1: \documentclass[twocolumn,showpacs,showkeys,preprintnumbers,amsmath,amssymb,refcheck]{revtex4}
2: \usepackage{graphicx}
3: \usepackage{dcolumn}
4: \usepackage{bm}
5:
6: \newcommand{\bv}[1]{\mbox{\boldmath$#1$}}
7:
8: \begin{document}
9:
10: \preprint{ }
11: % quant-ph.yasushi.8809
12: % PaperId: quant-ph/0503067, PaperPassword: zmzmb (access still password restricted)
13: \title{
14: Hamiltonian of Homonucleus Molecules for NMR Quantum Computing}
15:
16: \author{Yasushi Kondo$^{1}$, Mikio Nakahara$^{1}$, Kazuya Hata$^{1}$,
17: and Shogo Tanimura$^{2}$\\}
18:
19: \affiliation{%
20: $^{1}$Department of Physics, Kinki University,
21: Higashi-Osaka 577-8502, Japan\\
22: $^2$Graduate School of Engineering, Osaka City University,
23: Sumiyoshi-ku, Osaka 558-8585, Japan\\
24: }%
25:
26: \date{\today}
27:
28: \begin{abstract}
29:
30: We derive the Hamiltonian in the rotating frame for NMR quantum computing
31: with homonucleus molecules as its computational resource. The Hamiltonian thus
32: obtained is different from conventional
33: Hamiltonians that appear in literature. It is shown that
34: control pulses designed for heteronucleus spins can be
35: translated to pulses for homonucleus spins
36: by simply replacing hard pulses by soft pulses with properly chosen pulse
37: width. To demonstrate the validity of our Hamiltonian, we conduct several
38: experiments employing cytosine as a homonucleus molecule.
39: All the experimental results
40: indicate that our Hamiltonian accurately describes the dynamics of the
41: spins and that the conventional Hamiltonian fails.
42: Finally we use our Hamiltonian for
43: precise control of field inhomogeneity compensation
44: with a pair of $\pi$-pulses.
45:
46:
47: \end{abstract}
48:
49: \pacs{03.67.Lx, 82.56.Jn}
50:
51: \keywords{NMR quantum computer,
52: homonucleus molecule,
53: Deutsch-Jozsa algorithm,
54: compensating pulse}
55:
56: \maketitle
57:
58: \section{Introduction}
59:
60: Quantum computation currently attracts a lot of attention
61: since it is expected to solve some of computationally
62: hard problems for a conventional digital computer~\cite{ref:1}.
63: Numerous realizations of a quantum computer have been proposed to date.
64: Among others, a liquid-state NMR
65: (nuclear magnetic resonance)
66: quantum computer is regarded as
67: most successful. Early experiments demonstrated
68: quantum teleportation~\cite{nmr1},
69: quantum search algorithm~\cite{nmr2},
70: quantum error correction~\cite{nmr3}, and
71: simulation of a quantum mechanical system~\cite{nmr4}.
72: Undoubtedly, demonstration of Shor's factorization algorithm~\cite{VSB01}
73: is one of the most remarkable achievements in NMR quantum computation.
74: Although the number of admissible qubits in a liquid-state NMR quantum
75: computer is suspected to be limited up to about ten due to poor
76: spin polarization at a room temperature,
77: a liquid-state NMR quantum computer
78: is one of few quantum computers that are capable of running
79: nontrivial quantum algorithms thanks to well established NMR technology.
80:
81: Since the number of qubits within heteronucleus spins
82: is practically limited to two or three,
83: the use of homonucleus spins is inevitable
84: if we try to equip an NMR with a large number of qubits.
85: It should be pointed out, however, that
86: liquid-state NMR of homonucleus molecules
87: is still poorly understood and literature dealing with this subject often
88: lacks solid ground.
89: Although the product operator formalism \cite{ernst} has been
90: extensively employed to implement quantum algorithms with
91: an NMR quantum computer,
92: people overlooked significance of the genuine Hamiltonian.
93: Actually there is a subtle difference between
94: the conventionally used Hamiltonian
95: and the proper Hamiltonian for homonucleus spins.
96: It is, therefore, urgently required to establish
97: theoretical foundation underlying
98: a liquid-state NMR quantum computer with homonucleus molecules.
99:
100: Suppose we would like to implement a quantum algorithm whose unitary
101: matrix representation is $U_{\rm alg}$.
102: If the Hamiltonian $ H $ depends on the control parameters, which
103: we write collectively as $ \gamma(t) $,
104: the time evolution operator is given by
105: \begin{equation}
106: U[\gamma(t)] = {\mathcal T} \exp\left[-i
107: \int_0^T H(\gamma(t)) dt\right],
108: \end{equation}
109: where ${\mathcal T}$ stands for the time-ordering product.
110: We use the natural unit in which $\hbar = 1 $.
111: Optimal control of the quantum computer requires
112: a control function $ \gamma(t) $
113: that produces the specified quantum algorithm
114: $ U[\gamma(t)] = U_{\rm alg} $
115: in the shortest possible time $T$.
116: Recently, numerical scheme to find the optimal control
117: has been worked out for fictitious Josephson junction qubits,
118: where polygonal paths in the parameter space has been
119: utilized~\cite{qaa4,qaa6}.
120: For time-optimal control of an NMR quantum computer,
121: another method employing the Cartan decomposition of SU($ 2^n $)
122: has been proposed~\cite{ref:kg}
123: and has been demonstrated
124: experimentally with a two-qubit heteronucleus molecule~\cite{qaa}.
125: We note that exact optimal control has been found
126: for holonomic quantum computation
127: in an idealized situation~\cite{qaa3.5}.
128:
129: This paper has three aims:
130: (1) to provide the theoretical foundation for an NMR quantum computer
131: with homonucleus molecules,
132: (2) to show that any pulse sequence designed for heteronucleus molecules
133: can be translated into that for homonucleus molecules, and
134: (3) to demonstrate experimentally that our Hamiltonian accurately describes the
135: dynamics of the spins.
136: For these purposes, we carefully examine the
137: Hamiltonians for NMR spin dynamics.
138: Although the Hamiltonian for a homonucleus molecule
139: is the same as the one for a heteronucleus molecule in the laboratory frame,
140: the former looks quite different from the latter in a rotating frame.
141:
142: This paper is organized as follows.
143: In section II
144: we study Hamiltonians of homonucleus as well as heteronucleus molecules.
145: We carefully examine how they are transformed in a rotating frame
146: and what is appropriate approximation to be employed. Surprisingly,
147: our resulting Hamiltonian is different from the conventional Hamiltonian.
148: In section III, we conduct several experiments to verify our analysis
149: by taking cytosine as an example of homonucleus molecules.
150: We execute the Deutsch-Jozsa algorithm, execute the pulse sequence
151: for pseudo-pure state preparation, and verify the robustness of two-qubit
152: entangling operations. As an application of the correct form of the
153: Hamiltonian we implement field inhomogeneity
154: compensation using a pair of $\pi$-pulses in section IV.
155: Section V is devoted to conclusions and discussion.
156:
157: \section{Hamiltonian in Rotating Frame}
158:
159: In this section, we write down
160: the Hamiltonian of spin dynamics in the laboratory frame
161: and transform it to the one in a rotating frame.
162: Although
163: the Hamiltonian for a homonucleus molecule
164: has the same form as the one for a heteronucleus molecule
165: in the laboratory frame,
166: Hamiltonians in a rotating frame differ from each other.
167:
168: We restrict ourselves within two-qubit molecules for simplicity.
169: Generalization to molecules with more qubits is straightforward.
170: As an example of heteronucleus molecules,
171: we refer to $^{13}$C-labeled chloroform.
172: The qubits are spins of $^{13}$C and H nuclei.
173: We take cytosine solved in D$_2$O as an example of homonucleus molecule.
174: The qubits are spins of two hydrogen nuclei (protons) in this case.
175:
176: \subsection{Heteronucleus molecule}
177:
178: \subsubsection{Experimental setup}
179:
180: A liquid-state NMR
181: consists of three parts as described in \cite{NMR_textbook}.
182: The first part is magnetic coils;
183: a superconducting coil to generate a homogeneous static magnetic field
184: and a normal conducting coil
185: to generate temporally controlled field gradients.
186: The second part contains resonance circuits
187: for applying radio frequency (rf) magnetic fields to the sample.
188: They are also used to pick up rf signals from the sample.
189: The third part is an assembly of electronic circuits to
190: feed rf pulses into the resonance circuits
191: and to detect the signals picked up by the coils.
192:
193: The NMR setup for heteronucleus molecules
194: is shown schematically in Fig.~\ref{hetero}.
195: The first and second spins have
196: respective resonance frequencies
197: $ \omega_{0,i} $ ($ i=1,2 $),
198: which are also called Larmor frequencies.
199: Their resonance frequencies are widely different for heteronucleus
200: molecules under consideration.
201: Hence two sets of resonance circuits and assembly of electronic circuits
202: are required.
203: The large difference of the resonance frequencies,
204: $ \Delta \omega_0 = \omega_{0,2} - \omega_{0,1} $,
205: allows us to address each spin individually with a short pulse.
206:
207: \begin{figure}[tb]
208: \includegraphics[bb=0 0 470 630,width=8cm]{hetero.eps}
209: \caption{
210: NMR setup for heteronucleus molecules.
211: \label{hetero}
212: }
213: \end{figure}
214:
215: The oscillator $i$ $(i=1,2) $
216: in the third part generates an rf electric wave with frequency
217: $ \omega_{{\rm rf},i} $.
218: The sequencer $i$ modulates the rf wave
219: to shape a designed pulse.
220: A typical temporal duration of a pulse,
221: which is called the pulse width,
222: is of the order of 10~$\mu$s.
223: The rf pulses are amplified and fed into the resonance coil $i$,
224: which generates rf magnetic fields
225: applied to the sample in the test tube.
226: Precession of spins in molecules appears as
227: rotation of magnetization of the sample
228: and induces a signal at the coil $i$.
229: The receiver $i$ detects the signal.
230: The directional coupler
231: prevents transmission of the rf pulse from
232: the amplifier to the receiver.
233:
234: \subsubsection{Heteronucleus molecule in rotating frame}
235: \label{h_hetero_rf}
236:
237: The two-qubit Hamiltonian in the laboratory frame is
238: \begin{equation}
239: \label{H in lab}
240: H = H_0 + H_{{\rm rf},1} + H_{{\rm rf},2}.
241: \end{equation}
242: Here the system Hamiltonian $H_0$ is defined as~\cite{rmp_chuang}
243: \begin{equation}
244: H_0 =
245: - \omega_{0,1} I_z \otimes I
246: - \omega_{0,2} I \otimes I_z
247: % \nonumber\\ & &
248: + \!\! \sum_{k=x,y,z} \!\! J I_k \otimes I_k,
249: \label{system H}
250: \end{equation}
251: where $I_k = \sigma_k/2$,
252: $\sigma_k$ being the $k$-th Pauli matrix, and
253: $I$ is the unit matrix of dimension two.
254: The first two terms in $ H_0 $ describe free precession of the spins
255: in a static magnetic field
256: while the third term describes the intramolecule spin interaction
257: with coupling strength $ J $.
258:
259: $ H_{{\rm rf}, i} $ ($i=1,2$) represents the action
260: of the rf magnetic field generated by the coil $ i $
261: and hence is called the control Hamiltonian.
262: Their explicit forms are
263: \begin{eqnarray}
264: H_{{\rm rf},1}
265: \!\! &=& \!\!
266: - 2 \omega_{1,1} \cos( \omega_{{\rm rf},1} t - \phi_1)
267: ( I_x \otimes I + g I \otimes I_x),
268: \label{h_rf_lab1}
269: \\
270: H_{{\rm rf},2}
271: \!\! &=& \!\!
272: - 2 \omega_{1,2} \cos( \omega_{{\rm rf},2} t - \phi_2)
273: ( g^{-1} I_x \otimes I + I \otimes I_x).
274: \;\;
275: \label{h_rf_lab2}
276: \end{eqnarray}
277: Here,
278: the amplitude of the rf pulse $ \omega_{1,i} $,
279: the frequency of the pulse $ \omega_{{\rm rf},i} $
280: and the phase of the pulse $\phi_i$
281: are controllable parameters.
282: We may assume, without loss of generality, that
283: the rf field is applied along the $x$-axis in the laboratory frame.
284: In the above equations
285: we introduced the ratio of resonance frequencies of the two nuclei,
286: \begin{equation}
287: g = \frac{\omega_{0,2}}{\omega_{0,1}}.
288: \label{g}
289: \end{equation}
290:
291: We shall examine the transformation law of the Hamiltonians
292: from the laboratory frame to a rotating frame.
293: The spin dynamics in the laboratory frame
294: is governed by the Liouville equation
295: \begin{equation}
296: i \frac{d \rho}{dt} = [ H, \rho ],
297: \end{equation}
298: where $\rho$ is the density matrix of the system under
299: consideration.
300: The unitary operator
301: \begin{equation}
302: \label{rot}
303: U =
304: \exp ( -i \omega_{{\rm rot},1} I_z t) \otimes
305: \exp ( -i \omega_{{\rm rot},2} I_z t)
306: \end{equation}
307: transforms $ \rho $ into
308: the density matrix $ \tilde{\rho} $ in the rotating frame as
309: \begin{equation}
310: \tilde{\rho} = U \rho \, U^{\dagger}.
311: \end{equation}
312: Note that we can choose the rotation angular velocities
313: $ \omega_{{\rm rot},i} $ ($i=1,2$) arbitrarily.
314: %We can take different values as
315: %$ \omega_{{\rm rot},1} \ne \omega_{{\rm rot},2} $.
316: The time evolution of the system is now governed by
317: \begin{eqnarray}
318: i \frac{ d \tilde{\rho} }{dt} = [ \tilde{H}, \tilde{\rho} ]
319: \end{eqnarray}
320: with the transformed Hamiltonian
321: \begin{eqnarray}
322: \tilde{H}
323: &=&
324: U H U^{\dagger}
325: - i U \frac{d}{dt}U^{\dagger}
326: = \tilde{H}_0
327: + \tilde{H}_{{\rm rf},1}
328: + \tilde{H}_{{\rm rf},2}.
329: \end{eqnarray}
330: Here the transformed system Hamiltonian is
331: \begin{eqnarray}
332: \label{h0_r}
333: \tilde{H}_0
334: &=&
335: U H_0 U^{\dagger} - iU \frac{d}{dt}U^{\dagger}
336: \nonumber\\
337: &=&
338: - ( \omega_{0,1} - \omega_{{\rm rot},1} ) I_z \otimes I
339: \nonumber \\ &&
340: - ( \omega_{0,2} - \omega_{{\rm rot},2} ) I \otimes I_z
341: + J I_z \otimes I_z
342: \nonumber \\
343: && + \left(
344: \begin{array}{cccc}
345: 0 & 0 & 0 & 0 \\
346: 0 & 0 & \frac{J}{2}e^{i \Delta \omega_{\rm rot} t} & 0 \\
347: 0 & \frac{J}{2}e^{-i \Delta \omega_{\rm rot} t} & 0 & 0 \\
348: 0 & 0 & 0 & 0
349: \end{array}
350: \right),
351: \end{eqnarray}
352: where
353: $ \Delta \omega_{\rm rot}
354: \equiv \omega_{{\rm rot},2} - \omega_{{\rm rot},1} $.
355: The transformed control Hamiltonians
356: $ \tilde{H}_{{\rm rf},i} $ will be given later.
357: If we take the frame co-moving with each spin, which has
358: the angular velocities $ \omega_{{\rm rot},i} = \omega_{0,i} $,
359: the first two terms in Eq.~(\ref{h0_r}) vanish.
360: In the case of heteronucleus molecules,
361: the condition $ \left| \Delta \omega_0 \right| \gg J $
362: is always satisfied
363: and thus the matrix elements in the last line also vanish
364: after averaging over time. For example,
365: $ \left| \Delta \omega_0 \right|/2 \pi \sim 400$~MHz while
366: $ J /2 \pi\sim 200$~Hz for
367: $^{13}$C-labeled chloroform at 11~T, for which
368: $\left| \Delta \omega_0 \right| / J \sim 10^6$.
369: Therefore $ \tilde{H}_0 $ is well approximated by
370: \begin{equation}
371: \label{h0_r2}
372: \tilde{H}_0 = J I_z \otimes I_z.
373: \end{equation}
374:
375: When the resonance and co-rotating conditions
376: $ \omega_{{\rm rf},i} = \omega_{0,i} = \omega_{{\rm rot},i} $
377: are satisfied,
378: the control Hamiltonians in the rotating frame
379: \begin{equation}
380: \tilde{H}_{{\rm rf},i} =
381: U H_{{\rm rf},i} U^{\dagger}
382: \end{equation}
383: are approximately given as
384: %%%%%%%% CHOICE OF PHI ? %%%%%%%%%%%%
385: \begin{eqnarray}
386: \tilde{H}_{{\rm rf},1} &=&
387: - \omega_{1,1}
388: ( \cos \phi_1 \, I_x \otimes I
389: + \sin \phi_1 \, I_y \otimes I ),
390: \label{h_rf_r1}
391: \\
392: \tilde{H}_{{\rm rf},2} &=&
393: - \omega_{1,2}
394: ( \cos \phi_2 \, I \otimes I_x
395: + \sin \phi_2 \, I \otimes I_y )
396: \label{h_rf_r2}
397: \end{eqnarray}
398: after dropping terms rapidly oscillating with
399: frequencies $ 2 \omega_{0,i} $ and $ \Delta \omega_0 $.
400: Note that the factor 2 in front of $ \omega_{1,i} $
401: in Eqs.~(\ref{h_rf_lab1}) and (\ref{h_rf_lab2})
402: has disappeared
403: in Eqs.~(\ref{h_rf_r1}) and (\ref{h_rf_r2}).
404: This is physically understood as discussed in \cite{NMR_textbook};
405: a linearly polarized rf magnetic field oscillating with frequency $ \omega_{\rm rf} $
406: is a superposition of two circularly polarized fields
407: with frequencies $ \pm \omega_{\rm rf} $
408: and the effect of the component with $- \omega_{\rm rf}$ is averaged to vanish.
409: It is also important to notice that a pulse with frequency
410: $ \omega_{{\rm rf},i} $ influences only the spin $i$
411: and does not affect the other spin in the rotating frame.
412: This is because $ \left| \Delta \omega_0 \right |$
413: is much larger than the inverse of the typical pulse width
414: $ \sim 1/(10 \, \mu s) \sim 100 $~kHz and
415: hence the rf pulse resonating with one spin
416: does not have spectral component which affects the other spin.
417:
418: In conclusion, the Hamiltonian for a heteronucleus molecule
419: in resonant magnetic fields
420: % of respective spin
421: is
422: \begin{eqnarray}
423: \label{hetero_ham}
424: \tilde{H}
425: &=& J I_z \otimes I_z
426: \nonumber\\
427: & & - \omega_{1,1}
428: ( \cos \phi_1 \, I_x \otimes I + \sin \phi_1 \, I_y \otimes I )
429: \nonumber\\
430: & & - \omega_{1,2}
431: ( \cos \phi_2 \, I \otimes I_x + \sin \phi_2 \, I \otimes I_y )
432: \end{eqnarray}
433: in the rotating frame that has
434: the angular velocities $ \omega_{{\rm rot},i} = \omega_{0,i} $.
435:
436:
437: \subsection{Homonucleus molecules}
438:
439: \subsubsection{Experimental setup}
440:
441: \begin{figure}[tb]
442: \includegraphics[bb=0 270 470 660,width=8cm]{homo2.eps}
443: \caption{
444: NMR setup for homonucleus molecules.
445: % with two oscillators for individual qubits.
446: \label{homo2}
447: }
448: \end{figure}
449: The NMR setup for homonucleus molecules
450: is shown schematically in Fig.~\ref{homo2}.
451: Because
452: the difference of the resonance frequencies
453: $ \Delta \omega_0 = \omega_{0,2} - \omega_{0,1} $
454: is not large compared to $\omega_{0, i}$ in this case,
455: a common resonance circuit and a power amplifier can be used to
456: control both spins. For cytosine in D$_2$O, for
457: example, we find
458: $ \left| \Delta \omega_0 \right| / 2 \pi \sim 765.0$~Hz
459: while $ \omega_{0,i} /2 \pi \sim 500$~MHz.
460: Although the difference $ \Delta \omega_0 $ is small,
461: it still allows us to address respective spins individually
462: provided that the pulse width is sufficiently long.
463:
464: The oscillator $ i $ generates a continuous rf electric wave with
465: frequency $ \omega_{{\rm rf},i} = \omega_{0,i} $.
466: The sequencer shapes the continuous wave into pulses.
467: When addressing the two spins simultaneously,
468: a typical pulse width is of the order of 10~$\mu$s.
469: On the other hand, when addressing them individually,
470: a typical pulse width is of the order of
471: $ 2 \pi/ \left| \Delta \omega_0 \right| \sim 1$~ms.
472: The rf pulses from the two sequencers are mixed and amplified.
473: The coil generates magnetic fields and picks up signals from the sample
474: and the receiver detects the signals.
475: Due to close resonance frequencies $\omega_{0, i}$,
476: only one set of resonance circuit and receiver is necessary
477: for homonucleus molecules.
478:
479:
480: \subsubsection{Homonucleus molecule in rotating frame}
481:
482: The Hamiltonian for homonucleus molecule in the laboratory frame
483: has the identical form to
484: the Hamiltonian for a heteronucleus molecule (\ref{H in lab}).
485: Even for homonucleus molecule
486: the condition $ \left| \Delta \omega_0 \right| \gg J $
487: is satisfied in general.
488: For example, in the case of cytosine in D$_2$O,
489: $ \left| \Delta \omega_0 \right| / 2 \pi \sim 765.0$~Hz
490: while $ J /2 \pi \sim 7.1$~Hz, and thus
491: the above condition is satisfied.
492: Therefore, the approximation
493: used in the derivation
494: of the system Hamiltonian (\ref{h0_r2}) for a heteronucleus molecule
495: is also applicable
496: to derivation of that for a homonucleus molecule.
497: Thus the system Hamiltonian of a homonucleus molecule
498: takes the form
499: \begin{equation}
500: \label{h0_homo}
501: \tilde{H}_0 = J I_z \otimes I_z.
502: \end{equation}
503: in the co-rotating frame of each spin.
504:
505: The control Hamiltonian $ \tilde{H}_{{\rm rf},i} $
506: describes the action of the resonant magnetic field
507: in the frame rotating with angular velocity
508: $ \omega_{{\rm rot},i} = \omega_{0,i} = \omega_{{\rm rf},i} $.
509: Corresponding Hamiltonian
510: $ \tilde{H}_{{\rm rf},i} $ for homonucleus molecule
511: is considerably more complicated
512: even when terms rapidly oscillating with frequencies
513: $ 2 \omega_{0,i} $
514: and $ \omega_{0,1} + \omega_{0,2} $ are averaged out as
515: \begin{eqnarray}
516: \tilde{H}_{{\rm rf},1}
517: &=& - \omega_{1,1}
518: \Big[
519: \cos \phi_1 \, I_x \otimes I
520: + \sin \phi_1 \, I_y \otimes I
521: \nonumber \\
522: && + g % \, \omega_{1,1}
523: \cos ( \Delta \omega_0 t + \phi_1) I \otimes I_x
524: \nonumber \\
525: && + g % \, \omega_{1,1}
526: \sin ( \Delta \omega_0 t + \phi_1) I \otimes I_y
527: \Big],
528: \label{h_rf_r_homo1}
529: \\
530: \tilde{H}_{{\rm rf},2}
531: &=& - \omega_{1,2}
532: \Big[
533: \cos \phi_2 \, I \otimes I_x
534: + \sin \phi_2 \, I \otimes I_y )
535: \nonumber \\
536: && + g^{-1} % \, \omega_{1,2}
537: \cos ( -\Delta \omega_0 t + \phi_2) I_x\otimes I
538: \nonumber \\
539: && + g^{-1} % \, \omega_{1,2}
540: \sin ( -\Delta \omega_0 t + \phi_2) I_y \otimes I
541: \Big].
542: \label{h_rf_r_homo2}
543: \end{eqnarray}
544:
545: If we further assume that the pulse width $\tau$ are
546: long enough so that even slowly oscillating terms
547: in Eqs.~(\ref{h_rf_r_homo1}) and (\ref{h_rf_r_homo2}),
548: which contain $ \Delta \omega_0 $,
549: are averaged out, then
550: Eqs.~(\ref{h_rf_r_homo1}) and (\ref{h_rf_r_homo2})
551: reduce to Eqs.~(\ref{h_rf_r1}) and (\ref{h_rf_r2}).
552: Simultaneously, we can tune the pulse width $\tau$ short enough
553: ($J \tau \ll 1$)
554: so that the spin-spin interaction (\ref{h0_homo}) is negligible
555: while pulses are applied. Therefore we conclude that
556: an arbitrary pulse sequence designed for
557: heteronucleus molecules works
558: for homonucleus ones provided that
559: all the hard pulses are replaced by soft pulses
560: whose pulse width $ \tau $ satisfies the condition
561: $ 2 \pi/ \left| \Delta \omega_0 \right| < \tau \ll 2 \pi/ J $.
562: We will demonstrate this consequence experimentally in the next section,
563: where we set $\tau = 4 (2 \pi/ \left| \Delta \omega_0 \right|) = 5.229$~ms
564: $\ll 2 \pi/ J =140.8$~ms.
565:
566: \subsection{Conventional Hamiltonians}
567:
568: Here we make comparison between the Hamiltonians derived in the previous
569: subsection and the Hamiltonian for homonucleus molecules
570: used in literature.
571: Conventionally the system Hamiltonian
572: \begin{equation}
573: \label{h0_homo3}
574: \tilde{H}_{{\rm conv},0} =
575: - \Delta \omega_0 I \otimes I_z + J I_z \otimes I_z
576: \end{equation}
577: is used to describe
578: spin dynamics
579: in a static magnetic field in a rotating frame \cite{cory2}.
580: Apparently, it differs from our Hamiltonians (\ref{h0_homo}).
581:
582: We suspect that
583: the Hamiltonian (\ref{h0_homo3}) may be derived
584: from the original system Hamiltonian (\ref{system H})
585: via transformation from the laboratory frame
586: to the frame rotating with a common angular velocity
587: $ \omega_{{\rm rot},1} = \omega_{{\rm rot},2} = \omega_{0,1} $.
588: We will show, however, that this choice does not yield the Hamiltonian
589: (\ref{h0_homo3}) in the rotating frame.
590:
591: If we take a frame that rotates with
592: a common angular velocity equal to
593: %$ \omega_{{\rm rot},1} = \omega_{{\rm rot},2} =
594: $\omega_{0,1} $
595: for both spins,
596: transformation operator (\ref{rot}) becomes
597: \begin{equation}
598: \label{rot1}
599: U_{\rm com} =
600: \exp( - i \omega_{0,1} I_z t) \otimes
601: \exp( - i \omega_{0,1} I_z t).
602: \end{equation}
603: Then the system Hamiltonian (\ref{system H}) is transformed into
604: \begin{eqnarray}
605: \label{h0_homo2}
606: \tilde{H}_{{\rm com}, 0}
607: &=&
608: U_{\rm com} H_0 U_{\rm com}^{\dagger}
609: - i U_{\rm com} \frac{d}{dt} U_{\rm com}^{\dagger}
610: \nonumber\\
611: &=&
612: - \Delta \omega_0 I \otimes I_z
613: + J I_z \otimes I_z
614: \nonumber \\ &&
615: + \left(
616: \begin{array}{cccc}
617: 0 & 0 & 0 & 0 \\
618: 0 & 0 & \frac{J}{2} & 0 \\
619: 0 & \frac{J}{2} & 0 & 0 \\
620: 0 & 0 & 0 & 0
621: \end{array}
622: \right),
623: \end{eqnarray}
624: which does not agree with
625: the conventional Hamiltonian (\ref{h0_homo3}).
626:
627: Another system Hamiltonian in the laboratory frame
628: \begin{equation}
629: \label{h0_homo4}
630: H_{{\rm conv},0} =
631: - \omega_{0,1} I_z \otimes I
632: - \omega_{0,2} I \otimes I_z
633: + J I_z \otimes I_z
634: \end{equation}
635: is also sometimes employed in literature \cite{rmp_chuang,cory,raedt}
636: but this is also different from
637: the original system Hamiltonian (\ref{system H}).
638: We cannot take the Hamiltonian (\ref{h0_homo4}) as a correct one
639: since we cannot replace
640: $ \sum_k J I_k\otimes I_k $ by $ J I_z\otimes I_z $
641: in the laboratory frame.
642:
643: To illustrate the difference between our Hamiltonian
644: and the conventional Hamiltonian, let us consider the unitary gate
645: \begin{eqnarray}
646: \label{c-phase-gate}
647: U_{\rm E}
648: &=& \exp ( -i \pi I_z \otimes I_z )
649: \nonumber \\
650: &=& \left(
651: \begin{array}{cccc}
652: e^{-i \pi/4} & 0 & 0 & 0 \\
653: 0 & e^{i \pi/4} & 0 & 0 \\
654: 0 & 0 & e^{i \pi/4} & 0 \\
655: 0 & 0 & 0 & e^{-i \pi/4}
656: \end{array}
657: \right),
658: \end{eqnarray}
659: which is employed along with one-qubit operations
660: to implement the controlled-NOT
661: gate \cite{ref:1}.
662: We implement the gate $ U_{\rm E} $ from our system Hamiltonian
663: %$ \tilde{H}_0 $,
664: (\ref{h0_homo}), as
665: \begin{equation}
666: \label{UJ}
667: U_{J} (\pi/J)
668: = \exp( -i \pi \tilde{H}_0/J )
669: = \exp( -i \pi I_z \otimes I_z).
670: \end{equation}
671: In other words, we simply wait
672: for a time interval $T_J = \pi/J $
673: without applying any rf pulses.
674: % where we take Eq.~(\ref{h0_homo}) as $H_0^{rot}$.
675: Let us define the distance between $ U_{\rm E} $ and $ U_J(t) $ as
676: \begin{equation}
677: \| U_{\rm E} - U_J(t) \|
678: \equiv
679: \sqrt{ {\rm tr}
680: [ (U_{\rm E} - U_J(t))^{\dagger} (U_{\rm E} - U_J(t)) ] }.
681: \end{equation}
682: %It is easy to verify that the distance thus defined satisfies all
683: %the axioms of norm.
684: This is easily evaluated as
685: \begin{equation}
686: \| U_{\rm E} - U_{J}(t) \|
687: =
688: 2 \sqrt{2}
689: \sqrt{ 1 - \cos \frac{1}{4} (J t - \pi) }.
690: \label{error1}
691: \end{equation}
692: We observe that the distance vanishes at $ t=T_J$ so that
693: $ U_J(T_J) = U_{\rm E}$.
694: We note also that the distance remains close to zero in the vicinity
695: $t \sim T_J$.
696: This robust character of $ U_J(t) $ was clearly observed in our experiment
697: as shown in the next section.
698:
699: On the other hand,
700: if we replace $\tilde{H}_0$ in Eq. (\ref{UJ}) by the conventional
701: Hamiltonian (\ref{h0_homo3}),
702: the distance between $ U_{\rm E} $ and $ U_J (t) $ becomes
703: \begin{eqnarray}
704: \lefteqn{ \| U_{\rm E} - U_{{\rm conv} J}(t) \| }
705: \nonumber\\
706: &=& 2 \sqrt{2}
707: \sqrt{ 1-
708: \cos \Big( \frac{\Delta \omega_0 t}{2} \Big)
709: \cos \frac{1}{4} ( Jt - \pi )
710: }.
711: \label{error2}
712: \end{eqnarray}
713: Therefore, if the conventional Hamiltonian (\ref{h0_homo3})
714: were a correct one to describe the spin dynamics,
715: $ U_{{\rm conv}J} (t) $ would not coincide with $ U_{\rm E} $
716: at $ t = T_J$
717: and the distance
718: $ \| U_{\rm E} - U_{{\rm conv}J}(t) \| $
719: should oscillate in the vicinity of the $t \sim T_J$.
720: However,
721: such a rapid oscillation in time has not been observed in our experiment.
722:
723: \section{Experiments}
724: \label{exp}
725:
726: \subsection{Spectrometer and molecules}
727:
728: All the data were taken at room
729: temperature with a JEOL ECA-500 spectrometer \cite{jeol}, where
730: the hydrogen Larmor frequency is approximately 500~MHz.
731:
732: We used 0.6 mL, 23 mM sample of cytosine \cite{cytosine}
733: solved in D$_2$O.
734: The measured coupling strength is $J/2 \pi=7.1$~Hz
735: while the frequency difference is
736: $\left| \Delta \omega_0 \right| / 2 \pi = 765.0$~Hz.
737: The transverse relaxation time $T_2$ is measured to be
738: $\sim 1$~s for both hydrogen nuclei and
739: the longitudinal relaxation time $T_1$ is $\sim 7$~s.
740:
741: In order to measure the spin states,
742: we apply a reading pulse to only one spin, called spin 1,
743: and then obtained the spectrum by Fourier transforming the
744: free induction decay (FID) signal.
745: The state of the spin 1 is read from the sign of the peak in the spectrum
746: while the state of the other nucleus (spin 2) is
747: found from the peak position.
748:
749:
750: \subsection{Deutsch-Jozsa algorithm}
751:
752: The Deutsch-Jozsa (DJ) algorithm \cite{DJ} is one of the simplest
753: quantum algorithms
754: that illustrate the power of quantum computation
755: and has been implemented by several groups \cite{13c,cytosine}.
756: Let us consider a one-bit function $f: \{0, 1\} \to \{0, 1\}$.
757: For a two-qubit register,
758: there are only four possibilities for $f$, whose
759: explicit forms are
760: $f_1(0) = f_1(1) = 0$, $f_2(0) = f_2(1)= 1$, $f_3(0) = 0, f_3(1) = 1$
761: and $f_4(0) = 1, f_4(1) = 0$. The former two functions are said to be
762: ``constant'' while the latter two are ``balanced''.
763: With the DJ algorithm, we can tell whether a
764: given unknown function $f$ is constant or balanced
765: via only a single trial.
766:
767: Chuang {\it et al.} \cite{13c}
768: employed carbon-13 labeled chloroform, a
769: heteronucleus molecule, as a computational resource while
770: Jones and Mosca \cite{cytosine} used cytosine,
771: a homonucleus molecule, to execute the DJ algorithm.
772: Chuang {\it et al.} executed the DJ algorithm
773: using the pulse sequences shown in Table~\ref{dj_seq}.
774: According to our previous discussions, it should be possible to use
775: the pulse sequences of Chuang {\it et al.} for cytosine molecules
776: by simply replacing hard pulses with soft ones.
777:
778: The results of our quantum computations with cytosine are
779: summarized in Figs. \ref{dj_j} and \ref{dj_p}.
780: We started the computation with the thermal equilibrium state
781: since the DJ algorithm does not require a
782: pure initial state~\cite{13c}.
783:
784: \begin{table}[t]
785: \caption{
786: \label{dj_seq}
787: Control pulse sequences for the Deutsch-Jozsa algorithm
788: taken from the reference \cite{13c}. The functions $f_i(x)$ are
789: defined in the text.
790: Here, $[\frac{\pi}{2}]_{i}$ denotes
791: the $\pi/2$ pulses around the $i$-axes $(i=x, y, -x, -y)$.
792: The symbol $[\pi]$ denotes the $\pi$ pulse
793: around the $x$-axis. (1/$nJ$) denotes the two-qubit entangling operation
794: produced by turning off the rf pulses during the interval $2\pi/nJ$.
795: The pulse sequences are followed by a readout $\pi/2$ pulse
796: around the $x$-axis to the spin 1,
797: which is not shown in the Table.
798: The receiver detects the FID (free induction decay) signal
799: to measure the spin state.
800: }
801: \begin{tabular}{llllllll}
802: \hline
803: \hline
804: \multicolumn{1}{c}{Gate \hspace{3mm} } &
805: \multicolumn{7}{c}{Pulse sequence} \\
806: \hline
807: \hspace{2mm} $f_1$ &&&&&&& \\
808: spin 1 \hspace{4mm}
809: & $[\frac{\pi}{2}]_y$ &--&(1/4$J$)& -- &(1/4$J$) & --
810: & $[\frac{\pi}{2}]_{-y}$ \\
811: spin 2 & $[\frac{\pi}{2}]_{-y}$ &--&(1/4$J$)& $[\pi]$ &(1/4$J$)& $[\pi]$
812: & $[\frac{\pi}{2}]_y$ \\
813: \hline
814: \hspace{2mm} $f_2$ &&&&&&&\\
815: spin 1 & $[\frac{\pi}{2}]_y$ &-- &(1/4$J$) & -- &(1/4$J$) & --
816: & $[\frac{\pi}{2}]_{-y}$ \\
817: spin 2 & $[\frac{\pi}{2}]_{-y}$ &--&(1/4$J$)& $[\pi]$ &(1/4$J$)& --
818: & $[\frac{\pi}{2}]_y$ \\
819: \hline
820: \hspace{2mm} $f_3$ &&&&&&& \\
821: spin 1 & $[\frac{\pi}{2}]_y$ & -- & (1/2$J$) & $[\frac{\pi}{2}]_{-y}$
822: & $[\frac{\pi}{2}]_{-x}$& $[\frac{\pi}{2}]_y$
823: & $[\frac{\pi}{2}]_{-y}$ \\
824: spin 2 & $[\frac{\pi}{2}]_{-y}$ & $[\frac{\pi}{2}]_y$ & (1/2$J$) & $[\frac{\pi}{2}]_{-y}$
825: & $[\frac{\pi}{2}]_{x}$ &-- & $[\frac{\pi}{2}]_y$ \\
826: \hline
827: \hspace{2mm} $f_4$ &&&&&&&\\
828: spin 1 & $[\frac{\pi}{2}]_y$ &-- & (1/2$J$) & $[\frac{\pi}{2}]_{-y}$
829: & $[\frac{\pi}{2}]_{-x}$ &$[\frac{\pi}{2}]_y$ & $[\frac{\pi}{2}]_{-y}$ \\
830: spin 2 & $[\frac{\pi}{2}]_{-y}$ & $[\frac{\pi}{2}]_y$ & (1/2$J$) & $[\frac{\pi}{2}]_{-y}$
831: & $[\frac{\pi}{2}]_{-x}$ & -- & $[\frac{\pi}{2}]_y$ \\
832: \hline
833: \hline
834: \end{tabular}
835: \end{table}
836:
837:
838: \begin{figure}[b]
839: \includegraphics[bb=30 0 550 580,width=7.5cm]{dj_j.eps}
840: \caption{
841: The FID spectra of the spin 1 in cytosine
842: showing output of the DJ algorithms.
843: The sign of each peak indicates the state of the spin 1.
844: The location of each peak indicates the state of the spin 2.
845: When the initial state of the spin 2 is $|0\rangle$,
846: it causes a larger shift to the resonance frequency of the spin 1
847: and contributes to a left peak in each curve.
848: Then the sign of the left peak
849: discriminates whether $f_i$ is constant or balanced.
850: The left peak is
851: positive for $f_1$ and $f_2$ (constant) and
852: negative for $f_3$ and $f_4$ (balanced).
853: The numbers in the left side
854: are durations of the two-qubit operations in ms.
855: The correct duration of the two-qubit operations
856: is $T_J = 70.4$~ms.
857: Note that the spectra are insensitive to variation
858: of the two-qubit operation time.
859: \label{dj_j}}
860: \end{figure}
861:
862: \subsubsection{$J$-coupling time\label{J-c}}
863:
864: The DJ algorithm employs the $ J $-coupling unitary operator
865: $ U_J(t) $
866: with
867: $ t = 2 \pi / 4J = T_J/2$ or $ t = 2 \pi / 2J =T_J$
868: to entangle two spins.
869: The time durations for the two-qubit operations are
870: $ 2 \pi/4J + 2 \pi/4J = T_J $ for $f_1$ and $f_2$ and
871: $T_J$ for $f_3$ and $f_4$.
872: Thus the total execution time is $T_J$ for all four cases.
873:
874: As we discussed when we derived Eq. (\ref{error1}),
875: our Hamiltonian (\ref{h0_homo}) predicts that
876: $ U_J(t) $ does not deviate much from the desired unitary transformation
877: even when the gate operation time $ t $ deviates
878: from the correct value.
879: On the other hand, Eq. (\ref{error2}) tells us that
880: the conventional Hamiltonian (\ref{h0_homo3}) predicts that
881: $ U_J(t) $ sharply depends on the timing $ t $ and oscillates as
882: %It tells that $ U_J(t) $ fluctuates as
883: $ \cos (\Delta \omega_0 t/2) $.
884:
885: In experiment we executed the DJ algorithm
886: with various gate operation time $t$ in the vicinity of $T_J$
887: and observed how the resulting spectra depend
888: on $t$.
889: We employed the pulse sequences shown in Table~\ref{dj_seq},
890: in which all hard pulses used by Chuang {\it et al.} \cite{13c} were
891: replaced with Gaussian soft pulses with the pulse width 5.229~ms.
892:
893: The initial state of the molecules
894: is a thermal mixture of four states $|00\rangle$, $|01\rangle$,
895: $|10\rangle$, and $|11\rangle$.
896: The DJ algorithm does not work when the second qubit is $|1\rangle$
897: and fails to distinguish constant from balanced.
898: On the other hand,
899: it works regardless of the state of the first qubit.
900: In Fig.~\ref{dj_j}
901: the peaks with a smaller frequency shift (the right peaks)
902: are outputs from the initial states
903: $ | 01 \rangle $ or
904: $ | 11 \rangle $.
905: In this case the algorithm fails
906: to distinguish if $f_i$ is constant or balanced.
907: % On the other hand,
908: The peaks with a larger frequency shift (the left peaks)
909: in Fig.~\ref{dj_j}
910: are outputs from the initial states
911: $ | 00 \rangle $ or
912: $ | 10 \rangle $.
913: In this case the DJ algorithm successfully
914: tells us if $f_i$ is constant or balanced
915: by the sign of the peak (positive for $f_1$ and $f_2$ while negative for $f_3$
916: and $f_4$).
917:
918: We varied the $ J $-coupling time interval $ t $
919: in the range from 69.8~ms to 71.0~ms.
920: In other words,
921: $ \Delta \omega_0 t /2 $
922: was swept between
923: $ 26.7 \times 2 \pi$ and $ 27.2 \times 2 \pi $.
924: The exact duration to produce the designed unitary operator
925: correctly is 70.4~ms.
926: We observe from Fig.~\ref{dj_j} that
927: the spectra are not sensitive to variation of the time interval.
928: Therefore we concluded that our Hamiltonian (\ref{h0_homo})
929: accounts for the experimental results consistently.
930:
931: \subsubsection{Rf pulse width}
932:
933: In literature \cite{cytosine, raedt} it is recommended to use
934: soft pulses whose width is an integral multiple
935: of $ 2 \pi / \left| \Delta \omega_0 \right| $
936: in order to avoid undesirable effect caused by the term
937: $ \Delta \omega_0 I \otimes I_z $
938: in the conventional Hamiltonian (\ref{h0_homo3}).
939: Our discussion and experiment show that
940: this tuning is not necessary since
941: the relevant Hamiltonian (\ref{h0_homo}) does not contain
942: the term $ \Delta \omega_0 I \otimes I_z $.
943:
944: We executed the DJ algorithms shown in Table~\ref{dj_seq},
945: with different pulse width (5.229 and 6.217~ms).
946: In our setting,
947: the pulse width 5.229~ms is equal to
948: $ 4 \times 2 \pi / \left| \Delta \omega_0 \right|$,
949: while 6.217~ms is $ 4.76 \times 2 \pi / \left| \Delta \omega_0 \right|$.
950: The measured FID spectra of the spin 1 are shown in Fig.~\ref{dj_p}.
951: No significant changes in the spectra appeared
952: even if we tuned the pulse width to a fractional multiple of
953: $ 2 \pi/ \left| \Delta \omega_0 \right|$.
954: This result proves that
955: the pulse width need not be
956: an integral multiple of $ 2 \pi / \left| \Delta \omega_0 \right| $
957: to implement a given gate and obtain a reasonable spectrum.
958:
959:
960: \begin{figure}[b]
961: \includegraphics[bb=50 50 450 450,width=7.5cm]{dj_p.eps}
962: \caption{
963: The effect of variation of the pulse widths on the spectrum of the DJ algorithms.
964: The pulse widths for spins 1 and 2 are
965: (a) 5.229~ms and 5.229~ms,
966: (b) 6.217~ms and 6.217~ms, and
967: (c) 5.229~ms and 6.217~ms, respectively.
968: Observe that the spectra are insensitive to the variation of pulse width.
969: \label{dj_p}}
970: \end{figure}
971:
972: \section{Field Inhomogeneity Compensation}
973:
974: Here we discuss an experiment
975: to reveal the nature of the Hamiltonians
976: (\ref{h_rf_r_homo1}) and
977: (\ref{h_rf_r_homo2}),
978: which depict the action of the oscillating magnetic fields
979: on the spins.
980: It is common to employ the compensating pulse method to
981: suppress errors induced by field inhomogeneity.
982: We will show, by employing our Hamiltonian, that
983: the entangling operation with the $J$-coupling
984: is fragile in the presence of
985: the compensating pulses and a fine tuning of the gate operation time is
986: required.
987:
988: \subsection{$\pi$-pulse pair in J-coupling time}
989:
990: We have shown in the previous section that
991: it is not necessary to tune the $ J $-coupling time very accurately
992: since this operation is robust against small change of the
993: gate operation time.
994: However, if the system is under the influence of field inhomogeneity,
995: it may cause an error during the $ J $-coupling time.
996:
997: It is well known that this undesired effect caused by
998: field inhomogeneity
999: can be compensated by a series of hard
1000: $\pi$-pulse pair, whose width is of the order of 10~$\mu$s.
1001: The best known example may be
1002: the CPMG (Carr-Purcell-Meiboom-Gill) pulse sequence~\cite{NMR_textbook}.
1003: Let us apply this technique to $U_{J} (t) $.
1004: Then the pulse sequence for $ U_J(t) $ is replaced with
1005: \begin{equation}
1006: U_{J*}(t) \, : \, U_J(t/2)-[\pi]-U_J(t/2)-[\pi],
1007: \label{J star pulse}
1008: \end{equation}
1009: where time flows from left to right and
1010: $[\pi]$ denotes a hard pulse
1011: that rotates both spins by $\pi$ radian. We take the rotation axis
1012: to be
1013: the $x$-axis of one of the spins, for example the spin 1, while
1014: the rotation axis for the other spin, the spin 2, depends on the time when
1015: the $[\pi]$ is applied, according to the Hamiltonian (\ref{h_rf_r_homo1}).
1016: The pulse sequence is represented as the product of unitary matrices
1017: \begin{equation}
1018: U_{J*}(t) =
1019: U_{1,\pi}^{(2)} \, U_J(t/2) \, U_{1,\pi}^{(1)} \, U_J(t/2) ,
1020: \end{equation}
1021: where
1022: \begin{eqnarray}
1023: U_{1,\pi}^{(1)}
1024: & = &
1025: \exp ( -i \pi I_x \otimes I )
1026: \\ \nonumber
1027: & \times &
1028: \exp (
1029: -i \pi [
1030: \cos (\Delta \omega_0 \frac{t}{2}) I \otimes I_x
1031: + \sin (\Delta \omega_0 \frac{t}{2}) I \otimes I_y
1032: ] ) , \\
1033: U_{1,\pi}^{(2)}
1034: & = &
1035: \exp ( -i \pi I_x \otimes I )
1036: \\ \nonumber
1037: & \times &
1038: \exp \left(
1039: -i \pi [
1040: \cos (\Delta \omega_0 t) I \otimes I_x
1041: + \sin (\Delta \omega_0 t) I \otimes I_y
1042: ] \right) .
1043: \end{eqnarray}
1044: Note that
1045: we put the ratio of resonance frequencies
1046: $ g = \omega_{0,2} / \omega_{0,1} = 1 $
1047: for the homonucleus molecule.
1048: The resulting operator $ U_{J*} (t) $
1049: does not coincides with $ U_J (t) $.
1050: The distance between $ U_{J*}(t) $ and $ U_{\rm E}$ is evaluated as
1051: \begin{eqnarray}
1052: \label{error3}
1053: \lefteqn{
1054: \| U_{\rm E} - U_{J*}(t) \|
1055: }
1056: \nonumber\\
1057: &=& 2 \sqrt{2}
1058: \sqrt{1-
1059: \cos \Big( \frac{\Delta \omega_0 t}{2} \Big)
1060: \cos \frac{1}{4} (Jt - \pi)
1061: }.
1062: \end{eqnarray}
1063: The distance does not vanish generally
1064: because the two conditions,
1065: $ \cos ( \Delta \omega_0 t / 2 ) = \pm 1 $ and
1066: $ \cos ( J t/4 - \pi/4 ) = \pm 1 $,
1067: are rarely satisfied simultaneously.
1068: Moreover, the distance is very sensitive to $t$.
1069:
1070:
1071:
1072: \begin{table}[b]
1073: \caption{
1074: \label{tb_pps}
1075: Control pulse sequences to create the pseudo-pure state $|00\rangle$
1076: \cite{pps_j}.
1077: Here, $[\frac{\pi}{m}]_{i}$ denotes
1078: the $ \frac{\pi}{m} $ pulse around the $i$-axes $(i=x, y, -x, -y)$
1079: while
1080: (1/2$J$) denotes the two-qubit entangling operation
1081: implemented by turning off the rf pulses during the interval $2\pi/2J$.
1082: The symbol
1083: FG denotes application of a pulsed field gradient for spatial labeling.
1084: The pulse sequence is followed
1085: by a readout $\pi/2$-pulse around the $x$-axis to the spin 1,
1086: which is not shown in the Table.
1087: }
1088: \begin{tabular}{lcccccc}
1089: \hline
1090: \hline
1091: \multicolumn{1}{c}{} &
1092: \multicolumn{6}{c}{Pulse sequence} \\
1093: \hline
1094: spin 1 \hspace{3mm}
1095: & & FG & $[\frac{\pi}{4}]_x $&(1/2$J$) &$[\frac{\pi}{4}]_{-y}$&FG \\
1096: spin 2
1097: & $[\frac{\pi}{3}]_x$ & FG & &(1/2$J$) & & FG \\
1098: \hline
1099: \hline
1100: \end{tabular}
1101: \end{table}
1102:
1103: \subsection{Pseudo-pure state preparation}
1104:
1105: \begin{figure}[t]
1106: \includegraphics[bb=0 0 585 250,width=7.5cm]{pps.eps}
1107: % file 040801-015, 016, 012, 017, 018 == 125, 107, 108, 109, 124
1108: \caption{
1109: The effect of variation of the $J$-coupling time on the spectrum of the
1110: pseudo-pure state creation.
1111: The spectra in the left panel are measured
1112: without the $\pi$-pulse pair in the entangling operation.
1113: The spectra in the right panel are measured
1114: with the $\pi$-pulse pair.
1115: The numbers are the two-qubit operation time in ms.
1116: Note that the spectra
1117: generated with the $\pi$-pulse pair are
1118: sensitive to variation of the two-qubit operation time.
1119: \label{fig:pps}}
1120: \end{figure}
1121:
1122: An NMR quantum computer must be initialized to the pseudo-pure state
1123: $ |00 \rangle$ before executing a specific algorithm.
1124: This initialization procedure is implemented with
1125: the pulse sequence~\cite{pps_j} shown in Table~\ref{tb_pps}.
1126: In this section we examine the effect of the compensating $ \pi $ pulses
1127: on the initialization procedure.
1128:
1129: The initialization process contains the $ J $-coupling time,
1130: in which the two spins are entangled by the two-qubit operation $ U_J(t) $.
1131: We vary the gate operation time
1132: $t$ to introduce an operational error on purpose.
1133: It is possible, however, to apply a pair of $ \pi $-pulses to
1134: compensate this error while the $J$-coupling is under action as
1135: instructed in Eq. (\ref{J star pulse}).
1136: We have swept the gate operation time between 69.3~ms and 70.6~ms and
1137: the results are summarized in Fig.~\ref{fig:pps}.
1138:
1139: The spectra in the left panel of Fig.~\ref{fig:pps}
1140: were measured {\it without applying} the $ \pi $-pulse pair
1141: in the entangling operation.
1142: We observed that the spectra are robust
1143: against small variations of the gate operation time.
1144: The intense peaks with a larger frequency shift
1145: are signals from molecules in the $ |00\rangle $ state while
1146: the smaller peaks are error signals from
1147: small amount of molecules not in the $ |00 \rangle $ state.
1148:
1149: The spectra in the right panel %of Fig.~\ref{fig:pps}
1150: were measured with the $\pi $-pulse pair applied
1151: during the entangling operation.
1152: We observed that the spectra are very sensitive to the variation
1153: of the $ J $-coupling operation time,
1154: although it is still possible to adjust the operation time so that the
1155: desired pseudo-pure state is produced with a good precision.
1156: When the gate operation time was set at the correct value $ t = 70.3$~ms,
1157: the spectrum exhibited a sharper peak than those in the left panel.
1158: This result implies
1159: that the $ \pi $-pulse pair improved the quality of the initialized state.
1160: However, the spectra were fragile
1161: when the duration $t$ deviates from the correct value.
1162: For example, when the duration was set at $ t=69.7$~ms,
1163: the intense signal indicated that
1164: the most of molecules are in the $ |01 \rangle $ state, and
1165: not in the desired state $|00\rangle $.
1166:
1167: Thus we conclude that
1168: our Hamiltonian (\ref{h_rf_r_homo1}) and (\ref{h_rf_r_homo2})
1169: accurately account for the experimental results.
1170:
1171:
1172: \vspace{4ex}
1173: \section{Conclusions and Discussion}
1174:
1175: We have derived the relevant Hamiltonian for homonucleus molecules
1176: in NMR quantum computing and shown that any pulse sequence for
1177: a heteronucleus molecule may be translated into that for a homonucleus
1178: molecule by simply replacing hard pulses by soft pulses with
1179: a properly chosen pulse width.
1180: We have demonstrated that the NMR spectra in several experiments are
1181: accounted for with our Hamiltonian but not with the conventional
1182: Hamiltonian found in literature. It was shown in our experiments that
1183: the spectra are robust
1184: under small variations of the $J$-coupling operation time
1185: as well as of the rf pulse widths.
1186: Moreover, we provided the theoretical basis for
1187: field inhomogeneity compensation by a pair of hard $\pi$ pulses
1188: during the entangling operation and verified it experimentally.
1189:
1190: Generalization of the present work to molecules with more spins
1191: is straightforward. It is easy to find proper pulse sequence, either
1192: numerically \cite{qaa} or by Cartan decomposition \cite{ref:kg, warp},
1193: once an exact form of the
1194: Hamiltonian is obtained. Theoretical analysis as well as experiments
1195: on these subjects are under progress and will be published elsewhere.
1196:
1197: \section*{Acknowledgments}
1198:
1199: We would like to thank Manabu Ishifune for sample preparation,
1200: Toshie Minematsu for assistance in
1201: NMR operations and Katsuo Asakura and Naoyuki Fujii of JEOL
1202: for assistance in NMR pulse programming.
1203: MN would like to thank partial supports of Grant-in-Aids for
1204: Scientific Research from the Ministry of Education,
1205: Culture, Sports, Science and Technology, Japan, Grant No.~13135215 and
1206: from Japan Society for the Promotion of Science, Grant No.~14540346.
1207: ST is partially supported by the Ministry of Education, Grant No.~15540277.
1208:
1209: \begin{thebibliography}{99}
1210: \bibitem{ref:1} M.\ A.\ Nielsen and I.\ L.\ Chuang,
1211: {\it Quantum Computation and Quantum Information}
1212: (Cambridge University Press, Cambridge, 2000).
1213: \bibitem {nmr1}
1214: M.\ A.\ Nielsen, E.\ Knill, and R.\ Laflamme,
1215: Nature {\bf 396} 52 %-55
1216: (1998).
1217: \bibitem {nmr2}
1218: I.\ L.\ Chuang, N.\ Gershenfeld, and M.\ Kubinec,
1219: Phys.\ Rev.\ Lett.\ {\bf 80} 3408 %-2411
1220: (1998).
1221: \bibitem {nmr3}
1222: D.\ G.\ Cory, M.\ D.\ Price, W.\ Maas, E.\ Knill, R.\ Laflamme,
1223: W.\ H.\ Zurek, T.\ F.\ Havel, and S.\ S.\ Somaroo.
1224: Phys.\ Rev.\ Lett.\ {\bf 81} 2152 %-2155
1225: (1998).
1226: \bibitem {nmr4}
1227: S.\ Somaroo, C.\ H.\ Tseng, T.\ F.\ Havel, R.\ Laflamme,
1228: and D.\ G.\ Cory,
1229: Phys.\ Rev.\ Lett.\ {\bf 82} 5381 %-5384
1230: (1999).
1231: \bibitem {VSB01} L.\ M.\ K.\ Vandersypen, M.\ Steffen, G.\ Breyta, C.\
1232: S.\ Yannonl, M.\ H.\ Sherwood, and I.\ L.\ Chuang,
1233: Nature {\bf 393} 143 %-146
1234: (1998).
1235: \bibitem{ernst}
1236: R.\ R. Ernst, G.\ Bodenhausen, and A.\ Wokaun,
1237: {\it Principles of Nuclear Magnetic Resonance in One and Two
1238: Dimensions} (Oxford University Press, Oxford, 1991).
1239:
1240: % quantum algorithm acceleration
1241: \bibitem{qaa4} A.\ O.\ Niskanen, J.\ J.\ Vartiainen, and M.\ M.\ Salomaa,
1242: Phys.\ Rev.\ Lett.\ {\bf 90}, 197901 (2003).
1243:
1244: \bibitem{qaa6} J.\ J.\ Vartiainen, A.\ O.\ Niskanen, M.\ Nakahara,
1245: and M.\ M.\ Salomaa,
1246: % Implementing Shor's algorithm on Josephson charge qubits
1247: Phys. Rev. A {\bf 70}, 012319 (2004).
1248:
1249: \bibitem{ref:kg} N.\ Khaneja, R.\ Brockett, and S.\ J.\ Glaser,
1250: Phys. Rev. A {\bf 63}, 032308 (2001).
1251:
1252: \bibitem{qaa} M.\ Nakahara, Y.\ Kondo, K.\ Hata, and S.\ Tanimura,
1253: Phys.\ Rev.\ A {\bf 70}, 052319 (2004).
1254: \bibitem{qaa3.5} S.\ Tanimura, M.\ Nakahara, and D.\ Hayashi,
1255: % quant-ph/0406038, J.\ Math.\ Phys., to be published.
1256: J. Math. Phys. {\bf 46}, 022101 (2005).
1257:
1258: \bibitem {NMR_textbook} For example, see T.\ E.\ W.\ Claridge,
1259: {\it High-Resolution NMR techniques in Organic Chemistry}
1260: (Elsevier, Amsterdam, 2004).
1261:
1262: \bibitem {rmp_chuang} L.\ M.\ K.\ Vandersypen and I.\ L.\ Chuang,
1263: Rev.\ Mod.\ Phys.\ {\bf 76}, 1037 (2004).
1264:
1265: \bibitem{cory2}
1266: R.\ lamme, E.\ Knill, D.\ G.\ Cory, E.\ M.\ Fortunato,
1267: T.\ F.\ Havel, C.\ Miquel, R.\ Martinez, C.\ J.\ Negrevergne,
1268: G.\ Ortiz, M.\ A.\ Pravia, Y.\ Sharf, S.\ Sinha, R.\ Somma,
1269: and L.\ Viola, Los Alamos Science Number {\bf 27},
1270: 226 (2002).
1271:
1272: \bibitem{cory}
1273: D.\ G.\ Cory, R.\ Laflamme, E.\ Knill, L.\ Viola,
1274: T.\ F.\ Havel, N.\ Boulant, G.\ Boutis, E.\ Fortunato,
1275: S.\ Lloyd, R.\ Martinez, C.\ Negrevergne, M.\ Pravia,
1276: Y.\ Sharf, G.\ Teklemariam, Y.\ S.\ Weinstein, and
1277: W.\ H.\ Zurek, Fortschr.\ Phys.\ {\bf 40}, 875 (2000),
1278: J.\ A.\ Jones, Fortschr.\ Phys.\ {\bf 40}, 909 (2000).
1279:
1280: \bibitem{raedt} H.\ De Raedt, K.\ Michielsen, A.\ Hams, S.\ Miyashita,
1281: and K.\ Saito, Eur.\ Phys.\ J.\ B {\bf 27}, 15 (2002).
1282:
1283: \bibitem {jeol} http://www.jeol.com/nmr/nmr.html.
1284:
1285: \bibitem{cytosine} J.\ A.\ Jones, M.\ Mosca, and R.\ H.\ Hansen,
1286: J.\ Chem.\ Phys.\ {\bf 109} 1648 (1998).
1287:
1288: \bibitem{DJ} D.\ Deutsch, Proc.\ R.\ Soc.\ Lond.\ A {\bf 400}, 97 (1985),
1289: D.\ Deutsch and R.\ Jozsa, Proc.\ R.\ Soc.\ Lond.\ A {\bf 439}, 533 (1985).
1290:
1291: \bibitem{13c} I.\ L.\ Chuang, L.\ M.\ K.\ Vandersypen, X.\ Zhou,
1292: D.\ W.\ Leung, and S.\ Lloyd, Nature {\bf 393}, 143 (1998).
1293:
1294: \bibitem{pps_j}
1295: U.\ Sakaguchi, H.\ Ozawa, and T.\ Fukumi, Phys.\ Rev.\ A {\bf 61},
1296: 042313 (2000).
1297:
1298:
1299: \bibitem{warp} M.~Nakahara, J.~J.~Vartiainen, Y. Kondo, S.~Tanimura,
1300: and K. Hata, e-print quant-ph/0411153.
1301:
1302: \end{thebibliography}
1303:
1304: \end{document}
1305:
1306:
1307:
1308:
1309:
1310: