1: % Near-field Spectra of Quantum Well excitons
2: % with non-Markovian phonon scattering
3: % G. Mannarini, R. Zimmermann, revised January 2006
4: %----------------------------------------------------------------------
5:
6: % See the REVTeX 4 README file
7: % It also requires running BibTeX. The commands are as follows:
8: %
9: % 1) latex apssamp.tex
10: % 2) bibtex apssamp% 4) latex apssamp.tex
11: %
12: %
13: %##############################################################################
14: %\documentclass[11pt]{article}
15: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb,prb]{revtex4}
16: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb,prb]{revtex4}
17: %##############################################################################
18:
19: \usepackage{graphicx}% Include figure files
20: \usepackage{dcolumn}% Align table columns on decimal point
21: \usepackage{bm}% bold math
22: %\usepackage{feyn}% Feynman Diagrams
23:
24: % Hyperref:
25: %\usepackage[%
26: % pdftitle={Near-field Spectra of Quantum Well excitons with non-Markovian phonon scattering},
27: % pdfauthor={Mannarini, Gianandrea},
28: % pdfsubject={PRB},
29: % pdfkeywords,
30: % pdfpagemode=UseOutlines,
31: % pdftex=true, % bitte nicht ändern!
32: % plainpages=false, % bitte nicht ändern!
33: % hypertexnames=false, % bitte nicht ändern!
34: % pdfpagelabels=true, % bitte nicht ändern!
35: % hyperindex=true]{hyperref} % bitte nicht ändern!
36:
37: %\hypersetup{colorlinks=true, linkcolor=black, filecolor=black,
38: %urlcolor=black, citecolor=black}
39:
40: %\include{/home/mannarini/LaTeX_commands/command} % G.Mannarini's LaTeX commands
41: %\include{command}
42: % Author's shortcuts:
43: \newcommand{\be}{\begin{equation}}
44: \newcommand{\ee}{\end{equation}}
45: \newcommand{\bea}{\begin{eqnarray}}
46: \newcommand{\bean}{\begin{eqnarray}\nonumber}
47: \newcommand{\eea}{\end{eqnarray}}
48: \newcommand{\Eq}[1]{Eq.\,(\ref{#1})}
49: \newcommand{\Sect}[1]{Sect.\,(\ref{#1})}
50: \newcommand{\App}[1] {App.\,(\ref{#1})}
51: \newcommand{\Fig}[1]{Fig.\,\ref{#1}}
52: \newcommand{\Tab}[1]{Tab.\,\ref{#1}}
53: \newcommand{\bfi}{\begin{figure}}
54: \newcommand{\efi}{\end{figure}}
55: \newcommand{\bmi}{\begin{minipage}}
56: \newcommand{\emi}{\end{minipage}}
57: \newcommand{\la}{\left<}
58: \newcommand{\ra}{\right>}
59: \newcommand{\half}{\frac{1}{2}}
60: \newcommand{\ds}{\displaystyle}
61: \newcommand{\E}{{\cal E}}
62: \newcommand{\cG}{{\cal G}}
63: \newcommand{\e}{{\rm e}}
64: \newcommand{\K}{\mathbf{K}}
65: \newcommand{\ko}{\mathbf{k}_{0}}
66: \newcommand{\Q}{{\mathbf Q}}
67: \newcommand{\q}{{\mathbf q}}
68: \newcommand{\qp}{{\q_\parallel}}
69: \newcommand{\kp}{{k_\parallel}}
70: \newcommand{\kvecp}{{\mathbf{k}_\parallel}}
71: \newcommand{\R}{{\mathbf R}}
72: \newcommand{\Ra}{{\mathbf R_\alpha}}
73: \newcommand{\Rb}{{\mathbf R_\beta}}
74: \newcommand{\Rc}{{\mathbf R_\gamma}}
75: \newcommand{\Rd}{{\mathbf R_\delta}}
76: \newcommand{\re}{{\mathbf r}_e}
77: \newcommand{\rh}{{\mathbf r}_h}
78: \newcommand{\ro}{{\mbox{\boldmath $\rho$}}}
79: \newcommand{\roe}{{\mbox{\boldmath $\rho$}}_e}
80: \newcommand{\roh}{{\mbox{\boldmath $\rho$}}_h}
81: \newcommand{\muvec}{{\mbox{\boldmath $\mu$}}}
82: \newcommand{\chivec}{{\mbox{\boldmath $\chi$}}}
83: \newcommand{\nablavec}{{\mbox{\boldmath $\nabla$}}}
84: \newcommand{\Lz}{\overline{L_z}}
85: \newcommand{\DeltaR}{\Delta_{\mathbf R}}
86: \newcommand{\aB}{a_{\rm B}}
87: \newcommand{\om}{\omega}
88: \newcommand{\Om}{\Omega}
89: \newcommand{\wq}{w_{\mathbf q}}
90: \newcommand{\wk}{w_{\mathbf k}}
91: \newcommand{\Omq}{\Omega_{\mathbf q}}
92: \newcommand{\Omk}{\Omega_{\mathbf k}}
93: \newcommand{\DS}{\Delta_S}
94: \newcommand{\Dsh}{\Delta s/2}
95: \newcommand{\Dw}{\Delta\omega}
96: \newcommand{\Dwh}{\Delta\omega/2}
97: \newcommand{\eone}{\epsilon_1}
98: \newcommand{\ea}{\epsilon_{\alpha}}
99: \newcommand{\eaz}{\epsilon_{\alpha}^{(0)}}
100: \newcommand{\eb}{\epsilon_{\beta}}
101: \newcommand{\ebz}{\epsilon_{\beta}^{(0)}}
102: \newcommand{\eg}{\epsilon_{\gamma}}
103: \newcommand{\eeta}{\epsilon_{\eta}}
104: \newcommand{\etheta}{\epsilon_{\theta}}
105: \newcommand{\Ta}{\theta_{\alpha}}
106: \newcommand{\Da}{\Delta_{\alpha}}
107: \newcommand{\Db}{\Delta_{\beta}}
108: \newcommand{\psia}{\psi_\alpha(\mathbf{R})}
109: \newcommand{\psib}{\psi_\beta(\mathbf{R})}
110: \newcommand{\oma}{\omega_\alpha}
111: \newcommand{\omb}{\omega_\beta}
112: \newcommand{\mao}{m_{\alpha0}}
113: \newcommand{\mbo}{m_{\beta0}}
114: \newcommand{\mak}{m_{\alpha\k}}
115: \newcommand{\mako}{m_{\alpha\k_0}}
116: \newcommand{\maq}{m_{\alpha\q}}
117: \newcommand{\maqo}{m_{\alpha\q_0}}
118: \newcommand{\mbk}{m_{\beta\k}}
119: \newcommand{\Ma}{M_{\alpha}}
120: \newcommand{\Mb}{M_{\beta}}
121: \newcommand{\Na}{N_\alpha}
122: \newcommand{\Nb}{N_\beta}
123: \newcommand{\Naa}{N_{\alpha\alpha}}
124: \newcommand{\Nab}{N_{\alpha\beta}}
125: \newcommand{\Daa}{D_{\alpha\alpha}}
126: \newcommand{\Dab}{D_{\alpha\beta}}
127: \newcommand{\Pa}{P_\alpha}
128: \newcommand{\Pb}{P_\beta}
129: \newcommand{\g}{\mathbf{g}}
130: \newcommand{\ga}{\gamma_{\alpha}}
131: \newcommand{\Ga}{\Gamma_{\alpha}}
132: \newcommand{\Gb}{\Gamma_{\beta}}
133: \newcommand{\Go}{\Gamma_0(\omega)}
134: \newcommand{\hatTaq}{\hat{T}_{\alpha \q}}
135: \newcommand{\tildeTaq}{\tilde{T}_{\alpha \q}}
136: \newcommand{\hatTbq}{\hat{T}_{\beta \q}}
137: \newcommand{\tildeTbq}{\tilde{T}_{\beta \q}}
138: \newcommand{\tauh}{\tau/2}
139: \newcommand{\Dkh}{\Delta {\mathbf k}/2}
140: \newcommand{\Ba}{B^{\vphantom {\dagger }}_{\alpha}}
141: \newcommand{\Bad}{B^{\dagger}_{\alpha}}
142: \newcommand{\Bb}{B^{\vphantom {\dagger }}_{\beta}}
143: \newcommand{\Bbd}{B^{\dagger}_{\beta}}
144: \newcommand{\tabq}{t_{\alpha \beta}^{\q}}
145: \newcommand{\tabk}{t_{\alpha \beta}^{\k}}
146: \newcommand{\etal}{{\em et al.\,}}
147: \newcommand{\inte}{\int\!}
148: \newcommand{\Za}{Z_\alpha}
149: \newcommand{\Zb}{Z_\beta}
150: \newcommand{\ZR}{Z(\R)}
151: \newcommand{\Zaa}{Z_{\alpha\alpha}}
152: \newcommand{\Zab}{Z_{\alpha\beta}}
153: \renewcommand{\dag}{^{\dagger}}
154: \renewcommand{\Im}{\mbox{\sl Im\,}}
155: \renewcommand{\Re}{\mbox{\sl Re\,}}
156: \renewcommand{\k}{{\mathbf k}}
157: \renewcommand{\r}{{\mathbf r}}
158: \renewcommand{\H}{{\mathbf H}}
159: \renewcommand{\hom}{\hbar \omega}
160: % end of Author's shortcuts:
161:
162:
163: \begin{document}
164:
165: %\preprint{APS/123-QED}
166:
167: \title{Near-field spectra of quantum well excitons \\
168: with non-Markovian phonon scattering}
169:
170: \author{G. Mannarini}
171: \altaffiliation[Also at ]{Institut f\"ur Physik der Humboldt-Universit\"at
172: zu Berlin, Newtonstr. 15, 12489 Berlin, Germany}
173: \affiliation{National Nanotechnology Laboratory of CNR, via
174: Arnesano 16, 73100 Lecce, Italy}
175: \email{Gianandrea.Mannarini@unile.it}
176: \author{R. Zimmermann}%
177: \affiliation{Institut f\"ur Physik der Humboldt-Universit\"at
178: zu Berlin, Newtonstr. 15, 12489 Berlin, Germany}%
179:
180:
181: \date{\today}
182:
183: \begin{abstract}
184: The excitonic absorption spectrum for a disordered quantum well
185: in presence of exciton-acoustic phonon interaction is treated
186: beyond the Markov approximation. Realistic disorder exciton
187: states are taken from a microscopic simulation, and the
188: deformation potential interaction is implemented. The exciton
189: Green's function is solved with a self energy in second order
190: Born approximation. The calculated spectra differ from a superposition
191: of Lorentzian lineshapes by enhanced inter-peak absorption. This
192: is a manifestation of pure dephasing which should be possible to
193: measure in near-field experiments.
194: \end{abstract}
195:
196: \pacs{78.20.-e, 63.20.Ls, 73.43.Cd}% PACS, the Physics and Astronomy
197: % Classification Scheme.
198: %\keywords{Suggested keywords}%Use showkeys class option if keyword
199: %display desired
200: \maketitle
201: %--------------------------------------------------------------------------
202:
203:
204:
205:
206: \section{\label{sec:introduction} Introduction}
207:
208: Unique information on semiconductor nanostructures even beyond the
209: diffraction limit can be
210: obtained using near-field spectroscopy. This opens the way to detect fine
211: spectroscopic and structural details \cite{runge05}. Usual
212: selection rules for optical transitions are broken, which allows one
213: to observe also states which are dark in the far-field
214: \cite{hohenester05}. The present record experiment is a spatial
215: resolution of $30$\, nm which was achieved using a tapered
216: optical fiber and an InAs self-assembled quantum dot (QD) sample
217: \cite{matsuda02}. Even with a less demanding setup using a resolution
218: comparable to the wavelength of light, individual optical transitions
219: can be seen due to their spatial and spectral position.
220: Absorption spectra from quantum wells (QW)
221: with interface fluctuations could be measured in transmission
222: geometry through a metal aperture of $400$\, nm on a GaAs/AlGaAs
223: sample whose substrate was removed \cite{guest02}. This gives
224: access to single exciton state absorption lineshapes, with the
225: chance to detect deviations from a Lorentzian profile. Similar
226: lineshapes are seen in photoluminescence (PL). Among others,
227: Besombes \etal \cite{besombes01} found that the PL from a single
228: CdTe QD exhibits broad bands on the sides of the Lorentzian peak.
229: This has been also seen in Fourier-transformed four wave mixing
230: experiments by Borri \etal \cite{borri01,borri05}, who
231: systematically studied the importance of the broad band with
232: respect to the (Lorentzian broadened) zero phonon line (ZPL) as a
233: function of confinement energy in InAs/InGaAs QDs.
234:
235: However, using the standard Fermi's golden rule for the phonon
236: scattering, only a Lorentzian broadened ZPL is expected for the
237: absorption spectrum of a single exciton state. Phenomena which
238: go beyond the framework of Fermi's golden rule (such as the phonon
239: broad band) are often called quantum kinetic or non-Markovian
240: (i.e. memory) effects. The broad band can be understood in terms
241: of a non-perturbative coupling between electronic and phonon
242: degrees of freedom. This goes back to the Huang-Rhys theory of
243: $F$-centers \cite{duke65}, which considers intra-state (diagonal)
244: coupling to phonons only. This problem can be solved exactly with
245: the cumulant expansion (also known as Independent Boson Model
246: \cite{mahan}). Since off-diagonal level coupling is absent, the
247: ZPL remains unbroadened. Within this model, different
248: exciton-phonon couplings such as optical polar, acoustic
249: piezoelectric, and deformation potential have been studied
250: \cite{bondarev03}. This model was recently extended to include
251: off-diagonal coupling, successfully explaining why the ZPL get
252: broadened even if real phonon-assisted transitions to other
253: levels are not possible (pure dephasing)
254: \cite{muljarov04,muljarov05}. Apart from the cumulant expansion,
255: approaches based on various dynamical approximations have been
256: used, too. For instance, a 4th order selfconsistent summation of
257: diagrams allowed one to compare QW linewidths extracted from photon
258: echo measurements \cite{fan98, tak99}: It was found that at
259: elevated temperatures the linewidth broadening is mainly due to
260: pure dephasing. However, a simple quantum disk model for
261: computing the exciton wave functions was used in this case. The
262: correlation expansion for the carrier-phonon interaction which
263: was used in Ref.\onlinecite{krum02} to model the coupling to both
264: optical and acoustical phonons, is exact up to the 2nd order.
265: However, neither Coulomb or disorder effects were considered
266: there.
267:
268: A consistent theoretical treatment which describes non-Markovian
269: effects for {\em realistic}, disordered excitonic resonances from
270: semiconductor QWs is, to the best of our knowledge, still
271: missing. In this paper we face this problem accounting for both
272: diagonal and off-diagonal coupling to acoustic phonons via
273: deformation potential interaction. The excitonic self energy is
274: taken in second Born (2B) approximation and integrated numerically.
275: This approach reproduces both the real scattering processes
276: leading to the major ZPL broadening and the level-diagonal pure
277: dephasing responsible for the broad band. The resulting
278: (near-field) absorption spectrum exhibits apart from the
279: Lorentzian-broadened individual lines a superposition of broad
280: bands which results in an enhanced inter-peak absorption. We
281: suggest that this effect of non-Markovian dynamics can be
282: demonstrated in near-field measurements.
283:
284: The paper is organized as follows: The theory with a presentation
285: of different approximations for the self energy in
286: \Sect{sec:theory} is followed by numerical results in
287: \Sect{sec:results}, and the conclusions are given in
288: \Sect{sec:conclusions}. Finally, two appendixes contain more
289: technical details.
290:
291:
292: \section{\label{sec:theory} Theory}
293:
294: To model a realistic QW with disorder it is convenient
295: to start with disorder eigenstates of the exciton. If the
296: disorder strength is much smaller than both the level distance of
297: electron and hole confinement states and the exciton binding
298: energy, the total exciton wave function (in envelope function and
299: effective mass approximation) can be factorized as \cite{zim97}
300: %
301: \be \label{FactorWF}
302: \Psi_\alpha (\r_e,\r_h) =
303: u_e(z_e) \, u_h(z_h) \, \phi_{1s}(\roe-\roh) \, \psia \, ,
304: \ee
305: %
306: where confinement $u_a(z_a)$ in growth direction $z$, in-plane
307: exciton relative motion $\phi_{1s}(\ro)$, and in-plane
308: center-of-mass (COM) wavefunction $ \psia$ appear. Since the
309: first two functions are independent of disorder, they have to be
310: computed just once for the disorder-free QW. Then, it
311: suffices to solve a single particle equation for the COM exciton
312: states in two dimensions,
313: %
314: \be \label{COMmotion}
315: \left[ - \frac{\hbar^2}{2M} \DeltaR + V(\R) \right] \psia =
316: \ea \, \psia \, .
317: \ee
318: %
319: The potential $V(\R)$ accounts for interface disorder and is
320: spatially correlated on the scale of the exciton relative motion,
321: %
322: \be\label{V_def}
323: V(\R) =
324: \int \! d\R' \sum_{a=e,h} \eta_a^2 \, \phi_{1s}^2[\eta_a(\R-\R')] \,
325: \frac{dE_a}{dL_z} \, \Delta L_z(\R') \, ,
326: \ee
327: %
328: with mass ratios $\eta_e = M/m_h$ and $\eta_h = M/m_e$. The well
329: width fluctuations $\Delta L_z(\R)$ transfer into potential
330: fluctuations via the derivative of the ideal electron and hole
331: confinement energy $dE_a/dL_z$ \cite{zim_rev03}.
332: As zero of energy, we take the ideal 1s exciton transition energy
333: of the lowest sublevel transition hh1-e1 in a
334: GaAs/Al$_{0.3}$Ga$_{0.7}$As QW of given width $L_z$.
335:
336: Recently, Grochol \etal \cite{grochol05} have critically
337: questioned the factorization Ansatz Eq.(1). Calculations with the
338: full in-plane electron-hole wave function gave systematically
339: lower eigenergies, while the wave functions showed only minor
340: modifications compared to the factorization, at least in the
341: low-energy part of the spectrum. Since in the present paper, we
342: are interested in the low-energy tail of the absorption spectrum
343: with well localized states, the factorization can be safely used.
344:
345: The interaction with acoustic phonons is treated here as the main
346: source of scattering and dephasing. Optical phonon scattering
347: is not important here since, at temperatures below room
348: temperature, optical phonon absorption by the exciton states is
349: unlikely. Further, piezoelectric scattering is neglected since
350: the exciton wave functions are typically large in momentum space,
351: leading to small matrix elements for this kind of coupling
352: \cite{run_rev02, tak93}. Thus, only deformation potential
353: interaction with longitudinal acoustic phonons is relevant.
354:
355: This model leads to the following Hamilton operator for excitons
356: (creation operators $\Bad$) and acoustic phonons (creation
357: operators $a_\q\dag$ with three-dimensional momentum $\q$) with
358: (bulk like) dispersion $\wq=u_S\,|\q|$) :
359: %
360: \bean \label{ham} {\cal H} & = & \sum_\alpha \ea \, \Bad \Ba
361: \, + \, \sum_{\q} \hbar\wq \, a_\q\dag a_\q \\
362: & + & \sum_{\alpha \beta \q} \tabq (a_\q\dag + a_{-\q} ) \Bad \Bb \,.
363: \eea
364: %
365: The deformation potential matrix elements $\tabq$ are discussed
366: in \App{sec:app_me}. We have neglected the spin degree of freedom
367: which splits each localized exciton state into a bright doublet
368: via the exchange interaction. This splitting is driven by COM
369: anisotropy and has typical values of a few tens of $\mu$eV
370: \cite{zim_rev03}. It can be disregarded in calculating absorption
371: spectra with not too high resolution.
372:
373: Within linear response theory, the optical properties of the
374: system are given by the one-exciton Green's function (exciton
375: propagator)
376: %
377: \be G_{\alpha\beta}(t -t')= -i\langle{\cal T}\Ba (t) \Bbd(t') \rangle \ee
378: %
379: with time ordered finite-temperature expectation value. In
380: frequency space, the Green's function obeys the Dyson equation
381: (in what follows $\hbar=1$ is used, and the frequency $\omega$ is
382: understood to carry a small negative imaginary part, $\omega - i
383: 0$)
384: %
385: \be\label{lin_sys} \sum_{\eta} \left[ ( \omega -
386: \ea)\delta_{\alpha\eta} - \Sigma_{\alpha\eta}(\omega) \right]
387: G_{\eta\beta} (\omega) = \delta_{\alpha\beta} \, , \ee
388: %
389: introducing the exciton self energy matrix
390: $\Sigma_{\alpha\eta}(\omega)$. The absorption spectrum is
391: proportional to
392: %
393: \be \alpha(\omega) = \mathrm{Im}\sum_{\alpha \beta}m_\alpha^*\,
394: G_{\alpha\beta}(\omega)\, m_\beta \, . \ee
395: %
396: The optical interband matrix elements $m_\alpha$ are defined in
397: \App{sec:app_me}.
398:
399: For the self energy, we take the lowest order diagram for the
400: exciton-phonon interaction shown in \Fig{im:fig1}. It is called
401: 2nd order Born (2B) approximation since the interaction appears
402: just twice:
403: %
404: \be \label{self_basic} \Sigma_{\alpha\eta}(\omega) =
405: \sum_{\beta\q} \tabq t_{\beta \eta}^{-\q} \left[ \frac{1 +
406: n(\wq)}{\omega - \eb - \wq} + \frac{ n(\wq)}{\omega - \eb + \wq}
407: \right] \, . \ee
408: %
409: The phonons are assumed to be in equilibrium at the lattice
410: temperature $T$, $n(\wq)=1/(\mathrm{exp}(\wq/k_B T)-1)$. The two
411: terms in \Eq{self_basic} are related to phonon emission and
412: absorption. Note that the internal line in \Fig{im:fig1} is
413: understood to represent the bare exciton propagator. Taking
414: instead a dressed propagator with shifted quasiparticle energy
415: would be an improvement. Our calculations have shown that this
416: (acoustic) polaron shift for the localized exciton states is of
417: the order of a few $\mu$eV only, and we have chosen to neglect it
418: everywhere.
419:
420: An alternative derivation would start with the equations of
421: motion for the exciton propagator which couples to mixed
422: exciton-phonon operator expectation values (phonon-assisted
423: density matrix). The next equation in the hierarchy is then
424: decoupled in a way that two phonon operators can be combined into
425: the phonon occupation function. Eliminating the phonon assisted
426: density matrix from the equations leads to a differential
427: equation of the exciton propagator with a scattering being
428: essentially a time integration over the past. This kind of memory
429: in the scattering is fully equivalent to the frequency dependence
430: of the self energy in the diagrammatic approach. Further, the
431: non-diagonality of the self energy $\Sigma_{\alpha\eta}(\omega)$
432: leads to cross correlations among individual excitons.
433:
434: In the Markov approximation these correlations are neglected, but
435: more importantly the frequency dependence of the self energy is
436: dropped completely. In the time frame, this means to extract the
437: exciton propagator from the scattering integral at the latest
438: time, thus skipping any memory effect. Consequently, the temporal
439: dynamics of the propagator reduces to an exponential decay with a
440: constant damping $\gamma_\alpha^M$. In this paper we go beyond
441: this level of approximation, aiming at a full non-Markovian
442: description which includes the non-diagonality and the memory
443: (frequency dependence) in the self energy. For the sake of
444: comparison, we discuss in the following subsections the solution
445: of \Eq{lin_sys} at various levels of sophistication.
446:
447: The radiative contribution to the self energy (radiative damping)
448: will be neglected since it is tiny in thin QWs which have
449: strongly localized excitons \cite{zim_rev03}.
450:
451:
452: \subsection{Full non-Markovian level}\label{sec:theory_nm}
453:
454: The self energy \Eq{self_basic} can be written in a more compact
455: form by extending the energy integration to positive and negative
456: values,
457: %
458: \be \label{self_def} \Sigma_{\alpha\eta}(\omega) = \sum_\beta\int\!dE\,
459: n(E)\, \frac{J_{\alpha\eta}^\beta(E)}{\omega - \eb + E} \, ,
460: \ee
461: %
462: which accounts for both phonon absorption and emission. The
463: coupling function
464: %
465: \be \label{J_def} J_{\alpha\eta}^\beta(E) \equiv \mathrm{sgn}(E)
466: \sum_\q t_{\alpha\beta}^\q t_{\beta\eta}^{-\q} \, \delta(|E|
467: - \omega_\q) \,. \ee
468: %
469: can be viewed upon as a phonon density of states weighted by
470: coupling matrix elements. Its spectral range
471: decides which exciton states are coupled to the phonon bath. It
472: starts at least with a power of $E^3$ at small energies, as seen
473: in \App{sec:app_me}. Since the exciton wave functions are real valued, the
474: following symmetries hold: $J_{\alpha\eta}^\beta(E) =
475: J_{\eta\alpha}^\beta(E)$ and $J_{\alpha\alpha}^\beta(E) =
476: J_{\beta\beta}^\alpha(E)$. Thus, if we consider a system of $N$ exciton
477: states, $(N^3+N)/2$ components of the coupling function have to
478: be computed.
479: %
480: \bfi[t]
481: \centering
482: \includegraphics*[width=5cm]{fig1}
483: \caption{\label{im:fig1}Second order self energy diagram. The
484: arrow denotes the exciton propagator, while the wiggly line
485: stands for the phonon propagator. Each dot carries a coupling
486: element $t$.} \efi
487: %
488:
489: \subsection{Markov approximation}
490:
491: The Markov approximation for the exciton-phonon dynamics is
492: defined within our formalism by keeping only diagonal terms of
493: the self energy and evaluating them at the exciton resonance
494: (on-shell),
495: %
496: \be \label{M_self} \Sigma_{\alpha\eta}(\omega) \, \rightarrow \,
497: \delta_{\alpha\eta} \Sigma_\alpha^{M}\, , \quad \Sigma_\alpha^{M}
498: \equiv \Sigma_{\alpha\alpha} (\omega=\ea-i0) \,. \ee
499: %
500: Apart from the (small) polaron shift originating from the real
501: part, this level of approximation implies a constant damping
502: %
503: \be \label{markov_gamma} \gamma^M_\alpha/2 = \mathrm{Im}
504: \Sigma_\alpha^{M} = \pi \sum_{\beta} n(\eb-\ea) \,
505: J_{\alpha\alpha}^\beta(\eb-\ea) \, . \ee
506: %
507: (The factor 2 appears because $\gamma^M_\alpha$ is the decay rate
508: of the intensity, which is proportional to the squared exciton
509: propagator.) Thus, \Eq{lin_sys} reduces to a simple algebraic equation,
510: and the absorption comes out as
511: %
512: \bean
513: \alpha(\omega) &=& \sum_\alpha |m_\alpha|^2 \mathrm{Im}
514: \frac{1}{\omega - \ea - i\gamma^M_\alpha/2} \\
515: & = & \sum_\alpha |m_\alpha|^2 \frac{\gamma^M_\alpha/2}{(\omega
516: - \ea)^2 + (\gamma^M_\alpha/2)^2} \, , \eea
517: %,
518: which is a Lorentzian lineshape for each line. No broad bands are
519: obtained within the Markov approximation. The rate
520: $\gamma^M_\alpha$ is identical to the result of Fermi's golden
521: rule. We notice that within the Markov approximation, only
522: $(N^2+N)/2$ components of the coupling function have to be
523: computed.
524:
525:
526: \subsection{Single state limit}
527:
528: As already mentioned, it is not possible to account for broad
529: bands within the Markov approximation. In the single-state limit
530: (uncoupled exciton states), the Markov rates \Eq{markov_gamma}
531: vanish since the coupling function $J_{\alpha\alpha}^\alpha(E)$
532: tends strongly to zero at zero argument $E = \ea - \ea$. Thus,
533: in the absence of radiative damping, the absorption consists of
534: delta lines. However, if we do not perform the Markov
535: approximation, a composite absorption lineshape is obtained due
536: to the frequency dependence of the self energy
537: $\Sigma_{\alpha\alpha}(\omega)$. A broad band adds to the
538: unbroadened zero phonon line. Its weight with respect to the
539: total absorption area is $Z_\alpha=1/(1+S_\alpha)$, with the
540: Huang-Rhys factor $S_\alpha$ stemming from the self energy
541: derivative
542: %
543: \be\label{Huang-Rhys} S_\alpha = -\left.\frac{d {\mathrm Re}
544: \Sigma_{\alpha \alpha}(\omega)}{d\omega}\right|_{\omega=\ea-i0} =
545: \inte dE \, n(E)\, \frac{J_{\alpha\alpha}^\alpha(E)}{E^2} > 0 \,.
546: \ee
547: %
548: This result for $Z_\alpha$ agrees up to first order in $S_\alpha$
549: with the exact result from the independent Boson model, $Z_\alpha
550: = \mathrm{exp}(-S_\alpha)$. However, the single-state solution
551: neglects completely the inter-state scattering which leads to the
552: Lorentzian broadening of the lines.
553:
554:
555: \subsection{Semi-Markov approximation}\label{sec:theory_semi}
556:
557: In the full non-Markovian solution, both the Lorentzian
558: broadening of the ZPL and the broad bands are obtained using the
559: full self energy \Eq{self_def}. However, the numerical cost of
560: $O(N^3)$ coupling functions makes this treatment unfeasible for the
561: large number $N$ of excitons states covered by the focus of a typical
562: near-field experiment. Thus, either one has to select fewer
563: states, or one has to resort to another approximation which still
564: preserves the main features of the absorption spectrum.
565:
566: Aiming at this goal, we introduce the ``Semi-Markov
567: approximation" which neglects the off diagonal terms in the self
568: energy and keeps the frequency dependence only in the fully state
569: diagonal terms:
570: %
571: \bean \label{SM_self} \Sigma_{\alpha\eta}(\omega) & \rightarrow
572: & \delta_{\alpha\eta} \Sigma_\alpha^{SM}(\omega) \, ,\\ \nonumber
573: \Sigma_\alpha^{SM}(\omega) \! & \equiv & \!
574: \int\!dE\, \frac{n(E)\,J_{\alpha\alpha}^\alpha(E)}{\omega -\ea +E} \\ \nonumber
575: &+& \sum_{\beta\neq\alpha} \int\! dE\,
576: \frac{n(E)\,J_{\alpha\alpha}^\beta(E)}{\ea-i0-\eb+E} \,.\\
577: \eea
578: %
579: A linear expansion of $\Sigma_\alpha^{SM}(\omega)$ around the
580: exciton pole $\ea$ shows (cf. App.\,B) that the ZPL has a
581: reduced width
582: %
583: \be \label{gamma_red}
584: \gamma_\alpha^{SM} = \gamma^M_\alpha/(1+S_\alpha) \,.
585: \ee
586: %
587: The semi-Markov approximation is acceptable as long as the off
588: diagonal elements of the self energy are small with respect to
589: the differences between diagonal elements of the matrix problem
590: \Eq{lin_sys}. Since these are dominated by the exciton
591: eigenergies, the semi-Markov approximation is expected to deviate
592: markedly from the full non-Markovian treatment only in spectral
593: regions with a high density of overlapping states. Although much
594: less demanding in computer power (only $(N^2+N)/2$ coupling
595: functions needed - as in the Markov approximation), both broad
596: bands and broadening of the ZPLs are obtained in the semi-Markov
597: approximation.
598: %
599: %
600: \begin{table}[b]
601: \caption{\label{tab_mat_const}Material constants for bulk GaAs
602: and other parameters used in the numerical simulations. }
603: \begin{ruledtabular}
604: \begin{tabular}{llcc}
605: longitudinal sound velocity & $u_S$ & 5.33 & nm/ps \footnotemark[1] \\
606: bulk mass density & $ \rho_M$ & 5.37 & g/cm$^3$ \footnotemark[1]\\
607: def. pot. electrons & $D_c$ & -7000 & meV \footnotemark[1] \\
608: def. pot. holes & $D_v$ & +3500 & meV \footnotemark[1] \\
609: exciton mass & $ M $ & 0.300 & $m_0$ \footnotemark[2]\\
610: in-plane electron mass & $ m_e$ & 0.067 & $m_0$\\
611: in-plane hole mass & $m_h$ & 0.233 & $m_0$ \footnotemark[2]\\
612: QW width & $L_z$ & 5 & nm \\
613: disorder strength & $\sigma$ & 4.75 & meV \\
614: correlation length & $a_B$ & 9.9 & nm \\
615: grid size & $\Delta_x$& 1.65 & nm \\
616: \end{tabular}
617: \end{ruledtabular}
618: \footnotetext[1]{After Ref.\cite{sian01}}
619: \footnotetext[2]{After Ref.\cite{siar00}}
620: \end{table}
621: %
622: %
623: \bfi[t]
624: \centering
625: \includegraphics*[width=8.8cm]{fig2}
626: \caption{\label{im:fig2}(Color online) Probability amplitude of six exciton COM
627: wave functions in a 5\,nm wide GaAs/AlGaAs QW. The lines are
628: drawn at 50, 30, and 10\% of the peak probability of each wave
629: function. Notice that state 1 and state 6 do overlap.} \efi
630: %
631:
632:
633: \section{\label{sec:results} Results}
634:
635: We present now numerical solutions of the theory developed in
636: \Sect{sec:theory}. The Schr\"odinger equation \Eq{COMmotion} has
637: been discretized on a square grid with step size $\Delta_x$. An
638: uncorrelated potential corresponding to $(dE_a/dL_z)\Delta
639: L_z(\R)$ in \Eq{V_def} has been generated and convoluted with
640: the relative exciton wave function, using parameters of a 5 nm
641: wide GaAs/Al$_{0.3}$Ga$_{0.7}$As QW. The resulting disorder potential $V(\R)$ is
642: computed on the grid and has a correlation length close to the
643: exciton Bohr radius $a_B$ of this specific QW. The variance
644: $\sigma$ of this correlated potential has been adjusted to fit
645: experimental results from the speckle analysis \cite{koch02a}.
646: The diagonalization of \Eq{COMmotion} on a simulation mesh of
647: $128\times128$ grid points with periodic boundary conditions was
648: achieved via an ARPACK-based package. All used material constants
649: and simulation parameters are listed in \Tab{tab_mat_const}. In
650: particular the in-plane hole mass is taken from an excitonic
651: $k\cdot p$ calculation by Siarkos \etal \cite{siar00} which takes
652: into account the heavy-light hole mixing. An effective hole mass
653: of $m_h = 0.233\,m_0$ could be extracted which describes
654: satisfactorily the exciton COM motion for a wide range of quantum
655: well thicknesses. We have used this hole mass for the
656: effective-mass calculation in the present paper.
657:
658: In \Fig{im:fig2} we display the amplitudes $|\psia|^2$ of six
659: computed states, whose eigenergies are contiguous. These states
660: were selected by purpose of exhibiting various combinations of
661: oscillator strength and spatial overlap to other states. In
662: particular both quantities are large for $\psi_1(\R)$ and both
663: are small for $\psi_3(\R)$. State $\psi_5(\R)$ has large
664: oscillator strength and small overlap, while for $\psi_6(\R)$ the
665: contrary holds. Here, we are restricted to six states only since
666: we are going to compare the computationally most demanding
667: spectra using the full self energy matrix \Eq{self_def} with
668: other approximations.
669: %
670: \bfi[t]
671: \centering
672: \includegraphics*[width=8.8cm]{fig3}
673: \caption{\label{im:fig3}(Color online) Selected components of the
674: coupling function $C_{\alpha\eta}^\beta(E) \equiv
675: J_{\alpha\eta}^\beta(E)/E^3$ for some of the states of
676: \Fig{im:fig2}. The off-diagonal components have been magnified by
677: the factors given in brackets.} \efi
678: %
679:
680: The coupling function $J_{\alpha\eta}^\beta(E)$ is formed using
681: the eigenstates $\psia$ and represents the most time-consuming
682: part of the numerical task. The fully diagonal components
683: $J_{\alpha\alpha}^\alpha(E)$ are responsible for pure dephasing
684: in the single-state limit. These components, divided by $E^3$,
685: are approximately Gaussians with FWHM inversely proportional to
686: the localization length $\Lambda_\alpha$ of state $\alpha$ (full
687: curves in \Fig{im:fig3} for $\alpha=1$ and $\alpha=5$). Their
688: localization lengths (calculated from the inverse participation
689: ratio) are $16.4\,$ and $36.9\,$nm, respectively. Partially
690: off-diagonal terms $J_{\alpha\alpha}^\beta(E)$ with
691: $\beta\neq\alpha$ appear in the Markovian scattering rates
692: $\gamma_\alpha^M$ (\Eq{markov_gamma}) and are sensitive to the
693: spatial overlap between $\psia$ and $\psib$. In this sense,
694: $\psi_1(\R)$ overlaps with $\psi_6(\R)$ about 100 times more
695: effectively than with $\psi_5(\R)$, according to \Fig{im:fig3}.
696: Finally, fully off-diagonal terms $J_{\alpha\eta}^\beta(E)$ with
697: $\alpha$, $\beta$, $\eta$ all different account for deviations
698: between the full solution with self energy \Eq{self_def} and the
699: semi-Markov level with self energy \Eq{SM_self}. All nondiagonal
700: terms of the coupling function vanish at $E \rightarrow 0$ with
701: power $E^4$ due to the orthogonality of the exciton wave
702: functions.
703: %
704: \bfi[t]
705: \centering
706: \includegraphics*[width=8.8cm]{fig4}
707: \caption{\label{im:fig4}(Color online) Absorption lineshape for a
708: single state at $T=40\,$K (upper curves, red) and $T=10\,$K
709: (lower curves, blue). Results from the 2nd Born approximation
710: (solid lines) are compared with the exact solution (dashed) for
711: state $\alpha=1$ (left panel) and $\alpha=3$ (right panel) of
712: \Fig{im:fig2}. The localization lengths are $\Lambda_1=16.4\,$nm
713: and $\Lambda_3=23.3\,$nm. The arrows indicate the ZPLs, which
714: correspond also to the Markov limit of the absorption spectra.}
715: \efi
716: %
717:
718: Having computed the coupling functions, the self energy at the
719: desired level of approximation (single state, Markov,
720: semi-Markov, or full non-Markov ) can be easily generated. Using
721: this input, the Green's function and the absorption spectrum are
722: obtained. First we have checked the quality of our 2B treatment
723: against the exact solution (Independent Boson Model) which is
724: available in the single-state case only. In \Fig{im:fig4} we show
725: the absorption spectra for state $\alpha=1$ and $\alpha=3$ of
726: \Fig{im:fig2}. The broad band stems from phonon absorption (left
727: side of the ZPL) and phonon emission (right side). Therefore, the
728: asymmetry evolves into a more symmetric shape as temperature is
729: raised. There is very good agreement between 2B and exact results
730: at $T=10\,$K. However, at higher temperatures the broad band is
731: underestimated in the self energy approach.
732:
733: In \Fig{im:fig5} we display the absorption spectrum for the system
734: of six exciton sates. The Markovian spectrum (dashed) is
735: simply a superposition of Lorentzian lines, whose widths are set
736: by Fermi's golden rule, \Eq{markov_gamma}. Since the deformation
737: potential matrix elements $\tabq$ are sensitive to the spatial
738: overlap of the COM wave functions (see \App{sec:app_me}), states
739: 2,3,5 exhibit narrower lines compared to states 1,4,6 which have
740: a stronger overlap.
741: %
742: \bfi[t]
743: \centering
744: \includegraphics*[width=8.7cm]{fig5}
745: \caption{\label{im:fig5}(Color online) Absorption spectrum
746: calculated in the Markov approximation (black dashes), with
747: semi-Markov quality (blue dots), and with full non-Markovian
748: quality (red full line) at lattice temperatures of $T=10\,$K (top
749: panel) and $T=40\,$K (bottom panel). Notice the good agreement of
750: semi-Markov and full non-Markovian results. No focus function is
751: included. The triangles indicate the positions of the bare
752: exciton eigenergies.}
753: \efi
754: %
755: The full non-Markovian spectrum is characterized by enhanced
756: absorption between the single lines, as stressed by the
757: logarithmic representation of \Fig{im:fig5}. Furthermore a shift
758: of the peak energies is observed (polaron shift). In the low
759: energy part of the spectrum ($E\lesssim -3$\,meV) the energetic
760: spacing among the states is comparable to or larger than the
761: width of diagonal elements of the coupling function (cf.
762: \Fig{im:fig3}). Thus, sidebands are clearly noticed although they
763: overlap. They are about one order of magnitude larger than the
764: superposition of the Lorentzian tails in the Markovian spectrum.
765: Conversely, the energetic distance between states 5 and 6 is much
766: smaller than the width of the corresponding coupling function
767: $J_{\alpha\alpha}^\alpha(E)$. In this region the inter-peak
768: absorption is dominated by the Lorentzian tails of the ZPLs, and
769: the individual broad bands cannot be easily resolved. A careful
770: inspection reveals that the Lorentzian peak widths are smaller
771: than in the Markovian case. According to \Eq{gamma_red} and
772: \Eq{Huang-Rhys}, the width reduction scales with the Huang-Rhys
773: factor and is more important for stronger localized states and
774: higher temperatures, as observed (cf. e.g. the peak of state 1).
775: In \Fig{im:fig5} we also check the quality of the semi-Markov
776: approximation against the full solution. The agreement is
777: excellent at both $T=10\,$K and $T=40\,$K. Thus, pure dephasing
778: is dominated by state-diagonal processes. A similar conclusion
779: has been obtained in Ref.\,\onlinecite{tak99}.
780: %
781: \bfi[t]
782: \centering
783: \includegraphics*[width=8.8cm]{fig6}
784: \caption{\label{im:fig6}(Color online) Temperature dependent ZPL
785: widths $\gamma_\alpha$ for lines $1$ and $3$ of the full
786: non-Markovian (solid red lines) and the semi-Markov spectrum
787: (dotted blue lines) of \Fig{im:fig4}. The simplified expression
788: \Eq{gamma_red} (black solid lines) and the Markov rates
789: $\gamma_\alpha^M$ (black dashed lines) are also displayed for
790: comparison.} \efi
791: %
792:
793: In \Fig{im:fig6} the temperature dependence of the linewidths of
794: two selected states of \Fig{im:fig5} is studied. The reduced
795: Markov width \Eq{gamma_red} fairly describes both full solution
796: and semi-Markovian widths, as extracted from Lorentz fits in a
797: narrow spectral window around the ZPLs. This linewidth reduction
798: from Markov to non-Markov is a general feature and appears as
799: soon as the spectral weight of the quasi particle is reduced.
800: However, being a higher order effect in the coupling strength,
801: one may ask how an inclusion of higher order terms would alter
802: this finding. A consequent treatment of all diagrams up to 4th
803: order in the cumulant expansion by Muljarov \etal
804: \cite{muljarov05} has shown that the reduction is indeed a
805: pertinent feature. Its numerical value, however, might be
806: quantitatively not correct when restricting to the level of 2nd
807: Born approximation. In \Fig{im:fig6}, a sublinear dependence of
808: the linewidth on temperature is found for both states. This can
809: be easily explained by noting that the Huang-Rhys factor
810: increases nearly linear with $T$ (cf. \App{sec:app_zpl}).
811:
812: %
813: \bfi[t]
814: \centering
815: \includegraphics*[width=8.8cm]{fig7}
816: \caption{\label{im:fig7}(Color online) Top panel: ZPL weights
817: $Z_\alpha=1/(1+S_\alpha)$ (blue squares) and
818: $\Lambda_\alpha/(\Lambda_\alpha + bT)$ ratio (black circles) for
819: a set of 50 exciton states at $T=40\,$K. A good agreement is predicted from \Eq{Za_approx}
820: and it is found using $b=0.12\,$nm/K.
821: The dashed line is the constant ZPL weight within the Markov approximation. Bottom
822: Panel: Markovian (shadowed area) and semi-Markovian (blue solid
823: line) absorption spectrum for the same set of states at $T=40\,$K
824: using a focus of $50\,$nm.} \efi
825: %
826:
827: Having tested the features of the semi-Markov approximation in
828: \Fig{im:fig5} and \Fig{im:fig6}, we apply it to a system of many
829: exciton states. To come closer to the experimental situation of a
830: near-field measurement, we have included a Gaussian focus of
831: 50\,nm width (FWHM of field intensity). Using the 50 lowest
832: energy eigenstates of a disorder realization for a 5\,nm wide QW
833: we obtain the spectrum in \Fig{im:fig7} up to $E=0\,$meV which is
834: the exciton energy in an ideal QW of the same width. In a near
835: field geometry, the visibility of a state depends not only on its
836: oscillator strength, but on the overlap with the illumination
837: spot as well, as seen from \Eq{ma_fact}. Thus, several states in
838: the range $-3\,$meV$<E<-2\,$meV appear especially strong in the
839: present simulation. The absorption background between the peaks
840: can be clearly seen on a linear scale in this spectral region. In
841: the high energy tail ($E\approx 0\,$meV) there is practically no
842: difference between Markov and semi-Markov spectra. Even on a
843: logarithmic scale both spectra are nearly identical (not shown).
844: These observations can be explained considering the localization
845: length $\Lambda_\alpha$ and the spectral separation of the
846: exciton states. $\Lambda_\alpha$ is increasing across the
847: spectrum from $\Lambda_2=17.2\,$nm (the ground state is dark in the actual illumination geometry)
848: up to $\Lambda_{50}=54.1\,$nm.
849: Via the Huang-Rhys factor, this leads to an increasing ZPL
850: weight (cf. \App{sec:app_zpl}) which goes from $Z_2=0.77$ up to
851: $Z_{50}=0.92$, as seen in the top panel of \Fig{im:fig7}. Thus,
852: the importance of the non-Markovian broad bands (proportional to
853: $1-Z_\alpha$) decreases with photon energy. However, the
854: observability of this effect strongly depends on the spectral
855: separation of the resonances, as discussed for \Fig{im:fig5}.
856: The spectrally close doublet at about $-4.2\,$meV e.g. does not
857: show a distinct broad band, in spite of having a short
858: localization length (large $S_\alpha$). Thus, for observing
859: non-Markovian effects in the absorption spectrum from QWs, it is
860: necessary to position the near-field focus in such a way that
861: states with small localization length and large spectral
862: separation are excited.
863:
864: A non-Markovian computation of the absorption which includes all
865: optically active exciton states is at the moment beyond our
866: computational possibilities. We expect somewhat broader ZPL peaks
867: since more states would be available for phonon-mediated
868: scattering. However, the broad band absorption between peaks is
869: not expected to be modified as it is due to intra-state
870: relaxation (first term on the rhs of \Eq{SM_self}). Thus, a
871: calculation of the absorption spectrum using all states will not
872: reveal new features as compared to what we display in
873: \Fig{im:fig7}, at least within the semi-Markov approximation.
874:
875:
876:
877:
878: \section{\label{sec:conclusions} Conclusions}
879:
880: In conclusion, we have developed a theoretical description of the
881: exciton-acoustic phonon scattering in disordered QWs which goes
882: beyond the Markov approximation. We have considered deformation
883: potential coupling with acoustic phonons and realistic exciton
884: states resulting from a microscopic simulation. A second order
885: self energy approach is used for computing the absorption
886: spectrum from the exciton propagator. At the Markov level the
887: absorption is a superposition of Lorentzian peaks with widths
888: defined by Fermi's golden rule, which results from
889: phonon-assisted scattering between different states. At the full
890: non-Markovian level, virtual scattering events on the same state
891: are considered as well (pure dephasing). These processes manifest
892: itself in broad absorption bands around Lorentzian peaks (ZPL)
893: whose widths are smaller than according to Fermi's golden rule.
894: We have also discussed a semi-Markov approximation in which a
895: diagonal self energy is employed, but the important frequency
896: dependence of the diagonal term is preserved. This allows to reduce
897: the numerical task from $O(N^3)$ to $O(N^2)$, where $N$ is the
898: number of exciton states, while recovering all the important
899: features of the complete solution. We have shown that the
900: non-Markovian effect increases with a shorter localization length
901: of the exciton wave function, and for low spectral density of
902: states. A near-field setup could demonstrate these predictions
903: experimentally.
904:
905: \begin{acknowledgments}
906: Support from DFG in the frame of Sfb 296 is gratefully
907: acknowledged. G.~M. is thankful to Egor Muljarov (Berlin) for
908: discussions and to Roberto Cingolani (Lecce) for support from
909: CNR.
910: \end{acknowledgments}
911:
912: \appendix
913: \section{\label{sec:app_me} Interaction matrix elements and coupling function}
914:
915: The matrix element for optical interband transitions between an
916: exciton wave function $\Psi_\alpha(\r_e \r_h)$ and a light field
917: with momentum $\q$ and field envelope $E(\r)$ is given by
918: %
919: \be\label{dip_me_def} m_\alpha = \mu_{\mathrm{cv}} \inte d\r\,
920: e^{-i\q\cdot\r}\, E(\r)\, \Psi_\alpha (\r,\r) \ee
921: %
922: with the interband dipole moment $\mu_{\mathrm{cv}}$ of the band
923: edge states. The factorization \Eq{FactorWF} and transmission
924: normal to the QW plane ($\qp=0$) results in
925: %
926: \be\label{ma_fact}
927: m_\alpha = \mu_{\mathrm{cv}} \phi_{1s}(0) O_{eh} \inte d\R\, E(\r)\, \psia \,,
928: \ee
929: %
930: with $\phi_{1s}(0)$ as the relative exciton wave function at
931: the origin, and $O_{eh}$ the confinement overlap integral $O_{eh}
932: = \inte dz\, u_e(z) \,u_h(z)$.
933:
934: Within the same factorization Ansatz, the deformation potential
935: matrix element of Ref.\,\onlinecite{tak85} takes the form
936: %
937: \bean \label{t_ab_q} t_{\alpha\beta}^\q & =&
938: \sqrt{\frac{\hbar\wq}{2 u_S^2 \rho_M \, V }} \, (\psi_\alpha
939: \psi_\beta)_\qp \times \\ \nonumber
940: & & \Big[ D_c K_e (q_z) \chi( \qp/\eta_e) - D_v K_h (q_z) \chi(\qp/\eta_h) \Big] \\
941: \eea
942: %
943: with deformation potential constants $D_a$, sound velocity
944: $u_S$, mass density $\rho_B$, and normalization volume $V$.
945: Further, we have defined Fourier transforms of the squared
946: confinement and relative wave functions, and the COM overlap
947: %
948: \bean
949: K_a(q_z) &=& \int \! dz\, u_a^2(z)\, e^{-i q_z z} \nonumber \\
950: \chi (\q_\parallel) &=& \int \! d\ro\, \phi_{1s}^{2}({\mathbf{\rho}})\,
951: e^{-i\q_\parallel \cdot {\mathbf{\rho}}} \\
952: (\psi_\alpha \psi_\beta)_\qp &=& \inte d\R\, \psia e^{-i\qp\cdot\R} \psib
953: \,.\nonumber
954: \eea
955: %
956: In particular, for Gauss confinement states and a hydrogen-like
957: relative motion we obtain:
958: %
959: \bean
960: K_{a}(q_z) &=& \mathrm{exp}\left(-\half(q_z L_a)^2\right) \\
961: \chi (\q_{\parallel }) &=& \left( 1 +
962: (q_\parallel a_{B}/2)^2 \right)^{-3/2} \, ,
963: \eea
964: %
965: where $L_e=1.69\,$nm and $L_h=1.34$\,nm. These two values result
966: from a variational determination of the confinement levels in the
967: 5 nm wide GaAs/AlGaAs QW. Other material constants for GaAs are
968: listed in \Tab{tab_mat_const}.
969:
970: Since the exciton states $\psia$ are numerically generated in
971: Cartesian coordinates, we proceed with fast Fourier
972: transformation along $q_x$, $q_y$, $q_z$ and integrate
973: accordingly \Eq{J_def} as
974: %
975: \bea\label{J_cartesian}
976: \nonumber J_{\alpha\eta}^{\beta}(E) &=& \frac{V}{(2\pi)^3}
977: \int\! dq_x dq_y dq_z \, t_{\alpha\beta}^\q t_{\eta\beta}^{-\q} \, \times\\
978: && \delta(|E|-\hbar u_S \sqrt{q_x^2+q_y^2+q_z^2} \, .) \eea
979: %
980: The delta function is used to integrate over $q_z$. Special care
981: is taken to deal with the resulting square root singularities at
982: the remaining integration limits. Each matrix element $t$
983: contains a square-root of phonon energy, thus
984: $J_{\alpha\eta}^{\beta}(E)$ starts at least with a power of
985: $E^3$.
986:
987:
988:
989: \section{\label{sec:app_zpl}ZPL width and weight}
990:
991: In the semi-Markov approximation, the absorption is given by
992: %
993: \be\label{SM_absorption} \alpha(\omega) = \mathrm{Im} \sum_\alpha
994: \frac{|m_\alpha|^2} {\omega - \ea - \Sigma_\alpha^{SM}(\omega)} \,. \ee
995: %
996: For evaluating the ZPL width, we expand in \Eq{SM_absorption} the
997: self energy around the pole $\ea$. The real part at the resonance
998: (polaron shift) $\mathrm{Re}\Sigma_\alpha^{SM}(\ea) $ is small
999: and can be dropped, but its derivative has to be kept as
1000: %
1001: \be -\left. \frac{d}{d\omega} \mathrm{Re}
1002: \Sigma_\alpha^{SM}(\omega)\right|_{\omega = \ea-i0} = S_\alpha
1003: \, . \ee
1004: %
1005: The imaginary part is taken as $
1006: \mathrm{Im}\Sigma_\alpha^{SM}(\ea-i0) = \gamma^M_\alpha/2$.
1007: %
1008: Putting this into \Eq{SM_absorption} we end up with an absorption
1009: lineshape
1010: %
1011: \be \alpha(\omega\approx\ea) = \sum_\alpha \frac{|m_\alpha|^2}{1+S_\alpha} \,
1012: \mathrm{Im} \frac{1}{\omega -\ea - i \gamma_\alpha^{SM}/2} \,,
1013: \ee
1014: %
1015: which has a reduced ZPL weight of $1/(1+S_\alpha)$ and a reduced
1016: width $\gamma_\alpha^{SM}= \gamma_\alpha^{M}/(1+S_\alpha)$.
1017:
1018: At high temperatures, the phonon occupation can be replaced by
1019: $n(E) \rightarrow E/k_B T$. Together with the Gaussian shape of
1020: the diagonal coupling factor (see \Fig{im:fig3}) we can evaluate
1021: the Huang-Rhys factor \Eq{Huang-Rhys} in closed form
1022: %
1023: \be S_\alpha \approx \inte dE \frac{k_B T}{E} E^3 A\,
1024: \mathrm{exp}\left[- \left(\frac{\Lambda_\alpha E}{
1025: u_s}\right)^2\right] \frac{1}{E^2} = b \frac{T}{\Lambda_\alpha}
1026: \, . \ee
1027: %
1028: Thus, the ZPL weight is approximately given by
1029: %
1030: \be\label{Za_approx}
1031: Z_\alpha = \frac{1}{1+S_\alpha} \approx \frac{\Lambda_\alpha}{\Lambda_\alpha+ bT} \,.
1032: \ee
1033: %
1034: This approximation is getting better for less localized states, which
1035: can be seen in the top panel of \Fig{im:fig7}.
1036:
1037:
1038:
1039:
1040: \bibliography{PRB}
1041:
1042: \begin{thebibliography}{22}
1043: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi
1044: \expandafter\ifx\csname bibnamefont\endcsname\relax
1045: \def\bibnamefont#1{#1}\fi
1046: \expandafter\ifx\csname bibfnamefont\endcsname\relax
1047: \def\bibfnamefont#1{#1}\fi
1048: \expandafter\ifx\csname citenamefont\endcsname\relax
1049: \def\citenamefont#1{#1}\fi
1050: \expandafter\ifx\csname url\endcsname\relax
1051: \def\url#1{\texttt{#1}}\fi
1052: \expandafter\ifx\csname urlprefix\endcsname\relax\def\urlprefix{URL }\fi
1053: \providecommand{\bibinfo}[2]{#2}
1054: \providecommand{\eprint}[2][]{\url{#2}}
1055:
1056: \bibitem[{\citenamefont{Runge and Lienau}(2005)}]{runge05}
1057: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Runge}} \bibnamefont{and}
1058: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Lienau}},
1059: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{71}},
1060: \bibinfo{pages}{035347} (\bibinfo{year}{2005}).
1061:
1062: \bibitem[{\citenamefont{Hohenester et~al.}(2005)\citenamefont{Hohenester,
1063: Goldoni, and Molinari}}]{hohenester05}
1064: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Hohenester}},
1065: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Goldoni}}, \bibnamefont{and}
1066: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Molinari}},
1067: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{95}},
1068: \bibinfo{pages}{216802} (\bibinfo{year}{2005}).
1069:
1070: \bibitem[{\citenamefont{Matsuda et~al.}(2002)\citenamefont{Matsuda, Saiki,
1071: Nomura, Mihara, and Aoyagi}}]{matsuda02}
1072: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Matsuda}},
1073: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Saiki}},
1074: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Nomura}},
1075: \bibinfo{author}{\bibfnamefont{M.}~\bibnamefont{Mihara}}, \bibnamefont{and}
1076: \bibinfo{author}{\bibfnamefont{Y.}~\bibnamefont{Aoyagi}},
1077: \bibinfo{journal}{Appl. Phys. Lett.} \textbf{\bibinfo{volume}{81}},
1078: \bibinfo{pages}{2291} (\bibinfo{year}{2002}).
1079:
1080: \bibitem[{\citenamefont{Guest et~al.}(2002)\citenamefont{Guest, Stievater, Li,
1081: Cheng, Steel, Gammon, Katzer, Park, Ell, Thranhardt et~al.}}]{guest02}
1082: \bibinfo{author}{\bibfnamefont{J.~R.} \bibnamefont{Guest}},
1083: \bibinfo{author}{\bibfnamefont{T.~H.} \bibnamefont{Stievater}},
1084: \bibinfo{author}{\bibfnamefont{X.}~\bibnamefont{Li}},
1085: \bibinfo{author}{\bibfnamefont{J.}~\bibnamefont{Cheng}},
1086: \bibinfo{author}{\bibfnamefont{D.~G.} \bibnamefont{Steel}},
1087: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Gammon}},
1088: \bibinfo{author}{\bibfnamefont{D.~S.} \bibnamefont{Katzer}},
1089: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Park}},
1090: \bibinfo{author}{\bibfnamefont{C.}~\bibnamefont{Ell}},
1091: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Thranhardt}},
1092: \bibnamefont{et~al.}, \bibinfo{journal}{Phys. Rev. B}
1093: \textbf{\bibinfo{volume}{65}}, \bibinfo{pages}{241310(R)}
1094: (\bibinfo{year}{2002}).
1095:
1096: \bibitem[{\citenamefont{Besombes et~al.}(2001)\citenamefont{Besombes, Kheng,
1097: Marsal, and Mariette}}]{besombes01}
1098: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Besombes}},
1099: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Kheng}},
1100: \bibinfo{author}{\bibfnamefont{L.}~\bibnamefont{Marsal}}, \bibnamefont{and}
1101: \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Mariette}},
1102: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{63}},
1103: \bibinfo{pages}{155307} (\bibinfo{year}{2001}).
1104:
1105: \bibitem[{\citenamefont{Borri et~al.}(2005)\citenamefont{Borri, Langbein,
1106: Woggon, Stavarache, Reuter, and Wieck}}]{borri05}
1107: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Borri}},
1108: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Langbein}},
1109: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Woggon}},
1110: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Stavarache}},
1111: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Reuter}}, \bibnamefont{and}
1112: \bibinfo{author}{\bibfnamefont{A.~D.} \bibnamefont{Wieck}},
1113: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{71}},
1114: \bibinfo{pages}{115328} (\bibinfo{year}{2005}).
1115:
1116: \bibitem[{\citenamefont{Borri et~al.}(2001)\citenamefont{Borri, Langbein,
1117: Schneider, Woggon, Sellin, Ouyang, and Bimberg}}]{borri01}
1118: \bibinfo{author}{\bibfnamefont{P.}~\bibnamefont{Borri}},
1119: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Langbein}},
1120: \bibinfo{author}{\bibfnamefont{S.}~\bibnamefont{Schneider}},
1121: \bibinfo{author}{\bibfnamefont{U.}~\bibnamefont{Woggon}},
1122: \bibinfo{author}{\bibfnamefont{R.~L.} \bibnamefont{Sellin}},
1123: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Ouyang}}, \bibnamefont{and}
1124: \bibinfo{author}{\bibfnamefont{D.}~\bibnamefont{Bimberg}},
1125: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{87}},
1126: \bibinfo{pages}{157401} (\bibinfo{year}{2001}).
1127:
1128: \bibitem[{\citenamefont{Duke and Mahan}(1965)}]{duke65}
1129: \bibinfo{author}{\bibfnamefont{C.~B.} \bibnamefont{Duke}} \bibnamefont{and}
1130: \bibinfo{author}{\bibfnamefont{G.~D.} \bibnamefont{Mahan}},
1131: \bibinfo{journal}{Phys. Rev.} \textbf{\bibinfo{volume}{139}},
1132: \bibinfo{pages}{A1965} (\bibinfo{year}{1965}).
1133:
1134: \bibitem[{\citenamefont{Mahan}(1990)}]{mahan}
1135: \bibinfo{author}{\bibfnamefont{G.~D.} \bibnamefont{Mahan}},
1136: \emph{\bibinfo{title}{Many-Particle Physics}} (\bibinfo{publisher}{Plenum
1137: Press}, \bibinfo{year}{1990}).
1138:
1139: \bibitem[{\citenamefont{Bondarev et~al.}(2003)\citenamefont{Bondarev,
1140: Maksimenko, Slepyan, Krestnikov, and Hoffmann}}]{bondarev03}
1141: \bibinfo{author}{\bibfnamefont{I.~V.} \bibnamefont{Bondarev}},
1142: \bibinfo{author}{\bibfnamefont{S.~A.} \bibnamefont{Maksimenko}},
1143: \bibinfo{author}{\bibfnamefont{G.~Y.} \bibnamefont{Slepyan}},
1144: \bibinfo{author}{\bibfnamefont{I.~L.} \bibnamefont{Krestnikov}},
1145: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Hoffmann}},
1146: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{68}},
1147: \bibinfo{pages}{073310} (\bibinfo{year}{2003}).
1148:
1149: \bibitem[{\citenamefont{Muljarov and Zimmermann}(2004)}]{muljarov04}
1150: \bibinfo{author}{\bibfnamefont{E.~A.} \bibnamefont{Muljarov}} \bibnamefont{and}
1151: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zimmermann}},
1152: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{93}},
1153: \bibinfo{pages}{237401} (\bibinfo{year}{2004}).
1154:
1155: \bibitem[{\citenamefont{Muljarov et~al.}(2005)\citenamefont{Muljarov,
1156: Takagahara, and Zimmermann}}]{muljarov05}
1157: \bibinfo{author}{\bibfnamefont{E.~A.} \bibnamefont{Muljarov}},
1158: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Takagahara}},
1159: \bibnamefont{and}
1160: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zimmermann}},
1161: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{95}},
1162: \bibinfo{pages}{177405} (\bibinfo{year}{2005}).
1163:
1164: \bibitem[{\citenamefont{Fan et~al.}(1998)\citenamefont{Fan, Takagahara,
1165: Cunningham, and Wang}}]{fan98}
1166: \bibinfo{author}{\bibfnamefont{X.}~\bibnamefont{Fan}},
1167: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Takagahara}},
1168: \bibinfo{author}{\bibfnamefont{J.~E.} \bibnamefont{Cunningham}},
1169: \bibnamefont{and} \bibinfo{author}{\bibfnamefont{H.}~\bibnamefont{Wang}},
1170: \bibinfo{journal}{Solid State Comm.} \textbf{\bibinfo{volume}{108}},
1171: \bibinfo{pages}{857} (\bibinfo{year}{1998}).
1172:
1173: \bibitem[{\citenamefont{Takagahara}(1999)}]{tak99}
1174: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Takagahara}},
1175: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{60}},
1176: \bibinfo{pages}{2638} (\bibinfo{year}{1999}).
1177:
1178: \bibitem[{\citenamefont{Krummheuer et~al.}(2002)\citenamefont{Krummheuer, Axt,
1179: and Kuhn}}]{krum02}
1180: \bibinfo{author}{\bibfnamefont{B.}~\bibnamefont{Krummheuer}},
1181: \bibinfo{author}{\bibfnamefont{V.~M.} \bibnamefont{Axt}}, \bibnamefont{and}
1182: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Kuhn}},
1183: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{65}},
1184: \bibinfo{pages}{195313} (\bibinfo{year}{2002}).
1185:
1186: \bibitem[{\citenamefont{Zimmermann et~al.}(1997)\citenamefont{Zimmermann,
1187: Grosse, and Runge}}]{zim97}
1188: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zimmermann}},
1189: \bibinfo{author}{\bibfnamefont{F.}~\bibnamefont{Grosse}}, \bibnamefont{and}
1190: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Runge}},
1191: \bibinfo{journal}{Pure and Applied Chemistry} \textbf{\bibinfo{volume}{69}},
1192: \bibinfo{pages}{1179} (\bibinfo{year}{1997}).
1193:
1194: \bibitem[{\citenamefont{Zimmermann et~al.}(2003)\citenamefont{Zimmermann,
1195: Runge, and Savona}}]{zim_rev03}
1196: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zimmermann}},
1197: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Runge}}, \bibnamefont{and}
1198: \bibinfo{author}{\bibfnamefont{V.}~\bibnamefont{Savona}},
1199: \bibinfo{journal}{{\em Theory of resonant secondary emission: Rayleigh
1200: scattering versus luminescence} in: {Quantum Coherence, Correlation and
1201: Decoherence in Semiconductor Nanostructures}, edited by {T.~Takagahara.
1202: Elsevier Science}} (\bibinfo{year}{2003}).
1203:
1204: \bibitem[{\citenamefont{Grochol et~al.}(2005)\citenamefont{Grochol,
1205: Grosse, and Zimmermann}}]{grochol05}
1206: \bibinfo{author}{\bibfnamefont{M.} \bibnamefont{Grochol}},
1207: \bibinfo{author}{\bibfnamefont{F.}~\bibnamefont{Grosse}},
1208: \bibnamefont{and}
1209: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zimmermann}},
1210: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{71}},
1211: \bibinfo{pages}{125339} (\bibinfo{year}{2005}).
1212:
1213: \bibitem[{\citenamefont{Runge}(2002)}]{run_rev02}
1214: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Runge}}, \bibinfo{journal}{{\em
1215: Excitons in semiconductor nanostructures.} in: {Solid State Physics}, edited
1216: by {H. Ehrenreich and F. Spaepen. Academic Press, San Diego}}
1217: (\bibinfo{year}{2002}).
1218:
1219: \bibitem[{\citenamefont{Takagahara}(1993)}]{tak93}
1220: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Takagahara}},
1221: \bibinfo{journal}{Phys. Rev. Lett.} \textbf{\bibinfo{volume}{71}},
1222: \bibinfo{pages}{3577} (\bibinfo{year}{1993}).
1223:
1224: \bibitem[{\citenamefont{Siantidis et~al.}(2001)\citenamefont{Siantidis, Axt,
1225: and Kuhn}}]{sian01}
1226: \bibinfo{author}{\bibfnamefont{K.}~\bibnamefont{Siantidis}},
1227: \bibinfo{author}{\bibfnamefont{V.~M.} \bibnamefont{Axt}}, \bibnamefont{and}
1228: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Kuhn}},
1229: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{65}},
1230: \bibinfo{pages}{035303} (\bibinfo{year}{2001}).
1231:
1232: \bibitem[{\citenamefont{Siarkos et~al.}(2000)\citenamefont{Siarkos,
1233: Runge, and Zimmermann}}]{siar00}
1234: \bibinfo{author}{\bibfnamefont{A.}~\bibnamefont{Siarkos}},
1235: \bibinfo{author}{\bibfnamefont{E.}~\bibnamefont{Runge}},
1236: \bibnamefont{and}
1237: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zimmermann}},
1238: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{61}},
1239: \bibinfo{pages}{10854} (\bibinfo{year}{2000}).
1240:
1241: \bibitem[{\citenamefont{Kocherscheidt et~al.}(2002)\citenamefont{Kocherscheidt,
1242: Langbein, Mannarini, and Zimmermann}}]{koch02a}
1243: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Kocherscheidt}},
1244: \bibinfo{author}{\bibfnamefont{W.}~\bibnamefont{Langbein}},
1245: \bibinfo{author}{\bibfnamefont{G.}~\bibnamefont{Mannarini}},
1246: \bibnamefont{and}
1247: \bibinfo{author}{\bibfnamefont{R.}~\bibnamefont{Zimmermann}},
1248: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{66}},
1249: \bibinfo{pages}{161314(R)} (\bibinfo{year}{2002}).
1250:
1251: \bibitem[{\citenamefont{Takagahara}(1985)}]{tak85}
1252: \bibinfo{author}{\bibfnamefont{T.}~\bibnamefont{Takagahara}},
1253: \bibinfo{journal}{Phys. Rev. B} \textbf{\bibinfo{volume}{31}},
1254: \bibinfo{pages}{6552} (\bibinfo{year}{1985}).
1255:
1256: \end{thebibliography}
1257:
1258:
1259:
1260: \end{document}
1261: