1:
2: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
4:
5: % Some other (several out of many) possibilities
6: %\documentclass[preprint,aps]{revtex4}
7: %\documentclass[preprint,aps,draft]{revtex4}
8: %\documentclass[prb]{revtex4}% Physical Review B
9:
10: \usepackage{graphicx}% Include figure files
11: \usepackage{dcolumn}% Align table columns on decimal point
12: \usepackage{bm}% bold math
13:
14: %\nofiles
15:
16: \begin{document}
17:
18: \title{Concatenating dynamical decoupling with decoherence-free subspaces for quantum
19: computation}
20:
21: \author{Yong Zhang}\email{zhyong98@mail.ustc.edu.cn}\author{Zheng-Wei Zhou}\email{zwzhou@ustc.edu.cn} \author{Bo Yu}\author{Guang-Can Guo}
22: \address{Key Laboratory of Quantum Information,
23: University of Science and Technology of China, Chinese Academy of
24: Sciences, Hefei, Anhui 230026, China}
25:
26:
27: \begin{abstract}
28: A scheme to implement a quantum computer subjected to decoherence
29: and governed by an untunable qubit-qubit interaction is presented.
30: By concatenating dynamical decoupling through bang-bang (BB) pulse
31: with decoherence-free subspaces (DFSs) encoding, we protect the
32: quantum computer from environment-induced decoherence that results
33: in quantum information dissipating into the environment. For the
34: inherent qubit-qubit interaction that is untunable in the quantum
35: system, BB control plus DFSs encoding will eliminate its undesired
36: effect which spoils quantum information in qubits. We show how
37: this quantum system can be used to implement universal quantum
38: computation.
39: \end{abstract}
40:
41: \pacs{03.67.Pp, 03.67.Lx}
42: \maketitle
43:
44: \section{ Introduction}
45:
46: Quantum computation (QC) has become a very active field ever since
47: the discovery that quantum computers can be much more powerful
48: than their classical counterparts \cite{Shor1,Lloyd,Grover}.
49: Quantum computers act as sophisticated quantum information
50: processors, in which calculations are made by the controlled time
51: evolution of a set of coupled two-level quantum systems. Coherence
52: in the evolution is essential for taking advantage of quantum
53: parallelism, which plays an essential role in all quantum
54: algorithms. However, real physical systems will inevitably
55: interact with their surrounding environment. No matter how weak
56: the coupling that prevents an open system from being isolated, the
57: evolution of the system is eventually plagued by nonunitary
58: features such as decoherence and dissipation \cite{Gardiner}.
59: Quantum decoherence, in particular, is a purely quantum-mechanical
60: effect whereby the system loses its ability to exhibit coherent
61: behavior by getting entangled with the ambient degrees of freedom.
62: Decoherence stands as a serious obstacle common to all
63: applications, including QC, which rely on the capability of
64: maintaining and exploiting quantum coherence.
65:
66:
67: Recently, considerable effort has been devoted to designing
68: strategies able to counteract decoherence. Roughly speaking, three
69: classes of procedures are available to overcome the decoherence
70: problem. Two kinds of encoding methods of these strategies in the
71: field of quantum information are quantum error-correction codes
72: (QECCs) \cite{Shor2} and decoherence-free subspaces (DFSs, also
73: called error-avoiding codes) \cite
74: {Duan1,Duan2,Zanardi1,Zanardi2,Lidar1,Bacon,Kempe,Lidar2}, both
75: based on encoding the state into carefully selected subspaces of
76: the Hilbert space of the system. The main difference between the
77: two encoding strategies is that QECCs is an active strategy, in
78: which the encoding is performed in such a way that the various
79: errors are mapped onto orthogonal subspaces so that they can be
80: diagnosed and reversed, and DFSs instead provide a passive
81: strategy relying on the occurrence of specific symmetries in the
82: interaction with the environment, which guarantees the existence
83: of state space regions inaccessible to noise. The third strategy
84: can be termed dynamical decoupling or quantum ``bang-bang''(BB)
85: control \cite{Lorenza1,Lorenza2,Duan3,Vitali1,Zanardi3} after its
86: classical analog by using strong, fast pulses on quantum systems.
87: The basic idea is that open-system properties, specifically
88: decoherence, may be modified if a time-varying control field acts
89: on the dynamics of the system over time scales that are comparable
90: to the memory time of the environment. Dynamical decoupling has an
91: advantage over QECCs and DFSs, because it uses external pulses (BB
92: pulse) rather than requiring several physical qubits to encode one
93: logical qubit.
94:
95: Despite their promise to counteract decoherence in the process of
96: QC, QECCs and DFSs, in which ancillary physical qubits are
97: required for protecting quantum information, have their
98: disadvantage for the construction of a large scale quantum
99: computer, because the available physical resource is very exiguous
100: in the present quantum engineering. Dynamical decoupling does not
101: require an ancillary physical qubit to protect quantum
102: information, but entirely decoupling system from the environment
103: requires more complicated pulse operations. Moreover, the inherent
104: qubit-qubit interaction, which is vital to the implementation of
105: two-qubit gate, is assumed to be tunable in all the approaches
106: given above, but this will augment further the complexity of
107: quantum computer in microstructure. Our effort is devoted to
108: solving those problems mentioned above. In this work we present an
109: architecture of quantum computer with fixed coupling between
110: qubits. In our scheme, by concatenating dynamical decoupling and
111: DFSs encoding we can simultaneously overcome the effects from
112: decoherence and qubit-qubit interaction and realize the scalable
113: fault-tolerant QC.
114:
115:
116: The structure of the paper is as follows. In Sec. II, we review
117: dynamical decoupling by BB operations, and we show how to
118: counteract decoherence via encoding into DFSs and decoupling by BB
119: operations. In Sec. III we deal with the inherent qubit-qubit
120: interaction between physical qubits by BB operations. We show in
121: Sec. IV how the universal QC can be accomplished. Section V is for
122: discussion and concluding remarks.
123:
124: \section{Decoherence and Bang-Bang operation}
125:
126: We consider a two-level quantum system $S$ coupled to an arbitrary
127: bath $B$, which together form a closed system defined on the
128: Hilbert spaces
129: ${\cal H}$ $={\cal H}_S%TCIMACRO{\tbigotimes }
130: %BeginExpansion
131: \mathop{\textstyle \otimes }%
132: %EndExpansion
133: {\cal H}_B$ , ${\cal H}_S$ and ${\cal H}_B$ denoting $S$ and $B$
134: Hilbert spaces, respectively. The dynamics of the quantum system
135: $S$ coupled to a bath $B$ evolves unitarily under the Hamiltonians
136: \begin{equation}
137: H=H_S%
138: %TCIMACRO{\tbigotimes }
139: %BeginExpansion
140: \mathop{\textstyle \otimes }%
141: %EndExpansion
142: I_B+I_S%
143: %TCIMACRO{\tbigotimes }
144: %BeginExpansion
145: \mathop{\textstyle \otimes }%
146: %EndExpansion
147: H_B+H_{SB},
148: \end{equation}
149: where $H_S$, $H_B$, and $H_{SB}$ are the system, bath, and
150: interaction Hamiltonians, respectively. The interaction
151: Hamiltonians between the system and bath can be written as
152: \begin{equation}
153: H_{SB}=\sigma _x%
154: %TCIMACRO{\tbigotimes }
155: %BeginExpansion
156: \mathop{\textstyle \otimes }%
157: %EndExpansion
158: b_x+\sigma _y%
159: %TCIMACRO{\tbigotimes }
160: %BeginExpansion
161: \mathop{\textstyle \otimes }%
162: %EndExpansion
163: b_y+\sigma _z%
164: %TCIMACRO{\tbigotimes }
165: %BeginExpansion
166: \mathop{\textstyle \otimes }%
167: %EndExpansion
168: b_z. \label{e1}
169: \end{equation}
170: Here the $\sigma_\alpha$'s ($\alpha=x,y,z$) are the
171: spin-$\frac{1}{2}$ Pauli operators on physical qubit and the
172: $b_\alpha$'s are operators on the degrees of freedom of
173: environment. Due to the interaction Hamiltonian, the quantum
174: system will entangle with the environment so that the quantum
175: information encoded into quantum states irreversibly dissipates
176: into the environment, this is the so-called decoherence. The
177: objective of dynamical decoupling with BB operations used in our
178: scheme is to modify this unwanted evolution.
179:
180: The process of dynamical decoupling by BB operations, which
181: counteracts decoherence by applying sequences of strong and fast
182: pulses, serves for protecting the evolution of $S$ against the
183: effect of the interaction $H_{SB}.$ In the standard view of the
184: dynamical decoupling, a set of realizable BB operations can be
185: chosen such that they form a discrete (finite order) subgroup of
186: the full unitary group of operation on the Hilbert space of the
187: system. Denote the subgroup ${\cal G}$ and its elements $g_k$,
188: $k=0,1,\ldots ,\left| {\cal G}\right| -1$, where $\left| {\cal
189: G}\right| $ is the order of the group. The cycle time is
190: $T_c=\left| {\cal G}\right| \Delta t$, where $\left| {\cal
191: G}\right| $ is now also the number of pulse operations, and
192: $\Delta t$ is the time that the system evolves freely between
193: operations under $U_0=\exp (-iHt)$. The evolution of the system
194: now is given by
195: \begin{equation}
196: U(T_c)=%
197: %TCIMACRO{\tprod }
198: %BeginExpansion
199: \mathop{\textstyle \prod }%
200: %EndExpansion
201: _{k=0}^{\left| {\cal G}\right| -1}g_k^{\dagger}U_0(\Delta
202: t)g_k\equiv e^{iH_{eff}T_c}
203: \end{equation}
204: $H_{eff}$ denotes the resulting effective Hamiltonian. Obviously,
205: to satisfy the above equation, it is required that the pulses in
206: the sequence are very fast and strong compared with the evolution
207: of Hamiltonian $H$, which is the origin of the name ``bang-bang''
208: operation. In this BB limit, the system will evolve under the
209: effective Hamiltonian
210: \begin{equation}
211: H\rightarrow H_{eff}=\frac 1{\left| {\cal G}\right| }%
212: %TCIMACRO{\tsum }
213: %BeginExpansion
214: \mathop{\textstyle \sum }%
215: %EndExpansion
216: _{k=0}^{\left| {\cal G}\right| -1}g_k^{\dagger}Hg_k\equiv
217: %TCIMACRO{\tprod }
218: %BeginExpansion
219: \mathop{\textstyle \prod }%
220: %EndExpansion
221: \nolimits_{{\cal G}}(H).
222: \end{equation}
223: The map $%
224: %TCIMACRO{\tprod}
225: %BeginExpansion
226: \mathop{\textstyle \prod }%
227: %EndExpansion
228: _{{\cal G}}$ commutes with all $g_k$ so that the action of the
229: controller over times longer than the averaging period $T_c$ only
230: preserves the set of operators which are invariant under ${\cal
231: G}$, thereby enforcing a ${\cal G} $ symmetrization of the
232: evolution of $S$ \cite{Viola}. Recently, a general result has been
233: established by Facchi \emph{et al}. \cite{Facchi}, which states
234: that dynamical decoupling can be accomplished by a sequence of
235: arbitrary (fast and strong) pulses and symmetry or group structure
236: is not necessary, and the above procedure of decoupling by
237: ``symmetrization'' arises as a special case. The main drawback of
238: BB pulse decoupling procedures is that the timing constraints are
239: particularly stringent. In fact, perfect decoupling from the
240: environment is obtained only in the infinitely fast control limit
241: \cite {Lorenza2,Vitali1,Vitali2}, but it has been established that
242: these decoupling schemes can be effective in a realistic situation
243: with control pulses with finite strength and time duration
244: \cite{Vitali1,Vitali3}.
245:
246: Now let us first present our approach to counteract decoherence.
247: For modifying the coupling induced by the Hamiltonian in Eq.
248: (\ref{e1}), we consider a single BB operation
249: $U_{z1}=\exp(-i\sigma_z\pi/2)=-i\sigma_z$, and when no pulses are
250: applied the unit operator $I$ denotes the operation on qubits.
251: Using the commutation relation for Pauli operators, we have
252:
253: \begin{equation}
254: U^\dagger_{z1}\sigma _xU_{z1}=\sigma _z\sigma _x\sigma _z=-\sigma
255: _x,
256: \end{equation}
257: \begin{equation}
258: U^\dagger_{z1}\sigma _yU_{z1}=\sigma _z\sigma _y\sigma _z=-\sigma
259: _y,
260: \end{equation}
261: \begin{equation}
262: U^\dagger_{z1}\sigma _zU_{z1}=\sigma _z\sigma _z\sigma _z=\sigma
263: _z.
264: \end{equation}
265: Thus after cycles of BB operations, we can obtain the effective
266: interaction Hamiltonian
267: \begin{equation}
268: H_{SB}\rightarrow
269: %TCIMACRO{\tprod }
270: %BeginExpansion
271: \mathop{\textstyle \prod }%
272: %EndExpansion
273: \nolimits(H_{SB})=\sigma _z%
274: %TCIMACRO{\tbigotimes }
275: %BeginExpansion
276: \mathop{\textstyle \otimes }%
277: %EndExpansion
278: b_z,
279: \end{equation}
280: which still introduces phase decoherence. In order to counteract
281: phase decoherence, we can encode quantum information into DFSs. We
282: use a well-known code \cite{Duan1,Duan2,Palma} which two physical
283: qubits encode a logical qubit,
284: \begin{equation}
285: \left| 0\right\rangle_L =\left| 0_11_2\right\rangle%
286: and \left| 1\right\rangle_L =\left|
287: 1_10_2\right\rangle. \label{e2}
288: \end{equation}
289: Here $i=1,2$ indexes physical qubits. For the system consisting of
290: two physical qubits, the BB operation on the two physical qubits,
291: correspondingly, can be defined as collective rotation:
292: $U_z=U_{z1}\otimes
293: U_{z2}=\exp(-i\sigma_1^z\pi/2)\otimes\exp(-i\sigma_2^z\pi/2)=-\sigma
294: _1^z\otimes \sigma _2^z$,and then $%TCIMACRO{\tprod }
295: %BeginExpansion
296: \mathop{\textstyle \prod }%
297: %EndExpansion
298: \nolimits(H_{SB})=(\sigma_1^z+\sigma_2^z)%
299: %TCIMACRO{\tbigotimes }
300: %BeginExpansion
301: \mathop{\textstyle \otimes }%
302: %EndExpansion
303: b_z$. Clearly, such encoding on a pair of physical qubits ensures
304: that the encoded states are decoherence-free for phase error only
305: if the disturbances from the environment around the system are
306: identical. In other words, the two qubits must be arranged so
307: close to each other that they undergo collective phase
308: decoherence. Here the DFSs encoding together with BB operations
309: serve for combating decoherence.
310:
311: In Refs. \cite{Byrd,lidar3}, Byrd and Lidar have proposed a
312: comprehensive encoding and decoupling solution to problems of
313: decoherence. Decoherence is first reduced by encoding a logical
314: qubit into two qubits, then completely eliminated by an efficient
315: set of decoupling pulse sequences, in which cycles of pairs of BB
316: pulses generated from the same exchange Hamiltonian are used to
317: eliminate errors other than dephasing. The quantum code in our
318: scheme is analogous to the one they have proposed for reducing
319: phase decoherence. Then we apply directly a kind of simple BB
320: pulse on a physical qubit to selectively decouple the system from
321: the environment, which reduces the complexity of pulse operation.
322: In our scheme untunable qubit-qubit interaction can be controlled
323: by BB operations as discussed in the following section.
324:
325: \section{Interaction and Bang-Bang operation}
326:
327: To realize QC, any universal quantum gates (quantum operations)
328: must include single-qubit gates and two-qubit gates. A traditional
329: way for the implementation of single-qubit and two-qubit gates
330: requires a control on two qubits level that is an ability to
331: ``switch on'' and to ``switch off'' interaction between qubits.
332: But an ``always on'' coupling can cause certain problems for
333: quantum information preservation and QC. For example, if the
334: interaction between two physical qubits in the code (\ref{e2}) is
335: Heisenberg exchange interaction \cite{Ruskai}, the computational
336: basis will always be flipped under the exchange Hamiltonian, which
337: spoils quantum information in qubits. In general, quantum
338: computers exploit control techniques \cite {Kane,Vrijen} to tune
339: the interaction between two physical qubits to avoid the undesired
340: effect of the coupling, and tunability of the interaction constant
341: is at the heart of many solid-state proposals, but this prove
342: extremely difficult to achieve experimentally. Recently, some
343: schemes of QC governed by always on interaction have been
344: presented \cite{Zhou,Benjamin,Zhang}. In our scheme, we discuss
345: the case that the interaction is always on and untunable, and we
346: exploit BB operations to selectively decouple two physical qubits.
347:
348: Now we consider the general exchange interaction between physical
349: qubits. The exchange interaction Hamiltonian in the system has the
350: form
351: \begin{equation}
352: H_I=J_x\sigma _1^x%
353: %TCIMACRO{\tbigotimes }
354: %BeginExpansion
355: \mathop{\textstyle \otimes }%
356: %EndExpansion
357: \sigma _2^x+J_y\sigma _1^y%
358: %TCIMACRO{\tbigotimes }
359: %BeginExpansion
360: \mathop{\textstyle \otimes }%
361: %EndExpansion
362: \sigma _2^y+J_z\sigma _1^z%
363: %TCIMACRO{\tbigotimes }
364: %BeginExpansion
365: \mathop{\textstyle \otimes }%
366: %EndExpansion
367: \sigma _2^z, \label{e3}
368: \end{equation}
369: where $J_a$'s, $(a=x,y,z)$ are exchange interaction constants.
370:
371: We first consider the case of a single logical qubit. Under the
372: self-exchange interaction, we find
373: \begin{equation}
374: H_I\left| 0\right\rangle _L=(J_x+J_y)\left| 1\right\rangle
375: _L-J_z\left| 0\right\rangle _L,
376: \end{equation}
377: \begin{equation}
378: H_I\left| 1\right\rangle _L=(J_x+J_y)\left| 0\right\rangle
379: _L-J_z\left| 1\right\rangle _L.
380: \end{equation}
381: Obviously, quantum information encoded will be spoiled by the
382: self-exchange interaction. We selectively decouple the two
383: physical qubits encoded into a logical qubit by introducing a
384: selective decoupling BB operation
385: $R_z=I_1\otimes\exp(-i\sigma_2^z\pi/2)=-iI_1\otimes\sigma_2^z$. We
386: obtain
387: \begin{equation}
388: R_z^{\dagger}\sigma _1^x%
389: %TCIMACRO{\tbigotimes }
390: %BeginExpansion
391: \mathop{\textstyle \otimes }%
392: %EndExpansion
393: \sigma _2^xR_z=\sigma _1^x%
394: %TCIMACRO{\tbigotimes }
395: %BeginExpansion
396: \mathop{\textstyle \otimes }%
397: %EndExpansion
398: \sigma _2^z\sigma _2^x\sigma _2^z=-\sigma _1^x%
399: %TCIMACRO{\tbigotimes }
400: %BeginExpansion
401: \mathop{\textstyle \otimes }%
402: %EndExpansion
403: \sigma _2^x,
404: \end{equation}
405: \begin{equation}
406: R_z^{\dagger}\sigma _1^y%
407: %TCIMACRO{\tbigotimes }
408: %BeginExpansion
409: \mathop{\textstyle \otimes }%
410: %EndExpansion
411: \sigma _2^yR_z=\sigma _1^y%
412: %TCIMACRO{\tbigotimes }
413: %BeginExpansion
414: \mathop{\textstyle \otimes }%
415: %EndExpansion
416: \sigma _2^z\sigma _2^y\sigma _2^z=-\sigma _1^y%
417: %TCIMACRO{\tbigotimes }
418: %BeginExpansion
419: \mathop{\textstyle \otimes }%
420: %EndExpansion
421: \sigma _2^y,
422: \end{equation}
423: \begin{equation}
424: R_z^{\dagger}\sigma _1^z%
425: %TCIMACRO{\tbigotimes }
426: %BeginExpansion
427: \mathop{\textstyle \otimes }%
428: %EndExpansion
429: \sigma _2^zR_z=\sigma _1^z%
430: %TCIMACRO{\tbigotimes }
431: %BeginExpansion
432: \mathop{\textstyle \otimes }%
433: %EndExpansion
434: \sigma _2^z\sigma _2^z\sigma _2^z=\sigma _1^z%
435: %TCIMACRO{\tbigotimes }
436: %BeginExpansion
437: \mathop{\textstyle \otimes }%
438: %EndExpansion
439: \sigma _2^z.
440: \end{equation}
441: So after cycles of BB operations, we obtain effective self-interaction $%
442: %TCIMACRO{\tprod}
443: %BeginExpansion
444: \mathop{\textstyle \prod }%
445: %EndExpansion
446: (H_I)=J_z\sigma _1^z%
447: %TCIMACRO{\tbigotimes }
448: %BeginExpansion
449: \mathop{\textstyle \otimes }%
450: %EndExpansion
451: \sigma _2^z$, which is equivalent to Ising interaction; the encoded states $%
452: \left| 0_L\right\rangle $ and $\left| 1_L\right\rangle $ in Eq.
453: (\ref{e2}) are degenerate under the effective self-interaction.
454: Therefore, if we store information in these states, no evolution
455: whatsoever is present. In other words, for the untunable exchange
456: interaction quantum information is stabilized by means of BB
457: control and quantum encoding.
458:
459: Until now, we have introduced two BB operationS $U_z$ and $R_z$.
460: As already noted, the two BB operations are used on qubitS 1 and 2
461: to counteract decoherence and undesired interaction. Actually, the
462: pulse operations $R_z=I_1\otimes\exp(-i\sigma_2^z\pi/2)$ only act
463: on physical qubit 2. For physical qubit 1, only the pulse
464: operation $\sigma_1^z$ has an effect on the decoherence. But there
465: are two kinds of pulse operations in $U_z$ and $R_z$ effected on
466: qubit 2 to selectively eliminate not only qubit-qubit interaction
467: but also qubit-environment interaction. In other words, the number
468: of pulse operations on qubits 1 and 2 is dissimilar. Because we
469: apply the same pulse operations ($\sigma^z$) on every physical
470: qubit, the time intervals $\Delta t_1$ on qubit 1 and $\Delta t_2$
471: on qubit 2 are different too. This implies that we have applied a
472: kind of nonsynchronous pulse operations to overcome
473: environment-induced decoherence and unwanted coupling between
474: physical qubits.
475:
476: Let us now show how to devise nonsynchronous pulse operations for
477: decoupling different interactions. We can elaborately devise a set
478: of programmed pulse operations in which the time intervals of the
479: BB operations on two qubits are varied according to the program.
480: In our scheme, unitary pulse operations are $U_z$ and $R_z$ as
481: given above. Here we assume that the BB operation $U_z$ begins at
482: time ${t}_0 $=0 and devise the time interval between two pulse
483: operations is constant $\Delta t$. Then we devise the BB operation
484: $R_z$ begins at time ${t}_0+\Delta t/2$ and the time interval is
485: $\Delta t$ too. So the time intervals between a pair of pulses on
486: qubits 1 and 2 have the relation $\Delta t_1=2\Delta t_2$. In fig.
487: 1 we focus on the evolution of the $y$ ingredient in Hamiltonian
488: $H_{SB}$ under the cycles of BB pulses. (The same conclusion
489: adapts to the $x$ ingredient in $H_{SB}$.) $T_1=2\Delta t_1$ and
490: $T_2=2\Delta t_2$ denote the cycle time of decoupling operations
491: on qubits 1 and 2, respectively. After cycles of pulse operations,
492: the total effect of error operators ($Y$ in the figure) on qubits
493: 1 and 2, respectively, is zero in the cycles time $NT_i$
494: ($i=1,2$), here $N$ and $N_i (i=1,2,3)$ given in the following are
495: positive integer. This implies that decoherence on qubits 1 and 2
496: is hold back. In addition, by similar analysis, we find that for
497: the self-interaction between qubits 1 and 2, the total effect of
498: the error operator
499: $J_x\sigma^x_1\otimes\sigma^x_2+J_y\sigma^y_1\otimes\sigma^y_2$ is
500: also eliminated in the cycles time $T=N_1T_1=N_2T_2$, so in $y$
501: axis qubits 1 and 2 are decoupled. The result shows that the
502: programmed BB pulse operations can eliminate or selectively
503: eliminate not only qubit-environment interaction but also
504: qubit-qubit interaction. This gives us a very heuristic solution
505: to elimination of undesired coupling. The method of decoupling
506: with programmed unsymmetrical pulse operations may be of great
507: benefit to the implementation of QC in many complicated
508: circumstances.
509:
510: In the above discussion, we present a dynamical decoupling scheme
511: based on group averaging formulation. It is noteworthy that for
512: the two-qubit system the operation set $\{I,U_z,R_z\}$ has no
513: group structure, which accords with the result of Ref.
514: \cite{Facchi}.
515:
516: We still need to show how the interaction between two logical
517: qubits influences the encoded states of logical qubits. The
518: exchange interaction in Eq. (\ref{e3}) between two logical qubits
519: will induce unwanted flow of quantum information between two
520: logical qubits. This will inevitably result in the failure of the
521: preservation of quantum information and QC. In our scheme, quantum
522: computer is constructed in a one-dimensional array of physical
523: qubits. Now, we introduce new logical qubits $L_2$ and $L_3$ (See
524: Fig. 2). For logical qubit $L_2$, two selective decoupling BB
525: operations are chosen as $U_x$ and $R_x$, here $U_x=U_{x3}\otimes
526: U_{x4}=\exp(-i\sigma_3^x\pi/2)\otimes\exp(-i\sigma_4^x\pi/2)=-\sigma
527: _3^x\otimes \sigma _4^x$ and $R_x=I_3\otimes
528: U_{x4}=-iI_3\otimes\sigma_4^x$. Then, we can obtain the effective
529: interaction Hamiltonian
530: $%TCIMACRO{\tprod }
531: %BeginExpansion
532: \mathop{\textstyle \prod }%
533: %EndExpansion
534: \nolimits(H_{SB})=(\sigma_3^x+\sigma_4^x)%
535: %TCIMACRO{\tbigotimes }
536: %BeginExpansion
537: \mathop{\textstyle \otimes }%
538: %EndExpansion
539: b_x$
540: and the effective self-interaction $%
541: %TCIMACRO{\tprod}
542: %BeginExpansion
543: \mathop{\textstyle \prod }%
544: %EndExpansion
545: (H_I)=J_x\sigma _3^x%
546: %TCIMACRO{\tbigotimes }
547: %BeginExpansion
548: \mathop{\textstyle \otimes }%
549: %EndExpansion
550: \sigma _4^x$. Accordingly, two encoded states of $L_2$ encoded in
551: DFS can be written as
552: \begin{equation}
553: \left| 0\right\rangle_{L_B} =\frac12(\left|
554: 0_3\right\rangle+\left| 1_3\right\rangle)(\left|
555: 0_4\right\rangle-\left| 1_4\right\rangle)\label{e4},
556: \end{equation}
557: \begin{equation}
558: \left| 1\right\rangle_{L_B} =\frac12\left| 0_3\right\rangle-\left|
559: 1_3\right\rangle)(\left| 0_4\right\rangle+\left| 1_4\right\rangle)
560: \label{e5},
561: \end{equation}
562: where the subscript $B$ denotes the method of decoupling and
563: encoding for logical qubit $L_2$. Similarly, two selective
564: decoupling subgroups of logical qubit $L_3$ are chosen as $U_y$
565: and $R_y$, here $U_y=U_{y5}\otimes
566: U_{y6}=\exp(-i\sigma_5^y\pi/2)\otimes\exp(-i\sigma_6^y\pi/2)=-\sigma
567: _5^y\otimes \sigma _6^y$ and $R_y=I_5\otimes
568: U_{y6}=-iI_5\otimes\sigma_6^y$, and then, the quantum code in DFS
569: will have the form
570:
571: \begin{equation}
572: \left| 0\right\rangle_{L_C} =\frac12(\left|
573: 0_5\right\rangle+i\left| 1_5\right\rangle)(\left|
574: 0_6\right\rangle-i\left| 1_6\right\rangle),
575: \end{equation}
576: \begin{equation}
577: \left| 1\right\rangle_{L_C} =\frac12(\left|
578: 0_5\right\rangle-i\left| 1_5\right\rangle)(\left|
579: 0_6\right\rangle+i\left| 1_6\right\rangle).
580: \end{equation}
581: Obviously, with selective decoupling and encoding into DFSs, $L_2$
582: and $L_3$ can overcome decoherence and unwanted internal
583: interaction as $L_1$ does .
584:
585:
586: Now, we focus on the coupling between logical qubits $L_1$ and
587: $L_2$ that is equivalent to the coupling between physical qubits 2
588: and 3. The inherent interaction Hamiltonian between qubit 2 and 3
589: has the form as shown in Eq. (\ref{e3}). For physical qubit 2, the
590: pulse operation is $\sigma_2^z$, then the evolution of the $x$ and
591: $y$ ingredients in Hamiltonian $H_{SB}$ is changed. For qubit 3,
592: the pulse operation is $\sigma_3^x$ which changes the evolution of
593: the $y$ and $z$ ingredients in Hamiltonian $H_{SB}$. Then, after
594: cycles of pulse operations in the time $T=N_2 2\Delta t_2=N_3
595: 2\Delta t_3$, we obtain $\prod(\sigma_2^x\otimes\sigma_3^x)=0$ and
596: $\prod(\sigma_2^z\otimes\sigma_3^z)=0$. So the evolution of the
597: $x$ and $z$ ingredients in Hamiltonian $H_{SB}$ is eliminated. As
598: far as the evolution of the $y$ ingredient is concerned, since
599: pulses effect on qubit 2 at the interval of $\Delta t_2$, but on
600: qubit 3 at the interval of $\Delta t_1$, here $\Delta t_1=2\Delta
601: t_2$, the evolution about $y$ axis on qubits 2 and 3 is
602: unsymmetrical, then $\prod(\sigma_2^y\otimes\sigma_3^y)=0$, i.e.,
603: the evolution of the $y$ ingredient in Hamiltonian $H_{SB}$ is
604: eliminated. This can also be illuminated by Fig. 1. To sum up,
605: with cycles of pulse operations, the effect of Hamiltonian
606: $H_{SB}$ between qubits 2 and 3 is eliminated. In other words,
607: $L_1$ is entirely decoupled from $L_2$. The same conclusion can be
608: drawn for logical qubits $L_2$ and $L_3$.
609:
610: We showed above that with BB pulse operations and quantum encoding
611: into DFS, the three logical qubits overcome not only
612: environment-induced decoherence but also unwanted inherent
613: interaction which is always on and untunable between physical
614: qubits. And we devise that the three logical qubits are effected
615: with three different BB operations so that every logical qubit is
616: decoupled from others. Then, we can construct a scalable quantum
617: computer with the three logical qubits as a unit of computation,
618: i.e., the quantum computer has the periodic structure
619: $AABBCCAABBCC\cdots$, where $AA$, $BB$, and $CC$ denote encoded
620: logical qubits analogous to $L_1$, $L_2$,and $L_3$, respectively.
621:
622: \begin{figure}[here]
623:
624: \includegraphics[width=8cm]{1.eps}
625:
626: \caption{the evolution of physical qubit 1 and 2 about the $y$
627: axis under Hamiltonian $H_{SB}$ and pulse operations. White and
628: black rectangles denote strong and fast pulse operations $U_z$ and
629: $R_z$, respectively. $T_1$ and $T_2$ denote the cycle time of
630: decoupling operations on qubit 1 and 2. $Y$=$\sigma_y$ is error
631: operator on physical qubit.}
632: \end{figure}
633:
634: \begin{figure}[here]
635: \includegraphics[width=8cm]{2.eps}
636:
637: \caption{Architecture of three logical qubits in the quantum computer.%
638: Each dot is a physical qubit and the dashed lines represent
639: interaction between qubits. Every logical qubit consists of two
640: physical qubits. Arrows with different colors denote different
641: pulse operations on logical qubit.}
642: \end{figure}
643:
644: \section{Quantum Computation}
645:
646: Our discussion so far has concentrated on the preservation of
647: quantum information. To carry out quantum information, we must
648: have the ability to manipulate encoded quantum information. Thus
649: we still need to show that universal QC can actually be performed
650: in our scheme. DiVincenzo shows that for any unitary
651: transformation on quantum states it is sufficient to apply (a) all
652: single-qubit rotations [SU(2)] together with (b) the two-qubit
653: controlled-NOT (CNOT) gate on any two logical qubits
654: \cite{DiVicenzo}.
655:
656: In our scheme, we assume that any single-qubit operations on
657: physical qubits are realizable at will by virtue of external
658: pulses. We can define logical operations (denoted by a bar) which
659: act on the encoded qubits. For example, $\overline{X}:\left|
660: 0_L\right\rangle \leftrightarrow \left| 1_L\right\rangle $. For
661: logical qubit $L_1$, $\overline{X}=(J_x\sigma _1^x\otimes\sigma
662: _2^x+J_y\sigma _1^y\otimes\sigma _2^y)/(J_x+J_y)$. Logical
663: $\overline{X}$ operation can be easily achieved by recoupling
664: qubits 1 and 2 with the interaction Hamiltonian as shown in Eq.
665: (\ref{e3}). We adjust the time intervals of pulses on qubits 1 and
666: 2 both to $\Delta t_3$, where $\Delta t_3=\Delta t_2/2=\Delta
667: t_1/4$. In other words, only synchronous collective BB pulses are
668: applied, which just eliminate the coupling from environment but
669: have no effect on qubit-qubit coupling $H_I$. Then, we have
670:
671: \begin{eqnarray}
672: e^{i\theta H_I}\left| i\right\rangle _L &=&e^{i\theta (J_x\sigma _1^x%
673: %TCIMACRO{\tbigotimes }
674: %BeginExpansion
675: \mathop{\textstyle \otimes }%
676: %EndExpansion
677: \sigma _2^x+J_y\sigma _1^y%
678: %TCIMACRO{\tbigotimes }
679: %BeginExpansion
680: \mathop{\textstyle \otimes }%
681: %EndExpansion
682: \sigma _2^y+J_z\sigma _1^z%
683: %TCIMACRO{\tbigotimes }
684: %BeginExpansion
685: \mathop{\textstyle \otimes }%
686: %EndExpansion
687: \sigma _2^z)}\left| i\right\rangle _L \nonumber \\
688: &=&e^{-i\theta J_Z}e^{i\theta (J_x+J_y)\overline{X}}\left|
689: i\right\rangle _L.
690: \end{eqnarray}
691: By the free evolution under the inherent interaction Hamiltonian,
692: we can easily accomplish logical $\overline{X}$ operation. We must
693: note that the time intervals of pulses on qubits 2 and 3 are still
694: unequal; this implies that after cycles of pulse operations in the
695: time $T=N_2 2\Delta t_3=N_3 2\Delta t_1$, qubit 2 remains
696: decoupled from qubit 3; logical $\overline{X}$ operation on $L_1$
697: therefore has no impacts on other logical qubits. We can also
698: implement logical $Z$ operation $\overline{Z}=(\sigma _1^z-\sigma
699: _2^z)/2$ by direct pulse operations on physical qubits, then
700: $\overline{X}$ and $\overline{Z}$ generate all encoded-qubit SU(2)
701: transformations.
702:
703:
704: By inspection of quantum codes of logical qubits $L_1$, $L_2$, and
705: $L_3$, we find that the DFSs of $L_2$ and $L_3$ can be obtained by
706: performing a unitary transformation on that of $L_1$. For example,
707: the transformation of DFSs between $L_1$ and $L_2$ is a Hadamard
708: transformation. Obviously, single-encoded-qubit operations, which
709: preserve the DFSs of $L_1$ and $L_2$, respectively, have the same
710: unitary transformation too. Then, by performing a transformation
711: on single-encoded-qubit gate given above, all single-encoded-qubit
712: operations [SU(2)] on $L_2$ and $L_3$ can be easily achieved (See
713: Table I).
714:
715:
716: Two-encoded-qubit CNOT gate seems to be more complicated, but in
717: our scheme it is very easy to accomplish the two-qubit gate. For
718: the convenience of discussion, let us assume that we want to do a
719: CNOT operation from logical qubit $L_1$ to $L_2$ in Fig. 2. To
720: obtain a two-qubit gate, we consider the imprimitive gate
721: $W=e^{i\theta \sigma ^z\otimes \sigma ^z}$, which is equivalent to
722: a controlled rotation about the $z$ axis \cite{Bremmer},
723:
724:
725: \begin{equation}
726: e^{i\theta \sigma ^z\otimes \sigma ^z}\equiv \left| 0\right\rangle
727: \left\langle 0\right|
728: %TCIMACRO{\tbigotimes }
729: %BeginExpansion
730: \mathop{\textstyle \otimes }%
731: %EndExpansion
732: I+\left| 1\right\rangle \left\langle 1\right|
733: %TCIMACRO{\tbigotimes }
734: %BeginExpansion
735: \mathop{\textstyle \otimes }%
736: %EndExpansion
737: e^{i2\left| \theta \right| \sigma ^z}
738: \end{equation}
739: Conjugated by single-qubit Hadamard $H=%
740: %TCIMACRO{\dfrac 1{\sqrt{2}}}
741: %BeginExpansion
742: {\displaystyle {1 \over \sqrt{2}}}%
743: %EndExpansion
744: \left[
745: \begin{array}{cc}
746: 1 & 1 \\
747: 1 & -1
748: \end{array}
749: \right] $ operation on the second qubit, $W$ can be used to
750: implement a CNOT:
751:
752: \begin{equation}
753: \text{CNOT}\equiv \left| 0\right\rangle \left\langle 0\right|
754: %TCIMACRO{\tbigotimes }
755: %BeginExpansion
756: \mathop{\textstyle \otimes }%
757: %EndExpansion
758: I+\left| 1\right\rangle \left\langle 1\right|
759: %TCIMACRO{\tbigotimes }
760: %BeginExpansion
761: \mathop{\textstyle \otimes }%
762: %EndExpansion
763: e^{i(\frac \pi 2)\sigma ^x}
764: \end{equation}
765: To implement a encoded CNOT between $L_1$ and $L_2$, we must
766: recouple the two logical qubits with an interaction in the form
767: $\overline{Z}_{L_1}\otimes\overline{Z}_{L_2}$. We perform a
768: unitary Hadamard transformation on $L_2$. In other words, we
769: change the BB pulses characterized by $-\sigma _3^x\otimes \sigma
770: _4^x$ and $-iI_3\otimes \sigma _4^x$ to the same with $L_1$, and
771: the quantum code in Eqs. (\ref{e4}) and (\ref{e5}) to the same
772: with that in Eq. (\ref{e2}), i.e.,
773: $|0\rangle_{L_B}\rightarrow|0\rangle_{L_A}=|0_31_4\rangle$ and
774: $|1\rangle_{L_B}\rightarrow|1\rangle_{L_A}=|1_30_4\rangle$. It
775: should be noted that $L_2$ and $L_3$ are still entirely decoupled
776: after the unitary transformation. Then the effective interactions
777: between qubits 1, 2, 3 and 4 all are in the form of Ising
778: interaction. In this system we assume that the interaction only
779: exists between any nearest-neighbor physical qubits. Obviously,
780: $\overline{Z}_{L_1}\otimes\overline{Z}_{L_2}=\sigma_2^z\otimes
781: \sigma_3^z$, two-encoded-qubit CNOT gate can be implemented by the
782: evolution under the effective interaction $\sigma_2^z\otimes
783: \sigma_3^z$ and single-qubit Hadamard operation conjugately
784: effected on a physical qubit. Similarly, we can implement CNOT
785: operation between $L_2$ and $L_3$.
786:
787:
788: As above, we showed that it is possible to perform all single- and
789: two-encoded-qubit operations by means of pulse operations and
790: evolution under inherent interaction. In our scheme, single- and
791: two-encoded-qubit operations do not influence decoupling
792: operations and preserve DFSs all the time, so quantum states
793: encoded with quantum information will not undergo decoherence,
794: then we implement universal, fault-tolerant QC.
795:
796:
797: \section{Discussion and Conclusion}
798:
799: In this paper we have presented a scheme of scalable quantum
800: computer governed by untunable exchange Hamiltonian. We combine
801: ideas from the theory of decoherence-free subspaces and BB control
802: to solve the problem of strong decoherence. Cycles of simple BB
803: pulses are used to selectively decouple the system from external
804: environment, then by encoding two physical qubits into a DFS, we
805: obtain full protection against strong decoherence. By
806: concatenating BB control with the DFSs encoding, our scheme
807: decreases the number of physical qubits required to counteract
808: decoherence. It is highly important for the physicist to reduce
809: the physical resource needed for implementation of scalable
810: quantum computer, because quantum computing resources available
811: are still a stringent requirement for practical quantum
812: engineering. Comparing with other decoupling scheme, in our scheme
813: only very simple BB pulses are applied which is easy to
814: accomplish.
815:
816: Furthermore, we have discussed the influence of an always on and
817: untunable interaction between physical qubits on the logical
818: qubits. The undesired effects of the internal interaction can be
819: eliminated via cycles of BB operations, which simplifies the
820: physical structure of quantum computer that is devised in a very
821: complicated manner for implementing the tunability of the coupling
822: strength in many QC proposals. By different unsymmetrical
823: decoupling operations, every logical qubit is entirely decoupled
824: from others. With direct pulse operations on physical qubits and
825: effective interaction, we can achieve all single- and
826: two-encoded-qubit gates for implementing universal QC. Moreover,
827: in our scheme all single- and two-encoded-qubit operations
828: preserve logical qubits in a DFS all the time, so we implement
829: universal, fault-tolerant QC.
830:
831: \section{ Acknowledgments}
832:
833: This work was funded by National Fundamental Research Program
834: (2001CB309300), National Natural Science Foundation of China under
835: Grants No.10204020, the Innovation funds from Chinese Academy of
836: Sciences, and also by the outstanding Ph. D thesis award and the
837: CAS's talented scientist award entitled to Luming Duan.
838:
839: \begin{references}
840:
841: \bibitem{Shor1} P. W. Shor, in Proceedings of the Thirty-Fifth Annual Symposium
842: on Foundations of Computer Science. Edited be S.Goldwsser(IEEE
843: Computer Society, New York, 1994), pp. 124-134
844:
845: \bibitem{Lloyd} S. Lloyd, Science {\bf 273}, 1073(1996)
846:
847: \bibitem{Grover} L. K. Grover, Phys. Rev. Lett. {\bf 79}, 325 (1997)
848:
849: \bibitem{Gardiner} C. W. Gardiner, Quantum Noise (Springer, Berlin, 1991)
850:
851: \bibitem{Shor2} P. W. Shor, Phys. Rev. A {\bf 52}, R2493 (1995); A. M. Steane,
852: Phys. Rev. Lett. {\bf 77}, 793 (1996); E. Knill and R. Laflamme, Phys. Rev. A {\bf %
853: 55}, 900 (1997).
854:
855: \bibitem{Duan1} L. M. Duan and G. C. Guo, Phys. Rev. Lett. {\bf 79}, 1953 (1997).
856:
857: \bibitem{Duan2} L. M. Duan and G. C. Guo, Phys.R ev. A {\bf 57}, 737 (1998).
858:
859:
860:
861: \bibitem{Zanardi1} P. Zanardi and M. Rasetti, Mod. Phys. Lett. B {\bf11},
862: 1085 (1997).
863:
864: \bibitem{Zanardi2} P. Zanardi and M. Rasetti,Phys. Rev. Lett.{\bf 79}%
865: ,3306 (1997).
866:
867: \bibitem{Lidar1} D. A. Lidar, I. L. Chuang, and K. B. Whaley, Phys. Rev. Lett. {\bf 81}, 2594 (1998).
868:
869: \bibitem{Bacon} D. Bacon, J. Kempe, D. A. Lidar, and K. B. Whaley, Phys.
870: Rev. Lett. {\bf 85}, 1758 (2000).
871:
872: \bibitem{Kempe} J. Kempe, D. Bacon, D. A. Lidar, and K. B. Whaley, Phys. Rev. A
873: {\bf 63}, 042307 (2001).
874:
875: \bibitem{Lidar2} D. A. Lidar, D. Bacon, J. Kempe, and K. B. Whaley, Phys. Rev. A
876: {\bf 63}, 022306 (2001).
877:
878: \bibitem{Lorenza1} L. Viola and S. Lloyd, Phys. A {\bf 58}, 2733 (1998).
879:
880: \bibitem{Lorenza2} L. Viola, E. Knill, and S. Lloyd, Phys. Rev. Lett. {\bf 82},
881: 2417 (1999).
882:
883: \bibitem{Duan3} L. M. Duan and G. C. Guo, Phys. Lett. A {\bf 261}, 139 (1999).
884:
885: \newpage
886:
887: \bibitem{Vitali1} D. Vitali and P. Tombesi, Phys. Rev. A {\bf 59}, 4178 (1999).
888:
889: \bibitem{Zanardi3} P. Zanardi, Phys. Lett. A {\bf 258}, 77 (1999).
890:
891: \bibitem{Viola} L. Viola, E. Knill, and S. Lloyd, Phys. Rev. Lett.{\bf 85}, 3520 (2000).
892:
893: \bibitem{Facchi} P. Facchi, D. A. Lidar, and S. Pascazio, e-print
894: quant-ph/03030132
895:
896: \bibitem{Vitali2} D. Vitali and P. Tombesi, Phys. Rev. A{\bf 65}, 012305 (2002).
897:
898: \bibitem{Vitali3} D. Vitali, J. Opt. B: Quant Semi. Opt. {\bf 4}, 337 (2002).
899:
900: \bibitem{Palma} G. M. Palma, K. A. Suominen, and A. K. Ekert, Proc. R. Soc. London,
901: Ser. A {\bf 452}, 567 (1996).
902:
903: \bibitem{Byrd} M. S. Byrd and D. A. Lidar, Phys. Rev. Lett. {\bf 89}, 047901
904: (2002).
905: \bibitem{lidar3} D. A. Lidar and L. A. Wu, e-print quant-ph/0302198
906:
907: \bibitem{Ruskai} M. B. Ruskai, Phys. Rev. Lett. {\bf 85}, 194 (2000).
908:
909: \bibitem{Kane} B. E. Kane, Nature {\bf 408}, 339 (2000).
910:
911: \bibitem{Vrijen} R. Vrijen, E. Yablonovitch, K. Wang, H. W. Jiang, \newline
912: A. Balandin, V. Roychowdhury, T. Mor, and D. DiVincenzo, Phys.
913: Rev. A {\bf 62}, 012306 (2000).
914:
915: \bibitem{Zhou} X. X. Zhou, Z. W.Z hou, G. C. Guo, and M. J. Feldman, Phys. Rev. Lett. {\bf 89}, 197903 (2002)
916:
917: \bibitem{Benjamin} S. C. Benjamin and S. Bose, Phys. Rev. Lett. {\bf 90}, 247901 (2003)
918:
919: \bibitem{Zhang} Y. Zhang, Z. W. Zhou, B. Yu, and G. C. Guo, J. Opt. B: Quant Semi.Opt. {\bf 5}, 309 (2003).
920:
921: \bibitem{DiVicenzo} D. P. DiVincenzo, Phys. Rev. A {\bf 51}, 1015 (1995)
922:
923: \bibitem{Bremmer} M. J. Bremner, C. M. Dawson, J. L. Dodd, A. Gilchrist, A. W. Harrow,
924: D. Mortimer, M. A. Nielsen, and T. J. Osborne, Phys. Rev.
925: Lett.{\bf 89}, 247902 (2002).
926:
927: \end{references}
928: \newpage
929: \begin{table*}
930:
931: \renewcommand{\arraystretch}{2.0}
932: \caption{Comparison of properties between logical qubit $L_1$,
933: $L_2$ and $L_3$.}
934: \begin{tabular}{ccccc} \hline\hline& $L_1$ & $L_2$ & $L_3$ \\\hline\hline
935: $U_\alpha$ & $-\sigma _1^z\otimes \sigma _2^z$ & $-\sigma
936: _3^x\otimes \sigma _4^x$ & $-\sigma _5^y\otimes \sigma _6^y$ \\
937: \hline $R_\alpha$ & $-iI_1\otimes \sigma _2^z$ &
938: $-iI_3\otimes \sigma _4^x$ & $-iI_5\otimes \sigma _6^y$ \\
939: \hline $\left| 0\right\rangle _L$ & $\left| 0_11_2\right\rangle $
940: & $\frac 12(\left| 0_3\right\rangle +\left| 1_3\right\rangle
941: )(\left| 0_4\right\rangle -\left| 1_4\right\rangle )$ & $\frac
942: 12(\left| 0_5\right\rangle +i\left| 1_5\right\rangle )(\left|
943: 0_6\right\rangle -i\left| 1_6\right\rangle )$ \\ \hline $\left|
944: 1\right\rangle _L$ & $\left| 1_10_2\right\rangle $ & $\frac
945: 12(\left| 0_3\right\rangle -\left| 1_3\right\rangle )(\left|
946: 0_4\right\rangle +\left| 1_4\right\rangle )$ & $\frac12(\left|
947: 0_5\right\rangle -i\left| 1_5\right\rangle )(\left|
948: 0_6\right\rangle +i\left| 1_6\right\rangle )$
949: \\ \hline $\overline{X}$ & $(J_x\sigma _1^x\otimes \sigma
950: _2^x+J_y\sigma _1^y\otimes \sigma _2^y)/(J_x+J_y)$ & $(J_y\sigma
951: _3^y\otimes \sigma _4^y+J_z\sigma _3^z\otimes \sigma
952: _4^z)/(J_y+J_z)$ & $( J_x\sigma _5^x\otimes \sigma _6^x+J_z\sigma
953: _5^z\otimes \sigma _6^z)/( J_x+J_z)$ \\ \hline $\overline{Z}$ &
954: $(\sigma _1^z-\sigma _2^z)/2$ & $(\sigma _3^x-\sigma _4^x)/2 $ &
955: $(\sigma _5^y-\sigma _6^y)/2$\\\hline\hline
956: \end{tabular}
957: \end{table*}
958:
959: \end{document}
960: