cond-mat0402428/dnp.tex
1: %\documentclass[preprint,prb,showpacs]{revtex4}
2: %\documentclass[twocolumn,showpacs,prb,amsmath,amssymb]{revtex4}
3: \documentclass[twocolumn,showpacs,preprintnumbers,prl]{revtex4}
4: 
5: \usepackage[dvips]{graphicx}
6: \usepackage[dvips]{color}
7: 
8: \usepackage{graphicx}
9: \usepackage{epsfig}
10: 
11: \begin{document}
12: \title{Selective Dynamic Nuclear Spin Polarization in Spin-Blocked Double-Dot}
13: \author{Changxue Deng}
14: \author{Xuedong Hu}
15: \affiliation{Department of Physics, University at Buffalo, SUNY,
16: Buffalo, NY 14260-1500}
17: \date{\today}
18: 
19: \begin{abstract}
20: We study the mechanism of dynamical nuclear spin polarization by hyperfine
21: interaction in spin-blocked double quantum dot system.  We calculate the
22: hyperfine transition rates and solve the master equations for the nuclear
23: spins.  Specifically, we incorporate the effects of the nuclear quadrupole
24: coupling due to the doping-induced local lattice distortion and strain.  Our
25: results show that nuclear quadrupole coupling induced by the 5\% indium
26: substitution can be used to explain the recent experimental observation of
27: missing arsenic NMR signal in the spin-blocked double dots.  
28: \end{abstract}
29: \pacs{85.35.Be, 76.60.-k, 03.67.Lx, 
30: }
31: \maketitle
32: %\vskip2pc]
33: %\narrowtext
34: 
35: In many metals and semiconductors, nuclei and conduction electrons form a
36: coupled nonlinear spin system through the hyperfine interaction.  Such a
37: coupling is well studied in bulk experimental phenomena such as dynamical nuclear
38: spin polarization \cite{OO}.  
39: %
40: Both electron and nuclear spins in semiconductor heterostructures have been
41: proposed as candidates for qubits in a scalable quantum information
42: processor.  There are also a variety of spintronic devices where electron
43: spins provide novel functionalities.  For both quantum computation and
44: spintronics, understanding the electron-nuclear spin interaction and the
45: resulting coupled dynamics is crucial because of its important roles in spin
46: manipulation 
47: %(particularly for the nuclear spins, which are difficult to
48: %access otherwise) 
49: and decoherence.  
50: %
51: %Thus the study of coupled dynamics of electron and nuclear spins is a key to
52: %both quantum computing and spintronics.  
53: %
54: Indeed, in the past decade there have been extensive experimental and
55: theoretical studies of electron-nuclei hyperfine interaction in semiconductor
56: nanostructures \cite{Gammon-science, Kawa, decoherence, Zoller, SB-NS}.  In
57: this Letter, we show that nuclear quadrupole coupling can be a critical
58: factor in this coupled dynamics, thus needs to be carefully accounted for in
59: the study of spin-based quantum computers and spintronic devices.
60: 
61: %It has been known for a long time that optically oriented electrons 
62: %can be utilized to polarize lattice nuclei in bulk semiconductors 
63: %\cite{OO} due to the very long spin-lattice relaxation time $T_1$. 
64: %The nuclei interacts with electrons in the conduction band through 
65: %the Fermi contact hyperfine interaction leading to a nonlinear 
66: %electron-nuclear spin system with strong coupling. Recently there 
67: %have been intensive studies of understanding electron-nuclei 
68: %hyperfine interaction in semiconductor nanostructures 
69: %\cite{SB-NS, Gammon-science, Gammon, Zoller} which hold some 
70: %promises for future spintronic and quantum computation devices.
71: 
72: Dynamical nuclear spin polarization has recently been demonstrated in a spin
73: blocked semiconductor double quantum dot made of Ga$_{0.95}$In$_{0.05}$As
74: \cite{SB-NS,SB,QD-rev}.  Coulomb interaction and Pauli principle dictates that
75: in a double quantum dot two-electron singlet and triplet states are split, so
76: that proper voltage offset and bias between the double dot leads to
77: significant suppression in the tunnel current due to occupied triplet state
78: (thus the so-called spin blockade regime, which has been suggested for single
79: spin detection, a crucial component of quantum information processing)
80: \cite{SB}.  One way to lift this blockade is to apply an appropriate magnetic
81: field,
82: % 
83: %when singlet-triplet spin-flip transition can be facilitated by the
84: %electron-nuclear spin hyperfine coupling as one of the polarized triplet
85: %state is energetically degenerate with the singlet state \cite{SB-NS}.  This
86: %selective transition can then dynamically polarize the nuclear spins in the
87: %double quantum dot.  
88: %
89: so that electron-nuclear spin hyperfine coupling can facilitate electron
90: spin-flip transitions, which in turn leads to dynamical nuclear spin
91: polarization as was shown in Ref.~\cite{SB-NS}.  The reported experimental
92: observations reveal several quite unexpected phenomena \cite{SB-NS}, one of
93: which is that 
94: %
95: %the oscillation period/amplitude is significantly reduced when nuclear
96: %magnetic resonance (NMR) is applied at the Ga$^{69}$ and Ga$^{71}$
97: %frequencies, 
98: %
99: while nuclear magnetic resonance (NMR) signals were observed at the
100: $^{69}\!$Ga and $^{71}\!$Ga frequencies, no comparable response has been
101: found from the $^{75}\!$As nuclei.  It seems that nuclear spin polarization
102: cannot be built up in $^{75}\!$As nuclei even though they have the largest
103: concentration and the largest hyperfine coupling with the electrons among the
104: three nuclear species.  
105: 
106: In this Letter we study the mechanism of dynamical nuclear spin polarization
107: by hyperfine interaction in a spin-blocked double quantum dot, taking into
108: account electron tunneling and nuclear quadrupole coupling.  We show that the
109: absence of $^{75}\!$As nuclear spin signal is due to the local strain created
110: by the presence of Indium atoms.  Specifically, the substitution of the 5\%
111: In for Ga breaks the crystal symmetry and induces a local lattice distortion
112: \cite{Quad0,Quad}.  The charge redistribution generates {\it static} electric
113: field gradients (EFG) at the surrounding lattice sites, which lead to finite
114: quadrupole coupling for the nuclei \cite{GaAs_no_quad}.  The randomly
115: distributed In atom only substitute one of the nearest neighbors of
116: As atoms, while the nearest neighbors of Ga atoms are always four As atoms. 
117: Therefore the local electric field gradients at the Ga nuclei sites are on
118: average much smaller than at the As sites.  We show that this difference of
119: EFG at the locations of As and Ga nuclei is the major origin of the
120: unexpected absence of As nuclear spin signal.  
121: %
122: %Specifically, we first calculate the nuclear spin levels accounting for
123: %Zeeman splitting and quadrupole coupling.  We then identify and calculate
124: %the nuclear spin transition rates due to hyperfine coupling, incorporating
125: %electron tunneling to satisfy the energy conservation requirements.
126: 
127: %{\it Spin blockade} (SB) has been observed in a vertical double 
128: %quantum dot system with single-electron tunneling characteristics 
129: %\cite{SB}. SB appears when the electrons in both dots have the 
130: %same spin polarization. A large  current suppression occurs as 
131: %a result of Pauli exclusion principle. However a leakage current 
132: %of order 1~pA through the cotunnleing channel in the spin-blockaded 
133: %region has been recently reported by the same group \cite{SB-NS}. 
134: %Two more unexpected results have been found at the singlet-triplet 
135: %crossing when a DC magnetic field (0.4-0.9 Tesla) is applied in 
136: %the 2D plane of the well in these experiments. The first of them 
137: %is the coherent current oscillation with a period of order 100 
138: %seconds in the same region due to back reaction of nuclear spins 
139: %\cite{SB-NS, nutation}; The other one is that the oscillation 
140: %period is significantly reduced when nuclear magnetic resonance 
141: %(NMR) is applied at the Ga$^{69}$ and Ga$^{71}$ frequencies, while 
142: %no comparable response has been found from the As nuclei albeit 
143: %it has the largest concentration. These results suggest that 
144: %the gallium nuclei has been dynamically polarized via the 
145: %hyperfine interaction. 
146: 
147: %A key point in these experiments is that the singlet state $|0,0\rangle =
148: %\frac{1}{\sqrt{2}}(| \uparrow \rangle | \downarrow\rangle - |\downarrow
149: %\rangle | \uparrow\rangle )$ and one of the triplet state, i.e., $|1,-1
150: %\rangle = |\downarrow \rangle |\downarrow \rangle$ \cite{note_ST} have the
151: %same energy at some intermediate magnetic field.  Hyperfine induced flip-flop
152: %scattering between these two-particle spin states could be driven on
153: %resonance because energy conservation is satisfied in the process.  This
154: %transition always offers nuclei with the same spin orientation so that the
155: %nuclei can reach to a non-vanishing polarization in equilibrium condition. 
156: %The process is irreversible since once the electronic system is in the
157: %singlet state electron will tunnel through the dots to leads.  It should be
158: %noted that either electron in the dots could be flipped to make a
159: %singlet-triplet transition. In this letter we will concentrate on nuclear
160: %spin dynamics, specifically we will study the steady-state nuclear spin
161: %polarization under various experimental conditions and propose the possible 
162: %relaxation mechanism that can be used to explain the origin of the absence of
163: %As NMR signal.
164: 
165: %In the samples (InGaAs/AlGaAs) that are used to conduct 
166: %the quantum dot transport experiments 5\% indium is introduced in the well to 
167: %lower the bottom of the conduction band to 32 meV below the Fermi level of
168: %the GaAs contacts \cite{QD-rev,SB-NS}. The substitution of 5\% In for Ga
169: %breaks the crystal symmetry and induce a local lattice distortion
170: %\cite{Quad0,Quad}.  The charge redistribution generates static electric
171: %field gradients (EFG) at the lattice sites.  Since all the nuclei in GaAs
172: %have spin $\frac{3}{2}$ with finite quadrupole moments, we expect that there
173: %should be strong quadrupole interactions which have been confirmed with NMR
174: %technique \cite{Quad0, Quad}. Quadrupole effects in Al$_x$Ga$_{1-x}$As have
175: %been studied extensively in optical pumping experiments \cite{chap5}.
176: %However this distortion has different influence on Ga and As nuclei since
177: %the two atoms are coupled by the valence bonds, the randomly distributed In
178: %atom only substitute one of the nearest neighbors of As atoms; on the other
179: %hand the nearest neighbors of Ga atoms are four As atoms, therefore the
180: %local electric field gradients at Ga nuclei are less sensitive to
181: %replacement of Ga atoms by In atoms.  We show that this difference of EFG at
182: %the locations of As and Ga nuclei is the major origin of the unexpected
183: %observation.
184: 
185: We first note that the static quadrupole interaction alone cannot generate any
186: transition among the spin eigenstates because there is no energy relaxation
187: mechanism.  This is different from the spin-lattice effect, where the EFG
188: is generated by the lattice oscillations.  There quadrupole spin relaxation
189: can be realized by absorbing and emitting a phonon simultaneously (i.e., a
190: Raman process) \cite{spin-lattice}.  In the system we consider, the nuclei
191: couple with the electrons, which in turn couple to the leads (reservoir)
192: by tunneling, so that energy conservation can be satisfied through charge
193: exchange between the dot and the leads.  
194: %
195: %We could associate with the electron a density of state (DOS), i.e. energy
196: %broadening, then calculate the transition rates of second-order processes
197: %induced by quadrupole and hyperfine interaction.  
198: %
199: In the following we first find the nuclear spin eigenstates including the
200: static quadrupole interaction, then evaluate the first-order transition rates
201: among these eigenstates due to hyperfine coupling.  
202: %
203: %In this Letter we apply the second approach, where quadrupole
204: %interaction is not treated as a perturbation. 
205: 
206: The static Zeeman and quadrupole Hamiltonian for a single nuclear spin is
207: %
208: \begin{equation}
209: H_{n} = -\omega \hat{I}_z + \nu ( \hat{I}_Z^2 - \frac{1}{3} \hat{I^2} ),
210: \label{eq-H_n}
211: \end{equation}
212: %
213: where $\omega$ is the nuclear spin Lamour frequency and $\nu$ is quadrupole
214: coupling strength that is proportional to the EFG at the locations of the
215: nuclei, $z$ is the direction of external magnetic field and $Z$ is the
216: principle axis of the largest EFG \cite{NMR}.  We have
217: neglected the asymmetrical effect of quadrupole interaction.  It was
218: determined experimentally that $\nu_{\text{As}} \approx \text{2.5 MHz}$ and
219: $\nu_{\text{Ga}}$ is much less than $\nu_{\text{As}}$ in In$_x$Ga$_{1-x}$As
220: for $x$ up to 0.01 \cite{Quad}.  For a typical external field $B_{\text{ext}}$ (0.8 Tesla)
221: applied in the experiment \cite{SB-NS}, we can assume $\nu_{\text{Ga}} \ll
222: \omega_{\text{Ga}} \approx 10.4 ~\text{and} ~8.2 ~\text{MHz}$ for $^{69}$Ga
223: and $^{71}$Ga nuclei.  Since our intention is to study the relation of As and
224: Ga nuclear polarization, we will not differentiate $^{69}$Ga and $^{71}$Ga
225: further in this Letter.  The small $\nu_{\text{Ga}}$ allows eigenstates $|\phi_i \rangle$
226: ($i$=-3/2,-1/2,1/2 or 3/2) of $H_n$ for Ga nuclei to be expressed
227: perturbatively in terms of the Zeeman spin eigenstates $|m \rangle$.
228: %
229: \begin{eqnarray}
230: | \phi_{\pm \frac{3}{2} } \rangle &=& |\pm \frac{3}{2} \rangle - b ~|\pm
231: \frac{1}{2} \rangle \mp c ~|\mp \frac{1}{2}\rangle , \nonumber \\
232: | \phi_{\pm \frac{1}{2} } \rangle &=& |\pm \frac{1}{2} \rangle + b ~|\pm
233: \frac{3}{2} \rangle \mp c ~|\mp \frac{3}{2} \rangle ,
234: \label{eq:eigen1}
235: \end{eqnarray}
236: %
237: where $b=\sqrt{3}\nu_{\text{Ga}}\text{sin}\theta \text{cos}\theta
238: /\omega_{\text{Ga}}$ and $c = \sqrt{3}\nu_{\text{Ga}}\text{sin}^2\theta
239: /4\omega_{\text{Ga}}$.  $\theta$ is the angle between $z$ and $Z$ axes.  In
240: these calculations $x$ axis is chosen to be in the plane of $z$ and $Z$. 
241: These new states $|\phi_i \rangle $ are not eigenvectors of $\hat{I}_z$
242: anymore, though their average spin polarizations $\langle \phi_i | \hat{I}_z |
243: \phi_i \rangle = i + {\it O}(\nu_{\text{Ga}}^2/\omega_{\text{Ga}}^2)$.
244: 
245: %which is very close to $i$.
246: 
247: Transitions can be induced by the hyperfine interaction
248: among the new energy eigenstates of the Ga nuclei.  The hyperfine
249: Hamiltonian is
250: %
251: \begin{equation}
252: H_h = \sum_i^{N} A_i |\psi({\bf R}_i)|^2 \left[ \hat{I}_z^{i} \hat{S}_z +
253: \frac{1}{2}(\hat{I}_+^{i} \hat{S}_- + \hat{I}_-^{i} \hat{S}_+ ) \right].
254: \label{eq:hyper}
255: \end{equation}
256: %
257: Here $N \sim 10^5$ is the number of nuclei in a quantum dot; $A$ is the
258: hyperfine coupling constant; $\psi({\bf R}_i)$ is the electron envelope
259: wavefunction at the $i$th nucleus.  Using the electron density at nuclear
260: sites \cite{Paget} we estimate $A \approx 46~\mu$eV  and 40 $\mu$eV for 
261: As and Ga nuclei respectively.  Hamiltonian~(\ref{eq:hyper}) describes the interaction of an
262: electron with an ensemble of $N$ nuclei in the dot.  
263: 
264: To calculate the transition rates given in Fig.~1 we must study how the system
265: couples to the environment to satisfy energy conservation.  In the
266: present case electron exchange between a dot and the leads is the most
267: important energy relaxation mechanism \cite{NSR}.  For triplet state
268: $|\downarrow \downarrow \rangle$ \cite{note_ST} the electron in the 1st (2nd) dot only
269: couples to the left (right) lead.  The electron tunnel Hamiltonian can be
270: expressed as
271: %
272: \begin{equation}
273: H_e = \sum_{k} \epsilon_k c_{k \downarrow}^+c_{k \downarrow}
274: + \epsilon_0 d_{\downarrow}^{+} d_{\downarrow}
275: + \sum_{k} \left [t_k c_{k \downarrow}^{+} d_{\downarrow} + \text{H.C.}
276: \right],
277: \label{eq-He}
278: \end{equation}
279: %
280: where $\epsilon_0$ is the renormalized single particle energy level.  
281: Since the Hamiltonian
282: $H_e$ is quadratic, exact solution of the retarded Green's function
283: $G_{\downarrow \downarrow}^{Ret} = -i \theta (t) \langle \{d_{\downarrow}(t),
284: d_{\downarrow}^{+}(0)\} \rangle$ can be obtained by using the equation of
285: motion method.  The density of state (DOS) follows from $\rho_T (\epsilon) =
286: -\text{Im} G_{\downarrow \downarrow}^{Ret} (\epsilon + i0^+)/ \pi$.  It is
287: straightforward to show that the electron triplet state is broadened
288: %
289: \begin{equation}
290: \rho_T(\epsilon) =  \frac{1}{2\pi} \frac{\Gamma}{(\epsilon - \epsilon_0)^2 +
291: \Gamma^2/4},
292: \label{eq:dos}
293: \end{equation}
294: %
295: where $\Gamma = 2\pi \sum_k |t_k|^2 \delta(\epsilon-\epsilon_k)$ is the level 
296: broadening for the triplet state. 
297: %
298: %For singlet state the first dot can also exchange electron with 
299: %right lead. The DOS of the electron in singlet state $\rho_S$ has to be
300: %modified from Eq.~(\ref{eq:dos}). However since the inter-dot coupling is
301: %much weaker than the tunneling coupling between the dot and the lead,
302: %$\rho_S$ can be approximated by Eq.~(\ref{eq:dos}) \cite{note_DOS}.
303: %
304: For the electron in the singlet state, the DOS should be
305: modified by a weak inter-dot coupling.  However, the major energy relaxation
306: channel is still the lead-dot tunnel coupling, so we can approximate the
307: singlet DOS with Eq.~(\ref{eq:dos}) as well. 
308: 
309: \begin{figure}[t]
310: \begin{center}
311: \epsfig{file=fig1.eps, width=4.5cm,height=3cm}
312: \caption{Schematic of hyperfine induced transitions among the energy
313: eigenstates of Hamiltonian $H_n$.  The downward arrows represent simultaneous
314: flip-flop scattering of nuclei and the electron in a triplet state.  $F$ and
315: $G$ are the nuclear spin relaxation rates induced by the effective hyperfine
316: magnetic field of the electron ($\frac{A}{N} I_z S_z$).}
317: \label{fig1}
318: \end{center}
319: \end{figure}
320: 
321: The first term in Eq.~(\ref{eq:hyper}) does not involve electron spin flip. 
322: However, it can induce transitions among the energy eigenstates due to 
323: mixture of different spin eigenstates in Eq.~(\ref{eq:eigen1}). The
324: transition rate between state $|\phi_i \rangle$ and $|\phi_j \rangle$
325: due to this term is given by the Fermi golden rule,
326: %
327: \begin{equation}
328: W_{i,j}^{z} = \frac{\pi A^2}{2\hbar N^2} |\langle \phi_i | \hat{I}_z | \phi_j
329: \rangle |^2 P(E_j - E_i),
330: \label{eq:Wz}
331: \end{equation}
332: %
333: where $E_i$ and $E_j$ are the eigenenergies of $H_n$.  The absorption power
334: $P$ is
335: %
336: \begin{equation}
337: P(\Delta) = \sum_{\epsilon_i,\epsilon_f} \rho (\epsilon_i)
338: \rho (\epsilon_f) \delta(\epsilon_f - \epsilon_i + \Delta),
339: \label{eq:P}
340: \end{equation}
341: %
342: where $\rho(\epsilon)$ is the DOS of either the electron singlet or the
343: triplet state.  Using Eq.~(\ref{eq:Wz}), the transition rates in
344: Fig.~(\ref{fig1}) can be calculated: $F$ = $6b^2A^2/(N^2 \hbar \Gamma)$ and
345: $G$ = $6c^2A^2/(N^2 \hbar \Gamma)$.  In deriving these expressions we have
346: neglected the nuclear energy difference of the initial and final state since
347: they are much smaller than the width of electronic energy broadening.  
348: 
349: The calculation of the transition rates due to the hyperfine flip-flop terms
350: in Eq.~(\ref{eq:hyper}), which correspond to the polarized triplet to singlet
351: electron transitions, is more complicated.  When the two electron states are
352: degenerate, the hyperfine flip-flop transition is on resonance, leading
353: to nuclear spin polarization.  However, the transition should quickly becomes off
354: resonance due to the effective nuclear magnetic field $B_{\text{N}}$
355: experienced by the electrons. 
356: %
357: %Regarding $A I_z S_z$ as an average electron spin energy, 
358: %
359: The effective nuclear magnetic fields can be estimated as
360: %
361: \begin{equation}
362: B_{\text{N}} = \gamma_{\text{As}} \frac{\langle \hat{I}_z \rangle
363: _{\text{As}}}{I} + \gamma_{\text{Ga}} \frac{\langle \hat{I}_z \rangle
364: _{\text{Ga}}}{I},
365: \label{eq:BN}
366: \end{equation}
367: %
368: where $\gamma_{\rm As} \approx -2.8 $ Tesla and $\gamma_{\rm Ga} \approx -1.3$
369: Tesla \cite{Paget}.  As and Ga nuclei have independent spin temperatures
370: because their mutual flip-flop process is largely suppressed due to different
371: magnetic moments.  Equilibrium in nuclear spin polarization will thus be
372: established in each of the nuclear species independently.  $B_{\rm N}$ can be
373: quite large for even moderate nuclear polarization.  The energy difference of
374: the singlet state and triplet state due to $B_{\text{N}}$ can be approximated
375: by $\Delta_{ST} = -g^*\mu_B B_{\text{N}}$ assuming the energy of the singlet
376: state is independent of magnetic field.  The flip-flop transition rates of
377: hyperfine interaction from state $|\phi_i \rangle $ to state $|\phi_j
378: \rangle$ is thus
379: %
380: \begin{equation}
381: W_{i,j}^{-} = \frac{f_T \pi A^2}{8\hbar N^2}
382: | \langle \phi_{j} | \hat{I}_{-} | \phi_i \rangle |^2 P(\Delta_{ST} + E_{j} -
383: E_i),
384: \label{eq:W-}
385: \end{equation}
386: %
387: where $f_T$ is a multiplication factor due to cotunneling and the
388: probability of the triplet state occupation (electrons can tunnel into three 
389: triplet states and one singlet state, but only one triplet is relevant to 
390: nuclear polarization). $f_T$ should be multiplied by another factor 0.5 because 
391: there are two dots and the probability of nuclear spin-flip process in each of the dots is
392: 50\%. We estimate $f_T \approx 0.05$.  
393: %
394: %Again ignoring the nuclear magnetic energy in the argument of $P$ 
395: %
396: We then find that $D$ in Fig.~(\ref{fig1}) is given as $D = f_T A^2 \Gamma
397: /N^2\hbar (\Delta_{ST}^2 + \Gamma^2)$.
398: 
399: The quadrupole interaction of As nuclei is much stronger than that of Ga
400: [$\nu_{\text{As}} \approx 2.5$ MHz is in the same order as the Zeeman energy
401: $\omega_{\text{As}} \approx 5.9$ MHz (at 0.8 Tesla)] and leads to a complete
402: mixing of the spin eigenstates, so that the perturbative wavefunctions
403: [Eq.~(\ref{eq:eigen1})] for Ga nuclei have to be replaced with numerical
404: solutions:
405: %  
406: %We can write the new energy eigenstates as
407: %linear combinations of $|m \rangle$:
408: %
409: \begin{equation}
410: |\phi_i \rangle = \sum_{m} u_{im} |m \rangle.
411: \label{eq:eigen2}
412: \end{equation}
413: %
414: The expectation value of $\hat{I}_z$ of these state is $\langle \phi_i |
415: \hat{I}_z | \phi_i \rangle = \sum_{m} |u_{im}|^2 m$.  For example, we find
416: that when $\alpha \equiv \nu_{\text{As}} / \omega_{\text{As}} =0.3$, 
417: $\langle \phi_i | \hat{I}_z | \phi_i \rangle $ = 1.419, 0.558, -0.528 and
418: -1.450 for $i$ = 3/2, 1/2, -1/2 and -3/2 respectively.  All the matrix
419: elements of $\hat{I}_z$ and $\hat{I}_-$ can be easily obtained from
420: Eq.~(\ref{eq:eigen2}), and the transition rates can be calculated numerically
421: using Eq.~(\ref{eq:Wz}) and (\ref{eq:W-}).
422: %
423: %$\langle \phi_i | \hat{I}_z  | \phi_j \rangle$ and $\langle \phi_i |
424: %\hat{I}_- | \phi_j \rangle$ can be inserted into Eq.~(\ref{eq:Wz}) and
425: %Eq.~(\ref{eq:W-}) to evaluate the transition rates numerically.  
426: %
427: One should keep in mind that the illustration of the various transitions in
428: Fig.~(\ref{fig1}) is only applicable to the perturbative analysis for Ga
429: nuclei, while for As nuclei all transitions are possible.  However the
430: concept of polarization ($D$) and depolarization ($F$ and $G$) processes
431: illustrated in Fig. (\ref{fig1}) should still be valid.
432: 
433: After obtaining all the transition rates $W^z$ and $W^-$ we can proceed to
434: construct the master equations \cite{NMR}
435: %
436: \begin{equation}
437: \frac{d p_i} {dt} = \sum_{j \ne i} W_{j,i} p_j - \sum_{j \ne i} W_{i,j} p_i,
438: \label{eq:rate}
439: \end{equation}
440: %
441: where $W_{i,j} = W_{i,j}^z +W_{i,j}^-$, and $p_{i}$ is the probability of 
442: nuclear spins in state $| \phi_{i} \rangle$.  In Eq.~(\ref{eq:rate}) we have
443: neglected the spin-lattice relaxation, which is very weak at liquid Helium
444: temperature.  Here we can use the spin temperature approximation
445: because the nuclear spin build-up time ($\approx$ 100 s) \cite{SB-NS} is much
446: longer than the transverse spin relaxation $T_2~ (\approx 10^{-4} ~{\rm s})$.
447: We solve for the steady-state polarization: $d p_i/dt = 0$.  
448: %Only three of the four equations are independent, the remaining equation is 
449: %the normalization condition $\sum_{i} p_{i} = 1$.  
450: These equations are highly nonlinear, as the
451: equilibrium polarizations depend on the transition rates, while to calculate
452: the transition rates $W^-$ one must know the effective nuclear magnetic
453: field, which itself depends on the nuclear spin polarization.  We solve the
454: set of equations self-consistently.  The average nuclear polarization is $P =
455: \sum_i p_i \langle \phi_i | \hat{I}_z | \phi_i \rangle = \langle \hat{I}_z
456: \rangle /I$.  We can then calculate the effective nuclear magnetic fields
457: $B_{\text{N}}$ with Eq.~(\ref{eq:BN}).  
458: %The updated transition rates are used
459: %to find the new nuclear spin polarizations until self-consistency is
460: %achieved.
461: 
462: \begin{figure}[t]
463: \begin{center}
464: \epsfig{file=fig2.eps, width=8.5cm, height=4.8cm}
465: \caption{ Ratio of Ga and As nuclei polarization as a function of the relative
466: quadrupole interaction strength $\alpha ~(=\nu_{\text{Ga}}/\nu_{\text{As}})$
467: for $\Gamma=10~\mu$eV (panel a) and $\Gamma=30~\mu$eV (panel b).
468: $\Gamma$ is the width of level broadening due to electron tunneling between
469: the lead and the dot. The three curves in each graph represent three different
470: As quadrupole coupling strengths $\nu_{\text{As}}/\omega_{\text{As}} $ 
471: = 0.1, 0.3 and 0.5. $\theta$ has chosen to be 45 degrees for both As and Ga
472: nuclei.
473: }
474: \label{fig2}
475: \end{center}
476: \end{figure}
477: 
478: Figure~\ref{fig2} shows the calculated ratio of Ga and As nuclei polarization
479: as a function of $\alpha$, which is taken as a free parameter in our
480: calculations.  We consider As quadrupole effects with three different
481: strengths.  $\nu_{\text{As}}/\omega_{\text{As}} = 0.3$ corresponds to
482: $\nu_{\text{As}} \approx 1.8$ MHz.  In this case the Ga nuclear polarization
483: is two orders of magnitude greater than that of As nuclei when $\alpha
484: \approx$ 0.01.  This difference can qualitatively explain the experimental
485: observation that the As NMR signal is missing in measuring the current
486: oscillation period as a function of the AC magnetic field frequency.  We also
487: find the ratio is very sensitive to electron level width $\Gamma$. 
488: Figure~\ref{fig2} shows that the polarization ratio in the case of
489: $\Gamma=30~\mu$eV is roughly $\frac{1}{5}$ of that in the case of
490: $\Gamma=10~\mu$eV.  Since $D \propto \Gamma/(\Delta_{ST}^2 + \Gamma^2 )$, the
491: triplet to singlet transition rate decreases much faster when $\Gamma$ is
492: smaller, resulting in larger polarization ratio.  
493: 
494: Figure~\ref{fig3} shows how Ga nuclear spin polarization changes with the
495: relative quadrupole coupling strength $\alpha$.  It is obvious from the two
496: panels that $P_{\text{Ga}}$ increases rapidly as $\alpha$ decreases from 0.1
497: to 0.01.  The Ga polarization saturates to unity for smaller $\alpha$, and
498: larger $\Gamma$ leads to larger Ga polarization.
499: 
500: \begin{figure}[t]
501: \begin{center}
502: \epsfig{file=fig3.eps, width=8.5cm, height=4.8cm}
503: \caption{ Ga nuclear spin polarization as a function of the relative
504: quadrupole interaction strength $\alpha$ for $\Gamma=10~\mu$eV (panel a) and
505: $\Gamma=30~\mu$eV (panel b). All the parameters are the same as those used in
506: Fig. 2.}
507: \label{fig3}
508: \end{center}
509: \end{figure}
510: 
511: %As an illustration to our calculation, we consider the dynamical nuclear
512: %polarization of nuclei with spin $I=\frac{1}{2}$.  Equation~(\ref{eq:rate})
513: %in steady state can be solved easily and we obtain
514: %
515: %\begin{equation}
516: %\frac{P_{\text{As}}}{P_{\text{Ga}}} 
517: %               = \frac{D(B_{\text{N}}) + 2F_{\text{Ga}}}
518: %                      {D(B_{\text{N}}) + 2F_{\text{As}}},
519: %\end{equation}
520: %
521: %where $D(B_{\text{N}})$ and $F$ have the same meaning as in the above
522: %discussion.  We have used the fact the hyperfine constants of Ga and As are
523: %approximately equal ($D_{\text{As}} \approx D_{\text{Ga}}$). In the limit of 
524: %$F_{\text{Ga}} \rightarrow 0$  and $D(B_{\text{N}}) \ll F_{\text{As}}$ 
525: %(i.e., off resonance), $P_{\text{As}}/P_{\text{Ga}} \approx
526: %D(B_{\text{N}})/2F_{\text{As}} \ll 1$. 
527: 
528: The physical picture of the master equation calculation is actually quite
529: straightforward.  Both Ga and As have a polarization rate and a
530: depolarization rate.  The As nuclear spin depolarization rate is much larger
531: than that of Ga due to the quadrupole coupling.  Thus there exists a regime
532: of polarization rate where nuclear spin polarization can be built up in Ga
533: but not in As.
534: 
535: %At this point we would like to clarify the relation of the present work and 
536: %some of the previous studies where As polarization has been detected in spite
537: %of the presence of quadrupole interactions (say in AlGaAs)
538: %\cite{Gammon-science}. 
539: %
540: In our study As depolarization rates are greatly enhanced because the energy
541: levels of the dots are broadened through tunnel coupling to the leads, and
542: that the singlet-triplet transition becomes off resonance once the
543: polarization of Ga nuclei is established.  In previous experiments concerning
544: dynamical nuclear polarization, optical pumping creates a constant
545: polarization rate for the nuclei with the help of hyperfine interaction; and
546: the relaxation process due to carrier recombination overshadows the effect of
547: quadrupole interaction, so that As polarization can still be built up
548: \cite{OO}.
549:  
550: In conclusion, we have studied the dynamical nuclear polarization mechanism in
551: spin-blockade double-dot system.  Specifically, we calculate the hyperfine
552: induced polarization rates and depolarization rates among the eigenstates of
553: the nuclear spins.  We show that the average spin polarization in steady
554: state of different nuclear species (Ga and As) could differ by two orders of
555: magnitude due to static quadrupole interaction induced by lattice
556: distortions.  Our results can thus be used to explain the recent NMR
557: experiments conducted in such system.  Our calculation suggests both caution
558: and promise.  Doping is widely used in semiconductors to tailor electronic
559: properties.  However, as we point out in this Letter, there can be unexpected
560: side effects in material properties, such as increased nuclear spin
561: relaxation in the present case.  Conversely, controlled doping can also be
562: used to differentiate parts of a system (such as Ga and As nuclei here), so
563: that selective operations become possible.  How to utilize the additional
564: control provided by doping while overcoming its negative effects can be
565: critical to future quantum information processors and/or spintronic devices.
566: 
567: We acknowledge helpful conversations with K. Ono, S. Tarucha, and S. Das
568: Sarma.  We thank partial financial support by ARDA and ARO.
569: 
570: \begin{thebibliography}{99}
571: 
572: \bibitem{OO} F. Meier, and B.P. Zakharchenya, {\em Optical Orientation}
573: (North-Holland, Amsterdam, 1984).
574: 
575: \bibitem{Gammon-science} D. Gammon {\it et al.}, 
576: %S.W. Brown, E.S. Snow, T.A. Kennedy, D.S. Katzer, and D. Park, 
577: Science {\bf 277}, 85 (1997).
578: 
579: %\bibitem{Gammon} D. Gammon {\em et al.}, Phys. Rev. Lett. {\bf 86}, 5176
580: %(2001).
581: 
582: \bibitem{Kawa} R.K. Kawakami {\em et al.}, Science {\bf 294}, 131 (2001). 
583: 
584: \bibitem{decoherence} A.V. Khaetskii and Yu.V. Nazarov, Phys. Rev. B {\bf 64},
585: 125316 (2001); J. Schliemann, A.V. Khaetskii, and D. Loss, {\it ibid.} {\bf
586: 66}, 245303 (2002); R. de Sousa and S. Das Sarma, {\it ibid.} {\bf 67},
587: 033301 (2003); {\it ibid.} {\bf 68}, 115322 (2003).
588: 
589: \bibitem{Zoller} A. Imamo\={g}lu, E. Knill, L. Tian, and P. Zoller,
590: Phy. Rev. Lett. {\bf 91}, 017402 (2003).
591: 
592: \bibitem{SB-NS} K. Ono and S. Tarucha, cond-mat/0309062.
593: 
594: \bibitem{SB} K. Ono, D.G. Austing, Y. Tokura, and S. Tarucha, Science {\bf
595: 297}, 1313 (2002).
596: 
597: \bibitem{QD-rev} L.P. Kouwenhoven, D.G. Austing, and S. Tarucha, Rep. Prog.
598: Phys. {\bf 64}, 701 (2001).
599: 
600: \bibitem{Quad0} E.H. Rhoderick, J. Phys. Chem. Solids {\bf 8}, 498 (1958).
601: 
602: \bibitem{Quad} W.E. Carlos, S.G. Bishop, and D.J. Treacy, Phy. Rev. B {\bf
603: 43}, 12512 (1991); 
604: %W.E. Carlos, S.G. Bishop, and D.J. Treacy, 
605: Appl. Phy. Lett. {\bf 49}, 528 (1986).
606: 
607: \bibitem{GaAs_no_quad} The crystal structure of {\it pure} GaAs is
608: zinc-blende with cubic symmetry.  No noticeable field gradient is
609: expected due to this symmetry.
610: 
611: \bibitem{spin-lattice} J.V. Kranendonk, Physica {\bf 20}, 781 (1954).
612: 
613: \bibitem{NMR} A. Abragam, {\em The Principles of Nuclear Magnetism}
614: (Oxford University Press, London, 1961); C.P. Slichter, {\em Principles of
615: Magnetic Resonance} (Spinger-Verlag, Berlin, 1996).
616: 
617: %\bibitem{nutation} S.I. Erlingsson, O.N. Jouravlev, and Y.V. Nazarov,
618: %cond-mat/0309069.
619: 
620: \bibitem{Paget} D. Paget, G. Lampel, B. Sapoval, and V.I. Safarov, Phys. Rev.
621: B {\bf 15}, 5780 (1977).
622: 
623: \bibitem{NSR} Y.B. Lyanda-Geller, I.L. Aleiner, and B.L. Altshuler, Phys. Rev.
624: Lett. {\bf 89}, 107602 (2002).
625: 
626: \bibitem{note_ST} We have assumed that the electron g-factor is negative and
627: the triplet is the ground state at zero magnetic field. One could equivalently
628: discuss the other triplet state $|1,1 \rangle = |\uparrow \rangle |\uparrow
629: \rangle $ if the singlet is the ground state when there is no magnetic field
630: \cite{Jpn}.
631: 
632: \bibitem{Jpn} T. Inoshita, K. Ono, and S. Tarucha, J. Phys. Soc. Jpn. {\bf 72}
633: 183 (2003).
634: 
635: 
636: %\bibitem{note_DOS} We have evaluated the exact form of DOS in this case.
637: %$\rho_S(\epsilon) = \frac{\Gamma}{2\pi} \frac{(\epsilon - \epsilon_0)^2 +
638: %t_c^2 + \Gamma^2/4}
639: %{\left [(\epsilon-\epsilon_0)^2 - t_c^2 - \Gamma^2/4 \right ]^2 + 
640: %\Gamma^2(\epsilon-\epsilon_0)^2}$. $t_c$ is the inter-dot coupling
641: %coefficient and 
642: %a symmetrical double-dot is considered for simplicity. When $t_c$ (a few tens
643: %$\mu$eV) 
644: %is small $\rho_S \approx \rho_T$.
645: 
646: %\bibitem{chap5} V.G. Fleisher and I.A. Merkulov, Chapter 5 in Ref. \cite{OO}. 
647: 
648: %\bibitem{spin-lattice} J.A. McNeil, and W.G. Clark, Phys. Rev. B {\bf 13}
649: %4705 (1976).
650: 
651: \end{thebibliography}
652: 
653: \end{document}
654: