1: \documentclass[aps,pra,preprint,showpacs,floatfix,amsmath]{revtex4}
2:
3: \usepackage{graphicx}
4: \usepackage{hyperref}
5: \usepackage{bm}
6:
7: \topmargin=0.0in
8:
9: \begin{document}
10: \title{Temporal Interferometry: A Mechanism for Controlling Qubit
11: Transitions During Twisted Rapid Passage with Possible
12: Application to Quantum Computing}
13:
14: \date{\today}
15:
16: \author{Frank Gaitan}
17: \email{gaitan@physics.siu.edu}
18: \affiliation{Department of Physics; Southern Illinois University;
19: Carbondale, IL 62901-4401}
20:
21: \begin{abstract}
22: In an adiabatic rapid passage experiment, the Bloch vector of
23: a two-level system (qubit) is inverted by slowly inverting an external
24: field to which it is coupled, and along which it is initially aligned.
25: In twisted rapid passage, the external field is allowed to twist around
26: its initial direction with azimuthal angle $\phi (t)$ at the same time
27: that it is inverted. For polynomial twist: $\phi (t) \sim Bt^{n}$.
28: We show that for $n \geq 3$, multiple avoided crossings can occur during
29: the inversion of the external field, and that these crossings give rise
30: to strong interference effects in the qubit transition probability.
31: The transition probability is found to be a function of
32: the twist strength $B$, which can be used to control the time-separation of
33: the avoided crossings, and hence the character of the interference.
34: Constructive and destructive interference are possible. The interference
35: effects are a consequence of the temporal phase coherence of the wavefunction.
36: The ability to vary this coherence by varying the temporal separation
37: of the avoided crossings renders twisted rapid passage with adjustable
38: twist strength into a temporal interferometer through which qubit transitions
39: can be greatly enhanced or suppressed. Possible application of this
40: interference mechanism to construction of \textit{fast fault-tolerant}
41: quantum CNOT and NOT gates is discussed.
42: \end{abstract}
43:
44: \pacs{03.67.Lx,07.60.Ly,31.50.Gh}
45:
46: \maketitle
47:
48: \section{\label{sec1}Introduction}
49:
50: Adiabatic rapid passage (ARP) is a well-known procedure for inverting
51: the Bloch vector of a two-level system (qubit) \cite{arp}.
52: This is accomplished by inverting an external field $\mathbf{F}(t)$
53: which couples to the qubit, and along which the qubit is initially
54: aligned. The field inversion is done on a time-scale that is large
55: compared to the inverse Rabi frequency $\omega_{0}^{-1}$ (viz.\
56: adiabatic), though small compared to the thermal relaxation time
57: $\tau$ (viz.\ rapid). In the usual case, $\mathbf{F}(t)$ remains
58: within a plane that includes the origin:
59: $\mathbf{F}(t) = b\,\hat{\mathbf{x}}+ at\,\hat{\mathbf{z}}$,
60: with $-T_{0}/2 < t < T_{0}/2$, and $\omega_{0}^{-1}\ll T_{0} \ll \tau$.
61:
62: ARP can be used to implement a NOT gate on the quantum state of a qubit.
63: If one identifies the computational basis states $|0\rangle$ and $|1\rangle$,
64: respectively, with the spin-up and spin-down eigenstates along the
65: initial direction of the external field $\mathbf{F}(-T_{0}/2)$, then ARP maps
66: $|0\rangle \leftrightarrow |1\rangle$ which is the defining operation of a
67: NOT gate. Occurrence of a transition during ARP corresponds to an error in the
68: NOT gate since the Bloch vector is not inverted, and thus
69: $|i\rangle \rightarrow |i\rangle$, ($i=0,1$). Thus we can identify the ARP
70: transition probability with the NOT gate error probability. For ARP,
71: the adiabatic nature of the inversion ensures that the error probability is
72: exponentially small. The price paid for this reliability, however, is an
73: extremely slow NOT gate.
74:
75: In twisted adiabatic rapid passage, the external field
76: is allowed to twist around its initial direction with azimuthal angle
77: $\phi (t)$ at the same time that it is adiabatically inverted:
78: $\mathbf{F}(t)
79: = b\cos\phi (t)\,\hat{\mathbf{x}} + b\sin\phi (t) \,\hat{\mathbf{y}} + at\,
80: \hat{\mathbf{z}}$. Reference~\onlinecite{bry} showed that for twisted ARP, the
81: exponentially small transition probability contains a factor $\exp
82: [\,\Gamma_{g}\, ]$ of purely geometric origin. The simplest case
83: where $\Gamma_{g}\neq 0$ corresponds to quadratic twist: $\phi (t) =
84: Bt^{2}$. Zwanziger et.\ al.\ \cite{zwa} were able to experimentally
85: realize ARP with quadratic twist and obtained results in agreement
86: with the predictions of Reference~\onlinecite{bry}.
87:
88: In this paper we will consider twisted rapid passage with polynomial twist,
89: $\phi (t) \sim Bt^{n}$, and we will focus exclusively on qubit inversions
90: done at \textit{non-adiabatic}\/ rates.
91: Although we will briefly consider quadratic twist in Section~\ref{sec2} as a
92: test case for our numerical simulations, our interest will \textit{not}\/ be
93: the geometric effect of Reference~\onlinecite{bry}. Instead, our primary focus
94: will be on establishing the existence of
95: multiple avoided crossings during twisted rapid passage when $n\geq 3$,
96: and with exploring some of their consequences. After
97: general considerations (Section~\ref{sec2}), we will explicitly
98: examine cubic ($n=3$) and
99: quartic ($n=4$) twist, and will provide clear evidence that the
100: multiple avoided crossings produce strong interference effects in the
101: qubit transition probability. The transition probability is shown to be
102: a function of the twist strength $B$, which can be
103: used to control the time-separation of the avoided crossings, and hence
104: the character of the interference (constructive or destructive).
105: Cubic and quartic twist are examined in Sections~\ref{sec3} and
106: \ref{sec4}, respectively. We shall see that interference between the
107: multiple avoided crossings can greatly enhance or suppress qubit
108: transitions. The interference effects are a direct consequence of the
109: temporal phase coherence of the wavefunction. The ability to vary this
110: coherence by varying the temporal separation of the avoided crossings renders
111: twisted rapid passage with adjustable twist strength into a temporal
112: interferometer through which qubit transitions can be controlled. It will be
113: shown that quartic twist can implement qubit inversion
114: non-adiabatically while operating at a fidelity that exceeds the threshold for
115: fault tolerant operation. Finally, in Section~\ref{sec5}, we
116: summarize our results and discuss possible application of this interference
117: mechanism to quantum computing. In particular, we describe how one might
118: use non-adiabatic rapid passage with quartic twist to construct a fast
119: fault-tolerant quantum CNOT gate.
120:
121: It is worth noting that experimental \textit{confirmation\/} of the work
122: described in this paper has recently been carried out by Zwanziger et.\ al.\
123: \cite{jwz}. They have realized \textit{non}-\textit{adiabatic} rapid
124: passage with both cubic and quartic twist, and have observed clear evidence of
125: constructive and destructive interference in the qubit transition probability
126: due to interference between the avoided crossings, with excellent agreement
127: between the experimental data and our numerical simulations. This experimental
128: work provides clear proof-of-principle for our thesis that controllable
129: quantum interference exists during twisted rapid passage. With this
130: thesis now experimentally confirmed, future research can focus on applying
131: this interference to the task of constructing fast fault tolerant quantum CNOT
132: and NOT gates.
133:
134: After this paper was submitted, previous work was brought to our
135: attention which also examined models of rapid passage in which more than
136: one avoided crossing is possible, and in which interference effects
137: were also considered \cite{lim,suo,joy}. These papers focused solely on the
138: adiabatic limit. Application of this adiabatic theory to the Zwanziger
139: experiment yields predictions that are in poor agreement with the experimental
140: results \cite{jwz}. This failing is no doubt a consequence of the non-adiabatic
141: character of this experiment whose results thus lies beyond the scope
142: of the adiabatic theory developed in these papers. In contrast, the work we
143: present below is principally interested in the non-adiabatic limit, and our
144: simulation results are in full agreement with experiment \cite{jwz}. We
145: also consider possible application of these interference effects to the
146: construction of \textit{fast} \textit{fault}-\textit{tolerant} quantum CNOT
147: and NOT gates. Ref.~\cite{lim,suo,joy} do not consider such applications.
148:
149: \section{\label{sec2}Twisted Rapid Passage}
150:
151: We begin by briefly summarizing the essential features of rapid
152: passage in the absence of twist. Twistless rapid passage describes
153: a wide variety of phenomena, ranging from magnetization reversal in
154: NMR, to electronic transition during a slow atomic collision. The
155: essential situation is that of a qubit which is Zeeman-coupled to a
156: background field $\mathbf{F}(t)$,
157: \begin{equation}
158: H(t) = \mbox{\boldmath $\sigma$}\cdot\mathbf{F}(t) =
159: \left( \begin{array}{cc}
160: at & b \\
161: b & -at
162: \end{array} \right) \hspace{0.1in} ,
163: \label{labham}
164: \end{equation}
165: with $\mathbf{F}(t) = b\,\hat{\mathbf{x}} + at\,
166: \hat{\mathbf{z}}$. This particular form for $\mathbf{F}(t)$
167: describes inversion of
168: the background field in such a way that it remains in the x-z plane
169: throughout the inversion. For simplicity, we assume $a,\: b > 0$
170: throughout this paper. The instantaneous energies
171: $E_{\pm}(t)$ are:
172: \begin{equation}
173: E_{\pm}(t) = \pm\sqrt{b^{2} + (at)^{2}} \hspace{0.1in} ,
174: \end{equation}
175: and an avoided crossing is seen to occur at $t=0$ where the energy
176: gap is minimum. The Schrodinger dynamics
177: for twistless rapid passage can be solved exactly for arbitrary
178: values of $a$ and $b$ \cite{lan,zen}, and yields the Landau-Zener
179: expression for the transition probability $P_{LZ}$:
180: \begin{equation}
181: P_{LZ} = \exp\left[\, -\frac{\pi b^{2}}{\hbar |a|}\,\right]
182: \hspace{0.1in} .
183: \label{plz}
184: \end{equation}
185:
186: \subsection{Twisted Rapid Passage and Multiple Avoided Crossings}
187:
188: In {\it twisted\/} rapid passage, the background field
189: $\mathbf{F}(t)$
190: is allowed to twist around its initial direction during the course
191: of its inversion: $\mathbf{F}(t) = b\cos\phi (t)\,\hat{\mathbf{x}}
192: +b\sin\phi (t)\,\hat{\mathbf{y}} + at\, \hat{\mathbf{z}}$. It proves
193: convenient to transform to the rotating frame in which the x-y
194: component of the background field is instantaneously at rest. This is
195: accomplished via the unitary transformation $U(t) = \exp [-(i/2)\phi
196: (t)\sigma_{z}]$. The Hamiltonian $\overline{H}(t)$ in this frame is:
197: \begin{equation}
198: \overline{H}(t)
199: = \mbox{\boldmath $\sigma$}\cdot\overline{\mathbf{F}}
200: = \left( \begin{array}{cc}
201: \left(\, at -\frac{\hbar}{2}\dot{\phi}\,\right)
202: & b \\
203: b &
204: -\left(\, at - \frac{\hbar}{2}\dot{\phi}\,\right)
205: \end{array} \right)
206: \hspace{0.1in} ,
207: \label{rotham}
208: \end{equation}
209: where $\overline{\mathbf{F}}(t) = b\,\hat{\mathbf{x}}+ (at-\hbar\dot{\phi}/2)
210: \,\hat{\mathbf{z}}$ is the background field as seen in the rotating frame,
211: and a dot over a symbol represents the time derivative of that symbol.
212: The instantaneous energy eigenvalues are $\overline{E}_{\pm}(t) = \pm
213: \sqrt{\left(\, at - (\,\hbar\dot{\phi}\,)/2\,\right)^{2} +
214: b^{2}}$. Avoided crossings occur when the energy gap is minimum,
215: corresponding to when
216: \begin{equation}
217: at - \frac{\hbar}{2}\,\frac{d\phi}{dt} = 0 \hspace{0.1in} .
218: \label{cond}
219: \end{equation}
220: For polynomial twist: $\phi_{n}(t) = c_{n}Bt^{n}$, where $B$ is the
221: twist strength. The dimensionless constant $c_{n}$ has been introduced
222: to simplify some of the formulas below. For later convenience, we
223: chose $c_{n} = 2/n$. For polynomial twist, it is easily checked that
224: eqn.~(\ref{cond}) always has the root:
225: \begin{equation}
226: t = 0 \hspace{0.1in} ,
227: \label{avdx1}
228: \end{equation}
229: and that for $n\ge 3$, eqn.~(\ref{cond}) also has the $n-2$ roots:
230: \begin{equation}
231: t = \left(\, {\rm sgn}\, B\,\right)^{\frac{1}{n-2}}\,
232: \left(\,\frac{a}{\hbar |B|}\,\right)^{\frac{1}{n-2}}
233: \hspace{0.1in} .
234: \label{avdx2}
235: \end{equation}
236: All together, equation~(\ref{cond}) has $n-1$ roots,
237: though only the real roots correspond to avoided crossings.
238: For quadratic twist ($n=2$), only eqn.~(\ref{avdx1}) arises. Thus, for
239: this case, only the avoided crossing at $t=0$ is possible. For $n\geq 3$,
240: along with the avoided crossing at $t=0$, real solutions to
241: eqn.~(\ref{avdx2}) also occur. The different possibilities for this situation
242: are summarized in Table~\ref{table1}.
243: \begin{table*}[h]
244: \caption{\label{table1}Classification of regimes under which multiple avoided
245: crossings occur for polynomial twist with $n \geq 3$.}
246: \begin{ruledtabular}
247: \begin{tabular}{lrl}
248: \multicolumn{3}{c}{1. $\underline{ sgn\, B = +1}$} \\
249: (a)\hspace{0.2in} $n$ odd; & $\:\:\: 2$ avoided crossings at: &
250: $\:\:\: t=0$ and $t=\left(a/\hbar B\right)^{1/(n-2)}$ \\
251: (b)\hspace{0.2in} $n$ even; & $\:\:\: 3$ avoided crossings at: &
252: $\:\:\: t=0$ and $t=\pm\left(a/\hbar B\right)^{1/(n-2)}$ \\
253: \multicolumn{3}{c}{2. $\underline{ sgn\, B = -1}$} \\
254: (a)\hspace{0.2in} $n$ odd; & $\:\:\: 2$ avoided crossings at: &
255: $\:\:\: t=0$ and $t=-\left(a/\hbar |B|\right)^{1/(n-2)}$ \\
256: (b)\hspace{0.2in} $n$ even; & $\:\:\: 1$ avoided crossing at: &
257: $\:\:\: t=0$ \\
258: \end{tabular}
259: \end{ruledtabular}
260: \end{table*}
261: We see that for polynomial twist with $n\geq 3$, multiple avoided crossings
262: always occur for positive twist strength $B$, while for negative
263: twist strength, multiple avoided crossings only occur when $n$ is odd.
264: Note that the time separating the multiple avoided crossings can be adjusted
265: by variation of the twist strength $B$ and/or the inversion rate $a$.
266:
267: \subsection{Brief Detour: Background on Quadratic Twist}
268:
269: As mentioned earlier, quadratic twist has already been examined in the
270: literature \cite{bry}. It is of interest here only because its dynamics can
271: be solved exactly, and thus allows us to test our numerical simulations
272: before proceeding to unexplored cases of twisted rapid
273: passage. For quadratic twist, $\dot{\phi}_{2} = 2Bt$. Inserting this into
274: eqn.~(\ref{rotham}) gives $\overline{\mathbf{F}}(t) = b\,\hat{\mathbf{x}}+
275: \overline{a} t\,\hat{\mathbf{z}}$, with $\overline{a}= a -\hbar |B|\,
276: (\mathrm{sgn}\, B)$.
277: Thus rapid passage with quadratic twist maps onto twistless rapid passage
278: with $a\rightarrow \overline{a}$. This allows us to obtain an exact result
279: for the transition probability $P_{2}$ for arbitrary values of $a$ and $b$
280: from eqn.~(\ref{plz}) with $a\rightarrow \overline{a}$:
281: \begin{equation}
282: P_{2} = \exp\left[\, -\frac{\pi b^{2}}{ \hbar |\, a - \hbar |B|\, (
283: \mathrm{sgn}\, B)\, | } \right] \hspace{0.1in} .
284: \label{tranprob}
285: \end{equation}
286: In the adiabatic limit, this reduces to $P_{2}=P_{LZ}\exp [\Gamma_{g}]$,
287: where $\Gamma_{g}=-\pi Bb^{2}/a^{2}$ is the geometric exponent discovered
288: in Ref.~\cite{bry}. Eqn.~(\ref{tranprob}) makes the interesting prediction
289: that a \textit{complete\/} quenching of transitions will occur when
290: $\mathrm{sgn}\, B = +1$ and $a = \hbar B$, while no such quenching is possible
291: for $\mathrm{sgn}\, B = -1$. Zwanziger et.~al.\ \cite{zwa} were able to
292: realize rapid passage with quadratic twist experimentally and confirmed the
293: existence of $\Gamma_{g}$, and the twist-dependent quenching of transitions.
294: We now show that our numerical simulation also reproduces these effects.
295:
296: \subsection{Simulation Details}
297:
298: The equations that drive the numerical simulation follow from the
299: Schrodinger equation in the non-rotating frame:
300: \begin{equation}
301: i\hbar\frac{\partial}{\partial t}\, |\psi\rangle = H(t)\, |\psi\rangle
302: \hspace{0.1in} ,
303: \label{schro}
304: \end{equation}
305: where $H(t) = \mbox{\boldmath $\sigma$}\cdot\mathbf{F}(t)$, and
306: $\mathbf{F}(t) = b\cos\phi (t)\,\hat{\mathbf{x}} + b\sin\phi (t)\,
307: \hat{\mathbf{y}} + at\,\hat{\mathbf{z}}$. To obtain these equations in
308: the adiabatic representation, we expand $|\psi (t)\rangle$ in the
309: instantaneous eigenstates $|E_{\pm}(t)\rangle$ of $H(t)$:
310: \begin{equation}
311: |\psi (t)\rangle = S(t)\; e^{-\frac{i}{\hbar}\int_{-T_{0}/2}^{t}\, d\theta\,
312: \left( E_{-}-\hbar\dot{\gamma}_{-}\right)}\;
313: |E_{-}(t)\rangle \; - \;
314: I(t)\; e^{-\frac{i}{\hbar}\int_{-T_{0}/2}^{t}\, d\theta\,
315: \left( E_{+} -\hbar\dot{\gamma}_{+}\right)}\;
316: |E_{+}(t)\rangle \hspace{0.1in} .
317: \label{expan}
318: \end{equation}
319: Here $\gamma_{\pm}(t)$ are the geometric phases \cite{geo} associated with
320: the energy-levels $E_{\pm}(t)$, respectively, and
321: \begin{equation}
322: \dot{\gamma}_{\pm}(t) = i\langle\, E_{\pm}(t)\, |\,\frac{d}{dt}\, |\,
323: E_{\pm}(t)\, \rangle =
324: i \langle\, E_{\pm}(t)\, |\,\dot{E}_{\pm}(t)\,\rangle
325: \hspace{0.1in} .
326: \label{berfaz}
327: \end{equation}
328: Substituting eqn.~(\ref{expan}) into (\ref{schro}), and using the
329: orthonormality of the instantaneous eigenstates, one obtains the
330: equations of motion for the expansion coefficients $S(t)$ and $I(t)$:
331: \begin{subequations}
332: \label{alleom}
333: \begin{eqnarray}
334: \frac{dS}{dt} & = & -\Gamma^{\ast}(t)\; e^{-i\int_{-T_{0}/2}^{t}\,d\theta\,
335: \delta
336: (\theta )}\; \vspace{0.15in}I(t) \hspace{0.1in} ,
337: \label{eoma} \\
338: \frac{dI}{dt} & = & \mbox{} \;\;\; \Gamma (t)\; e^{i\int_{-T_{0}/2}^{t}\,
339: d\theta\,
340: \delta (\theta )} \; S(t) \hspace{0.1in} . \label{eomb}
341: \end{eqnarray}
342: \end{subequations}
343: Here,
344: \begin{eqnarray}
345: \delta (t) & = & \frac{E_{+}(t) - E_{-}(t)}{\hbar} - \left(\,
346: \dot{\gamma}_{+}(t)
347: - \dot{\gamma}_{-}(t) \,\right) \hspace{0.1in} ,
348: \label{delta} \\
349: \Gamma (t) & = & \langle\, E_{+}(t)\, |\,\dot{E}_{-}(t) \,\rangle
350: \hspace{0.1in} ,
351: \label{Gamma}
352: \end{eqnarray}
353: and one can show that $\Gamma^{\ast}(t) = -\,\langle E_{-}(t)\, |\,
354: \dot{E}_{+}(t) \,\rangle$.
355: Eqns.~(\ref{alleom}) are the qubit equations of motion in the adiabatic
356: representation and include the influence of the geometric phase on the dynamics
357: through $\delta (t)$. In the case of twistless rapid passage, the geometric
358: phase vanishes, and eqns.~(\ref{alleom}) reduce to the well-known equations
359: of motion for a two-level system found in Ref.~\cite{thr}.
360: Eqns.~(\ref{alleom}) can be put in
361: dimensionless form if we introduce the dimensionless variables:
362: $\tau = (a/b)t$, $\overline{\Gamma} = (b/a)\Gamma$, and $\overline{\delta} =
363: (b/a)\delta$. Here $a$ and $b$ are the parameters that appear in the
364: background field $\mathbf{F}(t)$. One obtains:
365: \begin{subequations}
366: \label{alldeom}
367: \begin{eqnarray}
368: \frac{dS}{d\tau } & = & -\overline{\Gamma}^{\ast}\; e^{-i
369: \int_{-\tau_{0}/2}^{\tau }\, d\theta\,
370: \overline{\delta}(\theta )}\; I(\tau )
371: \vspace{0.25in}\hspace{0.1in} , \label{deoma} \\
372: \frac{dI}{d\tau } & = & \mbox{}\;\;\;\overline{\Gamma}\, e^{i
373: \int_{-\tau_{0}/2}^{\tau }\,
374: d\theta \,\overline{\delta}(\theta )}\; S(\tau )
375: \hspace{0.1in} , \label{deomb}
376: \end{eqnarray}
377: \end{subequations}
378: where $\tau_{0} = (a/b)T_{0}$ is the (dimensionless) time over which
379: the qubit evolves. For rapid passage, the qubit is initially in the
380: negative energy level $|E_{-}(-\tau_{0}/2)\,\rangle$. This corresponds to
381: the initial condition:
382: \begin{subequations}
383: \label{allinicond}
384: \begin{eqnarray}
385: S(-\tau_{0}/2) & = & 1 \hspace{0.1in} , \label{iniconda} \\
386: I(-\tau_{0}/2) & = & 0 \hspace{0.1in} . \label{inicondb}
387: \end{eqnarray}
388: \end{subequations}
389: Our numerical simulation integrates eqns.~(\ref{alldeom}) over the
390: time-interval $[\,-\tau_{0}/2,\;\tau_{0}/2\,]$ subject to initial
391: condition (\ref{allinicond}). From this we determine the asymptotic
392: transition probability $P$:
393: \begin{equation}
394: P = |I(\tau_{0}/2)|^{2} \hspace{0.1in} ,
395: \label{tranprob2}
396: \end{equation}
397: for $\tau_{0} \gg 1$. Later, we will need the $\tau$-values corresponding to
398: the avoided crossings. These are determined by rewriting eqns.~(\ref{avdx1})
399: and (\ref{avdx2}) in dimensionless form. To this end, we introduce
400: \begin{equation}
401: \eta_{\,n} = \frac{\hbar B \, b^{n-2}}{a^{n-1}} \hspace{0.1in} ,
402: \label{twststr}
403: \end{equation}
404: and recalling that $\tau = (a/b)t$, one easily obtains:
405: \begin{equation}
406: \tau = 0 \hspace{0.1in} ,
407: \label{davdx1}
408: \end{equation}
409: and
410: \begin{equation}
411: \tau = \left(\, sgn\,\eta_{\,n}\,\right)^{\frac{1}{n-2}}\,
412: \left[\,\frac{1}{|\eta_{\,n}|}\,\right]^{\frac{1}{n-2}}
413: \hspace{0.1in} .
414: \label{davdx2}
415: \end{equation}
416: The avoided crossings correspond to $\tau = 0$ and also, for $n\geq 3$,
417: the real solutions of eqn.~(\ref{davdx2}).
418:
419: \subsection{Simulation Test Case: Quadratic Twist}
420:
421: For quadratic twist $\phi_{2}(t) = Bt^{2}$. The instantaneous eigenvalues
422: and eigenvectors of $H(t)$ are easily found to be $E_{\pm}(t)=\pm E(t)$,
423: where $E(t)=\sqrt{b^{2}+(at)^{2}}$, and
424: \begin{equation}
425: |E_{+}(t)\,\rangle = \left( \begin{array}{c}
426: \cos\frac{\theta}{2} \\
427: \sin\frac{\theta}{2}\, e^{i\phi_{2}}
428: \end{array} \right) \hspace{0.5in} ;
429: \hspace{0.5in} |E_{-}(t)\,\rangle =
430: \left( \begin{array}{c}
431: \sin\frac{\theta}{2} \\
432: -\cos\frac{\theta}{2} \, e^{i\phi_{2}}
433: \end{array} \right) \hspace{0.1in} ,
434: \label{eigstates}
435: \end{equation}
436: with $\cos\theta = at/E$. From the eigenstates one obtains:
437: \begin{subequations}
438: \label{allmisc}
439: \begin{eqnarray}
440: \dot{\gamma}_{\pm}(t) & = & -\frac{\dot{\phi}_{2}}{2}\left(\, 1 \mp
441: \cos\theta\, \right) \hspace{0.1in} ;
442: \label{gamdot} \\
443: \Gamma (t) & = & \frac{\dot{\theta}}{2} - i\frac{\dot{\phi}_{2}}{2}
444: \sin\theta \hspace{0.1in} ; \label{biggam} \\
445: \delta (t) & = & \frac{2E}{\hbar} - \dot{\phi}_{2}\cos\theta
446: \hspace{0.1in} . \label{delt}
447: \end{eqnarray}
448: \end{subequations}
449: $\overline{\Gamma}(\tau )$ and $\overline{\delta}(\tau )$ are then
450: determined from eqns.~(\ref{biggam}) and (\ref{delt}) and are found
451: to depend parametrically on the dimensionless ``inversion rate''
452: $\lambda = \hbar a/b^{2}$ and the dimensionless ``twist strength''
453: $\eta_{\, 2} = \hbar B/a$. ``Inversion rate'' and ``twist strength'' are
454: placed in quotes as $\lambda$ does not depend solely on the inversion rate
455: $a$, nor $\eta_{\, 2}$ solely on the twist strength $B$. Crudely speaking,
456: $\lambda = 1$ can be thought of as the boundary separating adiabatic and
457: non-adiabatic inversion rates, with $\lambda > 1$ corresponding to
458: non-adiabatic inversion. Having determined
459: $\overline{\Gamma}(\tau )$ and $\overline{\delta}(\tau )$,
460: eqns.~(\ref{alldeom}) are integrated numerically using an adjustable
461: step-size fourth-order Runge-Kutta algorithm. To simplify comparison
462: of the numerical result for the transition probability with the exact
463: result $P_{2}$, we re-write eqn.~(\ref{tranprob}) in terms of $\lambda$
464: and $\eta_{\, 2}$. One finds:
465: \begin{equation}
466: P_{2} = \exp\left[\, -\frac{\pi}{\lambda}\,
467: \frac{1}{\left|\, 1 - \eta_{\, 2}\,\right| } \,\right]
468: \hspace{0.1in} .
469: \label{dtranprob}
470: \end{equation}
471: Figure~\ref{fig1} shows a representative plot of the transition probability
472: $P(\tau ) = |I(\tau )|^{2}$ versus $\tau$.
473: \begin{figure}[h]
474: \includegraphics[scale=0.5]{fig1.eps}
475: \caption{\label{fig1}Representative plot of transition probability
476: $P(\tau )$ for quadratic twist with $\lambda = 10.0$ and $\eta_{\, 2} = 2.5$.}
477: \end{figure}
478: It is clear for the Figure that
479: the transition occurs in the vicinity of the avoided crossing at $\tau = 0$.
480: Note also that $P(\tau )$ has a small oscillation about its asymptotic
481: value $P = \lim_{\tau\rightarrow\infty}\, P(\tau )$. To average out the
482: oscillation, $P(\tau )$ (for given $\lambda$ and $\eta_{\, 2}$) was
483: calculated for 10 different values of $\tau\gg 1$, and $P$ was identified with
484: the average.
485: Figures~\ref{fig2} and \ref{fig3} show our numerical results for $P$ for
486: various values of $\eta_{\, 2}$ for $\lambda = 10.0$ and $\lambda = 3.0$,
487: respectively. Also plotted in each of these Figures is the exact result
488: $P_{2}$ (eqn.~(\ref{dtranprob})).
489: \begin{figure}[h]
490: \includegraphics[scale=0.5]{fig2.eps}
491: \caption{\label{fig2}Numerical results for the asymptotic transition
492: probability $P$ versus $\eta_{\, 2}$ for quadratic twist with
493: $\lambda = 10.0$. Also plotted is the exact result $P_{2}$.}
494: \end{figure}
495: \begin{figure}[h]
496: \includegraphics[scale=0.5]{fig3.eps}
497: \caption{\label{fig3}Numerical results for the asymptotic transition
498: probability $P$ versus $\eta_{\, 2}$ for quadratic twist with
499: $\lambda = 3.0$. Also plotted is the exact result $P_{2}$.}
500: \end{figure}
501: Figures~\ref{fig2} and \ref{fig3} show that our numerical results are in
502: excellent agreement with the exact result $P_{2}$, and clearly show the
503: quenching of transitions at $\eta_{\, 2} = 1$, and the absence of quenching
504: for negative $\eta_{\, 2}$. The $\lambda$ values shown are purposely highly
505: non-adiabatic. We
506: see that the twist-induced quenching clearly persists into the non-adiabatic
507: regime, although the width of the quench decreases with increasing $\lambda$.
508: The agreement of our simulations with eqn.~(\ref{dtranprob}) at small
509: $\eta_{\, 2}$ indicates that our simulations also account for the geometric
510: factor $\exp\left[\Gamma_{g}\right]$ in $P_{2}$. Having established that
511: our numerical algorithm correctly reproduces the essential results of
512: rapid passage with quadratic twist, we go on to consider the unexplored
513: areas of rapid passage with higher order twist. Referring to
514: Table~\ref{table1}, we see that all cases with odd $n$ have 2 avoided
515: crossings. Cubic ($n=3$) twist corresponds to the simplest example
516: of odd-order twist, and it is examined in the following Section.
517: Similarly, quartic ($n=4$) twist is the simplest example of even-order twist,
518: and we examine it in Section~\ref{sec4}.
519:
520: \section{\label{sec3}Cubic Twist}
521:
522: Having successfully tested our numerical algorithm against the exact
523: results for quadratic twist, we go on to consider cubic twist for which
524: $\phi_{3}(t)=(2/3)Bt^{3}$, and $\eta_{\, 3} = \hbar B b/a^{2}$ (see
525: eqn.~(\ref{twststr})). As in Section~\ref{sec2}, the instantaneous
526: eigenvalues of $H(t)$ are $E_{\pm}(t) =\pm E(t)$, and the instantaneous
527: eigenstates are given by eqn.~(\ref{eigstates}) with $\phi_{2}(t)\rightarrow
528: \phi_{3}(t)$.
529: Eqns.~(\ref{allmisc}) again apply, however $\dot{\phi}_{2}(t)
530: \rightarrow \dot{\phi}_{3}(t)$, and $\overline{\Gamma}(\tau )$ and
531: $\overline{\delta}(\tau )$ are determined from eqns.~(\ref{biggam}) and
532: (\ref{delt}). Having determined $\overline{\Gamma}(\tau )$ and
533: $\overline{\delta}(\tau )$, eqns.~(\ref{alldeom}) can be numerically
534: integrated subject to the initial condition specified in
535: eqns.~(\ref{allinicond}).
536: Before examining results of that integration, we show in Figure~\ref{fig4}
537: a plot of the numerical results for the transition probability $P(\tau )$
538: for $\lambda = 5.0$ and $\eta_{\, 3} = 0$.
539: \begin{figure}[h]
540: \includegraphics[scale=0.5]{fig4.eps}
541: \caption{\label{fig4} Plot of the transition probability $P(\tau )$ for
542: twistless non-adiabatic rapid passage with $\lambda = 5.0$ and $\eta_{\, 3}
543: = 0$.}
544: \end{figure}
545: This corresponds to twistless non-adiabatic rapid passage, and we include
546: this plot for later comparison with related plots for cubic and quartic twist.
547: The asymptotic transition probability for this case is $P=0.533$. Thus, if we
548: were to use this example of twistless non-adiabatic rapid passage to implement
549: a fast NOT-operation on a qubit, the operation would be slightly more likely
550: to produce an inversion (bit-flip) error than not. We will show below that if
551: a small amount
552: of cubic twist is included, the bit-flip error probability can be reduced by
553: 2 orders of magnitude while still maintaining the non-adiabatic inversion
554: rate $\lambda = 5.0$. This substantial reduction in error probability is due
555: to destructive interference between the two avoided crossings that occur
556: during rapid passage with cubic twist.
557:
558: \subsection{Demonstration of Quantum Interference}
559:
560: From eqns.~(\ref{davdx1}) and (\ref{davdx2}), we see that cubic twist is
561: expected to have 2 avoided crossings at $\tau_{1} = 0$ and $\tau_{2}=
562: sgn\,\eta_{\, 3} /|\eta_{\, 3} |$.
563: Figures~\ref{fig5} and \ref{fig6} show $P(\tau )$ for $\lambda = 5.0$
564: and $\eta_{\, 3} = 0.02$ and $\eta_{\, 3} = -0.02$, respectively.
565: \begin{figure}[h]
566: \includegraphics[scale=0.5]{fig5.eps}
567: \caption{\label{fig5} The transition probability $P(\tau )$ for non-adiabatic
568: rapid passage with cubic twist with $\lambda = 5.0$ and $\eta_{\, 3} =
569: 0.02$.}
570: \end{figure}
571: \begin{figure}[h]
572: \includegraphics[scale=0.5]{fig6.eps}
573: \caption{\label{fig6} The transition probability $P(\tau )$ for non-adiabatic
574: rapid passage with cubic twist with $\lambda = 5.0$ and $\eta_{\, 3} =
575: -0.02$.}
576: \end{figure}
577: Figure~\ref{fig5} (\ref{fig6}) clearly shows the expected avoided crossings
578: at $\tau = 0$ and $\tau = 50$ ($-50$). It is also clear from these Figures,
579: and comparison with Figure~\ref{fig4}, that the avoided crossings are
580: constructively interfering, leading to an asymptotic transition probability
581: of $P=0.997$. Figures~\ref{fig7} and \ref{fig8} show $P(\tau )$ for
582: $\lambda = 5.0$ and $\eta_{\, 3} = 0.05$ and $-0.05$, respectively.
583: \begin{figure}[h]
584: \includegraphics[scale=0.5]{fig7.eps}
585: \caption{\label{fig7} The transition probability $P(\tau )$ for non-adiabatic
586: rapid passage with cubic twist with $\lambda = 5.0$ and $\eta_{\, 3} =
587: 0.05$.}
588: \end{figure}
589: \begin{figure}[h]
590: \includegraphics[scale=0.5]{fig8.eps}
591: \caption{\label{fig8} The transition probability $P(\tau )$ for non-adiabatic
592: rapid passage with cubic twist with $\lambda = 5.0$ and $\eta_{\, 3} =
593: -0.05$.}
594: \end{figure}
595: The avoided crossings in Figure~\ref{fig7} (\ref{fig8}) clearly occur at
596: $\tau = 0$ and $\tau = 20$ ($-20$) as expected. Here the avoided crossings
597: interfere destructively, with $P = 0.270$. Summarizing, we see
598: that: (1) two avoided crossings do occur during rapid passage with cubic
599: twist as predicted in Table~\ref{table1}; (2) the avoided crossings
600: produce interference effects in the asymptotic transition probability $P$
601: which can be controlled through variation of their separation; and (3) the
602: separation of the avoided
603: crossings $\Delta\tau_{ac}= |\tau_{2}-\tau_{1}|=1/|\eta_{\, 3}|$ can be
604: altered by varying $\eta_{\, 3} = \hbar B b/a^{2}$. We now consider
605: two possible applications of this interference effect.
606:
607: \subsection{Non-Resonant Pump}
608:
609: First, consider twistless adiabatic rapid passage with $\lambda = 0.5$
610: and $\eta_{\, 3} = 0$. Figure~\ref{fig9} show the transition probability
611: $P(\tau )$
612: for this case. The asymptotic transition probability is
613: $P = 1.87\times 10^{-3}$.
614: \begin{figure}[h]
615: \includegraphics[scale=0.5]{fig9.eps}
616: \caption{\label{fig9} The transition probability $P(\tau )$ for twistless
617: adiabatic rapid passage with $\lambda = 0.5$ and $\eta_{\, 3} = 0$. Note
618: the greatly
619: reduced vertical scale compared to previous figures.}
620: \end{figure}
621: Figure~\ref{fig10} shows $P(\tau )$ for adiabatic
622: rapid passage with cubic twist with $\lambda = 0.5$ and $\eta_{\, 3} = 0.04$.
623: \begin{figure}[h]
624: \includegraphics[scale=0.5]{fig10.eps}
625: \caption{\label{fig10} The transition probability $P(\tau )$ for
626: adiabatic rapid passage with cubic twist with $\lambda = 0.5$ and
627: $\eta_{\, 3} = 0.04$.}
628: \end{figure}
629: The asymptotic transition probability in this case is $P=0.996$! Thus, by
630: introducing a small amount of cubic twist, constructive interference between
631: the avoided crossings transforms adiabatic rapid passage into a non-resonant
632: pump for the qubit energy levels. Figures~\ref{fig5} and \ref{fig6} indicate
633: that, should it be desired, equally large transition probabilities are also
634: possible at faster inversion rates $\lambda$. It is worth pointing out that to
635: produce such a large transition probability using twistless non-adiabatic
636: rapid passage would require $\lambda = 784$ (see eqn.~(\ref{dtranprob})
637: with $\eta_{\, 2} = 0$) as opposed to $\lambda\sim 0.5-5.0$ when cubic twist
638: is exploited.
639:
640: \subsection{Transition Quenching}
641:
642: We now show that one can utilize the interference between avoided crossings
643: to strongly suppress qubit transitions during \textit{non}-\textit{adiabatic}
644: rapid passage with cubic twist. Figure~\ref{fig11} shows $P(\tau )$ for
645: $\lambda = 5.0$ and $\eta_{\, 3} = 4.577\times 10^{-2}$.
646: \begin{figure}[h]
647: \includegraphics[scale=0.5]{fig11.eps}
648: \caption{\label{fig11} The transition probability $P(\tau )$ for non-adiabatic
649: rapid passage with cubic twist with $\lambda = 5.0$ and
650: $\eta_{\, 3} = 4.577\times 10^{-2}$. Note the slightly reduced vertical
651: scale.}
652: \end{figure}
653: The asymptotic transition probability for this case is
654: $P = 3.44\times 10^{-3}$. This is to be compared with twistless rapid passage
655: with $\lambda = 5.0$ (Figure~\ref{fig4}) for which $P = 0.533$.
656: Destructive interference between the two avoided crossings has reduced the
657: transition probability $P$ by 2 orders of magnitude relative to the
658: twistless case shown in Figure~\ref{fig4}. Thus if we were to implement a
659: fast NOT-operation using non-adiabatic rapid passage with cubic twist at
660: $\lambda = 5.0$ and $\eta_{\, 3} = 4.577\times 10^{-2}$, we would obtain
661: (on average)
662: 1 bit-flip error per 291 NOT-operations. By comparison, twistless rapid
663: passage with $\lambda = 5.0$ would produce (on average) 1 bit-flip error
664: for every 2 NOT-operations. This result strongly suggest the value of
665: exploring whether this destructive interference between
666: avoided crossings during twisted rapid passage could be exploited to produce
667: fast reliable quantum NOT and CNOT logic gates. As striking as this result for
668: cubic twist is, we shall see in the following Section that quartic twist can
669: reduce the bit-flip error probability even more dramatically.
670:
671: \section{\label{sec4}Quartic Twist}
672:
673: For quartic twist $\phi_{4}(t) = (1/2)Bt^{4}$ and $\eta_{\, 4} = \hbar Bb^{2}
674: /a^{3}$. Avoided crossings are expected to occur at $\tau_{1} = 0$, and
675: at $\tau_{2} = \pm 1/\sqrt{\eta_{\, 4}}$ (when $sgn\,\eta_{\, 4} = +1$;
676: see eqn.~(\ref{davdx2}) and Table~\ref{table1}). Formally, the analysis of
677: quartic twist parallels
678: that of quadratic and cubic twist. With the substitution $\phi_{2}(t)
679: \rightarrow \phi_{4}(t)$, eqns.~(\ref{eigstates}) and (\ref{allmisc}) continue
680: to apply, and one determines $\overline{\Gamma}(\tau )$ and
681: $\overline{\delta}(\tau )$ from eqns.~(\ref{biggam}) and (\ref{delt}). Once
682: $\overline{\Gamma}(\tau )$ and $\overline{\delta}(\tau )$ are known,
683: eqns.~(\ref{alldeom}) can be integrated numerically subject to the initial
684: condition specified in eqns.~(\ref{allinicond}).
685:
686: \subsection{Demonstration of Quantum Interference}
687:
688: In Figure~\ref{fig12} we plot the transition probability $P(\tau )$ for
689: $\lambda = 5.0$ and $\eta_{\, 4} = 4.6\times 10^{-4}$.
690: \begin{figure}[!h]
691: \includegraphics[scale=0.5]{fig12.eps}
692: \caption{\label{fig12} The transition probability $P(\tau )$ for
693: non-adiabatic rapid passage with quartic twist with $\lambda = 5.0$ and
694: $\eta_{\, 4} = 4.6\times 10^{-4}$.}
695: \end{figure}
696: The expected avoided crossings at $\tau_{1}=0$ and $\tau_{2}=\pm 46.63$
697: are clearly visible. The asymptotic transition probability for this case is
698: $P = 0.88$. For twistless rapid passage with
699: $\lambda = 5.0$
700: (see Figure~\ref{fig4}), $P = 0.533$. Thus the avoided crossings in
701: Figure~\ref{fig12} are constructively interfering, leading to an enhancement
702: of the transition probability $P$. Figure~\ref{fig13} shows $P(\tau )$ for
703: quartic twist with $\lambda = 5.0$ and $\eta_{\, 4} = -4.6\times 10^{-4}$.
704: \begin{figure}[!h]
705: \includegraphics[scale=0.5]{fig13.eps}
706: \caption{\label{fig13} The transition probability $P(\tau )$ for
707: non-adiabatic rapid passage with quartic twist with $\lambda = 5.0$ and
708: $\eta_{\, 4} = -4.6\times 10^{-4}$.}
709: \end{figure}
710: This Figure clearly shows only one avoided crossing at $\tau_{1}=0$, as
711: expected for $sgn\,\eta_{\, 4}= -1$ (see Table~\ref{table1}). The asymptotic
712: transition probability in this case is $P = 0.533$ which equals
713: the result for twistless rapid passage with $\lambda = 5.0$
714: (Figure~\ref{fig4}) to the level of precision obtained in our calculation.
715:
716: Figure~\ref{fig14} plots $P(\tau )$ for $\lambda = 5.0$ and $\eta_{\, 4} =
717: 1.6\times 10^{-3}$.
718: \begin{figure}[!h]
719: \includegraphics[scale=0.5]{fig14.eps}
720: \caption{\label{fig14} The transition probability $P(\tau )$ for
721: non-adiabatic rapid passage with quartic twist with $\lambda = 5.0$ and
722: $\eta_{\, 4} = 1.6\times 10^{-3}$. Note the slightly reduced vertical scale.}
723: \end{figure}
724: The Figure clearly shows the expected crossings at $\tau_{1}=0$ and
725: $\tau_{2} = \pm 25.0$. The asymptotic transition probability is
726: $P = 6.93\times 10^{-4}$ and corresponds to destructive interference
727: relative to twistless rapid passage with $\lambda = 5.0$ (Figure~\ref{fig4}).
728: We do not include a plot of $P(\tau )$ for $\lambda = 5.0$ and $\eta_{\, 4} =
729: -1.6\times 10^{-3}$ as it is similar to Figure~\ref{fig13}: one avoided
730: crossing at $\tau_{1}=0$ and $P = 0.533$.
731:
732: Summarizing these results, we see that: (i) three (one) avoided crossings
733: (crossing) occur(s)
734: as predicted in Table~\ref{table1} when $sgn\,\eta_{\, 4} = +1\; (-1)$;
735: (ii) the avoided crossings produce interference effects in the
736: transition probability, with the character of the interference (constructive
737: or destructive) determined by the separation of the avoided crossings; and
738: (iii) the separation of adjacent avoided crossings is given by $\Delta\tau_{ac}
739: = |\tau_{2}-\tau_{1}| = 1/\sqrt{\eta_{\, 4}}$ ($sgn\,\eta_{\, 4}=+1$), and
740: it is controllable through variation of $\eta_{\, 4}=\hbar Bb^{2}/a^{3}$.
741:
742: \subsection{Non-Resonant Pump}
743:
744: Quartic twist does not appear to be as effective at pumping the qubit
745: energy-levels as cubic twist. Figure~\ref{fig15} shows $P(\tau )$ for
746: $\lambda = 0.5$ and $\eta_{\, 4} = 6.45\times 10^{-3}$.
747: \begin{figure}[!h]
748: \includegraphics[scale=0.5]{fig15.eps}
749: \caption{\label{fig15} The transition probability $P(\tau )$ for
750: adiabatic rapid passage with quartic twist with $\lambda = 0.5$ and
751: $\eta_{\, 4} = 6.45\times 10^{-3}$.}
752: \end{figure}
753: The expected avoided crossings at $\tau_{1}=0$ and $\tau_{2}=\pm 12.45$
754: are clearly visible, and the asymptotic transition probability is
755: $P = 0.20$. Although this is a 2 order of magnitude improvement over
756: twistless adiabatic rapid passage with $\lambda = 0.5$ (Figure~\ref{fig9}),
757: it falls well short of the transition probability $P=0.996$ easily obtainable
758: with cubic twist. In fact, for $\eta_{\, 4}<1$, $P\sim 0.20$ was among the
759: largest $P$-values we could find. If larger values of twist strength are
760: allowed, the largest transition probability we could find was $P=0.64$
761: at $\eta_{\, 4}=3.00$.
762:
763: \subsection{\label{sec4c}Transition Quenching}
764:
765: Quartic twist proves to be much more effective at quenching
766: transitions during non-adiabatic rapid passage than cubic twist.
767: Table~\ref{table2} gives the transition probabilities for quartic twist
768: pulses for which $\lambda = 5.00$ and $\eta_{4}$ lies in the interval
769: [$3.95\times 10^{-3}$, $4.04\times 10^{-3}$].
770: \begin{table}[!h]
771: \caption{\label{table2}Transition probabilities for quartic twist with
772: $\protect\lambda = 5.00$ and $\protect\eta_{4}$ in the range [$3.95\times
773: 10^{-3}$, $4.04\times 10^{-3}$]\vspace{0.3in}.}
774: \begin{ruledtabular}
775: \begin{tabular}{cc}
776: $\:\eta_{4} \;\; (\,\times 10^{-3}\, )\; $ & \, $P$ \\ \hline
777: 3.95 & $2.0\times 10^{-2} $ \\
778: 3.96 & $1.3\times 10^{-2} $ \\
779: 3.97 & $6.8\times 10^{-3} $ \\
780: 3.98 & $3.6\times 10^{-3} $ \\
781: 3.99 & $9\times 10^{-4} $ \\
782: 4.00 & $4\times 10^{-5} $ \\
783: 4.01 & $8\times 10^{-4} $ \\
784: 4.02 & $3.9\times 10^{-3} $ \\
785: 4.03 & $1.0\times 10^{-2} $ \\
786: 4.04 & $1.7\times 10^{-2} $ \\
787: \end{tabular}
788: \end{ruledtabular}
789: \end{table}
790: The essential thing to notice in Table~\ref{table2} is that for $\eta_{4}
791: = 4.00\times 10^{-3}$, the transition probability $P = 4\times 10^{-5}$.
792: This is significant for the following reason. It has been shown that a quantum
793: computation of arbitrarily long duration becomes possible if the quantum
794: logic gates used to implement the computation all have error probabilities
795: (per gate operation) which lie below the threshold $P_{ft}$ for fault tolerant
796: operation \cite{pre}. This threshold has been estimated to be $P_{ft} \sim
797: 10^{-4}-10^{-5}$ \cite{flt}. In terms of the gate fidelity $F = 1 - P$,
798: the more optimistic estimate for $P_{ft}$ gives $F_{ft} = 0.9999$. We see
799: that for $\lambda = 5.00$ and $\eta_{4} = 4.00\times 10^{-3}$, twisted rapid
800: passage with quartic twist gives a gate fidelity of $F = 0.99996$ which
801: exceeds the best case estimate for fault tolerant operation $F_{ft} = 0.9999$.
802: This fault tolerant performance is achieved while inverting the qubit at a
803: \textit{non}-\textit{adiabatic} rate.
804: The reader should note that the values $\lambda = 5.00$ and $\eta_{4} =
805: 4.0\times 10^{-3}$ can be realized with existing NMR technology (see
806: Section~\ref{sec5d}). Our analysis
807: raises the exciting possibility that non-adiabatic rapid passage with quartic
808: twist might provide a means of realizing \textit{fast fault-tolerant} NOT and
809: CNOT gates. The novelty of this prospect is the marriage of operational speed
810: with fault-tolerance.
811: This marriage of speed and reliability is a direct consequence of the
812: destructive interference which is possible between the 3 avoided crossings
813: that arise during rapid passage with quartic twist. Quantum CNOT gates
814: are ubiquitous in quantum computing and quantum error correction
815: \cite{uni,bar,qec}. Thus, determining how to implement them in a fast
816: fault tolerant manner is a potentially significant development for the field.
817:
818: \section{\label{sec5}Discussion}
819:
820: \subsection{Summary}
821:
822: It has been our aim in this paper to show that multiple avoided crossings can
823: arise during twisted rapid passage, and that by varying their time-separation,
824: interference effects are produced which allow for a direct control over
825: qubit transitions. This time-separation is controlled through the
826: (dimensionless) twist strength $\eta$, and the resulting interference can be
827: constructive (enhancing transitions) or destructive (reducing
828: transitions). For nth-order polynomial twist, $\eta_{\, n} = \hbar Bb^{n-2}/
829: a^{n-1}$, where $B$ is the (dimensionful) twist strength, $2b$ is the
830: energy-gap separating the qubit energy-levels at an avoided crossing, and $a$
831: is the inversion rate of the external field $\mathbf{F}(t)$ (see
832: Section~\ref{sec2}). The interference effects are a consequence of the
833: temporal phase coherence of the wavefunction. The ability to vary this
834: coherence by varying the temporal separation of the avoided crossings renders
835: twisted rapid passge with adjustable twist strength into a temporal
836: interferometer through which qubit transitions can be greatly enhanced or
837: suppressed. Cubic and quartic twist were explicitly considered in
838: this paper as they are, respectively, the simplest examples of odd-order and
839: even-order polynomial twist in which these interference effects are
840: expected to occur. Although we have focused on these two cases, we do not
841: mean to suggest that these pulses are the best of all possible twisted
842: rapid passage pulses. A search is currently underway for other twisted rapid
843: passage pulses that might produce stronger destructive interference, and hence,
844: faster, more fault tolerant quantum CNOT and NOT gates (see below for
845: further discussion). We have seen that this interference mechanism
846: can be used to pump qubit energy-levels, as well as to strongly quench
847: qubit transitions during \textit{non}-\textit{adiabatic} twisted rapid
848: passage. Although cubic twist proved to be more effective at pumping than
849: quartic twist, quartic twist was found to be much more effective at quenching
850: qubit transitions. We have seen that quartic twist allows qubit inversion
851: to be done both non-adiabatically and at fidelities that exceed the threshold
852: for fault tolerant operation. The marriage of operational speed with
853: reliability is a direct consequence of the destructive interference that is
854: possible between the 3 avoided crossings that can arise during rapid passage
855: with quartic twist.
856:
857: \subsection{Implementing Quantum CNOT Gate}
858:
859: We now describe a procedure for implementing a quantum CNOT gate using
860: twisted rapid passage in the context of liquid state NMR. If the
861: liquid has low viscosity, one can ignore dipolar coupling between the
862: qubits, and if the remaining Heisenberg interaction between the qubits is
863: weak compared to the individual qubit Zeeman energies, it can be
864: well-approximated by an Ising interaction \cite{chu}. Under these
865: conditions, the Hamiltonian (in frequency units) for the control ($c$) and
866: target ($t$) qubits is:
867: \begin{equation}
868: \frac{H_{ct}}{\hbar} = -\omega_{c}\, I^{c}_{z} -\omega_{t}\, I^{t}_{z}
869: + 2\pi J\, I^{c}_{z}\,I^{t}_{z} \hspace{0.1in} .
870: \end{equation}
871: Here $\omega_{c}$ ($\omega_{t}$) is the resonance frequency of the isolated
872: control (target) qubit; $J$ is the Ising coupling constant; and
873: $\omega_{c} > \omega_{t} > \pi J$. We choose the single qubit computational
874: basis states (CBS) to be $|0\rangle = |\uparrow\rangle$ and $|1\rangle =
875: |\downarrow\rangle$. Thus the two-qubit CBS are: $|00\rangle = |\uparrow
876: \uparrow\rangle$; $|01\rangle = |\uparrow\downarrow\rangle$; $|10\rangle =
877: |\downarrow\uparrow\rangle$; and $|11\rangle = |\downarrow\downarrow\rangle$,
878: and they are the eigenstates of $H_{ct}$. The energy-levels (in frequency
879: units) are shown in Figure~\ref{fig16}, where
880: \begin{equation}
881: \omega_{\pm} = \omega_{t} \pm \pi J \hspace{0.1in} .
882: \end{equation}
883: \begin{figure}[h]
884: \includegraphics[scale=0.5]{fig16.epsi}
885: \caption{\label{fig16} Energy-level structure appropriate for implementing a
886: quantum CNOT operation using twisted rapid passage. The corresponding energies
887: (in frequency units) appear to the right of the energy levels.}
888: \end{figure}
889: \noindent Given this energy-level structure, we can implement a quantum CNOT
890: operation on the two qubits by sweeping through the $\omega_{+}$ resonance
891: using
892: twisted rapid passage. Decoupling \cite{ch2} is used to switch off the
893: dynamics of the
894: control qubit so that only the target qubit responds to the rapid passage
895: pulse. Since the two states $|00\rangle$ and $|01\rangle$ are not resonant,
896: they do not respond to the twisted rapid passage pulse. Thus,
897: \begin{equation}
898: \begin{array}{l}
899: |00\rangle\rightarrow |00\rangle \\
900: |01\rangle\rightarrow |01\rangle \hspace{0.1in} .
901: \end{array}
902: \end{equation}
903: On the other hand, for the $|10\rangle$ and $|11\rangle$ states, the
904: combination of decoupling and sweeping through the $\omega_{+}$ resonance
905: means that only the target qubit has its spin flipped. Thus,
906: \begin{equation}
907: \begin{array}{l}
908: |10\rangle\rightarrow |11\rangle \\
909: |11\rangle\rightarrow |10\rangle \hspace{0.1in} ,
910: \end{array}
911: \end{equation}
912: and we see that this procedure implements a quantum CNOT operation on the
913: two qubits.
914:
915: \subsection{Experimental Realization}
916:
917: Because of the fundamental significance of quantum CNOT gates to quantum
918: computing and quantum error correction \cite{uni,bar,qec}, it is hoped that
919: the feasibility of using rapid passage with quartic twist to implement this
920: gate might be tested experimentally (see penultimate paragraph of
921: Section~\ref{sec1}). Experimental realization of polynomial
922: twist $\phi_{n}(t) = (2/n)Bt^{n}$ should be possible through an adaptation of
923: the procedure used by Zwanziger et.\ al.\ \cite{zwa} to realize quadratic
924: twist. Thus: (1) the driving rf-field
925: is linearly polarized along the x-axis in the lab-frame with $F_{x}(t)
926: = 2b\cos\phi_{rf}(t)$; (2) the resonance offset $at$ (see eqn.~(\ref{labham}))
927: is produced by linearly sweeping the detector frequency $\omega_{det}(t)$
928: through the resonance at the Larmor frequency $\omega_{0}$ such that
929: $\omega_{det}(t) = \omega_{0} + (2at/\hbar )$; and (3) twist is introduced by
930: sweeping the rf-frequency $\omega_{rf}(t) = \dot{\phi}_{rf}$ through the
931: resonance at $\omega_{0}$ in such a way that $\omega_{rf}(t) = \omega_{det} -
932: \dot{\phi}_{n}$. It is worth noting that the resonance condition
933: $\omega_{rf}(t) =\omega_{0}$ is identical to our existence condition
934: for avoided crossings, eqn.~(\ref{cond}). Note that in our paper
935: the external field inversion takes place over the time-interval ($-T_{0}/2$,
936: $T_{0}/2$); the external field crosses the x-y plane at $t=0$ and is
937: initially aligned along the $-\hat{\mathbf{z}}$ direction. The Appendix
938: provides a translation key which relates the theoretical parameters of this
939: paper
940: to the experimental parameters of the Zwanziger experiments \cite{jwz,zwa}.
941:
942: Before leaving the subject of experimental realization of rapid passage
943: with quartic twist, two further remarks are in order. First, to insure
944: that all qubits are inverted when a spread of resonance frequencies occurs,
945: it is necessary to require that the frequency sweep cover a large enough
946: interval that the entire spread of resonance frequencies is included in it.
947: This gaurantees that all qubits will have passed through resonance by the end
948: of the frequency sweep. Second, a range of rf field strengths can also be
949: accomodated so long as $aT/2 \gg b_{max}$. This condition insures that the
950: frequency sweep begins far from resonance for all rf field strengths, and that
951: transitions will continue to occur only near the avoided crossings. One
952: therefore anticipates that in this case also, the interference effects will
953: continue to occur as predicted.
954: For reasonably good samples, magnets, and rf sources these constraints can be
955: satisfied, and the interference effects presented above should be readily
956: observable. This is in fact what is found experimentally~\cite{jwz}.
957:
958: \subsection{\label{sec5d}Other Pulses}
959:
960: Having introduced twisted rapid passage with polynomial twist, and pointed out
961: the possible advantages of quartic twist for quantum computing, it is natural
962: to ask how quartic twist compares with the more familiar $\pi$-pulse which
963: can also be used to implement quantum CNOT and NOT gates. We begin by comparing
964: the inversion time for quartic twist with that of a comparable $\pi$-pulse.
965: We focus on quartic twist with $\lambda = 5.00$ and $\eta_{4}= 4.00\times
966: 10^{-3}$ as this choice of parameters yields a gate fidelity $F=0.99996$
967: (see Section~\ref{sec4c}) which exceeds the threshold for fault tolerant
968: gate operation $F_{ft}\sim 0.9999$. We now show that this case achieves fault
969: tolerant operation while simultaneously matching the inversion speed of a
970: $\pi$-pulse.
971: In the notation of Ref.~\cite{zwa}, the basic experimental parameters for
972: twisted rapid passage are $A$, $B$, $\omega_{1}$, and $T$, and they are
973: related to our theoretical parameters by eqns.~(\ref{A8}) and (\ref{A3}).
974: $T$ continues to denote the duration of the twisted rapid passage pulse.
975: Because a twisted rapid passage sweep must begin
976: far from the avoided crossings, $A$ and $\omega_{1}$ cannot be chosen
977: independently. In the rf-frame, the asymptotic effective magnetic field must
978: lie near the z-axis so that
979: $\tan\theta = \omega_{1}/A \sim 0.1$. Choosing $\omega_{1} = 4000$ Hz gives
980: $A = 4\times 10^{4}$ Hz. Both of these values can be achieved with existing
981: NMR technology. Writing $f = \omega_{1}/A$, eqn.~(\ref{A9}) gives
982: \begin{displaymath}
983: T_{4} = \frac{4}{f\omega_{1}\lambda} \hspace{0.1in} .
984: \end{displaymath}
985: With $\lambda = 5.00$, this gives
986: \begin{displaymath}
987: T_{4} = 2\; \textrm{msec} \hspace{0.1in} .
988: \end{displaymath}
989: By comparison, the inversion time for a $\pi$-pulse with rf-amplitude
990: $\omega_{1} = 4000$ Hz is $T_{\pi} = \pi /\omega_{1} = 0.8$ msec. Thus,
991: twisted rapid passage with quartic twist is clearly capable of matching the
992: inversion speed of a comparable $\pi$-pulse while still exceeding the
993: threshold for fault tolerant operation. On the other hand, the error
994: probability for a typical $\pi$-pulse is $P\sim 10^{-3}$ due to,
995: for example, inhomogenities in the rf field amplitude. This corresponds to a
996: fidelity $F \sim 0.999$ so that, unlike the equally fast quartic twist pulse
997: which acts
998: fault tolerantly, the $\pi$-pulse falls \textit{short\/} of the threshold for
999: fault tolerant operation $F_{th} \sim 0.9999$.
1000:
1001: We hope in the future to examine higher order versions of polynomial twist
1002: to determine whether they have more effective quenching and/or robustness
1003: properties than cubic and quartic twist. We have also done preliminary work on
1004: the interesting case of periodic twist: $\phi (t) = \pi\rho\sin\omega t$.
1005: As we have seen, polynomial twist only allows 1--3 avoided crossings to occur
1006: during rapid passage. One can show that periodic twist allows the number of
1007: avoided crossings that occur during rapid passage to be modified through
1008: variation of the twist amplitude $\rho$ and frequency $\omega$. We intend
1009: to explore how the interference effects considered here are modified when more
1010: than 3 avoided crossings can occur.
1011:
1012: \begin{acknowledgments}
1013: I would like to thank: (1) T. Howell III for continued support; and
1014: (2) the National Science Foundation for support provided through
1015: grant number NSF-PHY-0112335.
1016: \end{acknowledgments}
1017:
1018: \appendix
1019: \section{Connection Between Theory and Experiment}
1020:
1021: For ease of comparison with Refs.~\cite{zwa} and \cite{jwz}, we choose
1022: $\mathbf{F}(t) = -b\cos\phi_{n}(t)\hat{\mathbf{x}} -b\sin\phi_{n}(t)
1023: \hat{\mathbf{y}} + at\hat{\mathbf{z}}$ in eqn.~(\ref{labham}). The
1024: Hamiltonian in the detector frame is then
1025: \begin{displaymath}
1026: \frac{H(t)}{\hbar} = \frac{at}{\hbar}\,\sigma_{z} - \frac{b}{\hbar}\,
1027: \cos\phi_{n}(t)\,\sigma_{x} -
1028: \frac{b}{\hbar}\,\sin\phi_{n}(t)\,
1029: \sigma_{y} \hspace{0.1in} .
1030: \end{displaymath}
1031: Here $\phi_{n}(t) = (2/n)\mathcal{B}t^{n}$, and to avoid confusion with the
1032: notation of Ref.~\cite{zwa}, we have switched the symbol used for the
1033: twist strength in the main body of this paper: $B\rightarrow \mathcal{B}$.
1034: Transformation to the rf-frame is done using the unitary operator $U(t) =
1035: \exp [-(i/2)\,\phi_{n}(t)\,\sigma_{z}]$ so that $H(t)\rightarrow
1036: \overline{H}(t)$:
1037: \begin{eqnarray}
1038: \frac{\overline{H}(t)}{\hbar} & = & \left(\,\frac{at}{\hbar} -
1039: \frac{\dot{\phi}_{n}}{2}\,\right)\,
1040: \sigma_{z} - \frac{b}{\hbar}\,
1041: \sigma_{x} \nonumber \\
1042: & = & \left(\,\frac{2at}{\hbar} - \dot{\phi}_{n}\,\right)\, I_{z} -
1043: \frac{2b}{\hbar}\, I_{x} \hspace{0.1in} , \label{A1}
1044: \end{eqnarray}
1045: and $\mathbf{I} = \mathbf{\sigma}/2$.
1046:
1047: The experimental Hamiltonian in the rf-frame appears in eqn.~(12) of
1048: Ref.~\cite{zwa}:
1049: \begin{equation}
1050: \frac{\overline{H}_{ex}(t)}{\hbar} = \left(\,\dot{\phi}_{rf} - \omega_{0}\,
1051: \right)\, I_{z} - \omega_{1}\, I_{x}
1052: \hspace{0.1in} . \label{A2}
1053: \end{equation}
1054: Comparing eqns.~(\ref{A1}) and (\ref{A2}) gives
1055: \begin{equation}
1056: \omega_{1} = \frac{2b}{\hbar}
1057: \label{A3}
1058: \end{equation}
1059: and
1060: \begin{equation}
1061: \dot{\phi}_{rf} - \omega_{0} = \frac{2at}{\hbar} - \dot{\phi}_{n}
1062: \hspace{0.1in} . \label{A4}
1063: \end{equation}
1064: Integrating eqn.~(\ref{A4}) gives
1065: \begin{equation}
1066: \phi_{rf}(t^{\prime\prime}) = \int_{-T/2}^{t^{\prime\prime}}\,
1067: dt^{\prime\prime\prime}\,\left[\,\omega_{0} +
1068: \frac{2aT}{\hbar}\left(\,
1069: \frac{t^{\prime\prime\prime}}{T}\,\right) -
1070: 2\mathcal{B}T^{n-1}\,\left(\,
1071: \frac{t^{\prime\prime\prime}}{T}\,\right)^{n-1}
1072: \,\right] \hspace{0.1in} . \label{A5}
1073: \end{equation}
1074: In this paper, we have parameterized time such that $t^{\prime\prime\prime}
1075: \in [\, -T/2, T/2\,]$, and $T$ is the duration of the twisted rapid
1076: passage pulse. Defining
1077: \begin{displaymath}
1078: \tau = \frac{t^{\prime\prime\prime}}{T} + \frac{1}{2} \hspace{0.1in} ,
1079: \end{displaymath}
1080: it follows that $\tau\in [0, 1]$. Introducing $t^{\prime} =
1081: t^{\prime\prime\prime}+ T/2$ and $t = t^{\prime\prime} + T/2$, eqn.~(\ref{A5})
1082: becomes
1083: \begin{equation}
1084: \phi_{rf}(t) = \int_{0}^{t}\, dt^{\prime}\,\left[\,\omega_{0} +
1085: \left(\, \frac{2aT}{\hbar}\,\right)\,\left(\,\tau -
1086: \frac{1}{2}\,\right) - \frac{2\mathcal{B}T^{n}}{T}\,
1087: \left(\,\tau - \frac{1}{2}\,\right)^{n-1}\,\right]
1088: \hspace{0.1in} . \label{A6}
1089: \end{equation}
1090:
1091: As explained in the caption of Figure~2 of Ref.~\cite{zwa},
1092: $\dot{\phi}_{rf} = \dot{\phi}_{det} - \dot{\phi}_{n}$; with $\dot{\phi}_{det}
1093: = \omega_{0} + 2A(\tau - 1/2)$; and generalizing to polynomial twist,
1094: $\dot{\phi}_{n} = nB(\tau - 1/2)^{n-1}/T$, where $B$ is the symbol used in
1095: Ref.~\cite{zwa} for the twist strength. Plugging these expressions for
1096: $\dot{\phi}_{det}$ and $\dot{\phi}_{n}$ into $\dot{\phi}_{rf} =
1097: \dot{\phi}_{det} - \dot{\phi}_{n}$, and integrating gives
1098: \begin{equation}
1099: \phi_{rf}(t) = \int_{0}^{t}\, dt^{\prime}\,\left[\,\omega_{0} +
1100: 2A\left(\,\tau - \frac{1}{2}\,\right) - \frac{nB}{T}\,
1101: \left(\,\tau - \frac{1}{2}\,\right)^{n-1}\,\right]
1102: \hspace{0.1in} . \label{A7}
1103: \end{equation}
1104: Equating eqns.~(\ref{A6}) and (\ref{A7}) gives
1105: \begin{subequations}
1106: \label{A8}
1107: \begin{eqnarray}
1108: A & = & \frac{aT}{\hbar} \label{A8a} \\
1109: B & = & \frac{2\mathcal{B}}{n}T^{n} \hspace{0.1in} . \label{A8b}
1110: \end{eqnarray}
1111: \end{subequations}
1112: Using eqns.~(\ref{A3}) and (\ref{A8}) in the definition of $\lambda$ (see
1113: discussion following eqns.~(\ref{allmisc})) gives
1114: \begin{equation}
1115: \lambda = \frac{4\, |A|}{\omega_{1}^{2}T} \hspace{0.1in} . \label{A9}
1116: \end{equation}
1117: Using eqns.~(\ref{A3}), (\ref{A8}) and eqn.~(\ref{twststr}) with $n=3$ and $4$
1118: gives
1119: \begin{equation}
1120: \eta_{3} = \frac{3}{4}\,\frac{B\omega_{1}}{A^{2}T} \label{A10}
1121: \end{equation}
1122: and
1123: \begin{equation}
1124: \eta_{4} = \frac{B\omega_{1}^{2}}{2A^{3}T} \hspace{0.1in} ,
1125: \label{A11}
1126: \end{equation}
1127: respectively. The results of this Appendix give the connection between our
1128: theoretical parameters and the experimental
1129: parameters $A$, $B$, $\omega_{1}$, and $T$ of the Zwanziger experiments
1130: \cite{jwz,zwa}. In Section~\ref{sec5}, these formulas are used to calculate
1131: the inversion time for a twisted rapid passage pulse with quartic twist.
1132:
1133: \begin{thebibliography}{99}
1134: \bibitem{arp} A. Abragam, \textit{Principles of Nuclear Magnetism}
1135: (Oxford University Press, New York 1961).
1136: \bibitem{bry} M. V. Berry, Proc.\ R. Soc.\ Lond.\ A \textbf{430}, 405
1137: (1990).
1138: \bibitem{zwa} J. W. Zwanziger, S. P. Rucker, and G. C. Chingas,
1139: Phys.\ Rev.\ A \textbf{43}, 3232 (1991).
1140: \bibitem{jwz} J. W. Zwanziger, U. Werner-Zwanziger, and F. Gaitan, submitted
1141: to Chem.\ Phys.\ Lett.\ .
1142: \bibitem{lim} R. Lim, J. Phys.\ A \textbf{26}, 7615 (1993).
1143: \bibitem{suo} K-A Suominen and B. M. Garraway, Phys.\ Rev.\ A \textbf{45},
1144: 374 (1992); K-A Suominen, Opt.\ Commun. \textbf{93}, 126 (1992);
1145: K-A Suominen, B. M. Garraway, and S. Stenholm, Opt.\ Commun.\
1146: \textbf{82}, 260 (1991).
1147: \bibitem{joy} A. Joye, J. Phys.\ A \textbf{26}, 6517 (1993); A. Joye, G.
1148: Mileti, and C. E. Pfister, Phys.\ Rev.\ A \textbf{44}, 4280
1149: (1991).
1150: \bibitem{lan} L. Landau, Phys.\ Z. Sowjetunion \textbf{1}, 46 (1932)
1151: \bibitem{zen} C. Zener, Proc.\ R. Soc.\ Lond.\ A \textbf{137}, 696
1152: (1932).
1153: \bibitem{geo} A. Shapere and F. Wilczek, \textit{Geometric Phases in
1154: Physics\/} (World Scientific, New Jersey, 1989).
1155: \bibitem{thr} W. R. Thorson, J. B. Delos, and S. A. Boorstein, Phys.\
1156: Rev.\ A \textbf{4}, 1052 (1971).
1157: \bibitem{pre} J. Preskill, Proc.\ R. Soc.\ Lond.\ A \textbf{454}, 385 (1998).
1158: \bibitem{flt} P. Shor, in \textit{Proceedings of the 37th Symposium on the
1159: Foundations of Computer Science}, (IEEE Computer Society Press,
1160: Los Alamitos, CA 1996), pp.\ 56-65; D. Gottesman, PhD thesis
1161: preprint http://www.arxiv.org/quant-ph/9705052.
1162: \bibitem{uni} D. P. DiVincenzo, Proc.\ R. Soc.\ Lond.\ A \textbf{454}, 261
1163: (1998).
1164: \bibitem{bar} A. Barenco et.\ al.\ , Phys.\ Rev.\ A \textbf{52}, 3457 (1995).
1165: \bibitem{qec} P. W. Shor, Phys.\ Rev.\ A \textbf{52}, R2493 (1995);
1166: A. M. Steane, Proc.\ R. Soc.\ Lond.\ A \textbf{452}, 2551
1167: (1996); A. R. Calderbank and P. W. Shor, Phys.\ Rev.\ A
1168: \textbf{54}, 1098 (1996); D. Gottesman, Phys.\ Rev.\ A
1169: \textbf{54}, 1862 (1996); E. Knill and R. Laflamme, Phys.\
1170: Rev.\ A \textbf{55}, 900 (1997); see also A. M. Steane, in
1171: \textit{Introduction to Quantum Computation and Information},
1172: eds.\ H. K. Lo, S. Popescu, and T. Spiller (World Scientific,
1173: New Jersey, 1998).
1174: \bibitem{chu} N. A. Gershenfeld and I. L. Chuang, Science \textbf{275}, 350
1175: (1997).
1176: \bibitem{ch2} I. L. Chuang, in \textit{Introduction to Quantum Computation
1177: and Information}, eds.\ H. K. Lo, S. Popescu, and T. Spiller
1178: (World Scientific, New Jersey, 1998).
1179: \end{thebibliography}
1180: \end{document}
1181: