quant-ph0402102/qtl.tex
1: %\documentclass[preprint]{revtex4}
2: \documentclass[pre,twocolumn,showpacs]{revtex4}
3: \usepackage{graphicx}
4: %\input{psbox.tex}
5: 
6: % a few macros
7: \newcommand{\equ}{Eq.}
8: \newcommand{\eqs}{Eqs.}
9: \newcommand{\fig}{Fig.}
10: \newcommand{\figs}{Figs.}
11: \newcommand{\sect}{Sec.}
12: \newcommand{\rem}[1]{}
13: \newcommand{\const}{\text{constant}}
14: \newcommand{\imag}[1]{\text{Im}(#1)}
15: \newcommand{\imagb}[1]{\text{Im}[#1]}
16: \newcommand{\real}[1]{\text{Re}(#1)}
17: 
18: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
19: \begin{document}
20: 
21: \title{Simulating noisy quantum protocols 
22: with quantum trajectories} 
23: \author{Gabriel G. Carlo, Giuliano Benenti, Giulio Casati, and Carlos 
24: Mej\'\i a-Monasterio}
25: \email{gabriel.carlo@uninsubria.it}
26: \homepage{http://www.unico.it/~dysco}
27: \affiliation{Center for Nonlinear and Complex Systems, Universit\`a degli 
28: Studi dell'Insubria and Istituto Nazionale per la Fisica della Materia, 
29: Unit\`a di Como, Via Valleggio 11, 22100 Como, Italy}
30: \date{\today}
31: \pacs{03.65.Yz,03.67.Hk,03.67.Lx}
32: %03.65.Yz Decoherence; open systems; quantum statistical methods
33: %03.67.Hk Quantum communication
34: %03.67.Lx Quantum computation
35: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
36: 
37: %\widetext
38: \begin{abstract} 
39: %The noise processes induced by the environment are one of the major 
40: %problems that the implementation of quantum information
41: %protocols has to face. 
42: %Understanding and controlling them is essential in order to operate 
43: %a quantum computer successfully. 
44: The theory of quantum trajectories is applied to simulate the effects
45: of quantum noise sources induced by the environment 
46: on quantum information protocols. 
47: We study two models that generalize single qubit noise channels like 
48: amplitude damping and phase flip to the many-qubit situation. 
49: We calculate the fidelity of quantum information transmission 
50: through a chaotic channel using the teleportation scheme with 
51: different environments. 
52: In this example, we analyze the role played by the kind of collective 
53: noise suffered by the quantum processor during its operation.
54: We also investigate the stability of a quantum algorithm simulating 
55: the quantum dynamics of a paradigmatic model of chaos, the baker's map.
56: Our results demonstrate that, using the quantum trajectories approach, 
57: we are able to simulate quantum protocols in the presence of noise and 
58: with large system sizes of more than 20 qubits.
59: \end{abstract}
60: \maketitle
61: 										
62: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
63: \section{Introduction}
64: \label{sec:intro}
65: 
66: It would be highly desirable to implement quantum protocols 
67: using processors perfectly isolated from the environment, since 
68: this is one of the main sources of error in quantum computation 
69: (there are also system specific imperfections, but here we will 
70: only address the environmental problem).  
71: Unfortunately, this is not possible. Quantum hardware will 
72: naturally become entangled with the environment during its operation. 
73: Thus, if any hope of profiting from the benefits of quantum computation 
74: is to be kept, understanding and controlling quantum noise effects 
75: is essential. On the other hand, the study of open systems is of 
76: interest in several fields, from both theoretical and experimental 
77: points of view \cite{Gra1,Gra2,Gra3,Knight}.  
78: 
79: Factoring large integers in polynomial time has been the milestone 
80: discovery that set out quantum computation as a major research topic
81: \cite{shor}. 
82: Nevertheless, in the short term, few qubits quantum computers, 
83: this kind of calculations will be necessarily out of reach, 
84: since they involve large systems. Then, it is reasonable to focus on 
85: understanding the behavior of accessible first realizations. 
86: It is interesting to remark that, with a few tens 
87: of qubits, quantum simulations of systems studied 
88: in quantum chaos (like the quantum baker's map \cite{Sch}, 
89: the quantum kicked rotator \cite{Bertrand}, and the quantum sawtooth 
90: map \cite{Ben1}) would outperform any calculation that can be done 
91: with present day supercomputers. 
92: A first step in the simulation of quantum chaos models has been
93: the implementation of the quantum baker's map on a three-qubit
94: NMR-based quantum processor \cite{cory}.
95: 
96: With this situation in mind, it is natural to ask to what extent we 
97: can know and control the operability and the stability of a 
98: quantum computer of this size. Theoretical studies rely 
99: on evaluating the reduced density matrix of the system 
100: (obtained after tracing out the environment), often in 
101: terms of various different approximations.
102: It would therefore be desirable to give an answer for generic
103: quantum protocols and noise models, using exact calculations.  
104: In this work, we propose the numerical simulation of superoperators 
105: as a way to do it. By means of quantum trajectories techniques 
106: we can reach results for system sizes for which the implementation 
107: of chaotic maps becomes relevant for theoretical studies in the 
108: field of quantum chaos, having the chance to include any kind of 
109: environmental effects. 
110: 
111: Instead of solving the density matrix directly, quantum trajectories 
112: stochastically evolve the state vector of the system, 
113: and after averaging over many runs the same results for 
114: the outcomes of any observable are obtained. 
115: The use of quantum trajectories in the field of quantum information
116: has been pioneered by \cite{Schack,BarencoBrun}.
117: In Ref.~\cite{Car}, we applied this formalism to study the 
118: effect of a dissipative environment 
119: on the quantum  teleportation protocol \cite{BennettBrassard} 
120: through a large chain of qubits.
121: This situation models the transmission of quantum information 
122: through a chaotic quantum channel.
123: %Teleportation is one of the basic methods of quantum 
124: %communication, thus being one of the most important 
125: %protocols in quantum information theory. Using 
126: %teleportation, an arbitrary unknown qubit can be transmitted by means 
127: %of a shared entangled pair (an EPR pair where one particle 
128: %is with the sender ''Alice'' and the other with the receiver ''Bob''). 
129: %Performing local operations and sending 
130: %two bits of classical information 
131: %is enough to accomplish the task. 
132: Here we review this model and present new results for different 
133: noise channels. 
134: 
135: As mentioned before, one of the main near future applications 
136: of quantum computation is the simulation of dynamical systems 
137: that are of great interest in quantum chaos.
138: In this work we focus on the quantum baker's map, one of the 
139: most important examples of this kind of systems. 
140: This map is fully chaotic and its quantized version consists of 
141: conveniently selected quantum Fourier transforms. 
142: We model the environment through a phase flip channel 
143: and study the fidelity of quantum computing of the quantum
144: baker's map in this noisy environment. We note that 
145: the fidelity has already been computed experimentally with 
146: a three-qubit NMR quantum processor \cite{cory}.
147: Hence, for the design and construction of quantum 
148: hardware with a larger number of qubits, simulations 
149: like those performed in this paper will become essential. 
150: 
151: This paper is organized as follows. In Sec.~\ref{sec:QT}, we present 
152: a brief explanation of the theory of quantum trajectories and connect 
153: it with the quantum operations approach to the density matrix 
154: evolution. We also refer to the master equation formulation. 
155: Afterward, in Sec.~\ref{sec:noise}, we make a short description 
156: of the noise channels that we use in the calculations, paying special 
157: attention to the amplitude damping models. In Sec.~\ref{sec:tele}, 
158: we review the quantum teleportation protocol presented in \cite{Car}, 
159: focusing on the different dissipative processes. 
160: In Sec.~\ref{sec:Baker}, we study the behavior of the fidelity
161: of the quantum algorithm for the baker's map in the presence 
162: of a phase flip noise. 
163: Finally, in Sec.~\ref{sec:conclusion}, we present our 
164: conclusions and outlook.
165: 
166: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
167: \section{Master equation, superoperators, and quantum trajectories}
168: \label{sec:QT}
169: 
170: There is a close relationship among the master equation in Lindblad 
171: form, the superoperator formalism and the quantum trajectories theory. 
172: It is useful to review this relationship and to motivate the use of the 
173: quantum trajectory approach when studying open systems 
174: like quantum processors. 
175: We start with the special case of the master equation formulation 
176: of the problem and relate it with simulations using quantum trajectories 
177: techniques. Then, in Sec.~\ref{sec:QT:qops} we generalize this 
178: establishing the connection with the broader quantum operators formalism. 
179: 
180: %%%%
181: \subsection{Master equation and quantum trajectories}
182: \label{sec:QT:meqt}
183: 
184: Real systems interact with the environment, 
185: and, as mentioned before, these interactions are usually referred 
186: to as {\em quantum noise} \cite{Chuangbook,Preskillbook}. 
187: Several models account for different sorts of system-environment
188: interactions, the particular choice depending on the nature of 
189: the system and environment under consideration. 
190: Such open quantum systems in general cannot be described by a pure state, 
191: but rather by a mixed state. 
192: Their evolution takes density matrices to density matrices. 
193: This allows the evolution from system's pure states to mixed ones, 
194: and also, in some noise models, from mixed to pure states. 
195: In order to obtain the differential equation corresponding to this 
196: process, one assumes Markovian behavior, giving the evolution of the 
197: density operator with reference only to its state at present. 
198: This Markovian assumption neglects memory effects: it implies that 
199: the time needed for the environment to loose the information it 
200: received from the system is short enough in comparison with the time 
201: scale of the dynamics we perceive. Then, we are entitled to regard the 
202: information flow in only one direction, neglecting any kind of 
203: feedback \cite{Preskillbook}.
204: 
205: The Lindblad form of the master equation of this system-environment 
206: model in the Born-Markov approximation can be formally written as 
207: \cite{Lindblad}:
208: \begin{equation}
209: \dot \rho = {\cal L} [\rho], 
210: \end{equation}
211: with formal solution
212: \begin{equation}
213: \rho(t) = \exp({\cal L}t) [\rho(0)], 
214: \end{equation}
215: where ${\cal L}$ stands for the ``Lindblandian'' operator. 
216: In order to obtain explicit expressions we can trace out the 
217: environment, which gives
218: \begin{equation}
219: \dot \rho = -\frac{i}{\hbar} [H_s,\rho] - \frac{1}{2} \sum_{\mu} 
220: \{L_{\mu}^{\dag} L_{\mu},\rho\}+\sum_{\mu} L_{\mu} \rho 
221: L_{\mu}^{\dag},
222: \label{eq:lin2}
223: \end{equation}
224: where $L_{\mu}$ are the Lindblad operators 
225: ($\mu \; \epsilon \; [1,\ldots,{\cal M}]$, the number ${\cal M}$
226: depending on the noise model), 
227: $H_s$ is the system's Hamiltonian
228: and \{\,,\,\} denotes the anticommutator. 
229: The first two terms of this equation can be regarded as the evolution 
230: performed by an effective non-Hermitian Hamiltonian, $H_{\rm eff}=
231: H_s+iK$, with $K=-\hbar/2 \sum_{\mu}L_{\mu}^{\dag}L_{\mu}$. 
232: In fact, we see that 
233: \begin{equation}
234:  -\frac{i}{\hbar} [H_s,\rho] - \frac{1}{2} \sum_{\mu} 
235: \{L_{\mu}^{\dag} L_{\mu},\rho\} = -\frac{i}{\hbar} 
236: [H_{\rm eff} \rho - \rho H_{\rm eff}^{\dagger}],
237: \end{equation}
238: which reduces to the usual evolution equation for the 
239: density matrix in the 
240: case of $H_{\rm eff}$ being Hermitian.
241: The last term is responsible for the 
242: so-called {\em quantum jumps}. 
243: In this context the Lindblad 
244: operators $L_{\mu}$ are also named {\em quantum jump operators}. 
245: If the initial density matrix is in a pure state 
246: $\rho(t_0)=|\phi(t_0)\rangle \langle\phi(t_0)|$, 
247: after a time $dt$ evolves to the following statistical mixture: 
248: \begin{equation}
249: \rho(t_0+dt)=(1-\sum_{\mu}dp_{\mu}) 
250: \: |\phi_0\rangle \langle\phi_0| \: + 
251: \sum_{\mu} dp_{\mu}  |\phi_{\mu}\rangle \langle\phi_{\mu}|, 
252: \label{eq:stat}
253: \end{equation}
254: with the probabilities $dp_{\mu}$ defined by 
255: \begin{equation}
256: dp_{\mu}\!=\!\langle \phi(t_0)| L_{\mu}^{\dag} 
257: L_{\mu} |\phi(t_0)\rangle dt,
258: \end{equation} 
259: and the new states by 
260: \begin{equation}
261: |\phi_0\rangle = 
262: \frac{(\openone-i H_{\rm eff} dt/\hbar) |\phi(t_0)\rangle}
263: {\sqrt{1-\sum_{\mu} dp_{\mu}}} 
264: \label{eq:stat2}
265: \end{equation}
266: and
267: \begin{equation}
268:  |\phi_{\mu}\rangle = \frac{L_{\mu} 
269: |\phi(t_0)\rangle}{||L_{\mu} |\phi(t_0)\rangle||}. 
270: \end{equation}
271: 
272: Then, the {\em quantum jump picture} turns out to be clear; 
273: with probability 
274: $dp_{\mu}$ a jump occurs and the system is prepared in the state 
275: $|\phi_{\mu}\rangle$. With probability $1-\sum_{\mu} dp_{\mu}$ there 
276: are no jumps and the system evolves according to the effective 
277: Hamiltonian $H_{\rm eff}$ (normalization is included also in this case 
278: because the evolution is given by a non-unitary operator).
279: 
280: The numerical method we are going to use in order to 
281: simulate the master equation is usually known as the Monte Carlo 
282: Wave function approach \cite{DalibardCastin}. 
283: We start from a pure state $|\phi(t_0)\rangle$ 
284: and at intervals $dt$, smaller than 
285: the timescales relevant for the evolution of the density matrix, 
286: we perform the following evaluation. 
287: We choose a random number $\epsilon$ 
288: from a uniform distribution in the unit interval $[0,1]$. 
289: If $\epsilon < dp$, where $dp=\sum_{\mu} dp_{\mu}$, 
290: the system jumps to one of the states $|\phi_{\mu}\rangle$ 
291: (to $|\phi_1\rangle$ if $0 \le \epsilon \le dp_1$, to 
292: $|\phi_2\rangle$ if $dp_1 < \epsilon \le dp_1+dp_2$, and so on). 
293: On the other hand, if $\epsilon > dp$, the evolution with 
294: the non-Hermitian Hamiltonian $H_{\rm eff}$ takes place, ending 
295: up in the state $|\phi_0\rangle$. In 
296: both circumstances we 
297: renormalize the state. We repeat this process as many times as 
298: $n_{\rm steps}=\Delta t/dt$ where $\Delta t$ is the whole 
299: elapsed time during the evolution. 
300: Note that we must take $dt$ much smaller than the time scales relevant
301: for the evolution of the open quantum system under investigation.
302: In our simulations, $n_{\rm steps}$ will be 
303: proportional to the number of quantum gates involved in 
304: the corresponding protocol. Each realization 
305: provides a different {\em quantum trajectory} and a particular set 
306: of them (given a choice of the Lindblad operators) is an 
307: ``unraveling'' of the master equation. 
308: It is easy to see that if 
309: we average over different runs we recover the probabilities 
310: obtained with the density operator. In fact, given an operator 
311: $A$, we can write the mean value $\langle A \rangle_t={\rm Tr}
312: [A \rho(t)]$ as the average over ${\cal N}$ trajectories: 
313: \begin{equation}
314: \langle A \rangle_t = 
315: \lim_{{\cal N}\to \infty}
316: \frac{1}{{\cal N}} \sum_{i=1}^{{\cal N}} \langle \phi_i(t)| A 
317: | \phi_i(t) \rangle.
318: \end{equation}
319: 
320: The advantage of using the quantum trajectories method is clear 
321: since we need to store a vector of length $N$ ($N=2^n$ is 
322: the dimension of the Hilbert space, $n$ being the number
323: of qubits) rather than a $N \times N$ density matrix. 
324: Moreover, there is also an advantage in computation 
325: time with respect to density matrix direct calculations. 
326: We find that a reasonable 
327: amount of trajectories (we have used $100\le {\cal N} \le 400$ in 
328: all calculations, unless otherwise mentioned) is needed in 
329: order to obtain a satisfactory statistical convergence. 
330: 
331: This picture can be formalized by 
332: means of the stochastic Schr\"odinger equation \cite{Brun}
333: \begin{eqnarray}
334: |d \phi \rangle & = & -i H_s |\phi\rangle dt - \frac{1}{2} \sum_{\mu} 
335: (L_{\mu}^{\dag} L_{\mu} - \langle L_{\mu}^{\dag} L_{\mu} 
336: \rangle_{\phi}) |\phi\rangle \: dt \nonumber  \\ 
337:  & & + \sum_{\mu} 
338: \left( \frac{L_{\mu}}{\sqrt{\langle L_{\mu}^{\dag} L_{\mu} 
339: \rangle_{\phi}}}
340: -\openone \right) |\phi \rangle \: dN_{\mu}.
341: \label{stocsch}
342: \end{eqnarray}  
343: This is a stochastic nonlinear differential equation, where the 
344: stochasticity is due to the measurement results: 
345: we think that the environment is actually measured (as it is 
346: the case in indirect measurement models) or simpler, that 
347: the contact of the system with the environment produces an 
348: effect similar to a continuous measurement \cite{Zurek}. 
349: The nonlinearity is due to the renormalization of the state 
350: vector. 
351: The stochastic differential variables $dN_{\mu}$ are 
352: statistically independent and represent measurement outcomes. 
353: Their ensemble mean is given by 
354: $M[dN_{\mu}]=\langle L_{\mu}^{\dag} L_{\mu} \rangle_{\phi} dt$. 
355: The probability that the variable $dN_\mu$ 
356: is equal to $1$ during a given time step $dt$
357: is $\langle \phi L_{\mu}^{\dagger} L_{\mu} \phi \rangle \: dt$.
358: Therefore, most of the time the variables 
359: $dN_{\mu}$ are $0$ and as a consequence the system evolves 
360: continuously by means of the non Hermitian effective Hamiltonian. 
361: However, when a variable $dN_{\mu}$ is equal to $1$,
362: the corresponding term in equation (\ref{stocsch}) 
363: is the most significant. In these cases the quantum jump occurs. 
364: Note that there are also other possibilities in order to unravel 
365: the master equation such as the quantum state diffusion 
366: \cite{GisinPercival,Schack}, for example.
367: 
368: %%%%
369: \subsection{Quantum operations and quantum trajectories}
370: \label{sec:QT:qops}
371: 
372: There is a close connection between the master equation in 
373: Lindblad form and the quantum operations theory 
374: \cite{Chuangbook,Preskillbook} and 
375: we use this latter formalism as a comparison tool for the quantum 
376: trajectories calculations. 
377: Furthermore, it will become clear that the evolution 
378: of the density matrix of the system given by this method can 
379: be put on the same 
380: footing as the stochastic evolution model. In fact, the latter 
381: constitutes a Monte Carlo simulation of the former. 
382: 
383: We write the solution to 
384: \equ~(\ref{eq:lin2}) over an infinitesimal time $dt$
385: as a completely positive map:
386: \begin{equation}
387: \rho(t+dt)=\$[\rho(t)]=\sum_{\mu=0}^{{\cal M}} M_{\mu}(dt) \rho(t) 
388: M_{\mu}^{\dag}(dt),
389: \label{eq:infsolution}
390: \end{equation}
391: where, for $\mu=0$, we have $M_0=\openone - i H_{\rm eff} dt/\hbar$ and, for 
392: $\mu > 0$, $M_{\mu}=L_{\mu} \sqrt{dt}$, satisfying $\sum_{\mu=0}^{\cal M} 
393: M_{\mu}^{\dag} M_{\mu} = \openone$ to first order in $dt$. 
394: Equation (\ref{eq:infsolution}) is called the Kraus representation
395: (or the operator sum representation) of the superoperator $\$$
396: and the operators $M_\mu$ are known as operators elements
397: for the quantum operation $\$$.
398: It can be shown that, if the global evolution (system 
399: plus environment) is unitary, the Kraus operators 
400: satisfy the completeness relation $\sum_{\mu=0}^{\cal M} 
401: M_{\mu} M_{\mu}^{\dag} = \openone$ \cite{Chuangbook,Preskillbook}. 
402: Note that the superoperator $\$$ maps density matrices to density 
403: matrices, that is $\rho(t+dt)$ is Hermitian, has unit trace and 
404: is nonnegative if $\rho(t)$ satisfies these properties. 
405: 
406: The action of the quantum operation $\$$
407: can be interpreted as $\rho$ being randomly replaced 
408: by $M_{\mu} \rho M_{\mu}^{\dag} / {\rm Tr} (M_{\mu} \rho M_{\mu}^{\dag})$, 
409: with probability ${\rm Tr} (M_{\mu} \rho M_{\mu}^{\dag})$. Equivalently, 
410: the set $\{M_{\mu}\}_{\mu=1,...,{\cal M}}$ 
411: defines a Positive Operator Valued Measurement 
412: (POVM) with operators $E_{\mu}=M_{\mu}^{\dag} M_{\mu}$, 
413: that satisfy 
414: $\sum_{\mu=0}^{\cal M} E_{\mu}=\openone$. 
415: The outlined process is equivalent to performing a continuous 
416: measurement on the system (or an indirect measurement if the environment 
417: is actually measured) and shows the close
418: connection between the Kraus operators 
419: formalism and the quantum jumps picture. 
420: %In fact, the usual orthogonal measurements 
421: %are modeled by a set of projectors $P_{\mu}$ as $P_{\mu} |\phi \rangle 
422: %\langle \phi | P_{\mu} / \langle \phi | P_{\mu} | \phi \rangle$ 
423: %(satisfying $P_{\mu}=P_{\mu}^{\dagger}$, $P_{\mu} P_{\nu}= \delta_{\mu \nu} 
424: %P_{\mu}$ and $\sum_{\mu} P_{\mu} = \openone$). The description of 
425: %the state of a given system after measurements whose results 
426: %are unknown is embodied by a density 
427: %matrix $\rho'=\sum_{\mu} P_{\mu} \rho P_{\mu}$.
428: %But in the case of a system that is measured through or together with 
429: %its environment (indirect measurement) the analog of this 
430: %last formula is different. It is written in terms 
431: %of the positive operators set mentioned before. The probability of a 
432: %given outcome 
433: %$\sqrt{E_{\mu}} \rho \sqrt{E_{\mu}}^{\dagger}$ 
434: %($\sqrt{E_{\mu}} \equiv M_{\mu}$) becomes 
435: %${\rm Tr}(\rho E_{\mu})$ now.
436: %
437: %Different sets of classical probabilities and ensembles of states 
438: %can give the same density matrix. Then, for a given system we find 
439: %different ensemble decompositions, and as a consequence there is a 
440: %unitary freedom in the operator sum representation \cite{Chuangbook}. 
441: %Two sets of operation elements 
442: %(two quantum operations) are equivalent if and only if there is a unitary 
443: %matrix that transforms one into the other.
444: %This corresponds to 
445: %performing indirect measurements  in 
446: %another environmental basis, like for instance
447: %\begin{equation}
448: %\$(\rho)=\sum_{\nu} N_{\nu} \rho N_{\nu}^{\dag}, 
449: %\end{equation}
450: %where $N_{\nu}=\sum_\mu U_{\nu \mu} M_{\mu}$, with
451: %$U$ unitary matrix. 
452: 
453: We would like to mention that the quantum operations formalism 
454: is more general than the master equation approach. 
455: The operator sum formulation in differential form can 
456: be obtained from the master equation, but a general quantum process 
457: described in terms of an operator sum representation needs to 
458: be Markovian in order to be tractable with a master equation. 
459: This opens the possibility of studying a wide range of 
460: non Markovian phenomena using quantum trajectories simulations. 
461: 
462: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
463: 
464: \section{Noise channels}
465: \label{sec:noise}
466: 
467: There are several ways to model the interaction of a system 
468: with the environment. The most common examples found in the literature 
469: are the amplitude damping channel, the phase flip channel,
470: and the depolarizing channel 
471: \cite{Chuangbook,Preskillbook}.
472: % and the so-called arbitrary noise 
473: %\cite{Oh}. 
474: In this section, we give a brief description of two relevant noise
475: models (amplitude damping and phase damping) in the single qubit 
476: case, and then we generalize them to $n$-qubit systems.
477: 
478: Dissipation (energy loss) is one of the main features present 
479: in open quantum systems. Different phenomena like, for example, 
480: spontaneous atomic emission of a photon or spin systems approaching 
481: equilibrium can be modeled by this quantum operation, i.e., 
482: the amplitude damping. 
483: Considering the environment initially in the vacuum state 
484: $|0\rangle_e$, there is a probability $p$ that the excited state 
485: of the system decays and that the state of the environment 
486: changes from the vacuum to $|1\rangle_e$.  
487: \begin{eqnarray}
488: |0\rangle_s |0\rangle_e & \rightarrow & |0\rangle_s |0\rangle_e, 
489: \nonumber \\ 
490: |1\rangle_s |0\rangle_e & \rightarrow & \sqrt{1-p} \; |1\rangle_s, 
491: |0\rangle_e
492: +\sqrt{p} \; |0\rangle_s |1\rangle_e,
493: \label{damp1}
494: \end{eqnarray}
495: where $|0\rangle_s$ stands for the spin down (ground) state 
496: and $|1\rangle_s$ for the spin up (excited) state of the system,
497: and the indexes $s$ and $e$ denote the quantum states of the 
498: system and of the environment, respectively.
499: Tracing out the environment the corresponding Kraus operators are 
500: obtained: 
501: \begin{equation}
502: M_0= \left(
503: \begin{array}{cc}
504: 1 & 0 \\
505: 0 & \sqrt{1-p} \\
506: \end{array} \right), \quad
507: M_1 = \left(
508: \begin{array}{cc}
509: 0 & \sqrt{p} \\
510: 0 & 0 \\
511: \end{array} \right).
512: \label{Kraus1}
513: \end{equation}
514: The operator $M_1$ is responsible for the  
515: quantum jumps and $M_0$ for the continuous evolution. 
516: For the case of infinitesimal evolution operators  
517: (see Sec.~\ref{sec:QT}), a 
518: repeated application of this noise channel 
519: gives an exponential decay law
520: of the population of the state $|1\rangle$ \cite{Preskillbook}. 
521: Therefore, this evolution drives any intial (pure or mixed) 
522: state of the qubit to the pure state $|0\rangle$. 
523: Note that here and in the following the matrix representations 
524: of the single-qubit Kraus operators are written in the 
525: $\{|0\rangle,|1\rangle\}$ basis.
526: 
527: We also consider the phase flip channel (which is equivalent to 
528: the phase damping channel \cite{Chuangbook}) 
529: given by the following model
530: \begin{eqnarray}
531: |0\rangle_s |0\rangle_e & \rightarrow & \sqrt{1-p} \; |0\rangle_s 
532: |0\rangle_e + \sqrt{p} \; |0\rangle_s |1\rangle_e, \nonumber \\ 
533: |1\rangle_s |0\rangle_e & \rightarrow & \sqrt{1-p} \; |1\rangle_s 
534: |0\rangle_e - \sqrt{p} \; |1\rangle_s |1\rangle_e,
535: \end{eqnarray}
536: with Kraus operators
537: \begin{equation}
538: M_0=\sqrt{1-p} \; \left(
539: \begin{array}{cc}
540: 1 & 0 \\
541: 0 & 1 \\
542: \end{array} \right), \quad
543: M_1 = \sqrt{p} \; 
544: \left(
545: \begin{array}{cc}
546: 1 & 0 \\
547: 0 & -1 \\
548: \end{array} \right).
549: \label{Kraus2}
550: \end{equation}
551: This kind of noise can be thought as describing quantum 
552: information loss, in contrast with the 
553: previous model, which describes energy loss. 
554: This is a purely quantum mechanical process 
555: and has been extensively studied in the context of 
556: quantum to classical correspondence \cite{Zurek}.
557: 
558: In order to generalize the single qubit amplitude damping 
559: process to many qubits  we will follow two different 
560: points of view. 
561: In the first case we assume that a single damping probability 
562: describes the action of the environment, irrespective 
563: of the internal many-body state of the system. 
564: In the second approach, we assume that each qubit has 
565: its own interaction with the environment, independently of 
566: the other qubits. This makes the damping probability 
567: grow with the number of qubits that can perform the transition
568: $|1\rangle\to |0\rangle$.
569: Both models assume that only one qubit of the system can 
570: decay at a time. 
571: 
572: In the first case \cite{Car}, we have a probability $p$ 
573: for the system to perform one of the possible 
574: transitions, each of them being equally likely. 
575: This can be illustrated with a two-qubit example:
576: \begin{eqnarray}
577: |00\rangle_s |00 \rangle_e & \rightarrow & 
578: |00\rangle_s |00 \rangle_e, \nonumber \\
579: |01\rangle_s |00 \rangle_e & \rightarrow & 
580: \sqrt{1-p} \: |01\rangle_s |00 \rangle_e + 
581: \sqrt{p} \: |00\rangle_s |01\rangle_e, \nonumber \\
582: |10\rangle_s |00 \rangle_e & \rightarrow & 
583: \sqrt{1-p} \: |10\rangle_s |00 \rangle_e +
584: \sqrt{p} \: |00\rangle_s |10\rangle_e, \nonumber \\
585: |11\rangle_s |00 \rangle_e & \rightarrow & 
586: \sqrt{1-p} \: |11\rangle_s |00 \rangle_e +
587: \sqrt{p/2} \nonumber \\
588:  & & (|10\rangle_s |01\rangle_e + |01\rangle_s |10\rangle_e).
589: \end{eqnarray} 
590: 
591: A $n$-qubit state $|i_{n-1}\ldots i_0\rangle$ 
592: ($i_l=0,1$, with $0\le j \le n-1$) decays in the interval $dt$ 
593: with a probability $p=\Gamma dt /\hbar$. 
594: After this infinitesimal time, the possible states of the system 
595: are those in which the damping $|1\rangle \to |0\rangle$ has 
596: occurred in one of the qubits, the damping probability being the same 
597: for all the qubits.
598: We provide a compact expression for 
599: the Kraus operators $M_{\mu}$ in our model, following the formulation 
600: given in Sec.~\ref{sec:QT:qops} for the infinitesimal evolution. 
601: The matrix elements for the $k$th operator 
602: in the computational basis 
603: $|i\rangle \equiv |i_{n-1}\ldots i_0\rangle$, with 
604: $i\equiv\sum_{l=0}^{n-1} i_l 2^l$, are given by   
605: \begin{equation}
606: [M_{\mu}^{(n)}]_{i,j} = \left\{
607: \begin{array}{ll}
608:  \sqrt{\frac{\Gamma dt}{\hbar \sum_{l=0}^{n-1} i_l}}, & 
609: \;{\rm for}\; j\geq 2^{(\mu-1)}, \\
610:  &  i=j-2^{(\mu-1)}, \\
611:  &  i_l=0,1, \\ 
612:  & \\
613:  0  & \;{\rm otherwise},
614: \end{array} \right.
615: \label{amp1}
616: \end{equation}
617: where the superscript ${}^{(n)}$ in $M_{\mu}^{(n)}$ 
618: underlines the fact that we are dealing with $n$-qubit
619: Kraus operators.
620: There are $n$ operators $M_{\mu}^{(n)}$ ($\mu=1,\ldots,n$), 
621: where the index $\mu$ singles out
622: which qubit undergoes the transition $|1\rangle \to |0\rangle$.
623: The $M_0^{(n)}$ operator is
624: \begin{equation}
625: [M_{0}^{(n)}]_{i,j} = \left\{
626: \begin{array}{ll}
627: 1 & \; {\rm for}\; i=j=0, \\
628:  \sqrt{1-\frac{\Gamma dt}{\hbar}} & 
629: \;{\rm for}\; i=j \ne 0, \\ 
630:  & \\
631:  0  & \;{\rm otherwise}. 
632: \end{array} \right.
633: \end{equation}
634: Note that, if we replace in the above definition the square root by its 
635: first order approximation, we arrive at the same expression for 
636: the action of the effective Hamiltonian given 
637: in Sec.~\ref{sec:QT} (see the first term in the right hand side 
638: of \equ~(\ref{eq:stat}) and \equ~(\ref{eq:stat2})).
639: 
640: For example, starting from the four-qubit pure state
641: $\rho(t_0)=|1011\rangle\langle 1011|$, the action of the generalized 
642: amplitude damping channel leads, after a time $dt$, to 
643: the statistical mixture 
644: $$
645: \rho(t_0+dt)=\left(1-\frac{\Gamma dt}{\hbar}\right)
646: |1011\rangle\langle 1011|+
647: \frac{\Gamma dt}{3\hbar}(|0011\rangle\langle 0011|
648: $$
649: \begin{equation}
650: +|1001\rangle\langle 1001|+|1010\rangle\langle 1010|).
651: \label{ampex}
652: \end{equation}
653: 
654: This situation can be described as {\em branching process} or 
655: a {\em cascade} in the population 
656: of different classes of states (see, e.g., Ref.~\cite{Flam}). 
657: A class in this model is naturally defined as the collection of 
658: all the states of the system having the same number of
659: qubits in the ``up'' state (i.e., of $|1\rangle$ states).  
660: The first class is the one initially populated with probability $W_0$.
661: The second class, with associated probability $W_1$, corresponds to the 
662: states that are obtained from the initial state after the damping of 
663: one qubit. The third class, with probability $W_2$, includes 
664: the states that can be reached by the damping of one qubit 
665: in any of the states of the second class, and so on until reaching the 
666: ``ground state'' $|0\ldots 0\rangle$ of the system. 
667: This last state is also the only state that resides in  
668: the last class. This class is populated with probability 
669: $W_m$, the number $m$ being the maximum number of up spins in the 
670: initial state (if we start from a state 
671: $|i_{n-1}\ldots i_0\rangle$ of the computational basis, 
672: then $m=\sum_{l=0}^{n-1} i_l$). Note that $m\le n$. 
673: The set of probabilities $W_m$ 
674: can be obtained by means of the following differential equations:
675: \begin{eqnarray}
676: \frac{dW_0}{dt} & = & -\frac{\Gamma}{\hbar} \; W_0, \nonumber \\
677:  & \vdots & \nonumber \\
678: \frac{dW_k}{dt} & = & \frac{\Gamma}{\hbar} \; (W_{k-1} -  W_k), 
679: \nonumber \\
680:  & \vdots & \nonumber \\
681: \frac{dW_m}{dt} & = & \frac{\Gamma}{\hbar} \; W_{m-1},
682: \end{eqnarray}  
683: with solutions
684: \begin{eqnarray}
685: %W_0 & = & \exp{(-\Gamma t /\hbar)} \nonumber \\
686: W_k & = & \frac{(\Gamma t /\hbar)^k}{k!} \; 
687: \exp{\left(-\frac{\Gamma t}{\hbar}\right)},
688: \nonumber \\
689: W_m & = & 1 - \sum_{k=0}^{m-1} W_k.
690: \label{ADch0}
691: \end{eqnarray}
692: In Fig.~\ref{fig:cascade} we show a comparison among the probabilities 
693: for each class obtained with quantum trajectories simulations and 
694: the theoretical formulae, in the case of $n=6$.
695: We have evolved the initial state $|2^n-1\rangle=|1\ldots 1\rangle$ 
696: up to time $2n \; dt$, for different values of the
697: dimensionless damping rate 
698: $\gamma=\Gamma dt/\hbar$. 
699: This evolution is purely   
700: dissipative, without considering any other kind of system's  
701: dynamics. As can be seen from Fig.~\ref{fig:cascade}, 
702: we have a very good agreement between quantum 
703: trajectories numerical simulations and the theoretical 
704: predictions of the model. 
705: 
706: %\begin{figure}[ht]
707: \begin{figure}
708: \includegraphics[width=8.0cm,angle=0]{fig1.eps}
709: \caption[]{\footnotesize Probabilities $W_k$ for each 
710: class of states in terms of the dimensionless damping rate 
711: $\gamma=\Gamma dt /\hbar$, 
712: in a system with $n=6$ qubits, obtained after evolving the initial
713: state $|1\ldots 1\rangle$ up to time $2n dt$, under the noise 
714: model (\ref{amp1}).
715: Solid lines correspond to the exact formulae (\ref{ADch0}), 
716: and the numbers close to each curve indicate the corresponding 
717: class. Diamonds 
718: stand for quantum trajectories simulations  
719: (error bars are not shown since they are smaller than the
720: size of the symbols).}
721: \label{fig:cascade}
722: \end{figure}
723: 
724: The other generalization for the amplitude damping 
725: differs from the 
726: previous one in that the decay probability is now the 
727: same for each single qubit process, independently of 
728: the state of the system. Then, the decay probability 
729: for a states of the computational basis is 
730: proportional to the number of qubits in the up state.
731: This can be illustrated in the two-qubit 
732: case. In this example, the noise channel is described 
733: by the following unitary evolution formulae:
734: \begin{eqnarray}
735: |00\rangle_s |00 \rangle_e & \rightarrow & 
736: |00\rangle_s |00 \rangle_e, \nonumber \\
737: |01\rangle_s |00 \rangle_e & \rightarrow & 
738: \sqrt{1-p} \: |01\rangle_s |00 \rangle_e + 
739: \sqrt{p} \: |00\rangle_s |01\rangle_e, \nonumber \\
740: |10\rangle_s |00 \rangle_e & \rightarrow & 
741: \sqrt{1-p} \: |10\rangle_s |00 \rangle_e +
742: \sqrt{p} \: |00\rangle_s |10\rangle_e, \nonumber \\
743: |11\rangle_s |00 \rangle_e & \rightarrow & 
744: \sqrt{1-2p} \: |11\rangle_s |00 \rangle_e +
745: \sqrt{p} \nonumber \\
746:  & & (|10\rangle_s |01\rangle_e + |01\rangle_s |10\rangle_e).
747: \end{eqnarray} 
748: The Kraus operators $M_{\mu}^{(n)}$ for this model are given 
749: by the $n$-factor tensor product
750: \begin{equation}
751: M_{\mu}^{(n)}=\openone \otimes \ldots \otimes M_1 
752: \otimes \ldots \otimes \openone,
753: \label{KrausOps2}
754: \end{equation}
755: where $M_1$ is given by Eq.~(\ref{Kraus1})
756: and $\mu$ ($\mu=1,\ldots,n$) coincides with the 
757: position of $M_1$ in (\ref{KrausOps2}), that is $\mu$
758: singles out the qubit which decays from $|1\rangle$ to
759: $|0\rangle$. 
760: In the computational basis, the matrix representation
761: of the operator $M_0^{(n)}$ is given by 
762: \begin{equation}
763: [M_{0}^{(n)}]_{i,j} = \left\{
764: \begin{array}{ll}
765: 1 & \; {\rm for}\; i=j=0, \\
766:  \sqrt{1-\sum_{l=0}^{n-1} i_l \frac{\Gamma dt}{\hbar}} & 
767: \;{\rm for}\; i=j \ne 0, \\ 
768:  & \\
769:  0  & \;{\rm otherwise}. 
770: \end{array} \right.
771: \end{equation}
772: The evolution of the pure state 
773: $\rho(t_0)=|1011\rangle\langle 1011|$ is different 
774: from what obtained in the previous many-qubit
775: damping model (see Eq.~(\ref{ampex}). We have
776: $$
777: \rho(t_0+dt)=\left(1-\frac{3\Gamma dt}{\hbar}\right)
778: |1011\rangle\langle 1011|+
779: \frac{\Gamma dt}{\hbar}(|0011\rangle\langle 0011|
780: $$
781: \begin{equation}
782: +|1001\rangle\langle 1001|+|1010\rangle\langle 1010|).
783: \end{equation}
784: 
785: The cascade in the population of the different classes
786: $W_k$ is ruled by the following set of differential 
787: equations:
788: \begin{eqnarray}
789: \frac{dW_0}{dt} & = & -\frac{\Gamma}{\hbar} \; n_0 \; W_0, \nonumber \\
790:  & \vdots & \nonumber \\
791: \frac{dW_k}{dt} & = & \frac{\Gamma}{\hbar} \; (n_{k-1} \; W_{k-1} -  
792: n_k \; W_k), 
793: \nonumber \\
794:  & \vdots & \nonumber \\
795: \frac{dW_m}{dt} & = & \frac{\Gamma}{\hbar} \; n_{m-1} \; W_{m-1}
796: =\frac{\Gamma}{\hbar}{W_{m-1}},
797: \end{eqnarray}  
798: where $n_k$ is the number of qubits in the ``up'' state 
799: for the class $k$. Taking into account that $n_k=n_0-k=m-k$, the 
800: solutions are 
801: \begin{eqnarray}
802: W_k & = & \frac{n!}{n_k!} \; \sum_{i=0}^{k} 
803: \frac{(-1)^{(k-i)}}{i! \:(k-i)!} 
804: \; \exp{\left(-\frac{n_i \: \Gamma t}{\hbar}\right)},
805: \nonumber \\
806: W_m & = & 1 - \sum_{k=0}^{m-1} W_k.
807: \label{ADch1}
808: \end{eqnarray} 
809: In Fig.~\ref{fig:cascadech1}, we take the initial state 
810: $|2^n-1\rangle$ evolving it for a time $n \; dt$, for different 
811: values of the dimensionless damping rate $\gamma$. Again, only 
812: dissipation is considered and we have found a very good  
813: agreement with our analytical predictions. 
814: 
815: %\begin{figure}[ht]
816: \begin{figure}
817: \includegraphics[width=8.5cm,angle=0]{fig2.eps}
818: \caption[]{\footnotesize Same as in Fig.~\ref{fig:cascade}, 
819: but for the noise model (\ref{KrausOps2}) and 
820: after evolution up to time $n dt$.} 
821: \label{fig:cascadech1}
822: \end{figure}
823: 
824: With this same idea we generalize the phase flip channel. 
825: The Kraus operators are given 
826: by the $n$-factor tensor product of \equ~(\ref{KrausOps2}), 
827: where now $\mu$ ($\mu=1,\ldots,n$) coincides with the 
828: position of the matrix $M_1$ of \equ~(\ref{Kraus2}), instead 
829: of \equ~(\ref{Kraus1}). 
830: The matrix representation of the Kraus operator $M_0^{(n)}$ in
831: the computational basis is
832: \begin{equation}
833: [M_{0}^{(n)}]_{i,j} = \left\{
834: \begin{array}{ll}
835:  \sqrt{1- n \frac{\Gamma dt}{\hbar}} & 
836: \;{\rm for}\; i=j, \\ 
837:  & \\
838:  0  & \;{\rm otherwise}. 
839: \end{array} \right.
840: \end{equation}
841: We study the stability of a given state vector subjected
842: to this decoherence channel. Namely, we compute the fidelity
843: $F={\rm Tr}[\rho_0 \rho(t)]$ of a $n$-qubit system,
844: for a random initial state $|\psi_0\rangle$, as a function 
845: of $\gamma=\Gamma dt/\hbar$ ($\rho_0=|\psi_0\rangle\langle\psi_0|$ 
846: is the density matrix for the initial state and 
847: $\rho(t)$ is the density matrix of the system at time $t$).
848: All components of the initial random state 
849: vector $|\psi_0\rangle$ are taken to be random complex numbers 
850: of modulus $1/\sqrt{2^n}$. 
851: A combinatorial calculation provides an exact closed expression for 
852: the fidelity in this case. We obtain  
853: \begin{equation}
854: F(t)=\frac{1}{2^n}+ \frac{n!}{2^n} \: \sum_{i=1}^n  
855: \frac{1}{i! \: (n-i)!} 
856: \: \exp{\left(-\frac{2 i \: \Gamma t}{\hbar}\right)}.
857: \label{eq:noisefidelity}
858: \end{equation}
859: Note that this formula does not give a simple exponential 
860: fidelity decay, but a superposition of exponential decays
861: with different rates, since the decay rate depends on the 
862: number of up spins in the different states of the 
863: computational basis.
864: %as the decay rate $\gamma_i$ associated to a computational 
865: %basis state with $i$ spins up is given by $\gamma_i=2i\Gamma/\hbar$.
866: 
867: The agreement of the quantum trajectories simulations with this
868: theoretical formula is shown in Fig.~\ref{fig:fidelitych4}.
869: The semi log version of the theoretical curve is shown in the inset,
870: where we plot $F-F_\infty$, $F_\infty=1/2^n$ being the asymptotic
871: value of fidelity at $t=\infty$ or at $\Gamma=\infty$. 
872: We note that the asymptotic decay of $\bar F=F-F_{\infty}$
873: takes place with the lowest decay rate $\Gamma_1=2\Gamma/\hbar$.
874: of \equ~(\ref{eq:noisefidelity}).
875: 
876: %\begin{figure}[ht]
877: \begin{figure}
878: \includegraphics[width=8.5cm,angle=0]{fig3.eps}
879: \caption[]{\footnotesize Fidelity of a random initial 
880: state in terms of the dimensionless decay rate 
881: $\gamma=\Gamma dt/\hbar$, 
882: in a system with $n=6$ qubits, subjected to the generalized phase
883: flip channel up to time $2 n^2 dt$. 
884: The very good agreement between quantum 
885: trajectories simulations (circles with error bars) and 
886: theoretical predictions (solid line) is clearly seen. 
887: Here and in the following figures the error bars give the
888: size of the statistical error.
889: The inset shows the semi log version of the theoretical curve 
890: plot for the modified fidelity $\bar F=F-F_{\infty}$.}
891: \label{fig:fidelitych4}
892: \end{figure}
893: 
894: We note that analytical formulae for fidelity decay can also be 
895: found for other special initial conditions.
896: For instance, \equ~(\ref{eq:noisefidelity}) remains valid
897: when the initial state is a random superposition of computational
898: basis states with up to $n_1$ spins up, provided that we replace
899: $n$ with $n_1$ in this equation.
900: 
901: In the following, we will apply the amplitude damping models 
902: to study the fidelity of a generalized teleportation protocol 
903: and the phase flip channel for the case of a quantum 
904: computer implementation of the baker's map.
905: 
906: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
907: 									  
908: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
909: \section{Quantum teleportation}
910: \label{sec:tele}
911: 
912: Recently, there have been several publications 
913: focused on the investigation of fidelity of teleportation in the presence 
914: of a noisy environment \cite{Badziag,Bandyo,Oh,Verstraete}.
915: We study this problem in the situation presented in 
916: \cite{Car}, where a model of quantum teleportation through a 
917: noisy chain of qubits has been used. 
918: A schematic drawing of this quantum protocol is shown 
919: in Fig.~\ref{fig:scheme}.
920: 
921: %\begin{figure}[ht]
922: \begin{figure}
923: \includegraphics[width=8.5cm,angle=0]{fig4.eps}
924: \caption[]{\footnotesize Schematic drawing of the teleportation 
925: procedure that we study in the text. Alice sends one of the 
926: qubits of her pure Bell state to Bob. Meanwhile there is 
927: dissipation induced by the environment. 
928: In the figure, there is a chain of $n=9$ qubits and 
929: the third of the $n-2$ swap gates required by this quantum
930: protocol has been applied. Qubit $|\psi\rangle$ 
931: has to be teleported.}
932: \label{fig:scheme}
933: \end{figure}
934: 
935: Let us first recall this protocol in the ideal case, 
936: without environmental effects. 
937: We consider a chain of $n$ qubits, and assume that Alice 
938: can access the qubits located at one end of the chain, 
939: Bob those at the other end. Initially 
940: Alice owns an EPR pair (for instance we take the Bell 
941: state $|\phi^+\rangle = 
942: (|00\rangle+|11\rangle)/\sqrt{2}$), while the remaining 
943: $n-2$ qubits are in a pure state. Thus, the 
944: global initial state of the chain is given by  
945: \begin{equation}
946: \sum_{i_{n-1},\ldots,i_2} 
947: c_{i_{n-1},\ldots,i_2} 
948: |i_{n-1}\ldots i_2\rangle
949: \otimes \frac{1}{\sqrt{2}} (|00\rangle + |11\rangle),  
950: \label{phiin}
951: \end{equation}
952: where $i_k=0,1$ denotes the up or down state of the qubit $k$.
953: In order to deliver one of the qubits of the EPR pair to Bob,
954: we implement a protocol consisting of $n-2$ swap gates 
955: that exchange the states of pairs of qubits:  
956: $$
957: \sum_{i_{n-1},\ldots, i_2}
958: \frac{c_{i_{n-1},\ldots,i_2}}{\sqrt{2}} 
959: (|i_{n-1} \ldots i_2 0 0 \rangle + 
960: |i_{n-1} \ldots i_2 1 1 \rangle) 
961: $$
962: $$
963: \rightarrow
964: \sum_{i_{n-1},\ldots, i_2}
965: \frac{c_{i_{n-1},\ldots,i_2}}{\sqrt{2}} 
966: (|i_{n-1} \ldots 0 i_2 0 \rangle + 
967: |i_{n-1} \ldots 1 i_2 1 \rangle) 
968: \rightarrow 
969: $$
970: \begin{equation}
971: ...\rightarrow 
972: \sum_{i_{n-1},\ldots, i_2}
973: \frac{c_{i_{n-1},\ldots,i_2}}{\sqrt{2}} 
974: (|0 i_{n-1} \ldots i_2 0 \rangle + 
975: |1 i_{n-1} \ldots i_2 1 \rangle). 
976: \end{equation}
977: After that, Alice and Bob share an EPR pair, and therefore 
978: an unknown state of a qubit ($|\psi\rangle=a|0\rangle +
979: b|1\rangle$) can be transferred from Alice to Bob by 
980: means of the standard teleportation protocol 
981: \cite{BennettBrassard}. 
982: In this work, we take random coefficients 
983: $c_{i_{n-1},\ldots,i_2}$, that is they
984: have amplitudes of the order of $1/\sqrt{2^{n-2}}$ (to assure 
985: wave function normalization) and random phases. This ergodic 
986: hypothesis models the transmission of a qubit through a chaotic 
987: quantum channel.  
988: 
989: We assume that our quantum protocol is implemented
990: by a sequence of instantaneous and perfect swap gates, 
991: separated by a time interval $\tau$, during which the 
992: corresponding noise channels introduce errors. 
993: This means that, using quantum trajectories, 
994: dissipation is implemented by means of  
995: ``infinitesimal'' Kraus operators (see Sec.~\ref{sec:noise}), 
996: following the quantum jump numerical procedure outlined 
997: in Sec.~\ref{sec:QT:meqt}. 
998: We also assume that the only effect of the system's 
999: Hamiltonian $H_s$ is to generate the swap gates.
1000: 
1001: Let us call $\rho_k^{(n)}$ the density matrix of the whole chain
1002: of $n$ qubits after $k$ swap gates.
1003: Since the evolution of the density matrix in a time step
1004: $dt$ is given, in the Kraus representation, by   
1005: \equ~(\ref{eq:infsolution}), we can write the evolution
1006: from $\rho_{k-1}^{(n)}$ to $\rho_k^{(n)}$ as follows: 
1007: \begin{eqnarray}
1008: \rho_k^{(n)}=U_{\rm sw}^{k,k+1} \: 
1009: \left[ \sum_{\mu_1,...,\mu_k=0}^{\cal M} 
1010: M_{\mu_k}^{(n)}(dt)\cdots M_{\mu_1}^{(n)}(dt) 
1011: \right. 
1012: \nonumber
1013: \\ 
1014: \rho_{k-1}^{(n)} \: 
1015: \left.
1016: (M_{\mu_1}^{(n)})^\dagger (dt) \cdots 
1017: (M_{\mu_k}^{(n)})^\dagger (dt) 
1018: \right]
1019:  \: {U_{\rm sw}^{k,k+1}}^{\dag},
1020: \end{eqnarray}
1021: where $U_{\rm sw}^{ij}$ is the swap operator that exchanges 
1022: the states of the qubits $i$ and $j$ and
1023: $k=\tau/dt$ is the number of quantum noise
1024: operations between two consecutive swap gates.
1025: After $n-2$ swap gates, we obtain the final state 
1026: $\rho_f^{(n)}=\rho_{n-2}^{(n)}$.
1027: 
1028: After computing the evolution of the initial 
1029: state of the chain of $n$ qubits up to time $\Delta t= (n-2) \tau$, 
1030: the standard teleportation protocol is implemented 
1031: \cite{BennettBrassard}. The fidelity of teleportation 
1032: is defined by 
1033: \begin{equation}
1034: F=\langle\psi|\rho_B|\psi\rangle,
1035: \end{equation}
1036: where $|\psi\rangle=a|0\rangle +b|1\rangle$ is the 
1037: state to be teleported, and $\rho_B$ is the density matrix
1038: of Bob's qubit at the end of the teleportation protocol, 
1039: obtained after tracing over all the other qubits of the chain:
1040: $\rho_B={\rm Tr}_{0,...,n-2} [\rho_f^{(n)}]$. 
1041: 
1042: In the quantum trajectories method, we compute the fidelity as 
1043: \begin{equation}
1044: F = \lim_{{\cal N}\to \infty}
1045: \frac{1}{\cal N} 
1046: \sum_{i=1}^{\cal N} \langle \psi | (\rho_B)_i 
1047: | \psi \rangle,
1048: \end{equation}
1049: where $(\rho_B)_i$ is the reduced density matrix of 
1050: Bob's qubit, obtained from the wave vector of the trajectory 
1051: $i$ at the end of the quantum protocol. 
1052: If the final state of the chain is 
1053: $|\phi \rangle = \sum_{j=0}^{2^{n}-1} \alpha_j |j\rangle$ 
1054: (an arbitrary state), the state of the 
1055: whole system is 
1056: \begin{eqnarray}
1057: |\phi \rangle |\psi\rangle = & \frac{1}{\sqrt{2}} 
1058: \sum_{j'=0}^{2^{n-1}-1} 
1059: [(a \alpha_{2j'}+b \alpha_{2j'+1}) |j'\rangle |\phi^+ \rangle 
1060: \nonumber \\
1061:  & + (a \alpha_{2j'}-b \alpha_{2j'+1}) |j'\rangle |\phi^- \rangle .
1062: \nonumber \\
1063:  & + (b \alpha_{2j'}+a \alpha_{2j'+1}) |j'\rangle |\psi^+ \rangle  
1064: \nonumber  \\
1065:  & + (b \alpha_{2j'}-a \alpha_{2j'+1}) |j'\rangle |\psi^- \rangle],  
1066: \end{eqnarray}
1067: In the previous expression we have used the four (maximally 
1068: entangled) Bell states 
1069: ($|\phi^\pm\rangle = (|00\rangle \pm |11\rangle)/\sqrt{2}$ and
1070: $|\psi^\pm\rangle = (|01\rangle \pm |10\rangle)\sqrt{2}$), 
1071: corresponding to the two least significant bits 
1072: (the first bit of the chain and $|\psi\rangle$). 
1073: Then, as required by the teleportation protocol, 
1074: we perform a Bell measurement whose result determines one
1075: out of possible unitary transformations acting on Bob's qubit.
1076: Owing to quantum noise, Bob's qubit is entangled with 
1077: the rest of the chain. Therefore, we must trace over these
1078: qubits to obtain the reduced density matrix $\rho_B$ 
1079: describing the state of Bob's qubit. 
1080: If we measure $|\phi^+\rangle$ for instance, and define 
1081: $\tilde \alpha_{j'}=1/\sqrt{2} \: (a \; \alpha_{2j'}+b \;
1082: \alpha_{2j'+1})$, we arrive at the following expression:
1083: \begin{equation}
1084: \rho_{\rm B}= C
1085: \left(
1086: \begin{array}{ll}
1087: \sum_{j'=0}^{D-1} | \tilde \alpha_{j'} |^2 & 
1088: \sum_{j'=0}^{D-1} \tilde \alpha_{j'} 
1089: \tilde \alpha_{j'+D}^* \\
1090: \sum_{j'=0}^{D-1} \tilde \alpha_{j'}^* 
1091: \tilde \alpha_{j'+D} &
1092: \sum_{j'=D}^{2D-1} | \tilde \alpha_{j'} |^2 
1093: \end{array}
1094: \right),
1095: \end{equation}
1096: where $D=2^{n-2}$ and $C=1/\sum_{j'=0}^{2D-1} |\tilde \alpha_{j'}|^2$
1097: is a normalization constant.
1098: Then, the fidelity becomes 
1099: \begin{eqnarray}
1100: F= & C \big( |a|^2 \sum_{j'=0}^{D-1}|\tilde \alpha_{j'}|^2 + 
1101: |b|^2 \sum_{j'=D}^{2D-1} |\tilde \alpha_{j'}|^2 +
1102: \nonumber \\ 
1103:  & + \sum_{j'=0}^{D-1} 2 \real{\tilde \alpha_{j'} \tilde 
1104: \alpha_{j'+D}^* a^* b} \big).
1105: \end{eqnarray}
1106: In the special case $a=b=1/\sqrt{2}$, it reduces to
1107: \begin{equation}
1108: F= \frac{1}{2}+ C \sum_{j'=0}^{D-1} 
1109: \real{\tilde \alpha_{j'} \tilde 
1110: \alpha_{j'+D}^*} .
1111: \label{fidteleportation}
1112: \end{equation}
1113: Similar expressions are obtained when the Bell measurement
1114: gives outcomes $|\phi^{-}\rangle$, $|\psi^{+}\rangle$, or 
1115: $|\psi^{-}\rangle$.
1116: 
1117: Let us now briefly discuss the case in which we 
1118: directly evolve the density matrix 
1119: of the whole system by means of the master 
1120: equation (\ref{eq:lin2}). After the sequence of 
1121: swap gates and quantum noise operations, we 
1122: obtain the final density matrix $\rho_f^{(n)}$
1123: for the $n$-qubit chain. 
1124: The Bell measurement is performed by the operator
1125: \begin{equation}
1126: \frac{M_{\phi^+} (\rho_f^{(n)} \otimes \rho_{\psi}) 
1127: M_{\phi^+}^{\dag}}
1128: {{\rm Tr}[M_{\phi^+} (\rho_f^{(n)} \otimes \rho_{\psi})
1129: M_{\phi^+}^{\dag}]},
1130: \end{equation} 
1131: where 
1132: $\rho_\psi=|\psi\rangle\langle\psi|$ is the density matrix
1133: describing the state of the qubit to be teleported and 
1134: $M_{\phi^+}=\openone^{(n-1)} \otimes 
1135: |\phi^+\rangle \langle\phi^+|$ 
1136: ($\openone^{(n-1)}$ is the identity for the remaining 
1137: $n-1$ qubits of the chain).
1138: Then, we take the partial trace and obtain the fidelity
1139: $F={\rm Tr}(\rho_{\rm B} \rho_{\psi})=
1140: \langle\psi |\rho_{\rm B} |\psi\rangle$. 
1141: It is straightforward to see that the above procedure is 
1142: equivalent to take the partial trace (over the mediating 
1143: qubits of the chain) first and then proceed with low 
1144: dimensionality calculations.
1145: 
1146: We have investigated the effect of the two amplitude damping 
1147: models described in Sec.~\ref{sec:noise} on the fidelity
1148: of quantum teleportation.
1149: In Fig.~\ref{telech01}, we show the results of our numerical
1150: simulations for the special case in which the state to be
1151: teleported is $|\psi\rangle=(|0\rangle+|1\rangle)/\sqrt{2}$.
1152: In this case, in the limit of infinite chain ($n\to\infty$) 
1153: or of large damping rate, 
1154: the density matrix $\rho_{\rm B}$ describing the state
1155: of Bob's qubit becomes $\rho_{\rm B}=|0\rangle\langle 0|$. 
1156: Thus, the asymptotic value of fidelity is given by 
1157: $F_\infty=1/2$ and we plot the values of 
1158: $\bar F=F-F_{\infty}$, for the noise models 
1159: (\ref{amp1}) and (\ref{KrausOps2}), corresponding to the inset
1160: and the main figure in Fig.~\ref{telech01}, respectively.
1161: For both cases we have checked the accuracy of the quantum trajectory 
1162: simulations by reproducing the results with direct density matrix 
1163: calculations. This was possible only up to $n=10$ qubits.
1164: The exponential decay of fidelity in the case in which quantum
1165: noise is modeled by Eq.~(\ref{KrausOps2}) is in agreement with the
1166: theoretical formula 
1167: \begin{equation}
1168: F=\frac{1}{2}+\frac{1}{2}\exp{\left(-\frac{\Gamma t}{\hbar}\right)}=
1169: \frac{1}{2}+\frac{1}{2}\exp(-\gamma k),
1170: \label{noise2q}
1171: \end{equation}
1172: where $\gamma=\Gamma \tau /\hbar$ is the dimensionless 
1173: damping rate and $k=t/\tau=n-2$ measures the time in 
1174: units of quantum (swap) gates.
1175: To derive this theoretical formula, we observe that this 
1176: quantum noise model does not generate entanglement 
1177: between the two qubits of the Bell pair 
1178: and the others qubit of the chain. Therefore, 
1179: it is sufficient to study the evolution of the Bell state 
1180: $|\phi^+\rangle\langle \phi^+|$ under the noise model 
1181: (\ref{KrausOps2}). This evolution takes place inside
1182: a two-qubit Hilbert space of dimension $4$ and its 
1183: exact solution is given by Eq.~(\ref{noise2q}).  
1184: On the contrary, the quantum noise model (\ref{amp1})
1185: entangles these two qubits with the rest of 
1186: the chain. In this case, the fidelity decay is not 
1187: exponential. Unfortunately, we could not provide an 
1188: analytical derivation of $F(t)$ in this case.
1189:  
1190: %\begin{figure}[ht]
1191: \begin{figure}
1192: \includegraphics[width=8.5cm,angle=0]{fig5.eps}
1193: \caption[]{Fidelity $\bar F=F-F_{\infty}$ ($F_{\infty}=1/2$) of 
1194: the teleportation of 
1195: the state $|\psi\rangle=(|0\rangle+|1\rangle)/\sqrt{2}$ as a 
1196: function of the dimensionless damping rate 
1197: $\gamma=\Gamma \tau/\hbar$, for the amplitude damping 
1198: model (\ref{KrausOps2}). Circles and squares are the results of 
1199: the quantum trajectories calculations for chains with $n=10$ and 
1200: $n=20$ qubits, respectively. Triangles give the results 
1201: of the density matrix calculations at $n=10$. 
1202: Straight lines correspond to the theoretical result
1203: of Eq.~(\ref{noise2q}). 
1204: Inset: the same but for the noise model (\ref{amp1}).}
1205: \label{telech01}
1206: \end{figure}
1207: 
1208: 
1209: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1210: 
1211: \section{The quantum baker's map}
1212: \label{sec:Baker}
1213: 
1214: The quantum algorithm for the simulation of the quantum baker's
1215: map has been proposed by Schack \cite{Sch} and recently 
1216: implemented by means of a three-qubit NMR-based quantum processor
1217: \cite{cory}, where the fidelity of the quantum computation of
1218: the baker's map in one map step was measured. 
1219: 
1220: The baker's transformation is one of the prototype models of classical and
1221: quantum chaos \cite{Bala}. It maps the unit square 
1222: $0\le{q,p}<1$ onto itself according to
1223: \begin{eqnarray}
1224: q_{k+1} & = & 2 q_k - [2 q_k],\nonumber \\
1225: p_{k+1} & = & (p_k+[2q_k])/2, 
1226: \label{bakermap}
1227: \end{eqnarray}
1228: where $[x]$ stands for the integer part of $x$ and the index $k$ 
1229: denotes the number of map iterations. The action of this map 
1230: corresponds to compressing the unit square in the $p$ direction and
1231: stretching it in the $q$ direction, then cutting it along the $p$
1232: direction, and finally stacking one piece on top of the other
1233: (similarly to the way a baker kneads dough). Note that map 
1234: (\ref{bakermap}) is area preserving.
1235: The baker's map is a paradigmatic model of classical chaos.
1236: Indeed, it exhibits sensitive dependence on initial conditions, 
1237: which is the distinctive feature of classical chaos: any small 
1238: error in determining the initial conditions is 
1239: amplified exponentially in time. In other
1240: words, two nearby trajectories separate exponentially, with a rate 
1241: given by the maximum Lyapunov exponent $\lambda=\ln 2$.
1242: 
1243: The baker's map can be quantized following \cite{Bala}. 
1244: We introduce the position ($q$) and momentum ($p$) operators, 
1245: and denote the eigenstates of these operators by $|q_j\rangle$ and 
1246: $|p_k\rangle$, respectively. The corresponding eigenvalues
1247: are given by $q_j=j/N$ and $p_k=k/N$, with $j,k=0,\dots,N-1$, $N$ 
1248: being the dimension of the Hilbert space. Note that, to fit $N$ 
1249: levels onto the unit square, we must set $2\pi \hbar= 1/N$. 
1250: Therefore, the effective Planck's constant of the 
1251: system is $\hbar_{\rm eff}\propto 1/N$.
1252: We take $N=2^n$, where $n$ is the number of qubits used to simulate 
1253: the quantum baker's map on a quantum computer. 
1254: Note that $\hbar_{\rm eff}$ drops exponentially with the
1255: number of qubits and therefore the semiclassical regime
1256: $\hbar_{\rm eff}\ll 1$ can be reached with a small number of qubits.
1257: The transformation between the position basis
1258: $\{|q_0\rangle,\dots,|q_{N-1}\rangle\}$ and the momentum basis
1259: $\{|p_0\rangle,\dots,|p_{N-1}\rangle\}$ is performed by means of
1260: a discrete Fourier transform $F_n$, defined by the matrix elements
1261: \begin{equation}
1262:   \langle q_k | F_n | q_j \rangle \equiv
1263:   \frac1{\sqrt{2^n}} \, \exp\!\left( \frac{2\pi ikj}{2^n} \right) .
1264:   \label{dftbaker}
1265: \end{equation}
1266: It can be shown \cite{Bala} that the quantized baker's map may 
1267: be defined by the transformation 
1268: \begin{equation}
1269:   |\psi_{k+1}\rangle =
1270:   B \, |\psi_k\rangle = 
1271:   F_n^{-1}
1272:   \left[
1273:     \begin{array}{cc}
1274:       F_{n-1} & 0 \\
1275:       0 & F_{n-1}
1276:     \end{array}
1277:   \right] 
1278:   |\psi_k\rangle,
1279: \label{baker1k}
1280: \end{equation}
1281: where $|\psi_k\rangle$ denotes the wave vector of the system
1282: after $k$ map steps,  the matrix elements are to be  
1283: understood relative to the position basis
1284: $\{|q_j\rangle\}$ and $F_{n-1}$ is the discrete Fourier transform,
1285: defined by Eq.~(\ref{dftbaker}).
1286: 
1287: Since the discrete Fourier transform can be calculated on 
1288: a quantum computer using $O(n^2)$ elementary gates
1289: (see, e.g., Ref.~\cite{Chuangbook}), the simulation of one 
1290: step of the baker's map requires $O(n^2=(\log N)^2)$ gates 
1291: \cite{Sch} (more precisely, we need $(n-1)^2$ controlled phase-shift
1292: gates and $2n-1$ Hadamard gates). Therefore, it is exponentially 
1293: faster than the best known classical computation, which is based on
1294: the fast Fourier transform and requires $O(N\log N)$
1295: gates.   
1296: 
1297: In this section, we investigate the fidelity of the quantum computation
1298: of the baker's map in the presence of quantum noise.
1299: We consider the phase flip noise channel, generalized to the 
1300: $n$-qubit case as discussed in Sec.~\ref{sec:noise}. We  
1301: take an initial state $|\psi_0\rangle$ with amplitudes of the 
1302: order of $1/\sqrt{2^n}$ and random phases. 
1303: We perform the forward evolution of the baker's map 
1304: up to time $k$ (this evolution is driven, in the noiseless case, 
1305: by the unitary operator $B^k$, with $B$ given in Eq.~\ref{baker1k}), 
1306: followed by the $k$-step backward evolution (represented by 
1307: the operator $(B^\dagger)^k$). Due to quantum noise, the initial state 
1308: $|\psi_0\rangle$ is not exactly recovered and the final state
1309: of the system is described by a density matrix $\rho_f$.
1310: The fidelity of quantum computation is given by 
1311: $F=\langle \psi_0 |\rho_f |\psi_0 \rangle$.
1312: 
1313: We are able to work out the following theoretical formula for 
1314: the decay of fidelity induced by phase flip noise:
1315: \begin{equation}
1316: F = \exp\left(-n\gamma N_g \right)=
1317: \exp(-2\gamma n^3 k),
1318: \label{eq:Bakerfidelity}
1319: \end{equation}
1320: where $\gamma=\Gamma \tau/\hbar$ is the dimensionless decay rate
1321: (again, $\tau$ denotes the time interval between elementary 
1322: quantum gates) and 
1323: $N_g=2 n^2 k$ is the total number of elementary quantum gates 
1324: required to implement the $k$ steps forward evolution of the 
1325: baker's map, followed by $k$ step backward.
1326: To derive this formula, we first of all note that 
1327: \equ~(\ref{eq:noisefidelity}), obtained in the absence of 
1328: any quantum gate operation, gives, at short times, 
1329: the exponential decay
1330: \begin{equation}
1331: F(t)= \exp(-\frac{<\Gamma>t}{\hbar}).
1332: \end{equation} 
1333: Here the decay rate $<\Gamma>$ is obtained after averaging
1334: all the decay rates that appear in Eq.~(\ref{eq:noisefidelity}):
1335: \begin{equation}
1336: <\Gamma> = \frac{n!}{2^n} \: \sum_{i=0}^n  
1337: \frac{1}{i! \: (n-i)!} 
1338: 2 i \Gamma = n \Gamma.
1339: \end{equation}
1340: The effect of the chaotic dynamics of the baker's map is to 
1341: induce a fast decay of correlations, so that any memory of the 
1342: initial state is rapidly lost and the fidelity decay 
1343: remains exponential also at long times.
1344: Therefore, the condition for the validity of 
1345: formula (\ref{eq:Bakerfidelity}) is that the randomization 
1346: process introduced by chaotic dynamics takes place in a time
1347: scale shorter than the time scale for fidelity decay.
1348: 
1349: We can determine from Eq.~(\ref{eq:Bakerfidelity}) the time 
1350: scale up to which a reliable quantum computation of the 
1351: baker's map evolution in the presence of the phase flip
1352: noise channel is possible even without quantum error
1353: correction. The time scale $k_f$ at which $F$ drops below 
1354: some constant $A$ (for instance, $A=0.9$) is given by 
1355: \begin{equation}
1356: k_f=-\frac{\ln A}{2\gamma n^3}.
1357: \end{equation}
1358: The total number of gates that can be implemented up to this 
1359: time scale is given by 
1360: \begin{equation}
1361: (N_g)_f= 2 n^2 k_f =-\frac{\ln A}{\gamma n}.
1362: \label{ngrel}
1363: \end{equation}
1364:  
1365: Our theoretical expectations are confirmed by the
1366: numerical data of Figs.~\ref{B:fid} and \ref{B:fid2}. 
1367: In Fig.~\ref{B:fid}, we show the fidelity after 
1368: one map step, as a function of the dimensionless damping rate 
1369: $\gamma=\Gamma \tau/\hbar$.
1370: The numerical data of this figure show that the fidelity 
1371: drops exponentially with $\gamma$, in excellent agreement with 
1372: Eq.~(\ref{eq:Bakerfidelity}). 
1373: We point out that our theory predicts not only the 
1374: exponential fidelity decay but also the numerical value of the 
1375: decay rate.
1376: Finally, we show in Fig.~\ref{B:fid2} the number $k_f$ of 
1377: forward/backward map steps required to reach a fixed 
1378: value $A$ of the fidelity ($F=A=0.9$), as a function of 
1379: the number $n$ of qubits. This time scale decays as a power 
1380: law, $k_f = C/n^3$, in agreement with Eq.~(\ref{eq:Bakerfidelity}).
1381: Again, our theory predicts also the value of the prefactor 
1382: $C$, in excellent agreement with numerical data.
1383: 
1384: %\begin{figure}[ht]
1385: \begin{figure}
1386: \includegraphics[width=8.5cm,angle=0]{fig6.eps}
1387: \caption[]{Semi log plot of the fidelity as a function 
1388: of the dimensionless decay rate $\gamma$, 
1389: for the baker's map after one map step in the 
1390: presence of the phase flip channel described in the text. 
1391: Circles and squares correspond to quantum trajectories 
1392: simulations with ${\cal N}=500$ trajectories, at $n=10$ 
1393: and $n=20$ qubits, respectively. Triangles give the 
1394: results obtained by direct computation of the 
1395: density matrix evolution at $n=10$. 
1396: Solid lines stand for the theoretical prediction 
1397: of \equ~(\ref{eq:Bakerfidelity}), namely
1398: $F(\gamma)=\exp(-2\gamma n^3)$.}
1399: \label{B:fid}
1400: \end{figure}
1401: 
1402: %\begin{figure}[t]
1403: %\psyoffset=-16cm
1404: %\psxoffset=.0cm
1405: %\psboxto(9cm;9cm){fidelityTime.eps}
1406: %\begin{figure}[ht]
1407: \begin{figure}
1408: \includegraphics[width=8.5cm,angle=0]{fig7.eps}
1409: \caption[]{Logarithmic plot of the time scale $k_f$
1410: (measured in number of steps of the baker's map) 
1411: required to reach a fidelity value of $F=0.9$, 
1412: as a function of the number $n$ of qubits, 
1413: for $\gamma\approx 6.47\times 10^{-5}$
1414: (this value has been chosen by demanding that $F=0.9$
1415: after one forward/backward step of the baker's map
1416: at $n=21$).
1417: The straight line gives the theoretical curve 
1418: $T(n)=C/n^3$, with $C=-\ln(0.9)/(2\gamma)\approx 8.14$.} 
1419: \label{B:fid2}
1420: \end{figure}
1421: 
1422: As we have seen above, the number of gates that 
1423: can be reliably implemented without quantum error correction
1424: drops only polynomially with the number of qubits, 
1425: $(N_g)_f\propto 1/n$ (see Eq.~\ref{ngrel}). 
1426: We would like to stress that this dependence 
1427: should remain valid also in other 
1428: environment models that allow only one qubit at a time 
1429: to perform a transition, like the other noise channels
1430: descussed in Sec.~\ref{sec:noise}.
1431: Furthermore, we note that the time scales for fidelity 
1432: decay derived in this section and confirmed in the 
1433: baker's map model, are expected to be valid for any quantum 
1434: algorithm simulating dynamical systems in the regime 
1435: of quantum chaos.
1436: 
1437: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1438: 
1439: 
1440: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1441: 
1442: \section{Conclusion}
1443: \label{sec:conclusion}
1444: 
1445: We have studied two quantum protocols, i.e. a teleportation 
1446: scheme through a chain of qubits and a quantum algorithm for  
1447: the quantum baker's map. We have modeled different sorts of 
1448: environments in order to get a deeper insight of the stability 
1449: of quantum computation when different interactions with the environment 
1450: taken into account. Two kinds of generalized amplitude damping models 
1451: have been considered for the teleportation scheme, giving very different 
1452: behaviors. After a theoretical analysis we were able to understand the 
1453: origin of these differences. This reveals the importance of the details 
1454: of the environmental models in assessing the operability bounds of 
1455: quantum processors. 
1456: In the case of the baker's map simulation, we have chosen the 
1457: phase flip type of noise and we could verify our theoretical predictions
1458: for fidelity decay. 
1459: The results of this paper show that quantum trajectories are a very
1460: valuable tool for simulating noise processes in quantum information 
1461: protocols with a high degree of efficiency.   
1462: 
1463: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1464: \begin{acknowledgments}
1465: 
1466: This work was supported in part by the EC contracts 
1467: IST-FET EDIQIP and RTN QTRANS, the NSA and ARDA under
1468: ARO contract No. DAAD19-02-1-0086, and the PRIN 2002 
1469: ``Fault tolerance, control and stability in
1470: quantum information precessing''.
1471: 
1472: \end{acknowledgments}
1473: 
1474: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1475: 
1476: \begin{thebibliography}{99}
1477: 
1478: \bibitem{Gra1} 
1479: W.H. Zurek, Rev. Mod. Phys. {\bf 75}, 715 (2003).
1480: 
1481: \bibitem{Gra2} 
1482: H. Grabert, P. Schramm, and G.-L. Ingold, Phys. Rep. {\bf 168}, 115 (1988).
1483: 
1484: \bibitem{Gra3}
1485: T. Dittrich, P. H\"anggi, G.-L. Ingold, B. Kramer, G. Sch\"on, and 
1486: W. Zwerger, {\it Quantum transport and dissipation} 
1487: (Wiley, Weinheim, 1998). 
1488: 
1489: \bibitem{Knight}
1490: M.B. Plenio and P.L. Knight,
1491: Rev. Mod. Phys. {\bf 70}, 101 (1998).
1492: 
1493: \bibitem{shor}
1494: P.W. Shor, SIAM J. Sci. Statist. Comput., {\bf 26}, 1484 (1997).
1495: 
1496: \bibitem{Sch}
1497: R. Schack, Phys. Rev. A {\bf 57}, 1634 (1998).
1498: 
1499: \bibitem{Bertrand}
1500: B. Georgeot and D.L. Shepelyansky, Phys. Rev. Lett. {\bf 86}, 2890 (2001).
1501: 
1502: \bibitem{Ben1}
1503: G. Benenti, G. Casati, S. Montangero, and D.L. Shepelyansky, 
1504: Phys. Rev. Lett. {\bf 87}, 227901 (2001); 
1505: Phys. Rev. A {\bf 67}, 052312 (2003).
1506: 
1507: \bibitem{cory}
1508: Y.S. Weinstein, S. Lloyd, J. Emerson, and D.G. Cory,
1509: Phys. Rev. Lett. {\bf 89}, 157902 (2002).
1510: 
1511: \bibitem{Schack}
1512: R. Schack, T.A. Brun, and I.C. Percival, J. Phys. A {\bf 28}, 
1513: 5401 (1995).
1514: 
1515: \bibitem{BarencoBrun}
1516: A. Barenco, T.A. Brun, R. Schack, and T.P. Spiller, Phys. Rev. A {\bf 56},
1517: 1177 (1997).
1518: 
1519: \bibitem{Car}
1520: G.G. Carlo, G. Benenti, and G. Casati, Phys. Rev. Lett.
1521: {\bf 91}, 257903 (2003).
1522: 
1523: \bibitem{BennettBrassard}
1524: C.H. Bennett, G. Brassard, C. Cr\'epeau, R. Jozsa, A. Peres, and W.K. Wootters, 
1525: Phys. Rev. Lett. {\bf 70}, 1895 (1993).
1526: 
1527: \bibitem{Chuangbook}
1528: M.A. Nielsen and I.L. Chuang, {\it Quantum Computation and 
1529: Quantum Information} (Cambridge University Press, Cambridge, 2000).
1530: 
1531: \bibitem{Preskillbook}
1532: J. Preskill, {\em Lecture notes on Quantum Information 
1533: and Computation} 
1534: (available at www.theory.caltech.edu/people/preskill/ph229).
1535: 
1536: \bibitem{Lindblad}
1537: G. Lindblad, Commun. Math. Phys. {\bf 48}, 119 (1976);
1538: V. Gorini, A. Kossakowski, and E.C.G. Sudarshan, 
1539: J. Math. Phys. {\bf 17}, 821 (1976).
1540: 
1541: \bibitem{DalibardCastin}
1542: J. Dalibard, Y. Castin, and K. M\o lmer, 
1543: Phys. Rev. Lett. {\bf 68}, 580 (1992).
1544: 
1545: \bibitem{Brun}
1546: T.A. Brun, Am. J. Phys. {\bf 70}, 719 (2002);
1547: T. A. Brun, quant-ph/0301046.
1548: 
1549: \bibitem{Zurek}
1550: W.H. Zurek, Phys. Today, October 1991, 36 (1991);
1551: see also quant-ph/0306072.
1552: 
1553: \bibitem{GisinPercival}
1554: N. Gisin and I.C. Percival, J. Phys. A {\bf 25}, 5677 (1992).
1555: 
1556: \bibitem{Flam}
1557: V.V. Flambaum and F.M. Izrailev, Phys. Rev E {\bf 64}, 036220 (2001).
1558: 
1559: \bibitem{Badziag}
1560: P. Badziag, M. Horodecki, P. Horodecki, and R. Horodecki, 
1561: Phys. Rev. A {\bf 62}, 012311 (2000).
1562: 
1563: \bibitem{Bandyo}
1564: S. Bandyopadhyay, Phys. Rev A {\bf 65}, 022302 (2002).
1565: 
1566: \bibitem{Oh}
1567: S. Oh, S. Lee, and H-w Lee, Phys. Rev. A {\bf 66}, 022316 (2002).
1568: 
1569: \bibitem{Verstraete}
1570: F. Verstraete and H. Verschelde, Phys. Rev. Lett {\bf 90}, 097901 (2003).
1571: 
1572: \bibitem{Bala} N.L. Balazs and A. Voros, Ann. Phys. (N.Y.) {\bf 190}, 1 
1573: (1989); M. Saraceno and A. Voros, Physica D {\bf 79}, 206 (1994).
1574: 
1575: \end{thebibliography}
1576: 
1577: 
1578: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
1579: 
1580: \end{document}
1581: 
1582: 
1583: 
1584: 
1585: 
1586: