1: %\documentstyle[prb,aps,psfig]{revtex}

2: %\documentstyle[preprint,aps,psfig]{revtex}

3: \documentstyle[prb,aps,psfig,twocolumn]{revtex}

4: \begin{document} 

5: \title{Role of bound pairs in the optical properties of highly excited

6: semiconductors: a self consistent ladder approximation approach.}

7: 

8: \twocolumn[\hsize\textwidth\columnwidth\hsize\csname@twocolumnfalse\endcsname

9: \author{C. Piermarocchi} \address{Department of Physics, University of

10: California San Diego, La Jolla, CA 92093-0319} \author{F. Tassone}

11: \address{Scuola Internazionale Superiore di Studi Avanzati, via Beirut

12: 4, I-34014 Trieste, Italy} \date{\today} \maketitle %\draft

13: 

14: 

15: 

16: \begin{abstract}

17: Presence of bound pairs (excitons) in a low-temperature electron-hole

18: plasma is accounted for by including correlation between fermions at

19: the ladder level. Using a simplified one-dimensional model with

20: on-site Coulomb interaction, we calculate the one-particle

21: self-energies, chemical potential, and optical response.  The results

22: are compared to those obtained in the Born approximation, which does

23: not account for bound pairs.  In the self-consistent ladder

24: approximation the self-energy and spectral function show a

25: characteristic correlation peak at the exciton energy for low

26: temperature and density. In this regime the Born approximation

27: overestimates the chemical potential.  Provided the appropriate vertex

28: correction in the interaction with the photon is included, both ladder

29: and Born approximations reproduce the excitonic and free pair optical

30: absorption at low density, and the disappearance of the exciton

31: absorption peak at larger density.  However, lineshapes and energy

32: shifts with density of the absorption and photoluminescence peaks are

33: drastically different.  In particular, the photoluminescence emission

34: peak is much more stable in the ladder approximation. At low

35: temperature and density a sizeable optical gain is produced in both

36: approximations just below the excitonic peak, however this gain shows

37: unphysical features in the Born approximation. We conclude that at low

38: density and temperature it is fundamental to take into account the

39: existence of bound pairs in the electron-hole plasma for the

40: calculation of its optical and thermodynamic properties.  Other

41: approximations that fail to do so are intrinsically unphysical in this

42: regime, and for example are not suitable to address the problem of

43: excitonic lasing.

44: 

45: \end{abstract} \pacs{}

46: 

47: ]

48: 

49: 

50: %\begin{multicols}{2}

51: 

52: 

53: \section{Introduction} 

54: 

55: The binding of a gas of oppositely charged fermions into a bosonic gas

56: of bound pairs has a fundamental interest in condensed matter physics,

57: in problems ranging from the state of hydrogen to the nature of the

58: electron-hole plasma in Si and Ge. This pairing takes its origin from

59: the Coulomb correlation between the charged fermions, and strongly

60: affects the statistical and thermodynamic properties of these systems.

61: In semiconductor physics, the role of the Coulomb correlation and

62: pairing in the description of the electron-hole plasma in Si and Ge

63: has been extensively investigated. Refined descriptions of both the

64: ground state and the thermodynamic properties of this system have been

65: developed in the past thirty years. In particular, an excitonic

66: insulator ground state was first proposed and then shown to be less

67: stable than a simpler electron-hole liquid in Si and Ge, due to

68: band-structure effects.\cite{rice77} Thermodynamics of the

69: electron-hole gas in simpler crystals was also addressed by

70: Haug\cite{haug76}, and Zhu {\it et al.}\cite{zhu96} aiming at deriving

71: a complete phase diagram from variational approaches.

72: 

73: 

74: 

75: 

76: In this paper we will deal with direct gap semiconductors and we will

77: discuss the thermodynamics of the photoexcited electron and holes at

78: quasi-thermal equilibrium. Our main concern is to determine how the

79: inclusion of pairing effects in the theoretical description, as done

80: for the systems mentioned above, can affect predictions on the shape

81: of optical spectra. In the optical properties of semiconductors, the

82: effect of the Coulomb electron-hole correlation comes into play to

83: explain the characteristic excitonic absorption and emission. These

84: excitonic features can be described by including vertex corrections in

85: the interaction of the electron-hole pair with the

86: photon.\cite{elliott57} In particular, excitonic absorption is found

87: in an empty crystal, when a background electron-hole plasma is absent.

88: However, when carriers are electrically or optically injected, they

89: reach quasi-thermal equilibrium through scattering, and a background

90: electron-hole plasma is eventually formed. In this typical situation

91: the absorption is modified, and spontaneous emission

92: (photoluminescence) also takes place. Absorption and emission

93: give valuable information about the state of the plasma, but a theory

94: describing both the thermodynamics and optical properties of the

95: plasma is required in order to extract this information.

96: Semiconductor Bloch equations, and their evolutions, are certainly

97: among the most known and used theories for this

98: purpose.\cite{haugkoch} In these theories, the electron-hole plasma

99: has been originally treated at the Hartree-Fock level,\cite{haug84}

100: and eventually screening of the Coulomb interaction was also included

101: at various levels of approximation.\cite{haugkoch} However, none of

102: these refinements accounts for the existence of bound excitons in the

103: plasma. This is certainly a weak point at low temperatures, when

104: simple thermodynamic arguments show that condensation of electron-hole

105: gas into a gas of bound-electron hole pairs is favorable in a wide

106: range of densities.

107: 

108: 

109: In the following sections we will show the relevant changes in the optical

110: properties of the semiconductor when the existence of excitons at low

111: temperatures is taken into account. This is clearly a leap beyond the

112: standard description of the interaction of light with the

113: semiconductor as described above.  Our purpose is qualitative, but we

114: can still gain solid understanding of the underlying physics. In our

115: opinion at this stage of development it is not wise to aim at

116: improving also quantitatively existing theories which are already far

117: too involved.  We keep the problem simple enough so as to always control

118: and understand the resulting physics. We thus consider an accurately

119: solvable model, where the system is one-dimensional, Coulomb

120: interaction is on-site, and electron and holes are spin-polarized.

121: Excitonic correlation is also expected to be stronger in one dimension

122: than higher dimensions. Moreover, for the single-pair problem, exactly

123: a single bound state is obtained with the on-site interaction.  In

124: our description of the electron-hole plasma, we either include carrier

125: scattering at the Born level, which does not describe bound pairs in

126: the plasma, or we include enough correlation so as to describe the

127: pairing into bound states. For this purpose, ladder diagrams in the

128: electron-hole scattering kernel\cite{kraeft86,zimmermann88,pereira98}

129: are included, and the electron and hole self energies calculated with

130: it.\cite{tassone99} The resulting Dyson equations for the Green

131: functions are solved self-consistently. The approximation is therefore

132: called self-consistent ladder approximation (SCLA). Screening is

133: expected to be weak in one dimension,\cite{benner91} due to reduced

134: screening phase-space, and we neglect it altogether.  We remark that

135: in the SCLA we calculate the single particle propagators and

136: self-energies only, thus in this sense, the theory remains purely

137: fermionic. We do not map the model onto bound and un-bound states at

138: any time. However, it is possible to show analytically that at low

139: temperatures and densities, a bosonic gas of excitons is effectively

140: described, and that self-consistency effectively introduces scattering

141: between these excitons. When this analogy is extended to larger

142: densities and temperatures, where bound and free carriers are expected

143: to coexist, we understand that self-consistency also accounts for

144: multiple exciton-free carrier, and free carrier-free carrier

145: scattering. Thus, a large amount of correlation, well beyond that

146: described in simpler Born approximations is included in the SCLA.  We

147: remark that we also assume quasi-thermal equilibrium in the electron-hole

148: gas, which is well justified as the characteristic radiative

149: recombination (several hundreds of picoseconds) is much slower than

150: the typical scattering time in the density range considered. The

151: theory is however written so as to be readily extended to the

152: non-equilibrium situation. We finally show how to correctly calculate

153: photon absorption and emission (photoluminescence or PL) in a

154: conserving sense, i.e. respecting f-sum rules. We only neglect

155: polaritonic effects, which are very weak in one

156: dimension.\cite{tassone.93}

157: 

158: 

159: 

160: 

161: Within the considered models, we highlight important trends in optical

162: and thermodynamical properties.  First, Born approximations

163: overestimate the chemical potential at low temperatures when a

164: consistent fraction of pairs is bound into excitons. Second, energetic

165: stability of the excitonic emission peak as a function of density is

166: better described in the SCLA, even well beyond disappearance of

167: excitonic absorption peak at large excitation densities. Third,

168: the large excitonic {\em gain} at densities just below those

169: where excitonic absorption disappears is clearly incorrectly described

170: in simpler Born approximations.  As these trends have a clear physical

171: origin, they are expected to hold qualitatively even when more

172: realistic interaction potentials (eventually including screening) are

173: considered. We conclude that these results do have important

174: experimental and theoretical relevance, signaling the necessity of an

175: adequate description of the excited semiconductor in the

176: low-temperature region. For example, current description of the

177: electron-hole plasma in the semiconductor Bloch equations is clearly

178: insufficient to address excitonic (inversion-less) lasing.

179: 

180: 

181: 

182: The paper is organized as follows. In Sec. \ref{sec:method} we

183: introduce and explain how to implement the self consistent ladder

184: approximation, and the Markov-Born approximation and self-consistent

185: Born approximation.  Single particle properties are analyzed in

186: Sec. \ref{sec:single}, where we show how excitonic correlation appears

187: in the electron and hole spectral functions.  Chemical potentials and

188: self-energies are then compared. In Sec. \ref{sec:excpot} we

189: analytically show how the SCLA at low temperature and density

190: describes an interacting exciton gas. Optical properties of a

191: semiconductor quantum wire are calculated in Sec. \ref{sec:optical}

192: for the three approximations. Absorption and emission as functions of

193: the density are also presented, and discussed in this

194: section. Conclusions are drawn is Sec. \ref{sec:conclusions}.

195: 

196: 

197: 

198: \section{Born and self consistent ladder approximation}

199: \label{sec:method}

200: 

201: 

202: In this section we introduce the SCLA, the simpler Markov-Born

203: approximation (MBA) and self-consistent Born approximation (SCBA). We

204: assume thermal equilibrium in the electron-hole plasma, but, in order

205: to avoid analytic continuation problems, we do not resort to the

206: Matsubara Green function technique,\cite{mahan81} which becomes

207: numerically delicate when the spectral function shows a large number

208: of poles. Instead, we work directly in the real time/frequency

209: space. This has the additional advantage that the approach can be

210: readily extended to non-equilibrium.\cite{danielewicz} We thus

211: consider four single particle Green's Functions:\cite{haug96} the

212: retarded, advanced, lesser and greater Green's functions defined as

213: \begin{mathletters}

214: \label{eq:greens}

215: \begin{eqnarray} 

216: G^+(1,2)&=&-i \theta(t_1-t_2) \langle T [c(1) c^\dagger(2)]_+ \rangle,

217: \label{eq:g+} \\ G^-(1,2)&=&i \theta(t_2-t_1) \langle T [c(1)

218: c^\dagger(2)]_+ \rangle, \label{eq:g-}\\ G^<(1,2)&=&i \langle

219: c^{\dagger}(2) c(1) \rangle, \label{eq:g<}\\ G^>(1,2)&=&-i \langle

220: c(1) c^{\dagger}(2)\rangle . \label{eq:g>}

221: \end{eqnarray}

222: \end{mathletters}

223: Here $T$ is the time ordering operator, $1=(x_1,t_1,\sigma_1)$, where

224: $x_1$ is the position, $t_1$ the time, and $\sigma_1$ is the spin

225: index. $c(1)$ is the electron annihilation operator and $[~ ]_+$ is

226: the anticommutator.  Similar expressions for the hole Green's

227: functions hold, with $ d(1)$ and $ d^{\dagger}(1)$ the hole

228: annihilation and creation respectively.  Only two of the four

229: functions above are independent, and in particular, we will consider

230: the retarded and lesser (or correlation) functions. Formally, a single

231: Green's function $G(x_1,\tau_1,x_2,\tau_2)$, may be introduced for

232: compactness of notation, with $\tau_1$ and $\tau_2$ defined on the

233: Keldysh contour, having the two branches shown in

234: Fig. \ref{fig:keldysh}. Following simple rules, any equation with

235: times defined on the Keldysh contour can be written back in a set of

236: equations defined on the ordinary time axis for the above four Green

237: functions (see for instance Ref. \onlinecite{haug96}).

238: \begin{figure}

239: \centerline{\psfig{figure=ptfig1.eps,width=8truecm}}

240: \vspace{1cm}

241: \caption{The Keldysh contour. Upper branch, normal time ordering,

242: lower branch, reversed time ordering.}

243: \label{fig:keldysh}

244: \end{figure}

245: The statistical averages are done in the macro-canonical ensemble:

246: $\langle \dots \rangle={\rm Tr}\{\rho \dots\}$ with $\rho=

247: e^{-\beta(H-\mu_{e} N_e -\mu_{h} N_h)}/{\rm Tr} \{e^{-\beta(H-\mu_{e}

248: N_e-\mu_{h} N_h)}\}$.  Here $\beta=1/T$ and $H$ is the total

249: Hamiltonian of the interacting electron-hole system.  In principle,

250: electrons and holes have independent chemical potentials, and

251: densities. However, we are interested in the description of a laser

252: excited semiconductor where the density of electrons and holes is the

253: same, and charge is balanced.  We will also assume that the

254: temperature of the two components of the interacting gas is the

255: same. In fact, after excitation, thermalization of carriers is mainly

256: driven by exchange of energy with the phonon thermal bath which, at

257: equilibrium, leads to $T_e = T_h = T $.  Moreover, for simplicity we

258: consider the same electron and hole masses, giving $\mu_e=\mu_h=\mu$

259: and the same Green functions and self-energies. Extension of the

260: theory to different masses is straightforward. Finally, the Green

261: functions above depend only on the time difference $t_2-t_1$ as the

262: system is stationary, and on the relative distance $x_2-x_1$ as the

263: system is homogeneous.

264: 

265:  In this thermal regime, the retarded and lesser Green functions

266: defined above are also related by the Kubo Martin Schwinger relations

267: \cite{kadanoff62}

268: \begin{mathletters}

269: \label{eq:KMSfermi}

270: \begin{eqnarray}

271: G^<(k,\omega)=-2 \Im [G^+(k,\omega)] f(\omega-\mu)\\ G^>(k,\omega)=2

272: \Im [G^+(k,\omega)](1-f(\omega-\mu)),

273: \end{eqnarray}

274: \end{mathletters}

275: where we considered the Fourier transforms with respect to the

276: relative time $t_1-t_2$, and with respect to the relative position

277: $x_1-x_2$. The function $f(\omega)=[exp(\omega/T)+1]^{-1}$ is the

278: Fermi function. In the stationary case

279: $G^{+}(k,\omega,\sigma)={G^{-}}^*(k,\omega,\sigma)$ contain the

280: information on the spectral properties of the quasiparticles, given by

281: the one particle spectral function $A(k,\omega,\sigma)=-2 \Im

282: [G^+(k,\omega,\sigma)]$.  The correlation functions instead contain

283: information on the quasi particle occupation number

284: $G^<(k,t_1=t_2=t,\sigma)=N(k,t,\sigma)$, and

285: $G^>(k,t_1=t_2=t,\sigma)=1-N(k,t,\sigma)$, where $N(k,t\sigma)$ are

286: the occupation numbers.

287: 

288: The on-site Coulomb interaction reads:

289: \begin{equation}

290: V(x) = \pm a \delta(x)~,

291: \label{eq:delta}

292: \end{equation} 

293: the potential is repulsive for electron-electron and hole-hole, and

294: attractive for electron-hole interaction. An attractive delta-like

295: potential between the electrons and the holes gives one bound state

296: only for the pair problem, and $a$ can be chosen such to reproduce the

297: exciton binding energy measured in the limit of low density.  The

298: constant $a$ has the dimensions of $energy \times length$, the exciton

299: Bohr radius is given by $a_B=\hbar^2/(m a)$ and the exciton binding

300: energy is $E_b=m a^2/2 \hbar^2$ where $m$ is the reduced mass of the

301: electron-hole system. We will use the units $\hbar=a=m=1$ throughout

302: the paper. In these units $a_B=1$ and $E_b=0.5$. We remind that

303: typical values for semiconductor quantum wires are $a_B=10^{-6}$ cm

304: and $E_b=10$ meV.\cite{optwires} With the above effective potential of

305: Eq. (\ref{eq:delta}) the interaction Hamiltonian is

306: $$H_C = \frac{1}{2}\sum_{1} c^\dagger (1)c^\dagger(1')c(1')c(1)+$$

307: $$+ \frac{1}{2}\sum_{1} d^\dagger (1)d^\dagger(1')d(1')d(1)+$$

308: \begin{equation}

309: -\sum_{1}

310: c^\dagger(1)d^\dagger(1')

311: c(1)d(1')~.\label{eq:H}

312: \end{equation} 

313: Here $1'=(x_1,t_1,\sigma'\ne \sigma)$ for the electron-electron and

314: hole hole interaction, while $1'=(x_1,t_1,\sigma')$ for the

315: electron-hole interaction, as due to the fermionic nature of the

316: carriers, only interaction with opposite spin is allowed in the

317: intraband term for a contact potential.  In the following, we assume a

318: spin-polarized system of electrons and holes, so that we keep only the last

319: term describing the electron-hole interaction in the Hamiltonian

320: (\ref{eq:H}). Generalization to both spins is straightforward.

321: 

322: 

323:  

324: The Dyson equation for the single particle $G(1,2)$ propagators read

325: \begin{equation}

326: G(1,2)=G_0(1,2)+\int

327: d\bar 3~d\bar 4 G_0(1,\bar 3)\Sigma(\bar 3,\bar 4)G(\bar 4,2)~, $$

328: \label{eq:dyson}

329: \end{equation}

330: where time integrations are over the Keldysh contour.  $G_0$ is the

331: free propagator. The functional dependence of the self energy $\Sigma$

332: on the single particle Green's functions determines the degree of

333: approximation. In particular, it can be expressed through an

334: electron-hole scattering kernel $T(1,2;1',2')$:

335: \begin{equation}

336: \Sigma(1,2)= i \int d\bar 1'~d\bar 2'~T(1,\bar 1';2,\bar 2') G(\bar

337: 2',\bar 1') \label{eq:sigma}

338: \end{equation}

339: In the Born approximation,

340: $$T(1,1';2,2')=V(1,1')+$$

341: \begin{equation} +

342:  V(1,1')i G(1,2)G(1',2') V(2,2'), $$

343: \label{eq:tborn}

344: \end{equation}

345: where $V(1,1')= V(x_1-x_{1'}) \delta(t_1,t_{1'})$, and $\delta(\tau,\tau')$

346: is the Dirac delta function extended on the Keldysh contour.  The

347: self-energy $\Sigma(k,\omega)$ can be calculated at the single

348: particle pole $\omega = k^2/4 $. In this case we are neglecting memory

349: terms in the scattering and thus using a Markov approximation; this

350: corresponds to the MBA. In the SCBA, the same scattering kernel of

351: Eq. \ref{eq:tborn} is used, but the full energy dependent structure of

352: the resulting self-energy given in Eq. (\ref{eq:sigma}) is used in the

353: Dyson equations Eq. (\ref{eq:dyson}).  The problem is then solved

354: self-consistently, as explained later. In this case, memory effects,

355: beyond the Markov approximation, are also included.  In the SCLA, the

356: kernel $T(1,1';2,2')$ is the solution of the Bethe Salpeter equations

357: in the ladder approximation, which for a generic potential reads

358: $$T(1,1';2,2')= V(1,1') +$$

359: \begin{equation} 

360: +\int d\bar 3~d\bar 3'~V(1,1')i G(1,\bar 3)G(1',\bar 3') T(\bar 3,\bar 3';2,2')

361: \label{eq:bethesalpetereh}

362: \end{equation} 

363: For a $\delta$-like potential, $V(1,1')=-\delta(1,1')$ and $T$ depends

364: only on $2-1$. The reduced equation reads:

365: \begin{equation} 

366: T(1;2)= -1 - i \int d{\bar 3}~G(1,\bar 3)G(1,\bar 3) T (\bar 3;2)

367: \label{eq:bethesalpetereh2}

368: \end{equation}

369:  The Eqs. (\ref{eq:dyson}), (\ref{eq:sigma}) and

370: (\ref{eq:bethesalpetereh}) are schematically represented in

371: Fig. \ref{fig:diagrams} for both the SCBA and the SCLA.

372: \begin{figure}

373: \centerline{\psfig{figure=ptfig2.eps,width=8truecm}}

374: \caption{(a) The Dyson equation, thick lines are the dressed

375: one-particle Green function, thin lines the bare ones. (b) The

376: self-energy with the scattering kernel $T(1;2)$, for the on-site

377: Coulomb potential. (c) The scattering kernel $T(1;2)$ for the Born

378: approximations. (d) The scattering kernel $T(1;2)$ in the ladder

379: approximation, given by the Bethe-Salpeter equation.}

380: \label{fig:diagrams}

381: \end{figure}

382: 

383: 

384: In the self-consistent approximation we solve self-consistently

385: Eqs.(\ref{eq:dyson}) and (\ref{eq:sigma}) with the scattering kernels

386: given by Eq. (\ref{eq:tborn}) or (\ref{eq:bethesalpetereh}) in the

387: Born or SCLA case respectively.  We explain in details how numerical

388: self consistency for a fixed temperature and chemical potential $\mu$

389: has been implemented.

390: 

391: 

392: (i) We start with ${G^+}(k,\omega)$ of the form ${G^+}_{0}=(\omega-

393: k^2/4+i \gamma)^{-1}$, and ${G^<}_0(k,\omega)=-2 \Im

394: [{G^+}_{0}(k,\omega)]~f(\omega-\mu)$.

395: 

396: (ii) In order to take advantage of the local character of the

397: electron-hole interaction, we use the Fourier transform of the Green

398: functions of Eq. (\ref{eq:greens}), $G(1,2)$, which actually have only

399: a spatial dependence on $r=x_1-x_2$ as the system is homogeneous.  We

400: calculate the free pair propagator defined as

401: \begin{equation}\label{eq:hr}

402:  H(1;2)= i G(1,2) G(1,2)

403: \end{equation} 

404: The retarded and correlation parts are calculated passing from the

405: Keldysh contours to the usual time axis:

406: \begin{equation}\label{eq:hr<}

407:  H^<(r,t)= i G^<(r,t) G^<(r,t),

408: \end{equation}\begin{equation}

409:  H^+(r,t)= i G^+(r,t) G^+(r,t)+

410: \label{eq:hr+} 2 i G^+(r,t) G^<(r,t).

411: \end{equation}

412: The dependence on the relative distance has been explicitly

413: shown. The dependence on the relative propagation time $t=t_2-t_1$

414: stems from the stationary condition. The pair propagator is then

415: Fourier transformed to $(k,w)$ space.

416: 

417: (iii.a) For the Born approximations, we use Eq. (\ref{eq:tborn})

418: \begin{mathletters}

419: \begin{eqnarray}

420: T^+(q,\omega)=-1+H^+(q,\omega)\\ T^<(q,\omega)=H^<(q,\omega)

421: \end{eqnarray}

422:  \end{mathletters}

423: 

424: (iii.b) For the SCLA case, Eq. (\ref{eq:bethesalpetereh2}) is readily

425: solved as:

426: \begin{mathletters}

427: \begin{eqnarray}\label{eq:t+w}

428: T^+(q,\omega)=-(1+H^+(q,\omega))^{-1}~,\\

429: T^<(q,\omega)=|T^+(q,\omega)|^2 H^<(q,\omega)~.

430: \end{eqnarray}

431: \end{mathletters}

432: 

433: (iv) The electron self-energy is then calculated in real space as in

434: Eq. (\ref{eq:sigma}). The retarded and correlation functions, defined

435: on the usual time axis, read:

436: $$

437: \Sigma^<(r,t)= i T^<(r,t)G^>(-r,-t)=-i T^<(r,t)[G^>(r,t)]^*,$$

438: $$

439: \Sigma^+(r,t)= - i T^<(r,t) G^-(-r,-t) +i T^+(r,t) G^<(-r,-t)=

440: $$

441: $$= - i T^<(r,t) [G^+(r,t)]^*+ i T^+(r,t) [G^<(r,t)]^*.$$

442: 

443: (v) The new electron Green's functions are then calculated as

444: \begin{mathletters}

445: $$

446: G^+(k,\omega)=(\omega- k^2/4+\Sigma^+(k,\omega)+i\gamma)^{-1},$$$$

447: G^<(k,\omega)=-2 \Im [G^+(k,\omega)]~f(\omega-\mu).$$

448: \end{mathletters} 

449: 

450: 

451: 

452: The procedure is repeated through step (ii) until self-consistence is

453: reached. Fast Fourier transforms are performed on a finite grid of

454: 16384$\times$128 points for the frequency and wavevector domains

455: respectively.  The frequency and wavevector domain $(k,w)$ is

456: (-20,20)X(-3,3).  Due to the finite k-range, we obtain $E_b=0.4$. The

457: external $\gamma>0$ in (v) is adiabatically switched off during 

458: self consistency. In this way the imaginary part of the final $G$ is

459: provided by the interaction only. For the non-self-consistent Born

460: approximation (or Markov-Born), we use the frequency independent,

461: on-pole $\Sigma^+_{k}(\omega=k^2/4)$ in the Green function, and stop

462: the procedure at step (v).

463: 

464: All of the considered approximations are conserving in the sense of

465: Kadanoff and Baym:\cite{baym61} the total density, total momentum,

466: total energy and total angular momentum of the system are

467: conserved. This is a fundamental property of any approximation for the

468: self-energy, otherwise unphysical results may result.  We

469: will come back to this point again in the calculation of the optical

470: response of the system in the various approximations.

471: 

472:   The polarized electron-hole gas with a contact potential interaction

473: that we consider here can be mapped onto a single band Hubbard model

474: with spin 1/2 and attractive interaction, in the limit of infinite

475: width of the band, and infinite on-site interaction so as to produce

476: finite $E_b$ (continuum limit). For the Hubbard model with repulsive

477: interaction, the self-consistent ladder approximation is known as

478: fluctuation exchange approximation (FLEX).\cite{flex} Even though in

479: this repulsive case the opening of an Hubbard gap is not reproduced,

480: in the attractive case the gap is of different nature and is

481: notoriously well described within this approximation.\cite{kagan98}

482: 

483: 

484: 

485: \section{Single Particle properties}

486: \label{sec:single}

487: 

488: \begin{figure} 

489: \centerline{\psfig{file=ptfig3.eps,width= 9 cm}}  

490: \caption{Electron spectral function at $k=0$ as a function of the

491: total carrier density. $T=0.1$.}

492: \label{fig:spectral1}

493: \end{figure}

494: We plot in Fig. \ref{fig:spectral1} the electron spectral function at

495: $k=0$ for different densities, obtained in the SCLA. The temperature

496: $T=0.1\ll E_b=0.5$. For this particular figure we used the k-range

497: (-6,6) in the numerical solution. For $n\sim 0.001$ the electron

498: spectral function has a very narrow Lorentzian shape.  For larger

499: densities, a satellite structure appears in the low energy side of the

500: spectral function. This structure is located at the exciton binding

501: energy, below the main peak, and accounts for the correlation of the

502: electron at $k_e=0$ bound with holes in other $k_h$ states. Excitons

503: with all values of the center of mass wavevectors are involved in this

504: peak,\cite{zimmermann88} as we will also show in the next section.

505: This correlation structure is of course not present in the Born

506: approximations, which maintain their single quasi-particle peak

507: structure at any density. As the density rises, the relative weight of

508: the satellite structure in the SCLA spectral function with respect to

509: the main quasi-particle peak increases. Moreover, both structures

510: become broader. For densities $n>0.1$, it becomes difficult to

511: distinguish between the excitonic and the main peak, and a single

512: broad, red-shifted quasi-particle structure appears. The merging of

513: the excitonic peak with the single quasi-particle peak can be

514: interpreted as a Mott transition. We remark however that it is

515: difficult to give a rigorous estimate of the Mott density, because

516: broadening is gradual and a sharp threshold is not evident.

517: 

518: 

519: 

520: \begin{figure} 

521: \centerline{\psfig{file=ptfig4.eps,width=10 cm}}

522: \caption{Self energy $\Sigma (k=0,\omega) $ for $T=0.1$ and different

523: densities given in the Figure, in the self consistent Born

524: approximation (SCBA) and self-consistent ladder approximation

525: (SCLA).}

526: \label{fig:self}

527: \end{figure}

528: In Fig. \ref{fig:self} we show the imaginary part of the electron self

529: energy for different densities. We will also refer in the following to

530: this quantity as (energy dependent) broadening. In the SCLA

531: $\Im(\Sigma)$ shows two peaks corresponding to dephasing experienced

532: by an unbound or a bound state propagated in the system. Both peaks

533: increase with density. At low density, $n=0.01$, dephasing at the

534: exciton energy dominates, while at $n=0.05$ both peaks become

535: comparable.  In the SCBA, the peak of $\Im(\Sigma)$ at $-E_b$ is

536: obviously absent, as propagation of bound pairs is not allowed in the

537: theory. Instead at low density $n=0.01$, the broadening at the main

538: quasi-particle energy ($\omega=0$) is similar in the SCLA and

539: SCBA. Indeed, few excitons are expected in the plasma, and scattering

540: mainly originates from free carriers.  An heuristic understanding may

541: be obtained with the action mass law relating the concentration of

542: different chemical species in a reaction. In our case, electron+hole

543: $\leftrightarrow$ exciton, and the law states ${n_c}^2/n_X=n^*(T)$,

544: where $n_c$ and $n_X$ represent the density of unbound carriers and

545: excitons respectively. $n^{*}(T)$ is a cross-over density which

546: depends only on the temperature, and at T=0.1, $n^{*}\sim 0.005$ . For

547: $n < n^*$, $n_X \ll n$. Therefore, a small density of excitons is

548: expected at $n=0.01\sim n^*$. At this density the exciton-free carrier

549: contribution to free carrier broadening is weaker than the free

550: carrier-free carrier contribution.  At larger density, bound excitons

551: in the plasma become dominant, and broadening at the quasi-particle

552: peak becomes much larger in the SCLA than in SCBA. This trend is

553: observed up to $n=0.15$ indicating relevance of correlated states in

554: the plasma even at this large density. However, the heuristic

555: interpretation of the broadening based on the action mass law given

556: above becomes meaningless, as it is not possible to distinguish

557: between two chemical species anymore.  At the highest considered

558: density $n=0.4$, we clearly observe a dip to very small broadening in

559: the self-energy for SCBA case. This dip is exactly at the Fermi energy

560: and accounts for the blocking of the scattering at the Fermi level. In

561: fact, only at this large density the Fermi gas becomes degenerate

562: (Fermi energy much larger than T). In the SCLA, the broadening never

563: vanishes, but is only partially reduced at the Fermi level. This is

564: indicative of a more complicated structure of the electron-hole plasma

565: and of the broadening process, where Pauli blocking looses much of its

566: effectiveness, and is suggestive of a non-Fermi liquid behavior.  We

567: finally noticed that $\Re(\Sigma)$ is comparable to $\Im(\Sigma)$ at

568: any density in the SCLA, and thus remarked that the excitonic

569: satellite peak in the spectral function does not correspond to a zero

570: of $\omega-\Re(\Sigma(\omega))$. Indeed, we do not expect appearance

571: of another simple quasi-particle in the plasma. For $\omega\sim -E_b$,

572: and for $E_b>>|\Im(\Sigma)|$, we have $ A_{k=0}\sim

573: |\Im(\Sigma)|/|E_b|^2$. Thus, the correlation satellite in the

574: spectral function follows the peak of $\Im(\Sigma)$ at $\omega=-E_b$.

575: 

576: 

577: In Fig.\ref{fig:mucomp} we show the density dependence of the electron

578: (or hole) chemical potential as a function of the density for $T=0.1$,

579: in the three considered approximations. At very low density, $n \ll

580: 0.01$, the chemical potential is $ \mu \ll - 0.4 $, and its value is

581: similar in all approximations. In this case, we are describing free

582: electron-hole pairs, as also suggested by a law of mass action, which

583: gives a crossover density of about $n^*=0.005$ at T=0.1. In this case

584: the description of $\mu$ is rather good even in approximations that

585: neglect existence of bound states. 

586: \begin{figure} 

587: \centerline{\psfig{file=ptfig5.eps,width= 8 cm}}  

588: \caption{Chemical potential $\mu$ in the three approximations, at

589: T=0.1.  The short-dashed line shows half the chemical potential of

590: non-interacting bosons, with ground state energy of $-E_b$.  }

591: \label{fig:mucomp}

592: \end{figure} 

593: However, for $n > 0.01$, the

594: chemical potential in the Born approximations is much larger than for

595: the SCLA. In fact, in the SCLA we are also describing the fraction of

596: cold interacting excitons in the plasma.  As the ground state energy

597: of this part of the plasma is at about $-E_b$, the total energy of the

598: system is thus reduced with respect to that of a gas of unbound

599: particles. In order to clarify further this point, we plot on the same

600: graph half of the chemical potential of bosons at a ground state

601: energy of $-E_b$ and mass equal to the exciton mass. 

602: %(2 in our units). 

603: This is the chemical potential of a gas of electron (and

604: holes) completely bound into bosonic excitons. The halving comes from

605: the equilibrium condition $\mu_{X}=\mu_e+\mu_h=2 \mu$.  The chemical

606: potential of bosons compares reasonably well with the chemical

607: potential calculated in the SCLA, up to $\mu<-0.25$, $n<0.1$. Above

608: this limit, the chemical potential from the SCLA grows faster. There

609: are two reasons for this faster growth: first, the exciton gas is

610: repulsively interacting, second, the exciton gas eventually undergoes

611: a Mott transition and ionizes into free carriers. At densities

612: $n>0.2$, the SCLA and SCBA are indeed comparable, and at even larger

613: density, also the MBA approximation is reasonable. This shows that the

614: SCLA is a powerful tool for the investigation of the intermediate

615: regime of densities, where deviations from both the (non-interacting)

616: bosonic and pure fermionic models are important.

617: 

618: 

619: 

620: 

621: 

622: \section{Mapping the SCLA to a bosonic model at low density and temperature}

623: \label{sec:excpot}

624: 

625: 

626: 

627: In Fig.\ref{fig:spectral2} we plot $A(k,\omega)$ for $T$=0.043 and

628: n=0.02. We can observe the parabolic dispersion of the main

629: quasi-particle peak, broadened by the scattering (white region in the

630: plot). Broadening is larger at small $k$ due to the larger phase space

631: in one dimension. For $\omega<0$, we have the correlation structure

632: shown for $k=0$ in Fig. \ref{fig:spectral1}, which is also present at

633: larger $k$.  We remark that this region of correlated electrons

634: extends into a region of $k$ of the order of 1 (i.e. of

635: $a_B^{-1}$). Its shape (dispersion) is related to the exciton

636: wavefunction, as we show in the following.

637: 

638: \begin{figure} 

639: \centerline{\psfig{file=ptfig6.eps,width= 8. cm}}

640: \caption{Contour plot of $A_e(k,\omega)$ at $T=0.04$ and n=0.02.}

641: \label{fig:spectral2}

642: \end{figure}

643: 

644: In a low-density, low-temperature limit, we can start to consider

645: $G=G_0$, and analytically calculate the retarded pair propagator

646: $H^+(k,\omega)$ from Eq. (\ref{eq:hr+}), neglecting $G_0^<$, which is

647: of the order of the density. We obtain:

648: $$ H^{(0)+}(k,\omega)\sim

649: \frac{-1}{\sqrt{2}\sqrt{\omega-\frac{k^2}{8}+2 i \gamma}},$$ where

650: $\gamma>0$ is the usual regularization number. Solving for the

651: T-matrix with the Bethe-Salpeter equation (\ref{eq:t+w}), and

652: expanding around its pole we obtain:

653: $$T^{(0)+}(k,\omega)=\frac{1}{\omega+\frac{1}{2}-\frac{k^2}{8}+2 i

654: \gamma}.$$ Thus, the T-matrix at the lowest order for the contact

655: potential has the form of a free propagator, for particles at an

656: energy $-1/2+k^2/8$, having a mass which is twice the electron mass,

657: i.e. the electron plus hole mass. Expanding the Bethe-Salpeter

658: equation to higher order, we obtain:

659: $$T^{(1)+}(k,\omega)=T^{(0)+}(k,\omega) H^{(1)+}(k,\omega)

660: T^{(0)+}(k,\omega).$$ Therefore, we may interpret $T^{(0)+}(k,\omega)$

661: as the free exciton (boson) Green function, and $ H^{(1)+}(k,\omega)$

662: as the lowest order self-energy.\cite{haussman} In particular

663: \begin{equation}\label{eq:hr1}

664: H^{(1)}(1,2)=2 i G^{(1)}(1,2) G_0(1,2),\end{equation}

665: $$G^{(1)}(k,\omega)= G_0(k,\omega) \Sigma^{(0)}(k,\omega)

666: G_0(k,\omega),$$

667: \begin{equation}\label{eq:sigmar0}

668: \Sigma^{(0)}(1,2)= i T^{(0)}(1,2) G_0(2,1).\end{equation} As

669: $T^{(0)}(k,\omega)$ is peaked close to $\omega=-1/2$, we may neglect

670: $G_0^<(k,\omega)$ in this region, and use

671: $$\Sigma^{(0)+}(k,\omega)\sim i\int\frac{dqd\omega'}{(2\pi)^2}

672: T^{(0)<}(q,\omega') G_0^{+}(k+q,\omega+\omega').$$ Integrating around

673: $\omega'=-1/2$, we obtain:

674: $$\Sigma^{(0)+}_k(\omega)\sim

675: \frac{n}{\omega+\frac{1}{2}+\frac{k^2}{8}+i\gamma},$$ where $n$ is the

676: total density. This is the structure shown in Fig. \ref{fig:self} at

677: $\omega=-1/2$, which produces the satellite correlation peak in the

678: electron spectral function, shown in Fig. \ref{fig:spectral2}. We also

679: notice that this structure has {\em negative} dispersion, as also

680: apparent in Fig. \ref{fig:spectral2}.

681: 

682: \begin{figure}

683: \centerline{\psfig{figure=ptfig7.eps,width=8truecm}}

684: \caption{The perturbation expansion to lowest order for the boson

685: (exciton) self-energy, and the four-particle vertex $F^{(0)}$.  Thin

686: lines are the bare electron (or hole) propagator. The shaded

687: two-particle Green function is the bare exciton propagator.}

688: \label{fig:excexc}\end{figure}

689: 

690: We can now define an exciton-exciton interaction at low density, from

691: the Eqs. (\ref{eq:hr1}-\ref{eq:sigmar0}) above, which are pictorially

692: shown in Fig. \ref{fig:excexc}. We thus define a four-point

693: interaction kernel $F^{(0)}$, and the boson self-energy $H^{(1)}(1,2)$

694: is written as:

695: \begin{equation}

696: H^{(1)}(1,2)= \int d\bar{1}~ d\bar{2}~F^{(0)}(1,2;\bar{1},\bar{2})

697: T^{(0)}(\bar{2},\bar{1}). \end{equation} We can see from

698: Fig. \ref{fig:excexc} that $F^{(0)}$ represents the electron (or hole

699: exchange) in the scattering of the two excitons. Thus

700: \begin{equation}

701:  F^{(0)}(1,2;1'2') = 2 i G_0(1,2) G_0(1',2') G_0(1,2') G_0(1',2).

702: \end{equation} If we neglect retardation

703: (or memory) effects, we may define an instantaneous potential at the

704: exciton energy using $\tau_1=\tau_2$, $\tau_{1'}=\tau_{2'}$, and

705: calculating the resulting $F$ at the exciton frequency

706: $\omega_0=-1/2=E_B$. For simplicity, we also use $k,k'\sim 0$ for the

707: incoming excitons, and arrive to

708: $$F^{(0)}(q,\omega_0) \sim 2i \sum_{q',\omega'}

709: G_0^{+}(q',\frac{\omega_0}{2}+\omega')

710: G_0^{+}(-q',\frac{\omega_0}{2}-\omega')$$ \begin{equation}

711: G_0^{+}(q-q',\frac{\omega_0}{2}+\omega')

712: G_0^{+}(q'-q,\frac{\omega_0}{2}-\omega')=\end{equation}

713: $$=\frac{64}{q^2}\frac{2\sqrt{4+q^2}-(4-q^2)}{(4+q^2)^2}.$$ This

714: function rapidly drops to zero at $q>1$ as expected, and

715: $F^{(0)}(0,\omega_0)=6$.  This value is also obtained from standard

716: boson-boson exchange-interaction expressions (see

717: e.g. Ref. \onlinecite{haughanamura}) when only the electron-hole

718: interaction is used.  We show the full structure of

719: $F^{(0)}(q,\omega_0)$ In Fig. \ref{fig:vq}.

720: \begin{figure}

721: \centerline{\psfig{figure=ptfig8.eps,width=8truecm}}

722: \caption{ The exciton-exciton interaction potential $W_q=F^{(0)}(q,\omega_0)$.}

723: \label{fig:vq}

724: \end{figure} 

725: 

726: 

727: We are now in a position to discuss some of the qualitative changes

728: introduced by simplifying the Coulomb interaction to an on-site

729: one. In the limit of low density, we may in fact compare the

730: exciton-exciton interaction calculated above with that calculated

731: using a realistic Coulomb interaction, as a standard boson-boson

732: exchange expression at small momenta exists.\cite{tassone99xx} In a

733: realistic wire, a short-range cut-off is introduced by the finite size

734: of the electron and hole wavefunctions in the confinement directions.

735: A typical cut-off is of the order of the Bohr radius or

736: smaller. Typically, we have to consider tight confinements in order to

737: avoid participation of higher confined levels into the exciton

738: wavefunction. The Coulomb interaction is then reasonably represented

739: by a function $V_r=A/(r+r_0)$, where $r_0$ is a cut-off distance, and

740: $A$ is then chosen such that $E_b=-1/2$. 

741: %Using $r_0=1$, we have

742: %$A=0.65$ and a resulting Bohr radius of $0.4$ only. In this limit, the

743: %electron and hole wavefunctions are not tightly confined.

744: We used $r_0=0.1$, and $A=0.18$. In this case the resulting Bohr radius

745: is 1, indicating tight confinement of the carriers in the confinement

746: plane. The exciton-exciton exchange interaction at small $q$ is

747: calculated in the small $q \ll {a_B}^{-1}$ limit

748: $$W_{xx}=2\sum_{k,k'}V_r(k-k')\phi_{1s}(k)\phi_{1s}(k')$$

749: \begin{equation}

750: \left[|\phi_{1s}(k)|^2-\phi_{1s}(k)\phi_{1s}(k')\right]\sim

751: 1.4~, \label{eq:xxbose}

752: \end{equation}

753: where $\phi_{1s}(k')$ is the 1s wave-function of the exciton. The

754: positive term in Eq. \ref{eq:xxbose} is the exchange term due to the

755: attractive electron-hole interaction, while the negative term is due

756: to the electron-electron and hole-hole repulsive interactions.  In the

757: delta like-potential the negative term is absent because of

758: locality. In the long range case instead, the electron-electron (and

759: hole-hole) interaction largely cancel the electron-hole interaction in

760: the exchange integral, resulting in an exciton-exciton interaction

761: which is four times smaller when compared with the value of 6 obtained

762: for $F^{(0)}(q=0,\omega_0)$. We conclude that local interaction leads

763: to an over-estimation of the boson-boson interaction at a given

764: density, and therefore of broadening. However, this does not imply

765: that broadening effects are {\em qualitatively} different. Moreover,

766: we also stress that this overestimation does not concern at all the

767: comparison of the SCLA with the other simpler approximations

768: considered in this work, as all are carried out using the same

769: interaction potential.

770: 

771: 

772: 

773: \section{Optical Properties}

774: \label{sec:optical}

775: 

776: We now consider interaction of the electron-hole system with a

777: transverse electromagnetic field in the dipole approximation:

778: \begin{equation}

779: H_P=\sum_{{\bf q },q} C_{{\bf q},q} (a_{{\bf

780: q},q}P^{\dagger}_{q}+a^{\dagger}_{{\bf q},q}P_{q}),

781: \label{eq:Hp}

782: \end{equation}

783: where

784: $$ C_{{\bf q},q}=\frac{e p_{cv}}{m_0 c}\sqrt{\frac{4\pi \hbar c}{ q_t

785: V}} I_{\parallel}({\bf q}),

786: $$

787: is the dipole matrix element, which includes also the overlap integral

788: of the electromagnetic field with the confined carriers

789: $I_{\parallel}$.  Here ${\bf q }$ represents 2D wavevectors in the

790: plane orthogonal to the wire axis, $q$ are wavevectors along the wire,

791: while $q_t^2={\bf q}^2+q^2$, $m_0$ is the free electron mass, $c$ the

792: velocity of light, and $p_{cv}$ the momentum matrix element between

793: the conduction and hole bands.  $a_{{\bf q},q}$ is the photon

794: destruction operator, and $P_{q}$ is the polarization operator,

795: defined as $P_{q}=\sum_{k}d_{q-k}c_{k}$, i.e. local in real space in

796: the dipole approximation.  As the section of the wires is always much

797: smaller than the wavelength of light, we have $I_\parallel\sim

798: 1$. Moreover $q_t \sim E_{GAP}/(\hbar c)$ is almost constant, thus we

799: use arbitrary units from now on, with $C=1$. Actual values of

800: absorption and gain are easily calculated by using the appropriate

801: value of $C$ for the given material.

802: 

803: 

804: 

805: In the process of scattering of light from the system, an absorbed

806: photon creates a coherent electron-hole pair which propagates in the

807: crystal, interacts with the background plasma, then recombines, and a

808: photon is emitted.  When we introduce photon propagators, the

809: interaction process is represented by a photon self-energy. We will

810: not solve the photon Dyson equation, nor we will dress the electron

811: and hole propagation with the photon interaction. In other words, we

812: neglect both polaritonic effects, and higher order non-linear

813: interactions. In fact, we are restricting our attention to weak

814: external fields, i.e. assuming interaction of the plasma with the

815: photon field to be weak. Otherwise, it could be possible for the

816: system to be driven out of equilibrium, because the photon chemical

817: potential is zero and far below that of the elecron-hole plasma. This

818: equilibrium condition is well fulfilled in real systems, exception

819: made for the strongest laser fields.  Within this weak interaction

820: approximation, the photon self-energy is exactly given by the pair

821: correlation function

822: \begin{equation}

823: \Pi_q(\tau,\tau')= \langle T P^{\dagger}_{q}(\tau)

824: P_{q}(\tau')\rangle.\label{eq:pairprop}\end{equation} However, this

825: correlation function has to be calculated within some approximation,

826: and also using approximate propagators.  The resulting self-energy may

827: not be conserving. In particular, particle flux conservation at the

828: vertex translates in the longitudinal f-sum rule\cite{baym61}, which

829: has to be respected in any reasonable approximation.

830: 

831: In order to shed light on this important issue, we consider the

832: problem of interaction of photons with the plasma from a different

833: viewpoint. As we are only interested in the linear response to the

834: external electromagnetic field, the correct photon self-energy is also

835: given by the linear response of the plasma to a {\em classical}

836: electromagnetic field, i.e. considering the photon creation and

837: destruction operators in Eq. (\ref{eq:Hp}) as c-numbers.  Baym and

838: Kadanoff give a recipe to construct a conserving expression for this

839: response function in Ref.\onlinecite{baym61}. When an external

840: coherent electromagnetic field is applied to the plasma, coherence

841: develops in the plasma, and pair propagators have additionally to be

842: defined along with the single particle propagators. They are analogous

843: to anomalous propagators in the theory of superconductivity. The

844: electron-hole anomalous propagator reads

845: \begin{equation}\label{eq:pairprop2}

846: G_{eh}(1,2)=i\langle T d(2) c(1) \rangle. \end{equation} 

847: Times are defined on the Keldysh contour as usual. Another anomalous

848: propagator $G_{he}$ can be similarly defined for convenience, and

849: notation regrouped into a matrix one, like for Nambu propagators in

850: the problem of superconductivity.\cite{nambu60} Dyson

851: Eq. (\ref{eq:dyson}) is extended to include both 

852: anomalous propagators, and the external interaction $H_p$:

853: $$

854: \int d\bar{2}~G_0^{-1}(1,\bar{2})G(\bar{2},3)=\delta(1,3)+\int

855: d\bar{4}~\Sigma(1,\bar{4})G(\bar{4},3)+$$

856: \begin{equation}

857: +\int

858: d\bar{4}~\Sigma_{eh}(1,\bar{4})G_{he}(\bar{4},3)-A(1)G_{he}(1,3)\end{equation}

859: $$\int d\bar{2}~G_0^{-1}(1,\bar{2})G_{eh}(\bar{2},3)=\int

860: d\bar{4}~\Sigma(1,\bar{4})G_{eh}(\bar{4},3)+$$

861: \begin{equation}+\int d\bar{4}~\Sigma_{eh}(1,\bar{4})G(\bar{4},3)- A(1)G(3,1)

862: \label{eq:dysonanomal}\end{equation}

863: 

864: As a general rule for generating a conserving approximation, the self

865: energies have to be obtained from functional derivation of a

866: grand-potential with respect to the propagators. The normal

867: self-energies have been already introduced.  For the anomalous self

868: energy $\Sigma_{eh}(1,2)$, we need to generalize the expression of the

869: grand-potential including terms where the normal propagators are

870: replaced by the anomalous ones.  The anomalous propagators $G_{eh}$

871: are at least first order in the external potential $A$, as spontaneous

872: symmetry breaking is ruled out by the Mermin-Wagner theorem.  For

873: linear response, we need the anomalous self energy $\Sigma_{eh}$ up to

874: linear terms in the external potential only, thus at most linear terms

875: in the anomalous propagators. Consequently, we need at most quadratic

876: terms in these propagators in the grand-potential to obtain the

877: $\Sigma_{eh}$ from the functional derivative. We conclude that the

878: only and sufficient term to be considered is the anomalous Fock term,

879: which is the usual Fock term with anomalous propagators replacing

880: normal ones. As the approximation is conserving, we are also

881: guaranteed that absorption fulfills the f-sum rule\cite{baym61} for

882: any value of density and temperature.  The anomalous self-energy

883: obtained from the anomalous Fock term reads

884: \begin{equation}\Sigma_{eh}(1,2)=-i G_{eh} (1,2)+O(A^2).\label{eq:sigmaanomal}

885: \end{equation}

886: Normal $G$ are at least second order in the external potential $A$,

887: while both $G_{eh}$ and $\Sigma_{eh}$ are at least first order. The

888: anomalous Dyson Eq. (\ref{eq:dysonanomal}) to lowest order becomes:

889: $$\int d\bar{2}~G^{-1}(1,\bar{2})G_{eh}(\bar{2},3)=-i \int

890: d\bar{4}~G_{eh}(1,\bar{4})G(\bar{4},3)-$$

891: \begin{equation}

892: - A(1) G(1,3).

893: \label{eq:dysonanomal2}\end{equation}

894: Introducing the response function

895: \begin{equation}

896: \tilde \Pi (1,2)= -i \left.\frac{\delta

897: G_{eh}(1,1)}{A(2)}\right|_{A=0},

898: \end{equation}

899: and taking the first order variation of Eq. (\ref{eq:dysonanomal2}),

900: we obtain and equation for $\tilde \Pi (1,2)$:

901: $$

902: \tilde \Pi (1,2)= i G(1,2)G(1,2) -i\int d\bar{3}~ G(1,\bar{3})

903: G(1,\bar{3}) \tilde \Pi (\bar{3},2)=$$

904: \begin{equation}=

905: H(1;2) - \int d\bar{3}~ H(1;\bar{3}) \tilde \Pi (\bar{3},2),

906: \end{equation}

907: which is easily solved and gives

908: \begin{equation} 

909: \tilde \Pi (1,2)= H(1;2) +\int d\bar{3}~ d\bar{4}~ H(1;\bar{3}) T

910: (\bar{3};\bar{4}) H (\bar{4};2). \end{equation} It is straightforward

911: to show that $\tilde \Pi (1,2)$ obeys bosonic Kubo-Martin-Schwinger

912: relations, as both $H$ and $T$ do. In particular, we have:

913: \begin{eqnarray} 

914: \tilde \Pi^+(q,\omega)= H^+(q,\omega)\left|  T^+(q,\omega)\right|^2 \\ 

915: \tilde \Pi^<(q,\omega)= H^<(q,\omega)\left|  T^+(q,\omega)\right|^2.

916: \end{eqnarray}

917: Absorption $\alpha(\omega)$ and photoluminescence $PL(\omega)$ at

918: $q=0$ are derived as $\alpha(\omega)=-\Im [\Pi^+(0,\omega)]$ and

919: $PL(\omega)=-\Im [\Pi^<(0,\omega)]$ respectively.

920: 

921: We remark that higher order terms than the anomalous Fock one have to

922: be considered in the anomalous self-energy expansion when a {\em

923: finite} external field is present, such as in the case of a strong

924: laser field.  Anomalous Born terms calculated in the Markov

925: approximation have been considered in the

926: semiconductor Bloch equations by Lindberg and Koch in

927: Ref. \onlinecite{lindberg}, and named polarization-polarization

928: scattering terms. Here we do not further pursue such extensions, which

929: are clearly beyond the scope of the present paper. We only remark that

930: our approach is easily extended to such conditions, while keeping full

931: control of the conservation laws.

932: 

933: 

934: \subsection{Optical spectra and stability of the excitonic emission}

935: 

936: We show in Fig. \ref{fig:abs}, the absorption spectra in the normal

937: direction ($q=0$) at $T=0.2$, for different densities, in the MBA,

938: SCBA, and SCLA. In the absorption spectrum, we remark at low density

939: the characteristic excitonic peak and the continuum absorption. The

940: exciton linewidth is extremely narrow in the SCBA, and much broader in

941: the MBA. The SCLA linewidth is intermediate. In the SCBA there is a

942: very small broadening at the exciton energy, as no excitons are

943: represented in the theory, while in the MBA, the free carrier

944: broadening at $\omega=0$ is assumed. This is somewhat larger than the

945: broadening in the SCLA at $T=0.2$.  For large density $ n>0.2$, the

946: excitonic peak is bleached in all approximations, and a region of

947: negative absorption, i.e. optical gain, appears. The absorption

948: changes sign at $\omega=\mu_e+\mu_h=2\mu$, coinciding with the change

949: of sign of the Bose factor, and resulting in a well defined

950: -i.e. positive- photoluminescence through the Kubo-Martin-Schwinger

951: relation.

952: 

953: We plot the emission or PL spectra in Fig. \ref{fig:pl}. First we

954: notice the different scale for the MBA, where a smaller peak emission

955: is calculated even for the largest considered density.  For large

956: density $n>0.2$, we observe a saturation of the intensity. This can be

957: understood in terms of the fermionic nature of the carriers, when the

958: electron-hole plasma becomes degenerate. Second, we notice

959: exciton-like emission even when excitonic absorption is bleached in

960: all models.  In fact, vertex corrections place the free carrier

961: emission at the exciton energy, even when no bound excitons are

962: described in the gas. 

963: \begin{figure}

964: \centerline{\psfig{file=ptfig9.eps,width=11 truecm}}

965: \caption{Absorption spectra in the Markof-Born Approximation (MBA),

966: self consistent Born approximation (SCBA) and self consistent ladder

967: approimation (SCLA) for different densities indicated in the figure.}

968: \label{fig:abs}

969: \end{figure}

970: Eventually, only at very small density the free

971: carrier and excitonic emission become distinguishable, as we notice an

972: exponential emission shoulder at high energy $\omega>0$ in the SCLA

973: and SCBA, reminding of a fermionic emission tail at small degeneracy

974: (Boltzmann distribution). At larger density, the two features at

975: different energy merge into a unique one, due to increased

976: broadening. A single peak is also observed in experiments, where

977: finite noise in the data and large inhomogeneous broadening mask any

978: minor feature.\cite{optwires} For this reason, it is in practice very

979: delicate to establish the position of the band-gap (bottom of the

980: free-carrier bands) directly from emission data.\cite{piermarocchi99}

981: Third, we remark that the low-energy shoulder of the PL emission in

982: the MBA is Lorentzian because of the Markov approximation, and the

983: broadening largely over-estimated also at low energies. The linewidth

984: in the SCBA is instead underestimated at low-density, as for

985: absorption.

986: 

987: \begin{figure} 

988: \centerline{\psfig{file=ptfig10.eps,width=11 cm}}

989: \caption{Photoluminescence spectra in the MBA, SCBA and SCLA for different

990: densities indicated in the figure.}

991: \label{fig:pl}

992: \end{figure}

993: 

994: 

995:  An interesting -and directly observed- physical quantity, is the

996: energy of the PL peak. As remarked above, emission is excitonic due to

997: the correlation of the coherent electron-hole pair emitted. In

998: Fig. \ref{fig:stability} we plot the position of this peak in the PL

999: as a function of the carrier density, together with the band-gap

1000: renormalization (BGR) defined as twice the energy of the main peak of

1001: the spectral function. Interestingly, exciton bleaching in absorption

1002: appears at a density which is comparable to that where the band gap

1003: renormalization crosses the emission energy, i.e. $n\sim 0.15$. This

1004: can therefore be assigned as a Mott density, i.e. the energy where the

1005: binding energy becomes negligible with respect to broadening, thus

1006: effectively vanishing. There are two interesting features to be

1007: observed in Fig. \ref{fig:stability}: first, the band gap

1008: renormalization is negligible in the SCLA for $n<0.1$, second,

1009: emission energy is constant in this range of densities, and

1010: blue-shifting less than in the SCBA for larger density. Comparison

1011: with the MBA is vitiated by excessive broadening in this

1012: approximation. The simple Hartree-Fock approximation instead gives

1013: results which are similar to those obtained in the SCBA for the band

1014: gap renormalization and the PL emission peak, apart from more

1015: pronounced blue-shifts.\cite{tassone99} Stability of the emission peak

1016: is usually interpreted as a partial compensation of the self-energy

1017: and vertex corrections. As in the Hartree-Fock approximation the

1018: broadening effects are missing, we deduce that these effects are

1019: indeed relevant for the cancellation found in the SCBA and in the SCLA

1020: at high density, and explain the reduction of the blue-shift with

1021: respect to simple Hartree-Fock calculations.

1022: \begin{figure} 

1023: \centerline{\psfig{file=ptfig11.eps,width=8 cm}}

1024: \caption{The band gap renormalization (BGR) and emission peak energy

1025: in the SCBA and SCLA as indicated in the figure.}

1026: \label{fig:stability}

1027: \end{figure}

1028: 

1029: 

1030: 

1031: \subsection{Excitonic gain}

1032: 

1033: 

1034: Optical gain is usually considered in the regime of degenerate

1035: electron-hole plasma. In this regime, spectral functions can be

1036: assumed to be simple Lorentzians, and the absorption can be written

1037: as:

1038: \begin{equation}\label{eq:gainfc}

1039: \alpha(\omega)\propto\int \frac{dk}{2 \pi} \frac{1-2 f(k^2/4

1040: -\mu)}{(\omega-k^2/2)^2+ \gamma^2},

1041: \end{equation}

1042: where $\gamma$ is the spectral broadening. Then, negative absorption

1043: or gain occurs only when the Pauli blocking factor ($1-2 f(k^2/4

1044: -\mu)$) becomes negative, which necessarily requires a chemical

1045: potential above the band gap, or in other words, inversion. However,

1046: both vertex corrections, and deviations of the spectral function from

1047: simple Lorentzian should be taken into account at lower density and

1048: temperature. In this case,

1049: absorption can be written as:

1050: \begin{eqnarray}

1051: \alpha(\omega) \propto |T^{+}(q=0,\omega)|^2

1052: \int\frac{dk}{2\pi}\frac{d\omega^\prime}{2 \pi}\times

1053: \label{eq:abstherm}

1054: \end{eqnarray}$$

1055: \left[1-f(\omega-\omega^\prime-\mu)-f(\omega^\prime-\mu)\right]

1056: A_{k}(\omega-\omega^\prime)A_{k}(\omega^\prime)~,

1057: $$

1058: The term in square brackets

1059: $1-f(\omega-\omega^\prime-\mu)-f(\omega^\prime-\mu)$ is the

1060: generalization of the Pauli blocking factor for a system where the

1061: quasiparticles are described by arbitrary spectral functions, and

1062: clearly one of the Fermi functions refers to electron occupation, the

1063: other to hole occupation. This expression of the absorption is valid

1064: only in the case of short range potential, and for the long range case

1065: $T^+$ depends also on the relative momentum of the pair $k$, and

1066: becomes part of the integral kernel.  Using the definition of Fermi

1067: and Bose functions, the absorption can be recast in the form

1068: \begin{eqnarray}

1069: \alpha(\omega)=\frac{|T^{+}(q=0,\omega)|^2}{g(\omega-2 \mu)}

1070: \int\frac{dk}{2\pi}\frac{d\omega^\prime}{2\pi} \times\label{eq:absg}

1071: \end{eqnarray}$$\times

1072: f(\omega-\omega^\prime-\mu)f(\omega^\prime-\mu)

1073: A_{k}(\omega-\omega^\prime)A_{k}(\omega^\prime)~,$$ where

1074: $g(\omega-2\mu)$ is the Bose function.  Therefore, gain clearly occurs

1075: below twice the chemical potential $\mu$ (i.e. $\mu_e+\mu_h$), even

1076: when the chemical potential is {\em below} the band-gap. However, for

1077: it to be sizeable, the spectral functions must have non-negligible

1078: weight in this region of the spectrum.  In the MBA and SCBA case, this

1079: weight is clearly given by broadening effects alone, while in the SCLA

1080: it is also due to the presence of the excitonic correlation peak in

1081: the spectral function as shown in Fig.\ref{fig:spectral1}.  A sizeable

1082: enhancement of the gain is found in correspondence to the excitonic

1083: resonance because of the vertex correction, given by the $|T^+|^2$

1084: factor in Eq. (\ref{eq:absg}). When the $2\mu< -E_B$, both gain and

1085: excitonic absorption coexist. In this case, we talk of {\it excitonic

1086: gain}. The system is inversion-less in the sense of

1087: Eq. (\ref{eq:gainfc}) above. However, only in the SCLA we are

1088: describing gain due to the presence and scattering of excitons in the

1089: low temperature, low-density plasma, while in the other

1090: approximations, gain is purely related to  dynamical effects in the

1091: interaction vertex with the photon.

1092: 

1093: In Fig.\ref{fig:gain} we plot the absorption spectra close to the

1094: exciton resonance at $T=0.1\ll E_B$. For the considered densities

1095: $n<0.1$, gain coexists with the excitonic resonance in absorption.

1096: Only for the MBA at $n=0.1$, the absorption peak is completely shifted

1097: to higher energies, and we can not anymore speak of excitonic

1098: absorption. As usual, the broadening in the exciton spectral region is

1099: too small and unphysical in the SCBA, resulting in very sharp and

1100: unphysical features. Gain in the SCLA is instead over a larger

1101: spectral region, and its value is larger than what predicted in the

1102: MBA at large density $n>0.3$ shown in Fig.\ref{fig:abs}. Excitonic

1103: gain clearly originates from inclusion of excitons in the plasma for

1104: the SCLA. In fact, reasonable broadening is calculated in the exciton

1105: spectral region, and mainly originating from exciton-exciton

1106: scattering as shown in Sec. \ref{sec:excpot}. We conclude, that the

1107: SCLA indicates that sizeable excitonic gain can be obtained at

1108: moderate density and temperature, in a reasonably large spectral

1109: region of the order of a fraction of $E_B$.

1110: 

1111: \begin{figure} 

1112: \centerline{\psfig{file=ptfig12.eps,width=11 cm}}

1113: \caption{Absorption spectra at $T=0.1$ for the MBA, SCBA, and SCLA,

1114: and low densities indicated in the figure.}

1115: \label{fig:gain}

1116: \end{figure} 

1117: Excitonic gain has been widely studied in II-VI quantum wells

1118: \cite{kozlov96}.  However these are two-dimensional system, and

1119: excitonic correlations are expected to be less pronounced in this

1120: systems, so at this stage it is not reasonable to make even a

1121: qualitative comparison.  For quantum wires, sizeable gain at 10 K

1122: ($T\sim 0.05$) for an estimated carrier density well below the Mott

1123: has been recently claimed by Sirigu {\it et al}.\cite{sirigu99} This

1124: is a first indication that excitonic gain might be relevant in these

1125: systems. However, a qualitative and quantitative understanding of this

1126: experiment is clearly premature. From the experimental side,

1127: inhomogeneous broadening due to interface disorder has to be decreased

1128: well below the binding energy, while a more realistic Coulomb

1129: interaction has to be addressed by the theory.

1130: 

1131: 

1132: 

1133: 

1134: \section{Conclusions}

1135: \label{sec:conclusions}

1136: 

1137: We have presented a model that includes excitonic correlation in the

1138: description of a highly-excited semiconductor quantum wire. The model

1139: has been simplified using a short range potential, and considering a

1140: polarized gas.  Correlation has been calculated self-consistently at

1141: the ladder level (SCLA). We have shown that bound states appear as a

1142: low-energy correlation peak in the spectral function of electrons and

1143: holes.  We compared the results obtained with SCLA to the ones

1144: obtained within lower order approximations that do not include

1145: excitonic correlation (Born approximations).  Even though the model is

1146: purely fermionic, we have shown how it can be effectively mapped at

1147: low temperatures and density to a gas of interacting excitons. We have

1148: thus derived an analytical expression for the effective

1149: exciton-exciton interaction.  The linear optical properties of the

1150: system have been calculated including vertex corrections at the Fock

1151: level. This ensures the conservation of sum rules in the optical

1152: response. The excitonic absorption at low density ensues both in

1153: ladder and Born approaches, but the broadenings at the exciton energy

1154: are either too small or too large in the Born approximations,

1155: depending on whether frequency dependence of the broadening is

1156: included or not. Excitonic emission well beyond exciton bleaching is

1157: also predicted in all models, but the peak shifts are more pronounced

1158: in the Born approximations, showing that a better cancellation between

1159: the self-energy and vertex correction results from the introduction of

1160: excitons. We have also shown that sizeable excitonic gain can be

1161: predicted at low temperature and density, when the electron plus hole

1162: chemical potential is just below the exciton energy. However, its

1163: correct description must include exciton broadening from

1164: exciton-exciton scattering, and unphysical values are thus obtained in

1165: the Born approximations. These qualitative conclusions clearly hold

1166: even for more refined descriptions of the electron-hole plasma than

1167: the Born approximations considered here, when these descriptions do

1168: not account for excitons in the plasma. For example this is the case

1169: for the model recently presented by Das Sarma and Wang in

1170: Ref. \onlinecite{dassarma00}, which includes screening at the

1171: plasmon-pole approximation level, but no excitons in the electron-hole

1172: gas. The excitonic gain calculated using this approximation shows a

1173: divergent behavior close to $-E_b$, similar to what we have found for

1174: the SCBA. The extension of the proposed approach to a full

1175: non-equilbrium theory could contribute to a deeper understanding of

1176: many body effects in electron-hole systems, also in the perspective of

1177: ultrafast optical experiments than can investigate far-from

1178: equilibrium conditions. In this context a recent theory by Hannewald

1179: {\it et al.}\cite{hannewald00} provides a convincing step in this

1180: direction, yet remaining within a non-dynamical approach.  The last

1181: issue concern the intrinsic limitations of the SCLA. It has been shown

1182: that in order to describe correctly the very diluite excitonic limit

1183: additional diagrams must be added to the self consistent ladder

1184: approach. An analysis of this problem can be found in

1185: Ref.\onlinecite{pieri98}, where a generalized T matrix approximation

1186: for the fermionic self energy overcoming this problem has ben

1187: proposed.  In conclusion, we have clearly shown that in the definite

1188: and important physical region of $T< E_b$ inclusion of excitonic

1189: correlation in the electron-hole plasma is relevant and necessary, and

1190: that the simplified model presented in this paper is a well-understood

1191: starting point for this purpose. Thus, its further developments to

1192: address realistic systems in higher dimensions are well motivated.

1193: 

1194: 

1195: \acknowledgments We thank A. Quattropani, P. Schwendimann, V. Savona,

1196: C. Ciuti, and L. J. Sham for stimulating discussions. One of the

1197: authors (C. P.) acknowledges support by the Swiss National Foundation

1198: for the Scientific Research.

1199: 

1200: 

1201: 

1202: 

1203: \begin{references}

1204: 

1205: \bibitem{rice77} Excellent reviews of the subject may be found in

1206: B. I. Halperin and T. M. Rice, in {\it Solid State Physics}, edited by

1207: H. Ehrenreich, F. Seitz, and D. Turnbull (Academic, Ney York, 1977),

1208: Vol.21.; T. M. Rice, {\it ibidem} Vol. 32.

1209: 

1210: \bibitem{haug76} H. Haug, Z. Phys. B {\bf 24}, 351 (1976).

1211: 

1212: \bibitem{zhu96} X. Zhu, M. S. Hybertsen, and P. B. Littlewood,

1213: Phys. Rev. B {\bf 54}, 13 575 (1996).

1214: 

1215: \bibitem{elliott57} R. J. Elliott, Phys. Rev. {\bf 108}, 1384 (1957).

1216: 

1217: \bibitem{haugkoch} H. Haug and S.W. Koch, 'Quantum theory of the

1218: optical and electronic properties of semiconductors, (Singapore, World

1219: Scientific, 1990).

1220: 

1221: 

1222: \bibitem{haug84} H. Haug and S. Schmitt-Rink, Progr. Quantum

1223: Electronics {\bf 9}, 3 (1984).

1224: 

1225: \bibitem{kraeft86} W. D. Kraeft, D. Kremp, W. Ebeling, and G. R\"opke,

1226: {\it Quantum statistics of charged particle systems}, Akademie-Verlag,

1227: Berlin 1986.

1228: 

1229: \bibitem{zimmermann88} R. Zimmermann, {\it Many Particle theory of

1230: highly excited semiconductors}, Teubner Liepzig (1988).

1231: 

1232: 

1233: \bibitem{pereira98} M. Pereira and K. Henneberger, Phys. Rev. B {\bf

1234: 58}, 1064 (1998).

1235: 

1236: 

1237: \bibitem{tassone99} F. Tassone and C. Piermarocchi,

1238: Phys. Rev. Lett. {\bf 82}, 843 (1999).

1239: 

1240: \bibitem{benner91} S. Benner and H. Haug, Euro Phys. Lett. {\bf 16},

1241: 570 (1991).

1242: 

1243: 

1244: \bibitem{tassone.93} F. Tassone and F. Bassani, Phys. Rev. B {\bf 51}

1245: 16973 (1995).

1246: 

1247: \bibitem{mahan81} D. G. Mahan, {\it Many particle physics}, Plenum

1248: Press (1981).

1249: 

1250: \bibitem{danielewicz} P. Danielewicz, Annals of Physics {\bf 152},

1251: 239, (1984).

1252: 

1253: 

1254: \bibitem{haug96} H. Haug and A. P. Jahuo, {\it Quantum kinetics in

1255: transpot and optical properties of semiconductors}, Springer-Verlag,

1256: Berlin (1996).

1257: 

1258: 

1259: \bibitem{kadanoff62} L. P. Kadanoff and G. Baym, {\it Quantum

1260: statistical mechanics}, W. A. Benjamin Reading, (1962).

1261: 

1262: 

1263: \bibitem{optwires} R. Cingolani {\it et al.} Phys. Rev. B {\bf 48}

1264: 14331 (1993); C. Gr\'eus {\it et al.}, Euro Phys. Lett. {\bf 34} 213

1265: (1996); R. Ambigapathy {\it et al.} Phys. Rev. Lett. {\bf 78} 3579

1266: q(1997); R. Kumar {\it et al.} Phys. Rev. Lett. {\bf 81}, 2578 (1998).

1267: 

1268: 

1269: \bibitem{baym61}G. Baym and L. P. Kadanoff, Phys. Rev. {\bf 124} 287

1270: (1961).

1271: 

1272: 

1273: \bibitem{flex} N.E. Bickers, D.J. Scalapino, and S.R. White,

1274: Phys. Rev. Lett. {\bf 62}, 961, (1989).

1275: 

1276: 

1277: \bibitem{kagan98} B. Kyung, E. G. Klepfish, and P. E. Kornilovitch,

1278: Phys. Rev. Lett. {\bf 80} 3109 (1998); M. Y. Kagan, R. Fr\'esard,

1279: M. Capezzali, and H. Beck, Phys. Rev. B {\bf 57}, 5995 (1998);

1280: 

1281: \bibitem{haussman} R. Haussman, Z. Phys. B {\bf 91}, 291 (1993).

1282: 

1283: 

1284: \bibitem{haughanamura} E. Hanamura, H. Haug, Physics Reports {\bf

1285: 33C}, 209, (1977).

1286: 

1287: \bibitem{tassone99xx} F. Tassone and Y. Yamamoto, Phys. Rev. B {\bf

1288: 59}, 10 830 (1999).

1289: 

1290: \bibitem{nambu60} Y. Nambu, Phys. Rev. B {\bf 117}, 648 (1960).

1291: 

1292: \bibitem{lindberg} M. Lindberg, and S.W. Koch, Phys. Rev. B {\bf 38},

1293: 3342 (1988).

1294: 

1295: \bibitem{piermarocchi99} C. Piermarocchi {\it et. al.} Solid State

1296: Commun. {\bf 112}, 433 (1999).

1297: 

1298: 

1299: \bibitem{kozlov96} J. Ding {\it et al.}, Phys. Rev. Lett. {\bf 69},

1300: 1707 (1992); V. Kozlov {\it et al.} Phys. Rev. B {\bf 54}, 13932

1301: (1996).

1302: 

1303: \bibitem{sirigu99} L. Sirgiu, D. Y. Oberli, L. Degiorgi, A. Rudra, and

1304: E. Kapon, Phys. Rev. B {\bf 61} R10575 (2000).

1305: 

1306: 

1307: \bibitem{dassarma00} S. Das Sarma and D. W. Wang,

1308: Phys. Rev. Lett. {\bf 84} 2010 (2000).

1309: 

1310: \bibitem{hannewald00} K. Hannewald, S. Glutch, and F. Bechstedt,

1311: Phys. Rev. B {\bf 62} 4519 (2000).

1312: 

1313: \bibitem{pieri98} P. Pieri and G. C. Strinati, eprint arXiv

1314: cond-mat/9811166 (1998) \end{references} %\end{multicols}

1315: \end{document}

1316: