1: \documentclass[prl,10pt,twocolumn,showpacs]{revtex4}
2:
3: \usepackage{graphicx}
4: \usepackage{color}
5: \usepackage{epsfig}
6: \usepackage{amsmath}
7: \usepackage{amstext}
8: \usepackage{amssymb}
9:
10: \usepackage[running]{lineno}
11: \DeclareGraphicsExtensions{.eps,.ps}
12:
13:
14: \begin{document}
15:
16:
17:
18: \author{Mateusz Cholascinski}
19:
20: \author{Ravindra W. Chhajlany}
21:
22: \affiliation{Institute of Physics, Adam Mickiewicz University, 61-614
23: Pozna\'n, Poland}
24:
25: \date{\today}
26:
27: \title{ Stabilized parametric Cooper-pair pumping in a
28: linear array of coupled Josephson junctions}
29:
30:
31:
32: \begin{abstract}
33: We present an experimentally realizable stabilized charge pumping
34: scheme in a linear array of Cooper-pair boxes. The system design
35: intrinsically protects the pumping mechanism from severe errors,
36: especially current reversal and spontaneous charge excitation. The
37: quantum Zeno effect is implemented to further diminish pumping
38: errors. The characteristics of this scheme are considered from the
39: perspective of improving the current standard. Such an improvement
40: bears relevence to the closure of the so-called measurement triangle
41: % which
42: % enables the relation of the current, voltage and frequency by
43: % fundamental constants
44: (see D. Averin [Nature 434, 285 (2005)]).
45: \end{abstract}
46:
47: \pacs{74.81.Fa 03.65.Xp 73.23.-b 85.25.Cp}
48:
49: \maketitle
50: % \tableofcontents
51:
52:
53:
54: Adiabatic variation of parameters in a quantum system results in
55: evolution in which population of each of the energy levels is
56: conserved (adiabatic following). It is well known that, during such
57: evolution, the energy levels acquire two types of phases: time
58: dependent dynamical phases and time independent geometrical phases
59: determined only by the geometry of the path traversed in parameter
60: space. The property of adiabatic following can be also used to induce
61: transport through a system, {\it viz.} a slowly transported potential
62: well carries its bounded particles. \\ For closed paths, the initial
63: and final states (usually) differ only by a phase factor \cite{berry},
64: but the system evolution during the cyclic variation can still be
65: characterized by non-vanishing transport properties. This phenomenon,
66: known as parametric pumping, is observable in mezoscopic devices,
67: where during each pumping cycle a definite charge
68: % (number of electrons)
69: is transported through the system
70: \cite{pothier,pekola1,keller,pekola,niskanen,niskanen2}. Quantum pumps
71: are hoped to transform an AC signal of frequency $f$ into DC current
72: given by the relation $I=fQ$, where $Q=ne$ for normal devices, and
73: $Q=2ne$ for superconducting pumps ($n$ is the total charge transported
74: per cycle). If operated with perfect accuracy, quantum pumps could be
75: utilized to establish a standard of current. This is needed to close
76: the measurement triangle relating voltage, current and frequency by
77: fundamental constants \cite{zorin}. However, nonadiabatic
78: corrections, uncontrolled tunneling, and sensitivity of the devices to
79: parametric fluctuations smear the output signal. An intristic
80: stabilizing property, like quantization of magnetic flux, crucial for
81: the voltage standard, is absent in the parametric charge pumps. Thus
82: the only way to stabilize the performance is to optimize the design of
83: such devices. E.g. in Ref. \onlinecite{niskanen} enhanced control
84: over the tunelling amplitudes in the so-called Cooper pair ``sluice''
85: resulted in improved characteristics of the pumping scheme.
86:
87: In this Letter, we combine advantages of the controllable Josephson
88: couplings used in the ``sluice'' and the array-like design of earlier
89: proposals \cite{pekola,toppari}. Our scheme is analyzed in the
90: framework of adiabatic passage in an effective two-level system.
91: Electrostatic coupling of the Josephson junctions forming the arrays
92: results in increased separation of the operational subspace from the
93: excited states. The array design also suppresses the most severe
94: process affecting the pumping accuracy -- current reversal. This
95: suppression increases exponentially with the number of junctions in
96: the system. Since after each pumping sequence the system is to be
97: found in a known {\em charge} state, we also make use of the quantum
98: Zeno effect to diminish the population of unwanted states.
99:
100:
101: The core mechanism of the adiabatic passage can be understood using
102: the simplest nontrivial -- two-level system. Suppose that we start at
103: degeneracy (zero field in the spin-$1/2$ language) and spin pointing
104: ``up'' in the $z$ direction. Then the field is increased along the
105: spin direction and traverses a path in the $x-z$ plane which
106: approaches the degeneracy again from the negative $z$ direction. For
107: adiabatic pulses, the system remains in the ground state and follows
108: the field. However, at the end the spin points in the negative
109: $z$ direction. This simple example shows that although the system is
110: in a nondegenerate level during the passage and the traversed path in
111: the parameter space is closed the resulting transformation is highly
112: nontrivial as the spin direction is reversed. This is due to
113: the level crossing occurring at the beginning and end of the path.
114: For zero field the states are degenerate, and this freedom of defining
115: the orthogonal basis results in discontinuity of the energy
116: eigenstates.
117:
118: A Cooper-pair box with tunable Josephson coupling \cite{makhlin} [see
119: Fig.~\ref{fi:basic}(a)], described by the Hamiltonian $H= -1/2 B_z
120: \sigma _z - 1/2 B_x \sigma _x$, where $B_x \equiv J (\Phi ) = 2 J_0
121: \cos (\pi \Phi / \Phi_{0}) $, and $B_z = 4 E_C (1 - 2n_g)$, mimics a
122: spin-1/2 system. Here $J_0$ is the Josephson coupling of the junctions
123: in the SQUID (Superconducting Quantum Interference Device) coupling
124: the box to the reservoir, $\Phi $ and $\Phi_{0}$ the external applied
125: flux through the SQUID and flux quantum respectively, $E_C = e^2/ (2
126: C_g + 2 C_J)$ is the single-electron charging energy of the box and
127: $n_g = C_g V_g /2e$ is the dimensionless gate charge.
128: %(or dimensionless gate voltage tuning the electrostatic potential of
129: %the box).
130: The two-level approximation holds if the system is operated in
131: the charge regime ($J_0 \ll E_C$) and $n_g$ is close to $1/2$. Then
132: the only relevant states are characterized by zero or one Cooper pair
133: in the box, corresponding to spin parallel and antiparallel to the
134: $z$ direction respectively. Performing the procedure
135: described above on the system, we can thus transport a Cooper pair to
136: or from a reservoir into an empty box, depending on the sequence of
137: pulses.
138:
139: \begin{figure}[t] %% h=here, t=top,p=page of floats
140: \centerline{\resizebox{0.25\textheight}{!}{\rotatebox{0}
141: {\includegraphics{system0}\hspace{1cm}\includegraphics{system1}}}}
142: \caption{(a) Cooper pair box\cite{makhlin}. Electrostatic potential of
143: the superconducting island (black node) is controlled by the gate
144: voltage $V_g$. The Josephson tunelling amplitude $J$ is tuned by
145: external flux $\Phi $. (b) A Cooper-pair box with two SQUIDS.}
146: \label{fi:basic}
147: \end{figure}
148:
149: To induce transport through the Josephson-junction system we use two
150: SQUIDs which now serve as terminals [see Fig.\ref{fi:basic}(b)]. Using
151: one of the terminals to transport a Cooper pair into the box (stage
152: one) and another to transport the same pair to the reservoir (stage
153: two) we generate current $I = 2ef$. Exemplary pulses of a full cycle
154: are depicted in Fig.~\ref{fi:pulses}. No transport takes place during
155: parts I and IV with the system in a definite (zero) charge state, so
156: we can limit the cycle to the sequences II and III only. This
157: simplification minimizes the cycle duration and eliminates
158: degeneracies from the procedure (this system is identical to the
159: Cooper pair ``sluice'' \cite{niskanen}).
160:
161:
162:
163: Let us comment briefly on the possible sources of imperfections in the
164: pump performance. The nonadiabatic corrections leave the system in an
165: unknown superposition of charge states after the full cycles (instead
166: of a definite charge state). These corrections whilst usually small,
167: can accumulate over time. Fluctuations of external parameters affect
168: the scheme similarly, though fluctuations in external flux have
169: a much more severe effect than those in the gate voltage, especially
170: when the SQUIDs are not ideally symmetric. The terminals cannot then
171: be completely closed, due to residual Josephson coupling, and the
172: fluctuations induce transitions between charge states. The error
173: related to this feature is the hardest to eliminate -- since there is
174: nonzero tunneling to the reservoir via both terminals all the time.
175: Cooper pairs can accidentally be transported in the wrong direction,
176: leaving no trace of the error in cases when the the final charge state
177: is identical to the expected state. In the following we show that
178: the array design combined with Zeno projections strongly
179: suppresses these pumping errors.
180:
181:
182: \begin{figure}[t] %% h=here, t=top,p=page of floats
183: \centerline{\resizebox{0.3\textheight}{!}{\rotatebox{0}
184: {\includegraphics{simplepulses}}}}
185: \caption{A pulse sequence applied to the simple pump that induces flow
186: of one Cooper pair through the system. The exact shape of, and
187: area under the pulses do not influence the process as long as the
188: adiabaticity condition holds.}
189: \label{fi:pulses}
190: \end{figure}
191:
192:
193: The system we consider is shown in Fig.~\ref{fi:setup}(a). It is
194: composed of $N$ Cooper-pair boxes. Nearest neighbours are coupled with
195: dc-SQUIDs, leftmost and rightmost islands are coupled to the
196: reservoirs via other SQUIDs (terminals). If the number of Cooper
197: pairs on the islands is $\{n_1, n_2, \ldots, n_N\}$, the electrostatic
198: energy of the system $E_{n_1, n_2, \ldots, n_N} = \sum_i E_{c(i)}(n_i
199: - n_{g (i)})^2 + E_m^{i, i+1}(n_i - n_{g (i)})(n_{i+1} - n_{g
200: (i+1)})$. Here $E_{c(i)}$ are the charging energies of the islands
201: and $E_m^{i, i+1}$ is the energy of electrostatic coupling between two
202: neighboring islands, which is finite for finite capacitance of the
203: coupling SQUID. If the dimensionless gate charges, $n_{g(i)}= C_{g(i)}
204: V_{g(i)}/2e$ are close to $1/2$, the lowest energy states are
205: characterized by either zero or one Cooper pair on each island. With
206: this assumption we can again reduce the Hilbert space and map our
207: system to a finite anisotropic Heisenberg spin-$1/2$ chain in an
208: external magnetic field. The Hamiltonian
209: \begin{gather}
210: H = -{1\over 2} B_x^1 \sigma _x^1 - {1\over 2}B_x^N \sigma _x^N -
211: {1\over 2} \sum_{i=1}^{N} B_z^i \sigma _z^i
212: \\
213: + {1\over 2} \sum_{i=1}^{N-1} \left[\Delta_{i,i+1} \sigma _z^i
214: \sigma _z^{i+1} \right. \left. - J_{i, i+1} \left(\sigma _+^i
215: \sigma _-^{i+1} + \sigma _+^{i+1} \sigma _-^{i}\right) \right] \nonumber
216: \label{J42a}
217: \end{gather}
218: is characterized by constant electrostatic coupling amplitudes $\Delta
219: _{i,i+1}$ and tunable parameters -- $B_x^{1,N}$ - the Josephson
220: coupling of the leftmost and rightmost SQUIDs, $B_z^i$ - electrostatic
221: potential of the islands and $J_{i,i+1}$ - Josephson coupling between
222: neighbouring islands. In general the electrostatic coupling has a
223: finite range determined by the screening length $\lambda
224: =\sqrt{C_{J}/C_{g}}$ \cite{schoen}. As discussed below, we consider
225: the regime in which the electrostatic coupling is much weaker than the
226: on-site electrostatic energy. This implies that $\lambda \ll 1$ and we
227: can thus limit ourselves to nearest neighbour electrostatic coupling
228: only.
229:
230: \begin{figure}[t] %% h=here, t=top,p=page of floats
231: \centerline{\resizebox{0.25\textheight}{!}{\rotatebox{0}
232: {\includegraphics{system2}}}}
233: \centerline{\resizebox{0.28\textheight}{!}{\rotatebox{0}
234: {\includegraphics{pulses}}}}
235: \caption{(a) The stabilized Cooper-pair pump. It is operated similarly
236: to the single-island system -- identical pulse sequence is applied to
237: every second junction (island) (b). Single electron transistors (SET) serve
238: as the dephasing elements. They are biased with short voltage pulses
239: every time the pump should be in a definite charge state [lowest plot
240: in (b)].}
241: \label{fi:setup}
242: \end{figure}
243:
244:
245: Because of the electrostatic interaction, for $n_g^i$ close to $1/2$
246: the most favorable charge states are antiferromagnetic ($|010101\ldots
247: \rangle $ and $|101010\ldots \rangle $). The pumping procedure is a
248: simple generalization of the scheme described for one island and two
249: terminals. We start with one of the antiferromagnetic states and
250: transfer the charge of every island to the right by two sites to
251: achieve pumping. This can be implemented by applying identical pulse
252: sequence to every second island (and SQUID), {\it i.e.} $B_x^1 =
253: J_{2,3} = J_{4,5} = \ldots \equiv J_{eo}$, $J_{1,2} = J_{3,4} = \ldots
254: = B_x^N \equiv J_{oe}$, $B_z^1 = B_z^3 = \ldots \equiv B_z^o$, and
255: $B_z^2 = B_z^4 = \ldots \equiv B_z^e$ [see Fig.~\ref{fi:setup}(b)]
256: ($o,e$ denote odd and even islands respectively). Here we focus on
257: systems with {\em odd} number of islands -- then the number of
258: coupling SQUIDs is even, and exactly half of them are active during
259: each step of the procedure. This symmetry simplifies our analysis but
260: is not crucial.
261:
262: The procedure, even for the single-island realization, takes place in
263: nondegenerate ground state. Due to the avoided degeneracy, spontaneous
264: transitions to unwanted levels cost energy and are thus suppressed.
265: The chain design additionally provides the following advantage: on the
266: one hand an energy gap between the antiferromagnetic states and the
267: remaining subspace $\Delta E \geq \min \{\Delta _{i,i+1}\}$, while on
268: the other hand accidental transitions between the antiferromagnetic
269: states need simultaneous transport through $(N+1)/2$ junctions --
270: another effect which diminishes exponentially with $N$ (see
271: Fig.~\ref{fi:results}(b) and explanation below). The most evident
272: advantage of this approach is, however, elimination of transport in
273: the wrong direction. Let us analyze the mechanism ruling this process.
274: Suppose that we start with the state $|101010 \ldots \rangle $. During
275: the first part of the cycle we want to shift the charge configuration
276: by one site to the right, {\it i.e.} to arrive at $|010101\ldots
277: \rangle $. As mentioned before, the residual Josephson coupling of
278: closed SQUIDs leads to two ways of arriving at the same state. One
279: could anticipate interference effects here, but since the residual
280: coupling is a fluctuating quantity (controlled by fluctuating magnetic
281: field) we treat the process as incoherent. A simple analysis is based
282: on the assumption that a Cooper pair is transported through an open
283: (correct) SQUID with probability $p$, and through closed (wrong) SQUID
284: with the probability $q = 1-p$. The measure of this pumping error for
285: each island might be the ratio $q/p$. For a chain of $N$ junctions the
286: probability of correct shift becomes $\propto p^{N/2}$, and of
287: incorrect shift $\propto q^{N/2}$. Hence the pumping error
288: $(q/p)^{N/2}$ decreases {\em exponentially} with the number of
289: junctions in the chain. More rigorous analysis requires calculation of
290: the energy difference between the two lowest energy levels for $B_z^i
291: = 0$. Indeed, in this case the lowest energy eigenstates are symmetric
292: and antisymmetric superpositions of the antiferromagnetic states (with
293: some addition of other charge states), and the energy difference,
294: being the rate of phase progression, in the charge basis is equivalent
295: to the rate of transition. Fig.~\ref{fi:results}(a) shows the results
296: for various values of the maximum Josephson coupling of the SQUIDs.
297: $T$ is the rate calculated for the desired transitions only (assuming
298: no residual coupling), $t$ is calculated for the wrong path. Then,
299: the time of single operations should be much larger than $1/T$ for the
300: adiabatic condition to be satisfied. The numerical calculation is
301: performed for a realistically achievable residual coupling equal to
302: $1\%$ of the maximum Josephson coupling of the SQUIDs. The
303: values of the electrostatic coupling amplitudes, and maximum Josephson
304: couplings are for simplicity assumed identical for each
305: junction (this is, again, not crucial and our observations
306: are valid also for different settings). The pumping error vs. $N$ is
307: plotted in Fig.~\ref{fi:results}(a).
308: \begin{figure}[t] %% h=here, t=top,p=page of floats
309: \centerline{\resizebox{0.28\textheight}{!}{\rotatebox{0}
310: {\includegraphics{toverT}}}}
311: \centerline{\resizebox{0.28\textheight}{!}{\rotatebox{0}
312: {\includegraphics{Gaps}}}}
313: \caption{(a) Pumping error $t/T$ as a function of $N$. For
314: logarithmic vertical axis the dependence is linear, implying
315: exponential suppression of the error with increasing $N$. (b)
316: Transition rates between the charge states as a function of
317: $N$.}
318: \label{fi:results}
319: \end{figure}
320: The dependence is clearly exponential, and very weakly depends on the
321: value of the maximum Josephson coupling. However, the value of $T$
322: itself should not decrease too strongly with increasing $N$ as it
323: determines the adiabatic rates. From Fig.~\ref{fi:results}(b) we can
324: see that the ratio $J_{max}/ \Delta $ should be as close to $1$ as
325: possible to avoid the effects of exponential suppression. For this
326: case, $T/ \Delta \approx 1$, where $\Delta $ can be chosen large thus
327: yielding a very comfortable adiabatic condition. Note that even for
328: $J_{max}/ \Delta \approx 1$ we can still work in the reduced Hilbert
329: space, as long as the charging energies of individual islands are much
330: larger than $J_{max}$ (this means that the system should be operated
331: in the charging regime with weak electrostatic coupling). The
332: exponential suppression of the energy gap between the lowest states
333: implies that unwanted transitions between the antiferromagnetic states
334: also diminish very rapidly with increasing $N$ (note that this
335: suppression for very small, erroneous, values of Josephson coupling is
336: much stronger than for moderate values, expected for open terminals).
337: This of course does not mean that the optimal setup should be
338: represented by as long an array as possible. Most probably there is
339: some moderate number of junctions for which the pumping errors are
340: strongly suppressed, and the decoherence effects typical for
341: macroscopic system are yet not severe. However, examination of the
342: long-chain limit goes beyond our analysis and would be appropriate for
343: a separate publication.
344:
345:
346:
347: Although during the passage the state of the system is a superposition
348: of numerous charge configurations, when all Josephson couplings are
349: switched off (or rather intended to be) the pump should be in one of
350: the antiferromagnetic states. All other components (nonadiabatic
351: corrections, residual Josephson coupling effects) contribute to
352: pumping errors. The optimal solution would be to project the state
353: onto a charge state by performing charge measurement. Indeed, if
354: the measured state is very close to an eigenstate of the measured
355: observable, the measurement takes the state closer to the
356: eigenstate. If frequently repeated, the measurement can diminish the
357: probability of transition to other eigenstates arbitrarily. This
358: phenomenon, known as quantum Zeno effect \cite{misra}, can be utilized
359: also in our system. The charge measurement can be performed by a
360: single-electron transistor (SET) coupled capacitively to the
361: superconducting island \cite{makhlin2} [see
362: Fig.~\ref{fi:setup}(a)]. This measurement is characterized by
363: few time scales, the dephasing time $\tau _{\phi }$, during which the
364: off-diagonal density matrix elements (in the basis of the
365: measured observable) vanish, the measurement time, $\tau _{meas}$,
366: after which the information about the charge can be read out, and the
367: mixing time, after which the back action of the SET on the island
368: destroys the information about the state. The parameters can be
369: selected in such way that $\tau _{mix} \geq \tau _{meas} \geq \tau
370: _{\phi }$. Since the actual information about the state is not
371: important, and only the dephasing mechanism modifies the state, the
372: quantum Zeno projection can be realized within a relatively short
373: time. Moreover, even if the time available is too short for the
374: off-diagonal entries of the density matrix to vanish, partial
375: dephasing still brings the system closer to the desired charge
376: state. The system can be thus dephased each time the system is {\em
377: supposed to be} in a definite charge state by switching on the bias
378: voltage of the SETs for a very short time interval.
379:
380: In summary, the presented scheme provides an accurate pumping
381: procedure. The array design exponentially suppresses the most severe,
382: untracable error of reversed current flow and diminishes errors due to
383: undesired charge configurations. These errors can be further minimized
384: by repeated dephasing with single-electron transistors.
385:
386:
387:
388: The authors thank J.~Siewert, O.~Buisson, and S.~Paraoanu for
389: discussions. M.C. acknowledges support from Polish Ministry Grant No.
390: 1 P03B 067 30.
391:
392:
393:
394:
395: \begin{thebibliography}{13}
396:
397:
398:
399: \bibitem{pothier} H. Pothier, P. Lafarge, C. Urbina, D. Esteve, and
400: M.H. Devoret, Europhys. Lett. 17, 249 (1992).
401:
402: \bibitem{pekola1} J.P. Pekola, A.B. Zorin, and M.A. Paalanen,
403: Phys. Rev. B 50, 11255 (1994).
404:
405: \bibitem{keller} M.W. Keller, J.M. Martinis, N.N. Zimmerman,
406: A.H. Steinbach, Appl. Phys. Lett. 69, 1804 (1996).
407:
408: \bibitem{pekola} J. P. Pekola, J. J. Toppari, M. Aunola,
409: M. T. Savolainen, and D. V. Averin, Phys. Rev. B 60, R9931 (1999).
410:
411: \bibitem{niskanen} A. O. Niskanen, J. P. Pekola, and H. Sepp\"a, Phys. Rev.
412: Lett. 91, 177003 (2003).
413:
414: \bibitem{niskanen2} A. O. Niskanen, J. M. Kivioja, H. Sepp\"a, and J. P. Pekola
415: Phys. Rev. B 71, 012513 (2005)
416:
417: \bibitem{zorin} K. K. Likharev, and A. B. Zorin, J. Low
418: Temp. Phys. {\bf 59}, 347 (1985).
419:
420: \bibitem{toppari}J. J. Toppari, J. M. Kivioja, J. P. Pekola,
421: M. T. Savolainen, J. Low Temp. Phys. 136, 57 - 91 (2004)
422:
423: \bibitem{makhlin} Yu.~Makhlin, G.~Sch\"on, and A.~Shnirman, Nature
424: {\bf 398}, 305 (1999).
425:
426: \bibitem{schoen} R. Fazio, K-H. Wagenblast, C. Winkelholz,
427: G. Sch\"{o}n, Phys. B {\bf 222}, 364 (1996).
428:
429: \bibitem{misra} B.~Misra and E.~C.~G.~Sudarshan, J. Math. Phys. 18,
430: 756 (1977).
431:
432: \bibitem{makhlin2} Yu. Makhlin, G. Sch\"on, and A. Shnirman,
433: Rev. Mod. Phys. 73, 357 (2001)
434:
435: \bibitem[Berry (1984), M.~V.~Berry]{berry} M.~V.~Berry,
436: Proc. R. Soc. London, Ser. A {\bf 392}, 45 (1984).
437: \end{thebibliography}
438:
439:
440:
441:
442: \end{document}
443:
444:
445:
446:
447: