1: \documentclass[twocolumn,aps,prb,final,amsfonts,amssymb,
2: amsmath,floatfix,showpacs,superscriptaddress]{revtex4}
3:
4: \usepackage[]{graphicx}
5: \usepackage{bm}
6:
7: \newcommand{\etal}{\textit{et al.}}
8: \newcommand{\eg}{e.g., }
9:
10: \begin{document}
11:
12: \title{Millisecond-range electron spin memory in singly-charged InP
13: quantum dots}
14: \author{Bipul~Pal}
15: \email[E-mail: ]{bipulpal@sakura.cc.tsukuba.ac.jp}
16: \affiliation{Institute of Physics, University of Tsukuba, Tsukuba 305-8571,
17: Japan}
18: \author{Michio~Ikezawa}
19: \affiliation{Institute of Physics, University of Tsukuba, Tsukuba 305-8571,
20: Japan}
21: \author{Yasuaki~Masumoto}
22: \affiliation{Institute of Physics, University of Tsukuba, Tsukuba 305-8571,
23: Japan}
24: \author{Ivan~V.~Ignatiev}
25: \affiliation{Institute of Physics, University of Tsukuba, Tsukuba 305-8571,
26: Japan}
27: \affiliation{Institute of Physics, St. Petersburg State University,
28: St.-Petersburg 198504, Russia}
29: \date{\today}
30:
31: \begin{abstract}
32: We report millisecond-range spin memory of resident electrons in an
33: ensemble of InP quantum dots (QDs) under a small magnetic field of $0.1$~T
34: applied along the optical excitation axis at temperatures up to about 5~K.
35: A pump-probe photoluminescence (PL) technique is used for optical
36: orientation of electron spins by the pump pulses and for study of spin
37: relaxation over the long time scale by measuring the degree of circular
38: polarization of the probe PL as a function of pump-probe delay. Dependence
39: of spin decay rate on magnetic field and temperature suggests two-phonon
40: processes as the dominant spin relaxation mechanism in this QDs at low
41: temperatures.
42: \end{abstract}
43:
44: \pacs{78.67.Hc, 78.55.Cr, 78.47.+p, 72.25.Rb}
45:
46: \maketitle
47:
48: Electron spin in semiconductors may be suitable for use as a quantum memory
49: in quantum repeaters as semiconductors are capable of converting photons to
50: electrons (and holes) while transferring quantum information from photon
51: polarization to electron spin, and vice versa.~\cite{awschalombook} However,
52: a long spin relaxation time ($\tau_{s}$) is necessary to realize spin
53: quantum memory. In semiconductor quantum dots (QDs) $\tau_{s}$ up to a few
54: ms was theoretically predicted as a result of suppression of important spin
55: relaxation mechanisms due to 3D confinement and lack of energy
56: dispersion.~\cite{khaetskiiprb61,khaetskiiprb64,woodsprb66} This has
57: stimulated experimental investigations of long spin memory in
58: QDs.~\cite{cortezprl89,fujisawanature419,elzermannature430,%
59: kroutvarnature432,laurentphe20,coltonprb69,ikezawaprbsub}
60: For example, electron spin relaxation time of about 200~$\mu$s in InGaAs
61: vertical QDs~\cite{fujisawanature419} at $T \le 0.5$~K and zero external
62: magnetic field ($B$), about 0.85~ms in electrically gated GaAs lateral
63: QDs~\cite{elzermannature430} at $T<0.3$~K and $B=8$~T, and about 20~ms
64: in self-assembled InGaAs QDs~\cite{kroutvarnature432} at $T=1$~K and
65: $B=4$~T has been recently reported. Many of the previous results of long
66: spin memory in QDs were obtained in InGaAs and GaAs QDs under special
67: condition of low temperature ($\le 1$~K) and large magnetic field
68: ($\ge 4$~T).
69:
70: In this letter we report observation of optically created electron
71: spin-orientation surviving up to about 1~ms in an ensemble of singly
72: negatively charged InP QDs at $B=0.1$~T applied along the optical
73: excitation axis at $T \sim 5$~K. The sample consists of a single layer
74: of self-assembled InP QDs embedded between GaInP barriers. The average
75: base diameter (height) of the QDs is about 40~(5)~nm with an areal density
76: $\sim 10^{10}$~cm$^{-2}$. We use a pump-probe PL
77: technique~\cite{cortezprl89,coltonprb69,ikezawaprbsub} to study electron
78: spin orientation dynamics by measuring the circular polarization [defined
79: as $P=(I^{++}-I^{+-})/(I^{++}+I^{+-})$, where $I^{++(-)}$ is the PL
80: intensity for excitation with $\sigma^{+}$ probe and detection of
81: $\sigma^{+(-)}$ probe PL] of the probe pulse PL in presence of a
82: preexcitation by a pump pulse. Our experimental setup is schematically
83: shown in Fig.~\ref{setup}(a).
84:
85: A CW Ti:sapphire laser beam is split into pump and probe beams. Two
86: acousto-optic modulators (AOM) driven by programmable function generator
87: (PFG) generates pump and probe pulses with controllable
88: pulse width and delay ($\tau$) between them.
89: Glan-Thompson polarizers (GTP) and wave plates are used to control the
90: circular polarization of the pump and probe beams.
91: The PL signal is sent through a combination of a photo-elastic
92: modulator (PEM) and a GTP before dispersing in a monochromator and
93: detecting in a GaAs photomultiplier tube (PMT). The PMT output is connected
94: to a two-channel gated photon counter (GPC). The PEM acts as an oscillating
95: $\lambda/4$-plate and when combined with GTP, allows detection of PL
96: intensity in the $\sigma^{+}$ and $\sigma^{-}$ channels. The PEM frequency,
97: $f_{\text{P}}=42$~kHz, is reduced to $f_{\text{T}}=f_{\text{P}}/(n+0.5)$
98: (typically $n=40$ in our measurements) to trigger the PFG and
99: GPC.~\cite{kalevich} Thus, one probe pulse (and A-gate of the GPC) is
100: centered at the $+\lambda/4$ and the next probe pulse (and B-gate of the
101: GPC) is centered at the $-\lambda/4$ retardation peaks of the PEM
102: [Fig.~\ref{setup}(b)]. We typically use a GPC gate width of 5~$\mu$s,
103: while the pump (probe) pulse width is 60~(3)~$\mu$s, giving a pump (probe)
104: power density [$W_{\text{pump (probe)}}$] of about 0.5~(0.05)~W~cm$^{-2}$.
105: The low probe power density ensures that the pump-induced spin polarization
106: is not fully erased by the probe pulse.
107: The excitation energy is tuned to about 1.771~eV (below-barrier, QD excited
108: state excitation) and the QD ground state PL is detected at about 1.729~eV.
109: An external electric bias of $U_{\text{bias}}=-0.1$~V is applied to the sample.
110: We find that under this condition the PL polarization is negative and reaches
111: maximum.~\cite{ikezawaprbsub} A study of trionic quantum beats in this
112: sample~\cite{kozinprb65} showed that at $U_{\text{bias}} \approx -0.1$~V
113: each QD contains one resident electron on an average. This suggests that
114: the negative PL polarization arises from trionic state, as is discussed \eg
115: in Refs.~\onlinecite{cortezprl89,laurentphe20,kavokinpssa195,brackerprl94}.
116:
117: In our experiments, a $\sigma^{+}$ (or $\sigma^{-}$) polarized pump
118: induces $\downarrow$ (or $\uparrow$) spin orientation of the resident
119: electrons.~\cite{opticalorientation}
120: A probe pulse, variably delayed with respect to the
121: pump pulse, tests this pump-induced spin-orientation. The probe beam (always
122: $\sigma^{+}$ polarized) creates a hot trion with \emph{parallel}
123: [$\downarrow \! \downarrow$-QDs] or \emph{anti-parallel}
124: [$\uparrow \! \downarrow$-QDs] electron spins. After a flip-flop process in
125: $\downarrow \! \downarrow$-QDs shown schematically in Fig.~\ref{longspin}(a),
126: the probe PL polarization becomes negative,~\cite{ikezawaprbsub} while it
127: is positive for the $\uparrow \! \downarrow$-QDs [Fig.~\ref{longspin}(b)].
128: At any given $\tau$ the net probe PL polarization is determined by the ratio
129: of $\downarrow \! \downarrow$- and $\uparrow \! \downarrow$-QDs.
130:
131: We measure the probe PL polarization for (i) co-circularly polarized
132: pump-probe ($P_{\text{CO}}$) [pump creates more
133: $\downarrow \! \downarrow$-QDs] and (ii) cross-circularly polarized
134: pump-probe ($P_{\text{CR}}$) [pump creates more
135: $\uparrow \! \downarrow$-QDs] [pump and probe polarizations for the two
136: cases are indicated in Fig.~\ref{setup}(b)].
137: A small static magnetic field of $B=0.1$~T is applied parallel to the
138: optical excitation (and sample growth) axis to suppress the effect of
139: fluctuating nuclear magnetic field.~\cite{merkulovprb65,braunprl94}
140: Polarizations $P_{\text{CR}}$ and $P_{\text{CO}}$ as a function of
141: $\tau$ are shown in Fig.~\ref{longspin}(c).
142: The difference \mbox{$P_{\text{CR}}-P_{\text{CO}}$} is a good measure of
143: the pump induced spin orientation of the resident electrons.~\cite{foot1}
144: A semilogarithmic plot of \mbox{$P_{\text{CR}}-P_{\text{CO}}$} obtained
145: from Fig.~\ref{longspin}(c) shows that the spin memory decay is
146: nonexponential [Fig.~\ref{longspin}(d)]. Thus, a spin relaxation time
147: cannot be defined in a simple way. However, it is clear from this data
148: that the spin memory decays on a millisecond time-scale.
149:
150: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
151: \begin{figure}[htb]
152: \includegraphics[clip,width=6.5cm]{setup.eps}
153: \caption{\label{setup} Schematic of the experimental setup (a) and time
154: synchronization (b) of the PEM retardation, the probe pulses, and the
155: GPC gates.
156: }
157: \end{figure}
158: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
159:
160: The observed long-lived spin polarization could result from a dynamic
161: nuclear polarization which may appear under the experimental condition
162: used.~\cite{merkulovprb65,optorient,gammonprl86} However, in a recent
163: study~\cite{ikezawaprbsub} of this aspect we have shown that very small
164: effective magnetic field ($<0.02$~T) in InP QDs,~\cite{ignatievconf,%
165: dzhioevpss41} arising from dynamic nuclear polarization, is not
166: consistent with the large amplitude of PL polarization observed in this
167: sample. Thus, the long spin memory observed here should be related to
168: the lack of efficient spin decay path in QDs. To investigate the spin
169: relaxation mechanisms effective in this case we study temperature
170: and magnetic field dependence of the spin decay process.
171:
172: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
173: \begin{figure}[htb]
174: \includegraphics[clip,width=8.3cm]{leveldia.eps}
175: \caption{\label{longspin} Schematics of $\downarrow \! \downarrow$-QDs
176: (a) and $\uparrow \! \downarrow$-QDs (b). Probe
177: PL polarization for co- ($P_{\text{CO}}$) and cross- ($P_{\text{CR}}$)
178: circularly polarized pump-probe (c), and the difference
179: \mbox{$P_{\text{CR}}-P_{\text{CO}}$} (d) as a function of $\tau$.
180: }
181: \end{figure}
182: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
183:
184: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
185: \begin{figure}[htb]
186: \includegraphics[clip,width=6.6cm]{tdep.eps}
187: \caption{\label{tdep} Delay dependence of
188: \mbox{$P_{\text{CR}}-P_{\text{CO}}$} at a few $T$. $B=0.1$~T,
189: $U_{\text{bias}}=-0.1$~V, $W_{\text{pump (probe)}}=0.5$ (0.05)~W~cm$^{-2}$.
190: Dashed lines are stretched exponential fits (discussed in the text). Inset
191: shows spin decay rate ($\gamma_{s}$) as a function of $T$.
192: }
193: \end{figure}
194: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
195:
196: Figure~\ref{tdep} shows decay of \mbox{$P_{\text{CR}}-P_{\text{CO}}$}
197: at a few temperatures for $B=0.1$~T. A faster decay is seen with
198: increasing $T$. As noted earlier, the decay is nonexponential and
199: suggests a distribution of decay rates, which may arise due to
200: inhomogeneous environment and size-distribution of the
201: QDs.~\cite{ikezawaprbsub} Theoretical
202: analysis shows (see \eg Ref.~\onlinecite{chenlum102})
203: that a spread of the relaxation rate results in a nonexponential decay
204: of the form $\sim \exp[-(\gamma_{s} \tau)^{c}]$ (the so-called stretched
205: exponential function), where the parameter $c$ depends on the
206: physical processes causing the spread. We find that the function fits
207: our data very well (dashed lines in Fig.~\ref{tdep}) if we use $c$ as a
208: fitting parameter. Effective spin decay rate $\gamma_{s}$ obtained from
209: such fits is plotted in the inset of Fig.~\ref{tdep} as a function of $T$.
210: A rapid increase in $\gamma_{s}$ is seen for $T>8$~K. Such an increase is
211: expected for thermally activated spin relaxation due to the phonon-mediated
212: coupling of the ground and excited electron states (two-phonon Orbach
213: process).~\cite{abragambook} We find that the function
214: $\gamma_{s} \sim (\exp[\Delta E/k_{B}T]-1)^{-1}+\gamma_{0}$
215: ($\Delta E =$~activation energy, $k_{B} =$~Boltzmann constant, and
216: $\gamma_{0}$ stands for spin decay rate
217: arising from temperature independent relaxation mechanisms) describing this
218: process fits the data very well (solid line in the inset of Fig.~\ref{tdep}).
219: From the fit we obtain $\Delta E \approx 5$~meV. This value is smaller than
220: that obtained experimentally for electron level spacing of 15 meV in
221: Ref.~\onlinecite{kozinprb65}. This discrepancy is probably due to
222: the difference in QD sub-ensemble probed in the two cases.
223:
224: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
225: \begin{figure}[htb]
226: %\vspace{0.4 cm}
227: \includegraphics[clip,width=6.6cm]{bdep.eps}
228: \caption{\label{bdep} Delay dependence of
229: \mbox{$P_{\text{CR}}-P_{\text{CO}}$} at a few $B$. $T=2$~K,
230: $U_{\text{bias}}=-0.1$~V, $W_{\text{pump (probe)}}=0.5$ (0.05)~W~cm$^{-2}$.
231: Dashed lines are stretched exponential fits (discussed in the text). Inset
232: shows spin decay rate ($\gamma_{s}$) as a function of $B$.
233: }
234: \end{figure}
235: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
236:
237: We now present the magnetic field dependence of the spin decay
238: process. Decay of \mbox{$P_{\text{CR}}-P_{\text{CO}}$} at
239: a few values of $B$ at $T=2$~K is shown in Fig.~\ref{bdep}. We find
240: that the decay becomes increasingly faster with increase in $B$. An
241: effective spin decay rate obtained from stretched exponential fit to
242: the data is plotted as a function of $B$ in the inset of Fig.~\ref{bdep}.
243: Decay rate $\gamma_{s}$ is found to increase superlinearly with $B$.
244: Several possible mechanisms for such an increase are discussed in the
245: literature.~\cite{khaetskiiprb64,woodsprb66,erlingssonprb66} Magnetic
246: filed couples the higher energy states with nonzero orbital momentum to
247: the electron spin states split by the magnetic field (Zeeman splitting)
248: that allows a small admixture of the states of opposite spin to each Zeeman
249: sublevel.~\cite{khaetskiiprb64,woodsprb66} At low temperature this
250: enables spin-flip transition between Zeeman sublevels via participation
251: of acoustic phonons to dissipate energy (one-phonon resonant process).
252: With increasing $B$ the Zeeman splitting increases. Due
253: to higher density of resonant phonons at increased energy and more
254: efficient mixture of the states by the magnetic field, the spin relaxation
255: rate increases. Theoretical calculations~\cite{khaetskiiprb64,woodsprb66}
256: have predicted $\gamma_{s} \sim B^{5}$ at very low temperature and large
257: magnetic field if the spin-orbit interaction and the one-phonon scattering
258: dominate. However, for $T$ of about a few kelvin, the two-phonon nonresonant
259: (Raman) scattering may become important.~\cite{woodsprb66} In that case,
260: the magnetic field dependence is only determined by the admixture of the
261: excited states and becomes quadratic.~\cite{khaetskiiprb64} Our data in
262: Fig.~\ref{bdep}-inset can be fitted very well with
263: $\gamma_{s}=\alpha +\beta B^{2}$. This argues for the two-phonon scattering
264: as the main mechanism of acceleration of the spin relaxation in magnetic
265: field.
266:
267: The acceleration of spin relaxation could also result from hyperfine
268: interaction.~\cite{erlingssonprb66} However, this is unlikely in our case
269: due to very small nuclear spin polarization in the InP
270: QDs~\cite{ignatievconf,dzhioevpss41} we studied.
271:
272: In conclusion, we have observed long spin memory, persisting over 1~ms,
273: in an ensemble of singly negatively charged InP QDs at small magnetic
274: field (0.1~T) and at moderate temperature ($\sim 5$~K). Our data on the
275: magnetic field and temperature dependence of spin decay rate suggests
276: two-phonon scattering may be the dominant spin relaxation mechanism.
277: Long spin memory observed here in III-V semiconductor QDs is relevant
278: for quantum information communication and storage. Though our study is
279: made at about 0.7~$\mu$m wavelength ($\lambda$), III-V semiconductor
280: system can be easily adapted to $\lambda=1.3$ and 1.5~$\mu$m, suitable
281: for fiber optic communication.
282:
283: Authors thank I.~Ya.~Gerlovin and T.~Takagahara for fruitful discussions.
284: B.~Pal thanks INOUE Foundation for Science, Japan, for financial
285: support. This work is supported by the Grant-in-Aid for the Scientific
286: Research \#13852003 and \#16031203 from the MEXT, Japan.
287:
288: \begin{thebibliography}{99}
289: \bibitem{awschalombook} D.~D.~Awschalom \etal, \textit{Semiconductor
290: Spintronics and Quantum Computation},(Springer-Verlag, Berlin 2002).
291: \bibitem{khaetskiiprb61} V.~Khaetskii \etal, Phys. Rev. B \textbf{61},
292: 12639 (2000).
293: \bibitem{khaetskiiprb64} A.~V.~Khaetskii \etal, Phys. Rev. B \textbf{64},
294: 125316 (2001).
295: \bibitem{woodsprb66} L.~M.~Woods \etal, Phys. Rev. B \textbf{66},
296: 161318 (2002).
297: \bibitem{cortezprl89} S.~Cortez \etal, Phys. Rev. Lett. \textbf{89},
298: 207401 (2002).
299: \bibitem{fujisawanature419} T.~Fujisawa \etal, Nature \textbf{419},
300: 278 (2002).
301: \bibitem{elzermannature430} J.~M.~Elzerman \etal, Nature \textbf{430},
302: 431 (2004).
303: \bibitem{kroutvarnature432} M.~Kroutvar \etal, Nature \textbf{432},
304: 81 (2004).
305: \bibitem{laurentphe20} S.~Laurent \etal, Physica E \textbf{20},
306: 404 (2004).
307: \bibitem{coltonprb69} J.~S.~Colton \etal, Phys. Rev. B \textbf{69},
308: 121307(R) (2004).
309: \bibitem{ikezawaprbsub} M.~Ikezawa \etal, Phys. Rev. B (to be published).
310: \bibitem{kalevich} V.~K.~Kalevich \etal, Instrum. Exp. Techn. \textbf{21},
311: 199 (1978).
312: \bibitem{kozinprb65} I.~E.~Kozin \etal, Phys. Rev. B \textbf{65},
313: 241312(R) (2002).
314: \bibitem{kavokinpssa195} K.~V.~Kavokin, Phys. Status Solidi (a)
315: \textbf{195}, 592 (2003).
316: \bibitem{brackerprl94} A.~S.~Bracker \etal, Phys. Rev. Lett. \textbf{94},
317: 047402 (2005).
318: \bibitem{opticalorientation} Various mechanisms of optical orientation of
319: spins are discussed \eg in Refs.~\protect{\onlinecite{cortezprl89,%
320: brackerprl94,optorient}}. Possible mechanisms of optical orientation of
321: resident electron spins in InP QDs under study will be discussed elsewhere.
322: \bibitem{optorient} F.~Meier and B.~P.~Zakharchenya, \textit{Optical
323: Orientation} (North-Holland, Amsterdam, 1984).
324: \bibitem{merkulovprb65} I.~A.~Merkulov \etal, Phys. Rev. B \textbf{65},
325: 205309 (2002).
326: \bibitem{braunprl94} P.-F.~Braun \etal, Phys. Rev. Lett. \textbf{94},
327: 116601 (2005).
328: \bibitem{foot1} Effect of dynamic nuclear polarization (if present) is
329: expected to be the same for $P_{\text{CR}}$ and $P_{\text{CO}}$ and taking
330: their difference may eliminate the influence of dynamic nuclear
331: polarization in the range of our measurement. A direct relation between
332: \mbox{$P_{\text{CR}}-P_{\text{CO}}$} and the spin-orientation of resident
333: electrons will be discussed elsewhere.
334: \bibitem{gammonprl86} D. Gammon \etal, Phys. Rev. Lett. \textbf{86},
335: 5176 (2001).
336: \bibitem{ignatievconf} I.~V.~Ignatiev \etal, Proc. $13^{\text{th}}$ Int.
337: Symp. Nanostructures: Physics and Technology, St. Petersburg, Russia, 2005;
338: also available on-line at lanl.arXiv.org: cond-mat/0508698.
339: \bibitem{dzhioevpss41} R.~I.~Dzhioev \etal, Phys. Solid State
340: \textbf{41}, 2014 (1999).
341: \bibitem{chenlum102} R.~Chen, J. Lumin. \textbf{102-103}, 510
342: (2003) and Refs. therein.
343: \bibitem{abragambook} A. Abragam and B. Bleaney, \textit{Electron
344: Paramagnetic Resonance of Transition Ions}, (Clarendon Press, Oxford, 1970).
345: \bibitem{erlingssonprb66} S.~I.~Erlingsson \etal, Phys. Rev. B
346: \textbf{66}, 155327 (2002).
347: \end{thebibliography}
348: \end{document}
349: