1: \documentclass[aps,prb,twocolumn,amsmath,amssymb,floatfix,superscriptaddress]{revtex4}
2:
3: \usepackage{graphicx}
4: \usepackage{epsfig}
5: \usepackage{dcolumn}
6:
7: \newcommand{\I}{\mathrm{i}}
8: \newcommand{\e}{\mathrm{e}}
9: \renewcommand{\d}{\mathrm{d}}
10: \newcommand{\T}{\mathrm{T}}
11: \newcommand{\C}{\mathrm{C}}
12: \newcommand{\leads}{\mathrm{leads}}
13: \renewcommand{\k}{\mathbf{k}}
14: \newcommand{\hc}{\text{h.c.}}
15:
16: \begin{document}
17:
18: \title{ Spin qubits with electrically gated polyoxometalate molecules
19: }
20:
21:
22: \author{J\"org Lehmann}
23: \affiliation{Department
24: of Physics und Astronomy,
25: University of Basel,
26: Klingelbergstrasse~82,
27: CH-4056~Basel, Switzerland}
28: \author{Alejandro Gaita-Ari{\~n}o}
29: \affiliation{Department
30: of Physics und Astronomy,
31: University of Basel,
32: Klingelbergstrasse~82,
33: CH-4056~Basel, Switzerland}
34: \affiliation{
35: Instituto de Ciencia Molecular,
36: Universitat de Valencia,
37: Pol\'{i}gono de La Coma, s/n,
38: E-46980~Paterna, Spain}
39: \author{Eugenio Coronado}
40: \affiliation{
41: Instituto de Ciencia Molecular,
42: Universitat de Valencia,
43: Pol\'{i}gono de La Coma, s/n,
44: E-46980~Paterna, Spain}
45: \author{Daniel Loss}
46: \affiliation{Department
47: of Physics und Astronomy,
48: University of Basel,
49: Klingelbergstrasse~82,
50: CH-4056~Basel, Switzerland}
51:
52: \date{\today}
53:
54: \begin{abstract} Spin qubits offer one of the most promising routes to
55: the implementation of quantum computers. Very recent results in
56: semiconductor quantum dots show that electrically-controlled gating
57: schemes are particularly well-suited for the realization of a universal
58: set of quantum logical gates. Scalability to a larger number of qubits,
59: however, remains an issue for such semiconductor quantum dots. In
60: contrast, a chemical bottom-up approach allows one to produce identical
61: units in which localized spins represent the qubits. Molecular
62: magnetism has produced a wide range of systems with tailored properties,
63: but molecules permitting electrical gating have been lacking. Here we
64: propose to use the polyoxometalate [PMo$_{12}$O$_{40}$(VO)$_2$]$^{q-}$,
65: where two localized spins-$\mathbf{1/2}$ can be coupled through the
66: electrons of the central core. Via electrical manipulation of the
67: molecular redox potential, the charge of the core can be changed. With
68: this setup, two-qubit gates and qubit readout can be implemented.
69: \end{abstract}
70:
71: \pacs{}
72:
73: \maketitle
74:
75: \section{Introduction}
76:
77: Quantum dots have often been termed artificial atoms and
78: molecules.\cite{vanderWiel2003a} Indeed, their tunability via
79: electrical gates has permitted one to reach into a regime where only a
80: single electron sits in each dot. A double dot then becomes the
81: analogue of a hydrogen molecule. The Heisenberg exchange coupling in
82: such a system can be manipulated by appropriately chosen electrical
83: pulse sequences, up to a point which has not yet been achieved for its
84: real counterpart.\cite{Burkard1999a} Recently, it has been
85: demonstrated that this technique allows the realization of the
86: fundamental one- and two-qubit quantum
87: gates~\cite{Barenco1995a,Loss1998a,Nielsen2000a} in such a
88: system.\cite{Petta2005a,Koppens2006a}
89:
90: Naturally, the question arises whether the successful schemes and
91: techniques of an electrical control of spins in quantum dots can be
92: transfered back to electron spins in single molecules. In view of the
93: recent experimental progress in the field of molecular electronics,
94: which demonstrates that magnetic transitions in atomic
95: chains~\cite{Hirjibehedin2006a} and single
96: molecules~\cite{Heersche2006a, Jo2006a} can be resolved in transport
97: measurements, such a goal should in principle be achievable. In
98: particular, the field of molecular
99: magnetism~\cite{Kahn1993a,Gatteschi2006a} has provided a plethora of
100: systems of almost arbitrary magnetic
101: functionality.\cite{Wernsdorfer1999a,Coronado2000a,Real2003a,Stepanow2004a,
102: ChemRev} Very recently, phase-coherence times of up to $3\,\mu s$
103: have been reported for such molecular nanomagnets after
104: deuteration.\cite{Ardavan2007a} Yet, a suitable system for an
105: electrical control of molecular qubits has still been missing. It is
106: this electrical control which is crucial for scalability and which
107: distinguishes this proposal from earlier ones also based on molecular
108: magnets.~\cite{Leuenberger2001a,Troiani2005a}
109:
110: Here we propose an experimental setup which permits the electrical
111: switching of the exchange interactions between two electron spins. The
112: system we choose for illustrating our proposal is
113: [PMo$_{12}$O$_{40}$(VO)$_2$]$^{q-}$,~\cite{Chen1996} a
114: polyoxometalate~\cite{ChemRev} which consists of a central
115: mixed-valence core based on the [PMo$_{12}$O$_{40}$]
116: Keggin~\cite{Keggin1933a} unit, able to act as an electron reservoir
117: accommodating a variable number of delocalized electrons hopping over
118: the Mo centers, capped by two vanadyl groups containing two localized
119: spins (cf.\ Fig.~\ref{fig:model})
120: %
121: \begin{figure}[ht]
122: \centering
123: (a) \raisebox{-4.5cm}{\includegraphics[width=0.4\linewidth]{fig1a}}
124: (b) \raisebox{-3.7cm}{\includegraphics[width=0.45\linewidth]{fig1b}}
125: \caption{(a) Schematic drawing of the polyoxometalate
126: [PMo$_{12}$O$_{40}$(VO)$_2$]$^{q-}$ separated by a tunnelling
127: barrier from a metallic substrate and contacted via a tunnel
128: coupling~$\Gamma$ to a tip at a potential $V_\mathrm{tip}$.
129: Indicated are the two localized spins $\mathbf{S}_\mathrm{L}$ and
130: $\mathbf{S}_\mathrm{R}$ of the V atoms in the center of the red
131: square pyramids. Depending on the redox state of the molecule, the
132: delocalized valence electrons of the Mo atoms in the center of the
133: blue octahedra form a spin $\mathbf{S}_\C$ or pair to a singlet.
134: The O anions are located at the vertices of the polyhedra. (b)
135: Ball-and-stick model of the polyoxometalate: O (grey), Mo (blue),
136: V (red), and P (yellow).}
137: \label{fig:model}
138: \end{figure}
139: %
140: The spins on these two (VO)$^{2+}$ units are weakly magnetically
141: coupled via the delocalized electrons of the central core. We shall
142: show how this magnetic coupling can be switched in an all-electrical
143: way and how this can be used for the implementation of a fundamental
144: two-qubit gate (providing entanglement), the so-called
145: square-root-of-swap $\sqrt{\text{\sc SWAP}}$.~\cite{Loss1998a}
146: Furthermore, we will detail how to use a variation of this procedure
147: for the readout of the final state of the two qubits.
148:
149: \section{Low-energy model of the polyoxometalate}
150:
151: For the description of the low-energy states of the polyoxometalate,
152: two cases have to be distinguished: For an \textit{even} number of
153: electrons on the mixed-valence Keggin core, their spins pair
154: antiferromagnetically to form a total spin~$0$ state. Then the system
155: can be modelled by the two spins $1/2$ on the vanadyl groups weakly
156: coupled via an indirect exchange mechanism mediated by the core
157: electrons. On the other hand, if the number of core electrons is
158: \textit{odd}, an unpaired spin $1/2$ on the core remains and one
159: obtains a set of three coupled spins $1/2$. Restricting ourselves to
160: two charge states differing by one electron, we can choose the
161: electron number of the even state as reference and write the
162: Hamiltonian in the form
163: \begin{equation}
164: \label{eq:H}
165: \begin{split}
166: H = {} & - J(n_\C)\,
167: \mathbf{S}_\mathrm{L}\cdot \mathbf{S}_\mathrm{R} - J_\C
168: (\mathbf{S}_\mathrm{L} + \mathbf{S}_\mathrm{R}) \cdot
169: \mathbf{s}_\C
170: \\ &
171: + (\epsilon_0 - e V_\mathrm{g})\, n_\C
172: + U n_\C(n_\C-1)/2
173: \,.
174: \end{split}
175: \end{equation}
176: Here, the first term describes the indirect-exchange coupling between
177: the left and right spins $\mathbf{S}_\mathrm{L}$ and
178: $\mathbf{S}_\mathrm{R}$ with a coupling constant~$J(n_\C)$ which
179: depends on the number of electrons $n_\C$---measured with respect to
180: the reference number---on the central core of the molecule. In the
181: case of an electron being localized on the core, its
182: spin~$\mathbf{s}_\C = (1/2)\sum_{\sigma\sigma'} d^\dagger_{\C\sigma}
183: \,\boldsymbol{\tau}_{\sigma\sigma'}\, d_{\C\sigma'}$ couples to the
184: left and right spins with coupling constant~$J_\C$. Here, the
185: operators $d_{\C,\sigma}$ ($d^\dagger_{\C,\sigma}$) destroy (create)
186: and electron on the central core and $\boldsymbol\tau$ is the vector
187: of the three Pauli matrices. The two last terms contain the orbital
188: energy~$\epsilon_0$ of the electron on the central core, which can be
189: shifted due to a gate potential~$V_\mathrm{g}$, and the molecule's
190: charging energy~$U$, respectively. The latter is assumed to be the
191: largest energy scale of the problem. We consider a situation where the
192: central core of the molecule is tunnel-coupled to one or more metallic
193: leads~$\ell$ permitting electrons to flow on and off the molecule with
194: tunnelling rates $\Gamma_\ell$. In the STM setup depicted in
195: Fig.~\ref{fig:model}(a), two leads, the tip and the metallic surface,
196: are present. We assume that an insulating layer---which still allows
197: electrons to tunnel through---between the surface
198: and the molecule leads to a molecule-surface coupling which is much
199: smaller than the typical energy scale~$J_\C$ of the molecular
200: Hamiltonian~\eqref{eq:H}. An applied bias
201: voltage~$V_\mathrm{b}=V_\mathrm{tip}$ then leads to a shift of the
202: molecular levels which to a very good approximation is described by
203: the linear relation $V_\mathrm{g}= \eta V_\mathrm{tip}$ with
204: $0<\eta<1$.~\cite{Datta1997a} Note that in a more sophisticated
205: three-terminal setup, the parameters $V_\mathrm{g}$ and $V_\mathrm{b}$
206: can be controlled independently.~\cite{Park1999a,Kubatkin2003a}
207:
208: \section{Ideal quantum-gate operation}
209:
210: The $\sqrt{\mathrm{SWAP}}$ operation is defined by
211: $|\chi,\lambda\rangle \rightarrow \left( | \chi,\lambda\rangle +
212: \mathrm{i}|\lambda,\chi\rangle\right)/(1+\mathrm{i})$, where
213: $\chi,\lambda = \uparrow, \downarrow$. For electrically confined
214: quantum dots, it can be physically implemented by an appropriately
215: chosen gate-voltage pulse which turns on the Heisenberg exchange coupling
216: between the qubits for a specific time (see below).\cite{Loss1998a}
217: In a molecular system, however, the exchange constants are fixed by
218: the chemical structure and cannot be directly changed. Thus, a more
219: sophisticated quantum-gate scheme has to be employed. We propose to
220: indirectly manipulate the coupling between the two qubits by changing
221: the occupation~$n_\C$ of the central core in the molecule described
222: above. This change in $n_\C$ can be induced by changing the gate
223: voltage $V_\mathrm{g}$, such that different charge sectors of the
224: molecule become stable (cf.\ Fig.~\ref{fig:gating_sequence}).
225: Chemically this amounts to a change in the redox potential of the
226: molecule. We will first discuss the basic idea of the quantum-gate scheme,
227: thereby assuming that the electron number~$n_\C$ can be changed in a
228: deterministic, externally controllable way. Later on, this assumption
229: will be relaxed and the full tunnelling dynamics due to the change in
230: $V_\mathrm{g}$ will be included.
231: %
232: \begin{figure}[ht]
233: \includegraphics[width=0.9\linewidth]{fig2}
234: \caption{Quantum-gate sequence for the
235: $\sqrt{\mathrm{SWAP}}$ operation and exchange coupling constants
236: during the corresponding gate phase. The gate is turned on for a
237: time $\tau_\mathrm{gate}.$}
238: \label{fig:gating_sequence}
239: \end{figure}
240:
241: Initially, one adjusts the gate voltage~$V_\mathrm{g}$ in such a way
242: that the stable configuration is given by an even number of electrons
243: on the central core ($n_\mathrm{C}=0$). We assume that the exchange
244: coupling $J_0=J(0)$ between the spins $\mathbf{S}_\mathrm{L}$ and
245: $\mathbf{S}_\mathrm{R}$ will then be very small and can be disregarded
246: for times much smaller than $\hbar/J_0$ (see also below). The quantum-gate
247: operation begins by changing the gate voltage such that the
248: $n_\mathrm{C}=1$ charge sector becomes stable. Then, on a time-scale
249: of the inverse tunnelling rate, an electron enters the central core.
250: After this has happened, the dynamics of the three-qubit system is
251: governed by the Hamiltonian~\eqref{eq:H} in the $n_\C=1$ subspace. The
252: spin-dependent part of this Hamiltonian becomes
253: \begin{equation}
254: \label{eq:H1_t}
255: H_1 = - (J_1-J_\C)\,
256: \mathbf{S}_\mathrm{L} \cdot \mathbf{S}_\mathrm{R}
257: -\frac{J_\C}{2} \,\mathbf{S}^2
258: \,,
259: \end{equation}
260: where
261: $\mathbf{S}=\mathbf{S}_\mathrm{L}+\mathbf{S}_\mathrm{R}+\mathbf{s}_\C$
262: is the total spin of the system and $J_1=J(1)$. Before considering the
263: time-evolution due to this Hamiltonian, we conclude our discussion of
264: the gate cycle: after a specific time~$\tau_\mathrm{gate}$, one
265: switches back to the initial gate voltage and the excess electron
266: tunnels off the central core again. After this has happened, the two
267: outer spins are, up to times much smaller than $\hbar/J_0$, decoupled
268: again.
269:
270: The only non-trivial dynamics is induced by the
271: Hamiltonian~\eqref{eq:H1_t}. The first term contains an effective
272: exchange coupling between the two spins. Up to an irrelevant phase
273: factor, its corresponding time-evolution yields the
274: $\sqrt{\mathrm{SWAP}}$ gate if the gate time $\tau_\mathrm{gate}$
275: fulfils~\cite{Loss1998a}
276: \begin{equation}
277: (J_1-J_\C) \, \frac{\tau_\mathrm{gate}}{\hbar} = \frac{\pi}{2} +
278: 2\pi \, n\,,
279: \label{eq:tswap}
280: \end{equation}
281: with $n$ being an arbitrary integer. The second term in
282: equation~\eqref{eq:H1_t} depends on the total spin of the three-spin-$1/2$
283: system. While this quantity is unknown, we can eliminate its influence
284: on the gate behavior if we restrict ourselves to stroboscopic gate
285: times $\tau_\mathrm{gate}$ for which the contribution
286: $\exp\left[\mathrm{i} (J_\C/2)\mathbf{S}^2 \tau_\mathrm{gate}/\hbar)\right]
287: $
288: to the time-evolution operator is trivial.
289: Evaluating the effect of the operator $\mathbf{S}^2$ in the eigenbasis
290: (see also equation~\eqref{eq:H1_d} below), this is the case for gate times given by
291: \begin{equation}
292: \tau_\mathrm{gate} = \frac{4\pi}{3} \frac{\hbar}{|J_\C|} \, m\,,
293: \label{eq:taun}
294: \end{equation}
295: with the arbitrary integer $m>0$. It is important to note that
296: relation~\eqref{eq:taun} is independent of the spin direction of the
297: additional electron. Equating conditions~\eqref{eq:taun} and
298: \eqref{eq:tswap}, we obtain the following relation between $J_1$ and
299: $J_\C$:
300: \begin{equation}
301: \label{eq:gatingcondition}
302: \frac{J_1}{|J_\C|} = \mathrm{sgn} J_\C +
303: \frac{3}{8} \frac{1-4n}{m}\,.
304: \end{equation}
305:
306: \section{Realistic quantum-gate behavior}
307:
308: In the preceding discussion we have assumed that the tunnelling
309: process of the electron is highly controlled, i.e., that we are
310: able to instantaneously switch between the $n_\C=0$ and $n_\C=1$
311: states. In reality, tunnelling is a probabilistic event occurring on a
312: mean time scale of the order of the inverse tunnelling rate. In order
313: to investigate to what extent this stochasticity affects our
314: analytical findings, we compare them with numerical results obtained
315: from a simulation of the incoherent quantum dynamics using the average
316: gate fidelity~$\mathcal{F}$ as figure of merit for the gate process
317: (see App.~\ref{sec:br} ). In Fig.~\ref{fig:fidelity}, we show this
318: quantity as a function of the gate time and the ratio $J_1/|J_\C|$ for
319: the case of a ferromagnetic coupling $J_\C>0$ (the antiferromagnetic
320: case behaves very similarly).
321: %
322: \begin{figure}[ht]
323: \centering
324: \includegraphics[width=0.9\linewidth]{fig3_small}
325: \caption{Average gate fidelity~$\mathcal{F}$ as a function of the
326: gate time $\tau_\mathrm{gate}$ and the ratio $J_1/|J_\C|$ for a
327: ferromagnetic $J_\C>0$. The conditions~\eqref{eq:tswap} and
328: \eqref{eq:taun} are indicated by solid lines. The parameters are:
329: $\hbar\Gamma = 5 |J_\C|$, $k_\mathrm{B}T = 0.001|J_\C|$, and
330: $V_\mathrm{g,on}=-V_\mathrm{g,off}=15|J_\C|$ (measured from the
331: charge-degeneracy point).}
332: \label{fig:fidelity}
333: \end{figure}
334: We focus on the strong-coupling regime $\hbar\Gamma\gg |J_\C|$, where
335: tunnelling proceeds rapidly compared to the internal coherent
336: dynamics. As expected from our analytical considerations, the average
337: gate fidelity assumes maxima whenever both conditions~\eqref{eq:tswap}
338: and \eqref{eq:taun} are fulfilled. We found the gate error
339: $1-\mathcal{F}$ to decrease with increasing tunnelling rate $\Gamma$. For
340: $\Gamma = 6 |J_\C|/\hbar$, for instance, we obtain a fidelity
341: $\mathcal{F}=0.99$ at the maximum indicated by the circle in
342: Fig.~\ref{fig:fidelity}.
343:
344: \section{Readout process}
345:
346: We now show that in a molecular system described by the
347: Hamiltonian~\eqref{eq:H}, it is possible to readout the quantum number
348: $S_0$ of the two outer spins by measuring the sequential tunnelling
349: current through the central dot. The current can be calculated using
350: the Bloch-Redfield equation approach (see App.~\ref{sec:br}). For a
351: qualitative understanding it is sufficient to consider the allowed
352: transitions, i.e., non-zero matrix
353: elements~$\langle\alpha'|d^\dagger_{\C,\sigma}|\alpha\rangle$, between
354: the eigenstates of the two different charge sectors $n_\C=0$ and
355: $n_\C=1$. The former are the eigenstates $|S_0, S_{0,z}\rangle$ of
356: the total spin
357: $\mathbf{S}_0=\mathbf{S}_\mathrm{L}+\mathbf{S}_\mathrm{R}$ of the two
358: outer spins. The Hamiltonian~\eqref{eq:H1_t} of the molecule with an
359: additional electron can be readily diagonalized by rewriting it up to
360: an irrelevant constant as
361: \begin{equation}
362: \label{eq:H1_d}
363: H_1 = -\frac{J_\C}{2} \,\mathbf{S}^2
364: - \frac{J_1-J_\C}{2}\,
365: \mathbf{S}_0^2
366: \,.
367: \end{equation}
368: Thus, in the basis of the simultaneous eigenstates $|S, S_0,
369: S_z\rangle$ of $\mathbf{S}^2$, $\mathbf{S}_0^2$ and $S_z$, the
370: Hamiltonian in the $n_\C=1$ subsector is already diagonal. For the
371: matrix elements~$\langle
372: S',S'_0,S'_z|d^\dagger_{\C,\sigma}|S_0,S_{0,z}\rangle$, we then obtain the
373: selection rules $S'=S_0\pm 1/2$, $S'_0 = S_0$, $S'_z = S_{0,z}+\sigma$. In
374: particular, we note that the quantum number $S_0$ stays invariant
375: under a complete sequential-tunnelling cycle. Since furthermore the
376: energy differences determining the tunnelling rates and thus the
377: current depend on the quantum number $S_0$, we find that the
378: (quasi-)stationary value of the current can be strongly dependent on
379: the initial preparation of $S_0$.
380:
381: This situation is depicted in Fig.~\ref{fig:readout}, where the
382: conductance peaks as a function of gate and bias voltage are shown
383: schematically for two different initial values of $S_0$.
384: \begin{figure}[ht]
385: \centering
386: \includegraphics[width=0.9\linewidth]{fig4}
387: \caption{Sketch of ground and excited state tunnel
388: transitions for $S_0=0$ (dashed line) and $S_0=1$ (solid line).}
389: \label{fig:readout}
390: \end{figure}
391: From the figure we see that by starting initially at zero
392: bias~$V_\mathrm{b}=0$ in the $n_\C=0$ state, we first reach, upon
393: increasing~$V_\mathrm{b}$, the allowed ground-state transition and
394: thus the onset of the sequential tunnelling current for the triplet
395: $S_0=1$. On the other hand, for a singlet preparation~$S_0=0$,
396: sequential tunnelling will only be possible at a voltage which is
397: higher by $(J_\C+2 J_1)/e$. At low enough tunnelling rate and
398: temperature, i.e., $\hbar\Gamma, k_\mathrm{B}T \ll J_\mathrm{C} + 2
399: J_1$, it is possible to distinguish these two cases and hence to
400: measure the quantum number $S_0$. As discussed before,
401: for an independent control of the two parameters $V_\mathrm{g}$ and
402: $V_\mathrm{b}$ in Fig.~\ref{fig:readout}, a three-terminal geometry is
403: required. In the STM setup discussed here, one will only move along a line in
404: Fig.~\ref{fig:readout}, which, however, is sufficient for the readout
405: scheme, if the molecule is in the $n_\mathrm{C}=0$ charge sector at
406: $V_\mathrm{b}=0$.
407:
408: Note that for an initial superposition with singlet and triplet
409: contributions, the readout process just described represents a
410: projective measurement. The first electron attempting to tunnel
411: determines the final outcome, i.e., the long-time current. The
412: average over many repeated measurements is described by the solution
413: of the master equation~\eqref{eq:master}.
414:
415: We remark that the singlet-triplet readout process requires a
416: substantially smaller tunnel coupling than the one for the
417: exchange-controlled quantum-gate operation. There are several ways how
418: to achieve this: Obviously, one could increase the molecule-tip
419: distance which leads to an exponentially strong change in the
420: tunnelling rate. Still, doing so on a very short time-scale might be
421: technically challenging. Another possibility would be to use that the
422: tunnel barrier---and, thus, in turn the coupling strength---depends on
423: both the gate- and the bias-voltage.~\cite{Wolf1989a} Depending on the
424: details of the contact, this might be enough to achieve the required
425: change in the tunnel coupling. Alternatively, one could transfer the
426: readout scheme to the so-called cotunnelling
427: regime,~\cite{Hirjibehedin2006a,Lehmann2006a} where transport proceeds
428: via virtual tunnelling on and off the molecule. Then, the onset of the
429: inelastic cotunnelling current is solely temperature-smeared, even in
430: the presence of a strong molecule-lead coupling. For measuring $S_0$
431: one would then need to apply a magnetic field, which will lead to a
432: splitting of the triplet state $S_0=1$ only. Thus, one will find only
433: for this state an inelastic cotunnelling step in the
434: conductance-voltage characteristics at the bias voltage corresponding
435: to the Zeeman energy.
436:
437: \section{Ab-initio modelling and implementation requirements}
438:
439: While the preceding discussion applies to any system described by a
440: Hamiltonian of the form~\eqref{eq:H}, we now return to the specific
441: case of the molecule [PMo$_{12}$O$_{40}$(VO)$_2$]$^{(q-)}$. A high
442: redox flexibility is a characteristical feature of polyoxometalates,
443: where cyclic voltammetry experiments show one- or two-electron
444: reversible redox peaks, depending on the system and on the conditions.
445: In particular, four compounds based on the XMo$_{12}$O$_{40}$ (X = Si,
446: Ge, P) Keggin structure (either alone or vanadyl-bicapped) have been
447: recently found to show reversible two-electron electrochemical
448: processes.~\cite{Shi2006}
449: %
450: We thus extract the exchange coupling strengths for
451: different charge states~$N$ of the molecule (see App.~\ref{sec:par}).
452: %
453: Identifying even and odd values of $N$ with $n_\C=0$ and $n_\C=1$,
454: respectively, we obtain the parameters in the Hamiltonian~\eqref{eq:H}
455: as given in Table~\ref{par} for $0\le N\le6$; for higher electronic
456: populations $J_0$, $J_1$ and $J_C$ are of comparable size. {
457: \squeezetable
458: \begin{table}[h]
459: \begin{ruledtabular}
460: \begin{tabular}{cc}
461:
462: \begin{tabular}{c|d|c}
463: $N$ & \multicolumn{1}{c|}{$J_0[\mathrm{meV}]$} & $E_\mathrm{gap}[\mathrm{meV}]$ \\
464: \hline
465: 0 & 0 & $>1000$ \\
466: 2 & 0.01 & $>100$ \\
467: 4 & 0.01 & $>100$ \\
468: 6 & 1.0 & $>100$ \\
469: \end{tabular}
470: \hspace{0.0cm}
471: &
472: \begin{tabular}{c|d|d|d|c}
473: $N$ & \multicolumn{1}{c|}{$J_\C[\mathrm{meV}]$} &
474: \multicolumn{1}{c|}{$\displaystyle\frac{J_1}{J_\C}$} &
475: \multicolumn{1}{c|}{$\alpha$}
476: & $\displaystyle\frac{E_\mathrm{gap}}{J_\C}$ \\
477: \hline
478: 1 & 1 & 0.01 & <0.1 & $>100$ \\
479: 3 & 1 & 0.1 & <0.1 & $\simeq 1$ \\
480: 5 & 1 & 0.1 & <0.1 & $>10 $ \\
481: \end{tabular}
482: \end{tabular}
483: \end{ruledtabular}
484: \caption{Orders of magnitude of the exchange coupling strengths in the
485: Hamiltonian~\eqref{eq:H} and estimates for the size of the
486: correction terms, the asymmetry parameter
487: $\alpha=(J_\mathrm{CL}-J_\mathrm{CR})/(J_\mathrm{CL}+J_\mathrm{CR})$
488: and the gap, $E_\mathrm{gap}$, between the low-energy states described by the
489: Hamiltonian~\eqref{eq:H} and further excited states, for different electronic
490: populations $N$. The charging energy $U$ is of the order of $1\,\mathrm{eV}$.} \label{par}
491: \end{table}
492: }
493:
494: It is worthwhile to consider how robust our predictions are against
495: deviations from the ideal experimental setup. Specifically, (I) the
496: orientation of the molecules may be difficult to control, (II) its
497: electronic structure, and, specifically, its ideal symmetry, may be
498: altered upon deposition on a surface, (III) the molecule might be
499: affected by counter ions in the surrounding, and (IV) higher-energy
500: states not described by the Hamiltonian~\eqref{eq:H} might be
501: energetically accessible.
502:
503: With respect to (I) we note that regardless of the orientation of the
504: molecule, the change in charge state is firmly expected to affect the
505: central Mo$_{12}$O$_{40}$ core and not the vanadyl groups, so we think
506: orientation will only have an additional effect on (II). As an
507: insulating thin layer separates the molecule from the metallic
508: substrate, the alteration in the electronic structure of the
509: polyoxometalate is expected to be minimal (details are discussed in
510: the App.~\ref{sec:par}). The main effect on the low-energy physics will
511: be to introduce an asymmetry in the couplings between the left and
512: right vanadyl groups and the Keggin structure, $J_\mathrm{CL}$ and
513: $J_\mathrm{CR}$, respectively. In Table~I, we give bounds for this
514: asymmetry, which can be quantified by the
515: parameter~$\alpha=(J_\mathrm{CL}-J_\mathrm{CR})/(J_\mathrm{CL}+J_\mathrm{CR})$.
516: Furthermore, we have verified that an asymmetry of up to $10\%$ does
517: still allow one to achieve a fidelity of $99\%$ in the regime $J_1\ll
518: J_\mathrm{CL}, J_\mathrm{CR}$ (for details see App.~\ref{sec:quant}). Concerning (III), cations needed to compensate the
519: negative charge of the polyoxometalate polyanions, as well as solvent
520: molecules, may be still present when the cluster is deposited. This
521: feature may be a potentially important addition on (II) which could be
522: minimized using an electrospray system to choose the mass/charge ratio
523: corresponding to an isolated polyoxoanion. Such techniques have
524: succesfully been applied to polyoxometalates and the peaks of the
525: isolated anions have been detected.\cite{Bonchio2003a,Mayer2004a}
526: Finally (IV), the full diagonalization of the effective Hamiltonian
527: also results in a set of excited states with energies $E\geq
528: E_\mathrm{gap}$ above those states described by the
529: Hamiltonian~\eqref{eq:H}: we also include these data in Table~I.
530:
531: Having these points in mind, we see that different charge states are
532: usable. If we consider for instance $N=4$ and $5$, one would need
533: molecule-lead coupling rates $\hbar\Gamma$ in the range of around
534: $0.1$ and $5 \, \mathrm{meV}$, for the gate and readout process,
535: respectively. For the gate procedure, voltage pulses of a precision on
536: the order of a few $\mathrm{ps}$ are required, which is not far from
537: what can be achieved with present technology. In fact, if $N=0$ and
538: $N=1$ are found to be chemically stable in the experimental
539: environment, their parameters will be even more suitable for our
540: purposes by allowing $10$ times longer pulse times and yielding
541: $J_0=0$ as well as $E_\mathrm{gap}\gg J_0, J_\C$.
542:
543:
544:
545:
546: \section{Conclusions}
547:
548: The present scheme defines a class of molecular systems: those where two
549: localized spins are coupled to a redox-active unit. The physically relevant
550: figures of merit are included in Table~\ref{par}; the chemical desiderata are
551: stability, facility of deposition on surfaces, and possibility of controlled
552: oligomerization. Indeed, scalability of the present scheme requires covalent
553: bonding and directed self-assembly of these logical building blocks.
554:
555: Molecules of this class can be found in several chemical families,
556: besides bicapped polyoxometalates. Tetrathiofulvalene derivatives,
557: polypyrrol, porphyrines/phthalocyanines, fullerene and single-wall
558: nanotubes are a few among a plethora of organic systems which can
559: reversibly lose or gain one electron, can be substituted with radical
560: groups, and have a rich and well-controlled chemistry. For example,
561: theoretical calculations on phthalocyanines point to the possibility of
562: the chemical tailoring of these molecules to enable exchange couplings
563: in the range required for the gate scheme described
564: here~\cite{Shultz1999}.
565:
566: While for an initial experimental realization with a single molecule,
567: an STM contacting scheme as discussed in the present article enables
568: the best control and is clearly favorable, a scalable method for a
569: molecular monolayer could be based, e.g., on a crossbar architecture
570: which already with current technology reaches very high
571: densities.\cite{Green2007a}
572:
573: In conclusion, we have proposed an experimental setup for
574: single-molecule all-electric two-qubit gate and readout which is
575: within reach of current technology. We have exemplified our scheme
576: using a mixed-valence polyoxometalate, for which the model
577: parameters have been calculated using an \textit{ab initio} approach.
578: The general principles
579: behind our proposal also apply to different classes of molecular
580: systems. The chemical design, synthesis and characterization of such
581: systems should open new routes to molecular spin-qubit quantum
582: computing.
583:
584:
585: \begin{acknowledgments}
586:
587: We thank B. Coish for discussions. Financial support by the EU RTN
588: QuEMolNa, the EU NoE MAGMANet, the NCCR Nanoscience, the Swiss NSF, the
589: Spanish MEC (MAT2004-3849) and the Generalitat Valenciana is
590: acknowledged.
591:
592: \end{acknowledgments}
593:
594: \appendix
595:
596: \section{Bloch-Redfield-equation approach and average gate fidelity}
597:
598: \label{sec:br}
599:
600: For the description of the molecular dynamics in the presence of the
601: molecule-lead coupling we employ a Bloch-Redfield-equation formalism.
602: This approach, which is valid in the sequential-tunnelling regime,
603: i.e., to lowest order in the tunnelling rates $\Gamma_\ell$, includes the
604: full coherent quantum dynamics of the spin
605: states~\cite{Engel2001a,Kohler2005a} as described by the density
606: matrix elements $\varrho_{\alpha\beta} = \mathrm{Tr}_\leads \langle
607: \alpha|\varrho|\beta\rangle$ after tracing out the leads. In the
608: usual Born-Markov approximation, we obtain
609: \begin{equation}
610: \label{eq:master}
611: \begin{split}
612: \dot{\varrho}_{\alpha\beta} = &
613: -\I \omega_{\alpha\beta} \varrho_{\alpha\beta}
614: + \frac{1}{2}
615: \sum_{\ell \alpha' \beta'} \!
616: \bigg\{\!\!
617: \left[W^\ell_{\beta' \beta \alpha \alpha'} + (W^\ell_{\alpha' \alpha \beta \beta'})^\ast\right]
618: \varrho_{\alpha'\beta'}\\
619: &\quad-
620: W^\ell_{\alpha \beta'\beta'\alpha'}\,
621: \varrho_{\alpha'\beta}
622: -
623: (W^\ell_{\beta \alpha'\alpha'\beta'})^\ast\,
624: \varrho_{\alpha\beta'}
625: \bigg\}\,,
626: \end{split}
627: \end{equation}
628: where we have introduced the frequencies
629: $\omega_{\alpha\beta}=(E_\alpha-E_\beta)/\hbar$ and the transition
630: rates $W^\ell_{\beta' \beta \alpha \alpha'} = W^{\ell+}_{\beta' \beta
631: \alpha \alpha'} + W^{\ell-}_{\beta' \beta \alpha \alpha'}$ due to
632: tunnelling across a molecule-lead contact $\ell$. The rates for
633: tunnelling of an electron on and off the molecule assume the form
634: $W^{\ell+}_{\beta' \beta \alpha \alpha'} = \Gamma_{\ell} \,
635: f(E_\alpha-E_{\alpha'}-\mu_\ell) \sum_{\sigma} \langle
636: \beta'|d_{\C,\sigma}|\beta\rangle\!\langle
637: \alpha|d_{\C,\sigma}^\dagger|\alpha'\rangle$ and $W^{\ell-}_{\beta'
638: \beta \alpha \alpha'} = \Gamma_{\ell} \,
639: [1-f(E_{\alpha'}-E_{\alpha}-\mu_\ell)] \sum_{\sigma} \langle
640: \beta'|d^\dagger_{\C\sigma}|\beta\rangle\!\langle
641: \alpha|d_{\C,\sigma}|\alpha'\rangle$, respectively. Here, we assume
642: that the lead electrons are described by a Fermi distribution
643: $f(\epsilon-\mu_\ell) = \{1+\exp[(\epsilon-\mu_\ell)/kT]\}^{-1}$ at
644: temperature~$T$ and electro-chemical potential~$\mu_\ell$. The
645: tunnelling current across contact $\ell$ can be determined from the
646: density matrix elements via $ I_\ell = e\, \mathrm{Re} \sum_{\alpha
647: \alpha' \beta} \big( W^{\ell-}_{\beta \alpha' \alpha' \alpha} -
648: W^{\ell+}_{\ell \beta \alpha' \alpha' \alpha}
649: \big)\varrho_{\alpha\beta}$.
650:
651: Using the Bloch-Redfield equation approach, we are able to simulate a
652: realistic gate cycle. This allows us to calculate the average gate
653: fidelity~\cite{Poyatos1997a} $\mathcal{F}=\overline{\mathrm{Tr}
654: \left[\rho_\mathrm{real}(t_\mathrm{f}) \rho_\mathrm{ideal}\right]}$
655: as quantitative measure for the gate quality. Here, the overline
656: indicates the average over 16 unentangled input states
657: $|\Psi_i\rangle_\mathrm{L} |\Psi_j\rangle_\mathrm{R}$, where
658: $|\Psi_1\rangle=|{\uparrow}\rangle$,
659: $|\Psi_2\rangle=|{\downarrow}\rangle$,
660: $|\Psi_3\rangle=(|{\uparrow}\rangle + |{\downarrow}\rangle)/\sqrt{2}$,
661: and $|\Psi_4\rangle=(|{\uparrow}\rangle + \I
662: |{\downarrow}\rangle)/\sqrt{2}$, $\rho_\mathrm{ideal}$ is the ideal
663: result of the $\sqrt{\mathrm{SWAP}}$ operation as defined above and
664: $\rho_\mathrm{real}(t_f)$ is the state of the system according to the
665: quantum dynamics~\eqref{eq:master} at the final time~$t_f$.
666:
667: \section{Evaluation of exchange parameters}
668:
669: \label{sec:par}
670:
671: For the evaluation of the exchange coupling constants and to check the
672: validity of our numerical assumptions, we used an effective
673: Hamiltonian approach. We considered the main one- and two-center
674: phenomena in mixed-valence systems, namely magnetic exchange, electron
675: transfer, Coulomb repulsion and orbital energy, which were
676: parametrized for [PMo$_{12}$O$_{40}$(VO)$_2$]$^{q-}$ (details will be
677: published elsewhere). Thus, we diagonalized a 14-site
678: $t$--$J$--$V$--$\epsilon$ effective Hamiltonian for different
679: electronic populations. We projected this full energy level scheme
680: into a subsystem containing only the localized spins in the
681: [V$^{IV}$O] groups, plus, for odd values of $N$, an unpaired spin in
682: the Mo$_{12}$O$_{40}$ Keggin structure.
683:
684: For $N$ \textit{even}, $J_0$ is given by the energy difference between the
685: lowest singlet and triplet states, while $E_\mathrm{gap}$ is the gap
686: to the next energy level. For $N$ \textit{odd}, the evaluation of
687: $E_\mathrm{gap}$ is equally direct. Obtaining the exchange parameters
688: is also straightforward when a symmetric molecular structure is assumed:
689: $J_\C$ and $J_1$ can then be readily calculated from the energy
690: differences between the first doublet and quadruplet states.
691: When deviations from symmetry are considered, the
692: determination of the three exchange parameters $J_\C$,
693: $J_\mathrm{CL}$ and $J_\mathrm{CR}$ requires the analysis of the
694: wavefunctions.
695:
696: The spatially-dependent microscopic parameters $t$, $J$, $V$, and
697: $\epsilon$ were obtained from ab-initio calculations of molecular
698: subclusters, which accounts for inhomogeneities in the molecule. The
699: experimental setup is expected to imply small variations in bond
700: distances and angles, as well as a different charge distribution in
701: the surrounding of the molecule. In order to assess how these
702: perturbations affect the parameters in the effective
703: Hamiltonian~\eqref{eq:H}, the calculations were repeated for a
704: reasonable range of deviations ($\pm 50\%$ in $J$ and $t$, and $\pm
705: 1$eV in $\epsilon$). Except for the case with $N=3$, where the excited
706: states not contained in the Hamiltonian~\eqref{eq:H} lie close in
707: energy, the parameters of our interest proved to be robust enough
708: within the orders of magnitude indicated in Table~\ref{par}.
709:
710:
711:
712: \section{Characteristic gate quantifiers}
713:
714: \label{sec:quant}
715:
716: Besides the average gate fidelity discussed in the main text, one can
717: consider further quantities for assessing the quality of a
718: quantum-gate operation~\cite{Poyatos1997a}. One is the so-called gate
719: purity
720: \begin{equation}
721: \label{eq:s1}
722: \mathcal{P} = \overline{\mathrm{Tr} \left[\rho_\mathrm{real}(t_\mathrm{f})^2\right]}\,,
723: \end{equation}
724: which is a measure for the decoherence of the gate. Here, the average
725: is over all 16 unentangled input states $|\psi_\mathrm{ue}(t_0)\rangle
726: = |\Psi_i\rangle_\mathrm{L}|\Psi_j\rangle_\mathrm{R}$, $i,j=1,\dots4$
727: (see App.~\ref{sec:br} for the definition of the $|\Psi_i\rangle$). For
728: an ideal $\sqrt{\mathrm{SWAP}}$ operation it reaches unity. Another
729: measure is the entanglement capability~$\mathcal{C}$ which is defined
730: as the smallest eigenvalue of the partial transpose of the density
731: matrix~\cite{Peres1996a} $\rho_\mathrm{real}(t_\mathrm{f})$ obtained
732: for all 16 unentangled input states just defined. Ideally, it assumes
733: a value of $-1/2$. Finally, the quantum degree of the gate is given
734: by
735: \begin{equation}
736: \label{eq:s3}
737: \mathcal{Q} = \, \max_{|\psi_\mathrm{me}\rangle, \rho_\mathrm{ue}(t_0)}
738: \langle\psi_\mathrm{me}| \rho_\mathrm{real}(t_\mathrm{f}) |\psi_\mathrm{me}\rangle\,,
739: \end{equation}
740: i.e., as the maximum of the overlap between all possible output states
741: obtained for all 16 unentangled input states and all maximally
742: entangled states~$|\psi_\mathrm{me}\rangle$, viz, the four Bell
743: states $(|{\uparrow}\rangle_\mathrm{L}|{\uparrow}\rangle_\mathrm{R}
744: \pm
745: |{\uparrow}\rangle_\mathrm{L}|{\uparrow}\rangle_\mathrm{R})/\sqrt{2}$
746: and $(|{\downarrow}\rangle_\mathrm{L}|{\uparrow}\rangle_\mathrm{R} \pm
747: |{\uparrow}\rangle_\mathrm{L}|{\downarrow}\rangle_\mathrm{R})/\sqrt{2}$.
748: In the optimal case, the quantum degree reaches $1/2$ for the
749: $\sqrt{\mathrm{SWAP}}$ gate.
750:
751: In Fig.~\ref{fig:S1}, we show these four gate quantifiers as a
752: function of the gating time~$\tau_\mathrm{gate}$ and the ratio
753: $J_1/|J_\mathrm{C}|$. The gate fidelity~$\mathcal{F}$ has already been
754: discussed in the main text. We observe that the gate
755: purity~$\mathcal{P}$ assumes almost unity for the stroboscopic gating
756: times~(4). This means that at these times, the additional
757: spin~$\mathbf{s}_\mathrm{C}$ and the spin $\mathbf{S}_0 =
758: \mathbf{S}_\mathrm{L}+\mathbf{S}_\mathrm{R}$ become disentangled again,
759: such that one recovers a pure state when removing the additional
760: electron. Furthermore, we find that all gate quantifiers except for
761: the gate fidelity overestimate the quality of the gate: They almost
762: assume their perfect values even for parameters where the gate
763: fidelity is much below unity. The quantum degree~$\mathcal{Q}$ turns
764: out to be completely independent of the gating time and the
765: exchange-coupling ratio.
766:
767: \begin{figure*}[ht]
768: \includegraphics[width=0.9\textwidth]{figs1_small}
769: \caption{Four average gate quantifiers (a) gate fidelity (b) purity
770: (c) entanglement capability, and (d) quantum degree each
771: normalized to the maximal value achievable for an ideal
772: $\sqrt{\mathrm{SWAP}}$ operation as a function of the gating
773: time~$\tau_\mathrm{gate}$ and the ratio $J_1/|J_\mathrm{C}|$. The
774: parameters are as in Fig.~3. The conditions (3) and (4) are
775: indicated by solid lines. Note that the color scale for the gate
776: fidelity is different from the one used in Fig.~3.}
777: \label{fig:S1}
778: \end{figure*}
779:
780:
781: The situation in the presence of an
782: asymmetry~$\alpha=(J_\mathrm{CL}-J_\mathrm{CR})/(J_\mathrm{CL}+J_\mathrm{CR})$
783: between the couplings $J_\mathrm{CL}$ and $J_\mathrm{CR}$ of the
784: central Keggin core to the left and right, respectively, vanadyl
785: groups is shown in Fig.~\ref{fig:S2}. We have chosen the upper bound
786: $\alpha=0.1$ from the results of our \textit{ab initio} calculations
787: (cf.\ Table I). We find for the gating times shown in
788: Fig.~\ref{fig:S2} that this noticeable aberration from perfect
789: symmetry does not strongly affect the gate quality for small and/or
790: negative values of the ratio $J_1/|\bar J_\mathrm{C}|$, where $\bar
791: J_\mathrm{C} = (J_\mathrm{CL}+J_\mathrm{CR})/2$ is the average
792: coupling of the central spin to the left and right spins.
793: In particular, we still find a fidelity better than $\mathcal{F}=0.99$ at the
794: maximum indicated by the circles in Fig.~\ref{fig:S2} for a tunnel
795: rate $\Gamma=11 |\bar J_\mathrm{C}|/\hbar$.
796: %
797: \begin{figure*}[ht]
798: \includegraphics[width=0.9\textwidth]{figs2_small}
799: \caption{Four average gate quantifiers (a) gate fidelity (b) purity
800: (c) entanglement capability, and (d) quantum degree each
801: normalized to the maximal value achievable for an ideal
802: $\sqrt{\mathrm{SWAP}}$ operation as a function of the gating
803: time~$\tau_\mathrm{gate}$ and the ratio $J_1/|\bar J_\mathrm{C}|$,
804: where $\bar J_\mathrm{C} = (J_\mathrm{CL}+J_\mathrm{CR})/2$ is the
805: average exchange coupling between the Keggin core and the
806: left and right vanadyl groups. An asymmetry
807: $\alpha=(J_\mathrm{CL}-J_\mathrm{CR})/(J_\mathrm{CL}+J_\mathrm{CR})=0.1$
808: of these exchange couplings is assumed. The other parameters are
809: as in Fig.~3. The conditions (3) and (4) are indicated by solid
810: lines.}
811: \label{fig:S2}
812: \end{figure*}
813:
814:
815: \begin{thebibliography}{10}
816:
817: \bibitem{vanderWiel2003a}
818: W.~G. van~der Wiel, S. {D}e Franceschi, J.~M. Elzerman, T. Fujisawa, S.
819: Tarucha, and L.~P. Kouwenhoven, Rev. Mod. Phys. {\bf 75}, 1 (2003).
820:
821: \bibitem{Burkard1999a}
822: G. Burkard, D. Loss, and D.~P. DiVincenzo, Phys. Rev. B {\bf 59}, 2070
823: (1999).
824:
825: \bibitem{Barenco1995a}
826: A. Barenco, C.~H. Bennett, R. Cleve, D.~P. DiVincenzo, N. Margolus, P. Shor, T.
827: Sleator, J.~A. Smolin, and H. Weinfurter, Phys. Rev. A {\bf 52}, 3457
828: (1995).
829:
830: \bibitem{Loss1998a}
831: D. Loss and D.~P. DiVincenzo, Phys. Rev. A {\bf 57}, 120 (1998).
832:
833: \bibitem{Nielsen2000a}
834: M.~A. Nielsen and I.~L. Chuang, {\em Quantum Computation and Quantum
835: Information} (Cambridge University Press, New York, 2000).
836:
837: \bibitem{Petta2005a}
838: J.~R. Petta, A.~C. Johnson, J.~M. Taylor, E.~A. Laird, A. Yacoby, M.~D. Lukin,
839: C.~M. Marcus, M.~P. Hanson, and A.~C. Gossard, Science {\bf 309}, 2180
840: (2005).
841:
842: \bibitem{Koppens2006a}
843: F.~H.~L. Koppens, C. Buizert, K.~J. Tielrooij, I.~T. Vink, K.~C. Nowack, T.
844: Meunier, L.~P. Kouwenhoven, and L.~M.~K. Vandersypen, Nature {\bf 442}, 766
845: (2006).
846:
847: \bibitem{Hirjibehedin2006a}
848: C.~F. Hirjibehedin, C.~P. Lutz, and A.~J. Heinrich, Science {\bf 312}, 1021
849: (2006).
850:
851: \bibitem{Heersche2006a}
852: H.~B. Heersche {\it et~al.}, Phys. Rev. Lett. {\bf 96}, 206801 (2006).
853:
854: \bibitem{Jo2006a}
855: M.-H. Jo {\it et~al.}, Nano Lett. {\bf 6}, 2014 (2006).
856:
857: \bibitem{Kahn1993a}
858: O. Kahn, {\em Molecular magnetism} (VCH, New York, 1993).
859:
860: \bibitem{Gatteschi2006a}
861: D. Gatteschi, R. Sessoli, and J. Villain, {\em Molecular nanomagnets} (Oxford
862: University Press, Oxford, 2006).
863:
864: \bibitem{Wernsdorfer1999a}
865: W. Wernsdorfer and R. Sessoli, Science {\bf 284}, 133 (1999).
866:
867: \bibitem{Coronado2000a}
868: E. Coronado, J.~R. Gal\'an-Mascar\'os, C.~J. G\'omez-Garc{\'{i}}a, and V.
869: Laukhin, Nature {\bf 408}, 447 (2000).
870:
871: \bibitem{Real2003a}
872: J.~A. Real, A.~B. Gaspar, V. Niel, and M.~C. {Mu\~noz}, Coord. Chem. Rev. {\bf
873: 236}, 121 (2003).
874:
875: \bibitem{Stepanow2004a}
876: S. Stepanow {\it et~al.}, Nature Materials {\bf 3}, 229 (2004).
877:
878: \bibitem{ChemRev}
879: Hill, C. L. Special thematic issue on polyoxometalates, Chem. Rev. \textbf{98},
880: 1--390 (1998).
881:
882: \bibitem{Ardavan2007a}
883: A. Ardavan, O. Rival, J.~J.~L. Morton, S.~J. Blundell, A.~M. Tyryshkin, G.~A.
884: Timco, and R.~E.~P. Winpenny, Phys. Rev. Lett. {\bf 98}, 057201 (2007).
885:
886: \bibitem{Leuenberger2001a}
887: M.~N. Leuenberger and D. Loss, Nature {\bf 410}, 789 (2001).
888:
889: \bibitem{Troiani2005a}
890: F. Troiani, M. Affronte, S. Carretta, P. Santini, and G. Amoretti, Phys. Rev.
891: Lett. {\bf 94}, 190501 (2005).
892:
893: \bibitem{Chen1996}
894: Q. Chen and C.~L. Hill, Inorg. Chem. {\bf 35}, 2403 (1996).
895:
896: \bibitem{Keggin1933a}
897: J.~F. Keggin, Nature {\bf 131}, 908 (1933).
898:
899: \bibitem{Datta1997a}
900: S. Datta, W. Tian, S. Hong, R. Reifenberger, J.~I. Henderson, and C.~P. Kubiak,
901: Phys. Rev. Lett. {\bf 79}, 2530 (1997).
902:
903: \bibitem{Park1999a}
904: H. Park, A.~K.~L. Lim, A.~P. Alivisatos, J. Park, and P.~L. McEuen, Appl. Phys.
905: Lett. {\bf 75}, 301 (1999).
906:
907: \bibitem{Kubatkin2003a}
908: S. Kubatkin, A. Danilov, M. Hjort, J\'er\^ome, C.~J.-L. Br\'edas, N.
909: Stuhr-Hansen, P. Hedeg{\aa}rd, and T. Bj{\o}rnholm, Nature {\bf 425}, 698
910: (2003).
911:
912: \bibitem{Wolf1989a}
913: E.~L. Wolf, {\em Principles of Electron Tunneling Spectroscopy} (Oxford Univ.
914: Press, New York, 1989).
915:
916: \bibitem{Lehmann2006a}
917: J. Lehmann and D. Loss, Phys. Rev. B {\bf 73}, 045328 (2006).
918:
919: \bibitem{Shi2006}
920: Z. Shi, J. Peng, C.~J. G\'omez-Garc\'{i}a, S. Benmansour, and X. Gu, J. Solid
921: State Chem. {\bf 179}, 253 (2006).
922:
923: \bibitem{Bonchio2003a}
924: M. Bonchio, O. Bortolini, V. Conte, and A. Sartorel, Eur. J. Inorg. Chem. {\bf
925: 4}, 699 (2003).
926:
927: \bibitem{Mayer2004a}
928: C.~R. Mayer, C. Roch-Marchal, H. Lavanant, R. Thouvenot, N. Sellier, J. Blais,
929: and F. Secheresse, Chem. Eur. J. {\bf 10}, 5517 (2004).
930:
931: \bibitem{Shultz1999}
932: D.~A. Shultz and K.~A. Sandberg, J. Phys. Org. Chem. {\bf 12}, 10 (1999).
933:
934: \bibitem{Green2007a}
935: J.~E. Green {\it et~al.}, Nature {\bf 445}, 414 (2007).
936:
937: \bibitem{Engel2001a}
938: H.-A. Engel and D. Loss, Phys. Rev. Lett. {\bf 86}, 4648 (2001).
939:
940: \bibitem{Kohler2005a}
941: S. Kohler, J. Lehmann, and P. H\"anggi, Phys. Rep. {\bf 406}, 379 (2005).
942:
943: \bibitem{Poyatos1997a}
944: J.~F. Poyatos, J.~I. Cirac, and P. Zoller., Phys. Rev. Lett. {\bf 78}, 390
945: (1997).
946:
947: \bibitem{Peres1996a}
948: A. Peres, Phys. Rev. Lett. {\bf 77}, 1413 (1996).
949:
950: \end{thebibliography}
951:
952: \end{document}
953: