1: \documentclass[aps,prl,amsmath,twocolumn]{revtex4}
2: %\documentstyle[aps,prl,amsmath,12pt]{revtex}
3: \usepackage{bm}
4: \usepackage{graphicx}
5: \begin{document}
6: \draft
7: \title{Qubit Rotation by STIRAP}
8:
9: \author{Z. Kis$^{(1, 2)}$ and F. Renzoni$^{(3)}$}
10:
11: \address{\ $^{(1)}$ Department of Physics, Royal Institute of Technology (KTH),
12: Roslagstullsbacken 21, SE-10691 Stockholm, Sweden}
13:
14: \address{\ $^{(2)}$ Research Institute for
15: Solid State Physics and Optics, H-1525 Budapest, P.O. Box 49, Hungary
16: }
17:
18: \address{\ $^{(3)}$ Laboratoire Kastler Brossel, D\'epartement de Physique de
19: l'Ecole Normale Sup\'erieure, 24 rue Lhomond, 75231, Paris Cedex 05, France}
20:
21: \date{\today{}}
22:
23:
24:
25: \begin{abstract}
26: We introduce a novel procedure for qubit rotation, alternative to the
27: commonly used method of Rabi oscillations of controlled pulse area. It is
28: based on the technique of Stimulated Raman Adiabatic Passage (STIRAP) and
29: therefore it is robust against fluctuations of experimental parameters.
30: Furthermore, our work shows that it is in principle possible to perform
31: quantum logic operations via stimulated Raman adiabatic passage. This opens
32: up the search for a completely new class of schemes to implement logic
33: gates.
34: \end{abstract}
35:
36: \pacs{PACS: 42.50.Hz, 03.65.Ta}
37:
38: \maketitle
39:
40:
41:
42: The growing interest in quantum computation stimulates the search for schemes
43: to prepare and manipulate quantum states \cite{Bouwmeester}. The basic
44: operation on a single qubit is the rotation. This is usually achieved via
45: coherent Rabi oscillations. For a two-level optical transition Rabi
46: oscillations are simply produced by direct coupling using a laser field. If
47: the qubit is formed by two ground states instead, Rabi flopping is then
48: implemented by Raman transitions via a virtual excited state, and a
49: fundamental logic gate has been demonstrated using this configuration
50: \cite{raman}. Fundamental quantum logic operations based on (conditional)
51: Rabi flopping have also been implemented in cavity QED experiments
52: \cite{haroche}. To perform qubit rotation utilizing Rabi flopping, a complete
53: control of the pulse area is required. An imperfect control leads obviously to
54: errors in the computation procedure.
55:
56: In this work, we introduce a novel procedure for qubit rotation. It is based
57: on the technique of Stimulated Raman Adiabatic Passage (STIRAP \cite{review})
58: and therefore it is robust against fluctuations of several
59: experimental parameters such as the pulse area. Furthermore, our proposal
60: shows that STIRAP is not only a powerful tool to transfer population between
61: quantum states \cite{review2}, including here the creation of coherent
62: superpositions \cite{marte,una98,vitanov99,renzoni,una01a} and entangled
63: states \cite{una01b}, but can also play a role in the implementation of a
64: quantum logic operation.
65:
66: Consider the generic linear superposition of two long-living atomic
67: states $|1\rangle$ and $|2\rangle$
68: \begin{equation}
69: |i\rangle\! = \!\alpha |1\rangle + \beta |2\rangle~.
70: \label{qubit}
71: \end{equation}
72: The state rotated about a unit vector $\bm n$ through an angle $\zeta$ reads
73: \begin{equation}
74: |f\rangle\! =\! \hat R_{\bm n}(\zeta)|i\rangle~,
75: \label{rotated}
76: \end{equation}
77: with $\hat R_{\bm n}(\zeta)$ being an element of the $SU_2$ group
78: defined as
79: \begin{equation}\label{rotop}
80: \hat R_{\bm n}(\zeta)\!=\!\exp\left(-i\frac{\zeta}{2}\,
81: {\bm n}\cdot\hat{\bm \sigma}\right)\!=\!\cos\frac{\zeta}{2}-
82: i{\bm n}\cdot\hat{\bm \sigma}\sin \frac{\zeta}{2}~,
83: \end{equation}
84: where ${\bm n}\!=\!(\sin\vartheta\cos\varphi, \sin\vartheta\sin\varphi,
85: \cos\vartheta )$, and $\hat{\bm\sigma}\!=\!(\sigma_x, \sigma_y, \sigma_z )$
86: are the Pauli's spin operators $\sigma_x\!=\!|1\rangle\langle
87: 2|+|2\rangle\langle 1|$, $\sigma_y\!=\!i(|2\rangle\langle 1|-|1\rangle\langle
88: 2|)$, $\sigma_z\!=\!|1\rangle\langle 1|-|2\rangle\langle 2|$.
89:
90: Mapping of the state (\ref{qubit}) into the state (\ref{rotated}) can be
91: achieved easily using a single STIRAP sequence, as described for example in
92: Ref.~\cite{renzoni}. However, in that scheme the Stokes and pump pulses which
93: link the different states must satisfy well defined relationships with the
94: coefficients of the initial and final superpositions. Therefore, that
95: procedure cannot be considered as a general rotation, because for a given
96: pulse sequence, not every state will be transformed in the same way. As it
97: will be shown here, the use of multiple STIRAP sequences allows indeed true
98: rotations, with the axis and angle of rotation being uniquely defined
99: by the parameters of the laser pulses.
100:
101: In our scheme, shown in Fig. \ref{scheme}, three ground states
102: $|1\rangle$, $|2\rangle$ and $|3\rangle$ are coupled via a
103: single excited state $|4\rangle$ by different laser fields.
104: We assume that due to their polarizations and frequencies, each
105: laser field drives only one transition.
106: The ground states $|1\rangle$ and $|2\rangle$ define our qubit, while
107: the state $|3\rangle$ is an auxiliary state which will be occupied only
108: in the intermediate phase of the rotation procedure.
109:
110: The detunings of the three laser fields are the same, i.e. the system
111: is at multiphoton resonance which is a necessary condition for
112: STIRAP. The Schr\"odinger equation of this system reads
113: \begin{equation}\label{sch1}
114: \frac{d}{dt}|\psi(t)\rangle\!=\!-\frac{i}{\hbar}{\hat
115: H}(t)|\psi(t)\rangle,
116: \end{equation}
117: where $|\psi(t)\rangle$ denotes the state vector of the four-level
118: system and the Hamiltonian $\hat H(t)$ is given in the interaction
119: picture and in the rotating-wave approximation (RWA) as
120: \begin{equation}\label{ham}
121: \hat H(t)\!=\!\hbar\Delta|4\rangle\langle 4|+\frac{\hbar}{2}
122: \sum_{i=1}^3(\Omega_i(t)|i\rangle\langle 4|+h.c.)
123: \end{equation}
124: Here $\Delta$ is the laser detuning. The pulsed Rabi frequencies $\Omega_1
125: (t)$ and $\Omega_2 (t)$ are taken to have essentially the same envelopes:
126: $\Omega_1(t)\!=\!\Omega (t)\cos\chi$, $\Omega_2\!=\!\Omega
127: (t)\exp{(i\eta)}\sin\chi$, where $\chi$ and $\eta$ are fixed angles. The
128: pulses 1,2 (i.e. the fields which yield the Rabi frequencies $\Omega_1(t)$ and
129: $\Omega_2(t)$) and 3 are delayed with respect to each other, however, for an
130: efficient STIRAP process they must have a significant overlap. In the
131: following $\Omega(t)$ will be taken as real.
132:
133: Our procedure consists of two STIRAP processes. In the first one, the fields 1
134: and 2 are the pump fields, whereas the field 3 plays the role of the Stokes
135: field. In this STIRAP process the field 3 has the same phase as that of
136: fields 1, 2. The pulses are applied in the counterintuitive order, i.e. the
137: Stokes pulse arrives before the pump ones. The transfer process can be
138: described as follows: The pump fields 1 and 2 define a dark (or noncoupled)
139: state $|NC\rangle $
140: \begin{equation}\label{nc}
141: |NC\rangle\!=\!-\sin\chi |1\rangle + e^{i\eta}\cos\chi |2\rangle~,
142: \end{equation}
143: in the subspace spanned by the states $\{ |1\rangle ,|2\rangle \}$
144: \cite{ari}. The orthogonal state (the coupled or bright state)
145: $|C\rangle $ is
146: \begin{equation}\label{cc}
147: |C\rangle\!=\!\cos\chi |1\rangle + e^{i\eta}\sin\chi |2\rangle~,
148: \end{equation}
149: from which the population can be transferred to state $|3\rangle$ if
150: all the three fields are on. By decomposing the initial superposition
151: $|i\rangle $ onto $|C\rangle $ and $|NC\rangle $,
152: \begin{equation}
153: |i\rangle = \langle NC|i\rangle |NC\rangle + \langle C|i\rangle |C\rangle
154: \end{equation}
155: with
156: \begin{mathletters}
157: \begin{eqnarray}
158: \langle NC|i\rangle &=& -\alpha\sin\chi + \beta e^{-i\eta}\cos\chi \\
159: \langle C|i\rangle &=& \alpha\cos\chi +\beta e^{-i\eta}\sin\chi
160: \end{eqnarray}
161: \label{coeff}
162: \end{mathletters}
163: we find that only the component along $|C\rangle $ will be
164: affected by the STIRAP process and eventually mapped to the target
165: state $|3\rangle $. Indeed, the Hamiltonian in Eq.~(\ref{ham})
166: can be rewritten in the form
167: \begin{equation}\label{hamnew}
168: \hat H(t)\!=\!\hbar\Delta|4\rangle\langle 4|+\frac{\hbar}{2}
169: (\Omega(t)|C\rangle\langle 4|+\Omega_3(t)|3\rangle\langle 4|+h.c.)~.
170: \end{equation}
171: Inserting Eq.~(\ref{hamnew}) into Eq.~(\ref{sch1}) one obtains a
172: Schr\"odinger equation which describes an ordinary STIRAP process in a
173: three-level system. The dark state $|\psi_D(t)\rangle$ of the Hamiltonian
174: in Eq.~(\ref{hamnew}) is given by
175: \begin{equation}\label{dark}
176: |\psi_D(t)\rangle\!=\frac{1}{\sqrt{\Omega^2(t)+\Omega^2_3(t)}}
177: (\Omega_3(t) |C\rangle - \Omega(t) |3\rangle)~.
178: \end{equation}
179: Using the same argument as in the case of STIRAP, it can be easily
180: shown that in the adiabatic limit the system is left, after the first
181: pulse sequence, in the superposition of the three ground states
182: \begin{equation}
183: |\psi\rangle \!= \langle NC|i\rangle |NC\rangle
184: - \langle C|i\rangle |3\rangle~,
185: \label{int}
186: \end{equation}
187: without populating the excited state during the evolution.
188: We recognize in Eq. (\ref{int}) that the component of $|i\rangle $ along
189: the noncoupled state $|NC\rangle $ is untouched, whereas the orthogonal
190: bright component is transferred to the target state $|3\rangle $.
191:
192: The second step of the rotation procedure is a reverse STIRAP process which
193: maps the state $|3\rangle$ back to the qubit subspace. In this STIRAP process
194: the phase of field 3 is in general different from that of fields 1, 2.
195: The pulses are applied in the reverse order with respect to the first step
196: of the rotation procedure: the pulses 1 and 2 (which play now the role of the
197: Stokes pulses) arrive before the pulse 3 (the pump). The fields 1 and 2 have
198: the same Rabi frequencies as in the first step of the procedure. In this way
199: the state $|\psi\rangle$ prepared by the first STIRAP process,
200: Eq.~(\ref{int}), is initially a dark state for the three laser pulses because:
201: (a) the state $|3\rangle$ is initially not coupled (counterintuitive pulse
202: order); (b) the state $|\psi\rangle$ has no component along $|C\rangle$ and
203: therefore it is decoupled from the fields 1,2. The darkness of $|\psi\rangle $
204: allows the implementation of the second STIRAP process although all the ground
205: states are initially populated. In this process the state $|3\rangle $ is
206: transferred back to the qubit subspace. More precisely, it will be mapped on
207: the coupled state $|C\rangle$ because the state $|NC\rangle $ is a decoupled
208: state also in this second STIRAP process. The component of the initial qubit
209: $|i\rangle $ along the coupled state $|C\rangle $ and the new component along
210: the same state obtained by mapping back the state $|3\rangle $ will differ by
211: a phase. This phase is determined by the phase difference $\delta$ between
212: the relative phase of the field 3 with respect to the fields 1,2 in the two
213: STIRAP processes. Clearly, for identical phases the system certainly goes
214: back to the initial state $|i\rangle $.
215: % Of interest for the rotation procedure introduced in this work is the choice
216: % of a phase shift $\pi$.
217: In the general case a similar calculation which yielded Eq.~(\ref{int}) shows
218: that in the adiabatic limit the component of $|\psi\rangle$ along $|3\rangle $
219: is mapped back onto the subspace $\{ |1\rangle , |2\rangle \}$ according to
220: \begin{eqnarray}
221: \langle 3|\psi\rangle|3\rangle &\to &
222: e^{-i\delta}\langle 3|\psi\rangle|C\rangle ~,
223: \end{eqnarray}
224: so that the final state is
225: \begin{equation}
226: |\psi_f\rangle \!=\! \langle NC|i\rangle |NC\rangle
227: +e^{-i\delta}\langle C|i\rangle |C\rangle~.
228: \end{equation}
229: By substituting the expressions (\ref{coeff}) for the coefficients and
230: taking into account Eqs.~(\ref{rotated}) and (\ref{rotop}), we obtain
231: \begin{eqnarray}\label{psif}
232: |\psi_f\rangle \!=\!e^{-i\delta/2} \hat R_{\bm n}(\delta)|i\rangle~,
233: \end{eqnarray}
234: where ${\bm n}\!=\!(\sin2\chi\cos\eta, \sin2\chi\sin\eta, \cos2\chi )$. Apart
235: from a global phase $-\delta/2$ the states $|i\rangle$ and $|\psi_f\rangle$
236: are connected through the rotation $\hat R_{\bm n}(\delta)$. If the qubit
237: is isolated, then this global phase is unimportant. If the qubit is part of a
238: larger system, e.g. there are several qubits which form a quantum computer,
239: then the global phase is clearly relevant, however, it may be incorporated
240: into the algorithm being implemented on the quantum computer.
241:
242: The geometric interpretation of the rotation procedure described above is
243: straightforward: the axis of rotation ${\bm n}$ is defined by the polar
244: angle $2\chi$ and the azimutal angle $\eta$, which characterize the relative
245: amplitude and phase of the laser pulses acting on the states $|1\rangle$ and
246: $|2\rangle$, respectively. In the Hilbert space the state vector $|NC\rangle$,
247: Eq. (\ref{nc}), is parallel with this axis, therefore, it is unaffected by
248: the rotation. The state vector $|C\rangle$, Eq. (\ref{cc}), is perpendicular
249: to the rotation axis, and it is rotated through an angle $\delta$ about
250: this axis.
251:
252: Our rotation procedure is sensitive to the relative phase and relative
253: amplitude of the pulses 1 and 2. Moreover, the method is also sensitive to the
254: phase $\delta$, the phase difference of the pulse 3 in the first and second
255: STIRAP processes. However, it is robust against the fluctuation of the pulse
256: shapes and pulse areas.
257:
258: Our analysis is supported by numerical calculations. In these calculations,
259: the envelope of the different pulsed Rabi frequencies has been taken as
260: Gaussian, although any sufficiently smooth pulse shape is suitable. The width
261: of the pulses and their overlaps have been chosen to satisfy well the usual
262: STIRAP conditions \cite{review}. For different choices of the initial state
263: of the qubit (\ref{qubit}) we have solved numerically the Schr\"odinger
264: equation (\ref{sch1}) and determined the time dependent state of the system
265: during the rotation procedure composed of two STIRAP processes. An example of
266: our calculations is shown in Fig.~\ref{numerics} where the populations of the
267: three ground states are shown together with the pulse sequence which effects
268: the qubit rotation. The time evolution in the case of identical phases for the
269: two STIRAP processes is also reported for comparison. In order to analyze
270: quantitatively the quality of our rotation procedure, we have calculated the
271: fidelity $F=|\langle\psi_f|\psi_{\infty}\rangle |^2$, where
272: $|\psi_{\infty}\rangle $ is the numerically calculated state after the
273: rotation procedure and $|\psi_f\rangle $ the state of the rotated qubit. We
274: have verified that the fidelity is unity within numerical error. We have also
275: verified that under usual STIRAP conditions the excited state occupation is
276: negligeable throughout the rotation procedure.
277:
278: In conclusion we have shown that it is possible to perform qubit rotations by
279: stimulated Raman adiabatic passage, and correspondingly, proposed a rotation
280: procedure. The procedure is composed of two STIRAP processes. The first step
281: consists in projecting the original qubit on two perpendicular states: the
282: dark and the bright states, as defined by the pump pulses. The component of
283: the qubit along the bright state is then transferred to an auxiliary state. In
284: the second STIRAP process, the auxiliary state is mapped back onto the bright
285: state, with a phase shift $\delta$ with respect to the initial component of
286: the qubit along this state. The resulting state corresponds indeed to a
287: rotation of the qubit, with the axis and angle of rotation determined
288: uniquely by the parameters of the laser fields.
289:
290: Furthermore, our work shows that it is in principle possible to perform
291: quantum logic operations via stimulated Raman adiabatic passage. This
292: opens up the search for new schemes for logic gates based on STIRAP.
293:
294:
295: We are indebted to Stig Stenholm for useful discussions and a careful reading
296: of the manuscript. F.R. is also grateful for his hospitality at the Royal
297: Institue of Technology (KTH) of Stockholm, where the main part of this work
298: has been done. Z. Kis acknowledges the financial support of the European Union
299: Research and Training Network COCOMO, contract HPRN-CT-1999-00129, the J\'anos
300: Bolyai program of the Hungarian Academy of Sciences, and the National
301: Scientific research Fund (OTKA) of Hungary under Contract No. T034484.
302:
303: \begin{references}
304: \bibitem{Bouwmeester} {D.~Bouwmeester, A.~Ekert and A.~Zeilinger
305: (eds)}, {\em The physics of quantum information} (Springer,
306: Berlin, 2000).
307: \bibitem{raman} C.~Monroe, D.M.~Meekhof, B.E.~King, W.M.~Itano, and
308: D.J.~Wineland, Phys.Rev. Lett. {\bf 75}, 4714 (1995).
309: \bibitem{haroche} A.~Rauschenbeutel, G.~Nogues, S.~Osnaghi, P.~Bertet,
310: M.~Brune, J.M.~Raimond, and S.~Haroche, Phys. Rev. Lett. {\bf 13},
311: 5166 (1999).
312: \bibitem{review} STIRAP has been recently reviewed in K.~Bergmann,
313: H. Theuer, and B.W.~Shore, Rev. Mod. Phys. {\bf 70}, 1003 (1998)
314: and N.V.~Vitanov, M.~Fleishhauer, B.W.~Shore, and K.~Bergmann,
315: Adv. At. Mol. Opt. Phys. (in press).
316: \bibitem{review2} N.V.~Vitanov, T.~Halfmann, B.W.~Shore, and K.~Bergmann,
317: Ann. Rev. Phys. Chem. {\bf 52}, 763 (2001).
318: \bibitem{marte} P.~Marte, P.~Zoller and J.L.~Hall, Phys. Rev. A {\bf
319: 44}, R4118 (1991).
320: \bibitem{una98} R.~Unanyan, M.~Fleischhauer, B.W.~Shore, K.~Bergmann,
321: Opt. Comm. {\bf 155}, 144 (1998).
322: \bibitem{vitanov99} N.V.~Vitanov, K.-A.~Suominen and B.W.~Shore,
323: J.Phys. B: At. Mol. Opt. Phys. {\bf 32}, 4535 (1999).
324: \bibitem{renzoni} F.~Renzoni and S.~Stenholm, Opt. Comm. {\bf 189}, 69
325: (2001).
326: \bibitem{una01a} R.G.~Unanyan, B.W.~Shore, and K.~Bergmann, Phys. Rev.
327: A {\bf 63}, 043401 (2001).
328: \bibitem{una01b} R.G.~Unanyan, B.W.~Shore, and K.~Bergmann, Phys. Rev.
329: A {\bf 63}, 043405 (2001).
330: \bibitem{ari} For a review of dark states, see E.~Arimondo, in
331: Progress in Optics ed. E. Wolf, vol. 35 (Elsevier, Amsterdam, 1996)
332: p. 257.
333: \end{references}
334:
335: \newpage
336:
337: \begin{figure}
338: \includegraphics[width=5cm]{fig1}
339: \caption{Interaction scheme for the rotation of a qubit by STIRAP.
340: The three ground states $|1\rangle$, $|2\rangle$ and $|3\rangle$
341: are coupled to the excited state $|4\rangle$ by three different laser
342: fields. The qubit is defined by the ground states $|1\rangle$ and
343: $|2\rangle$. State $|3\rangle$ is an auxiliary state occupied only in the
344: intermediate phase of the rotation procedure.}
345: \label{scheme}
346: \end{figure}
347:
348:
349: \begin{figure}
350: \includegraphics[width=17cm]{fig2}
351: \caption{
352: Time evolution of the ground state populations and pulse sequences for two
353: different STIRAP procedures. In the procedure shown in the left column, the
354: second STIRAP process is precisely the reverse of the first one (top - left)
355: and the system returns back to its original state (bottom - left). In the
356: procedure shown in the right column, in the second STIRAP process the phase
357: of the pulse 3 is shifted by $\delta\!=\!\pi$ with respect to the the pulses
358: 1 and 2 (top-right). In this case the system does not return back to its
359: original state, and the qubit is rotated through an angle $\pi$ about a unit
360: vector characterized by the polar angle $2\chi$ and the azimutal angle
361: $\eta$. Parameters of the calculation are: $\alpha\!=\! \cos\pi/5$,
362: $\beta\!=\!\sin\pi/5$, $\Delta\!=\!0$, $\chi\!=\!-\pi/12$, and $\eta\!=\!0$.
363: The pulses have Gaussian shape $\exp[ -(t\pm T/2 \pm t_0)^2/2\tau^2 ]$, with
364: $\tau\!=\!2$, $t_0\!=\!1.6$, $T\!=\!20$. The time and the Rabi frequencies
365: are measured in arbitrary units. }
366: \label{numerics}
367: \end{figure}
368:
369: \end{document}
370:
371: