_{\eta_0}^{\bar{\eta}}d\eta\;\frac{g^2(\eta)}{E_0-\eta}.
1065: \]
1066:
1067: \noindent The corresponding unperturbed states can be expanded as
1068:
1069: \[
1070: |\phi_0^{(\pm)}\rangle=a_0^{(\pm)}|0_b\rangle+\int_{\eta_0}^{\bar{
1071: \eta}}d\eta\;A^{(\pm)}(\eta)|\eta\rangle
1072: \]
1073:
1074: \noindent the expansion coefficients being given by
1075:
1076: \[
1077: |a_0^{(\pm)}|^2=\frac{1}{1+\frac{1}{4}\int_{\eta_0}^{\bar{\eta}}
1078: d\eta\;\frac{g^2(\eta)}{(E_0-\eta)^2}}\hspace{1cm}{\rm and}
1079: \hspace{1cm}A^{(\pm)}(\eta)=\pm\frac{g(\eta)}{E_0-\eta}\,a_0^{(\pm)}.
1080: \]
1081:
1082: \noindent and the overlap factor is
1083:
1084: \[
1085: \langle\phi_0^{(+)}|\phi_0^{(-)}\rangle=\frac{1-\frac{1}{4}\int_{
1086: \eta_0}^{\bar{\eta}}d\eta\;\frac{g^2(\eta)}{(E_0-\eta)^2}}{1+
1087: \frac{1}{4}\int_{\eta_0}^{\bar{\eta}}d\eta\;\frac{g^2(\eta)}{
1088: (E_0-\eta)^2}}.
1089: \]
1090:
1091: \noindent The behavior of this object involves the energy dependence
1092: of the coupling function $g(\eta)$ and of the solution $E_0$ of the
1093: dispersion equation. In the case when $g^2(\eta)=\gamma\eta^s$, with
1094: $0\leq s<1$ one finds, for fixed $e_0=\eta_0=0$ that $E_0$ approaches
1095: minus infinity and $|a_0^{(\pm)}|^2$ approaches unity as the high
1096: energy cut-off $\bar{\eta}$ increases. In the same limit the overlap
1097: also approaches unity and the transition rate out of the localized
1098: states approaches its free value. For values of $\bar{\eta}$ not too
1099: large in comparison with $\epsilon$ the overlap is smaller than one,
1100: leading to a quenching of this transition rate. These behaviors are
1101: supported by exact numerical results.
1102:
1103: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1104:
1105: \begin{thebibliography}{99}
1106:
1107: \bibitem{gen} A. O. Caldeira and A. J. Leggett, Phys. Rev. Lett {\bf
1108: 46}, 211 (1981).
1109:
1110: \bibitem{BNW} D. M. Brink, M. C. Nemes and D. Vautherin, Annals of
1111: Physics {\bf 147}, 171 (1983).
1112:
1113: \bibitem{legg} A. J. Leggett, S. Chakravarty, A. T. Dorsey,
1114: M. P. A. Fisher, A Garg and W. Zwrger, Revs. Mod. Phys. {\bf 59}, 1
1115: (1987).
1116:
1117: \bibitem{fano} U. Fano, Phys. Rev. {\bf 124}, 1866 (1961).
1118:
1119: \bibitem{livopt} H. Carmichael, {\it Lectures Notes in Physics -
1120: An Open Systems Approach to Quantum Optics}, Springer Verlag, {\bf m
1121: 18}, 26 (1991).
1122:
1123: \bibitem{meq} W. H. Louisell, Quantum Statistical Properties of
1124: Radiation Eds. John Wiley and Sons, Inc, 343 (1973).
1125:
1126: \bibitem{vankampen} N. V. Van Kampen, Kgl. Danske Videnskab. Selskab,
1127: Mat.-fys. Medd. {\bf 26}, No. 15 (1951).
1128:
1129: \end{thebibliography}
1130:
1131: \newpage
1132:
1133: \section*{Figure captions.}
1134:
1135: \noindent Fig. 1 - Tunneling probability $P(t)$, idempotency defect
1136: $\delta(t)$ (top) and tunneling rates $R_d(t)$, $R_u(t)$ and
1137: $\dot{P}(t)=R_d(t)+R_u(t)$ (bottom) in the single finite relaxation
1138: time, weak-coupling case with parameters $\Gamma=0.3$ and
1139: $\epsilon=1$. The time scale is defined so that $\hbar=1$.
1140:
1141: \vspace{.5cm}
1142:
1143: \noindent Fig. 2 - Same as Figure 1 in the single finite relaxation
1144: time, strong coupling case with parameters $\Gamma=3.$, $\epsilon=1$.
1145: Note the smaller depth of the time scale in this case.
1146:
1147: \vspace{.5cm}
1148:
1149: \noindent Fig. 3 - Same as Figure 1 in the case of two finite
1150: relaxation times with parameters $\Gamma=2$, $\Gamma'=.5$,
1151: $\epsilon=1$.
1152:
1153:
1154: \newpage
1155:
1156: \begin{figure}
1157: \centering
1158: \begin{minipage}[c]{.45\textwidth}
1159: \centering
1160: \caption{Tunneling probability $P(t)$, idempotency defect $\delta(t)$
1161: (top) and tunneling rates $R_d(t)$, $R_u(t)$ and $\dot{P}(t)=R_d(t)+
1162: R_u(t)$ (bottom) in the single finite relaxation time, weak-coupling
1163: case with parameters $\Gamma=0.3$ and $\epsilon=1$. The time scale is
1164: defined so that $\hbar=1$.}
1165: \label{fig1}
1166: \end{minipage}%
1167: \hfill
1168: \begin{minipage}[c]{.45\textwidth}
1169: \begin{center}
1170: \includegraphics[width=2.7in]{F1b.eps}
1171: \end{center}
1172: \end{minipage}
1173: \end{figure}
1174:
1175: \begin{figure}
1176: \centering
1177: \begin{minipage}[c]{.45\textwidth}
1178: \centering
1179: \caption{Same as Figure \ref{fig1} in the of single finite relaxation
1180: time, strong coupling case with parameters $\Gamma=3.$, $\epsilon=1$.
1181: Note the smaller depth of the time scale in this case.}
1182: \label{fig2}
1183: \end{minipage}%
1184: \hfill
1185: \begin{minipage}[c]{.45\textwidth}
1186: \begin{center}
1187: \includegraphics[width=2.7in]{F2b.eps}
1188: \end{center}
1189: \end{minipage}
1190: \end{figure}
1191:
1192: \begin{figure}
1193: \centering
1194: \begin{minipage}[c]{.45\textwidth}
1195: \centering
1196: \caption{Same as Figure \ref{fig1} in the case of two finite
1197: relaxation times, with parameters $\Gamma=2$, $\Gamma'=.5$,
1198: $\epsilon=1$.}
1199: \label{fig3}
1200: \end{minipage}%
1201: \hfill
1202: \begin{minipage}[c]{.45\textwidth}
1203: \begin{center}
1204: \includegraphics[width=2.7in]{F3b.eps}
1205: \end{center}
1206: \end{minipage}
1207: \end{figure}
1208:
1209: \end{document}
1210:
1211: