1: % as.tex

2: %\documentclass {revtex4}

3: \documentclass[prb,twocolumn,showpacs,showkeys,groupedaddress,amsmath,amssymb]{revtex4}% Physical Review B

4: %

5: %\usepackage{Bm}

6: \usepackage {amssymb}

7: \usepackage {amsmath}

8: %\usepackage[dvips]{epsfig}

9: \usepackage{graphics,amssymb}

10: \usepackage{natbib}

11: \usepackage{revsymb}

12: 

13: % Some other (several out of many) possibilities

14: %\documentclass[preprint,aps]{revtex4}

15: %\documentclass[preprint,aps,draft]{revtex4}

16: 

17: \usepackage{graphicx}% Include figure files

18: \usepackage{dcolumn}% Align table columns on decimal point

19: \usepackage{bm}% bold math

20: %

21: \newcommand{\DS}{\displaystyle}

22: 

23: \begin{document}

24: 

25: 

26: \title{Mechanisms of arsenic clustering in silicon\\}

27: % Force line breaks with \\

28: 

29: 

30: \author{F. F. Komarov}

31: \affiliation{Department of Physical Electronics, Belarusian State

32: University, 1 Kurchatov

33: Str., Minsk 220064, Belarus \\}%

34: %e-mail: KomarovF@bsu.by

35: 

36: \author{O. I. Velichko} %\email[]{oleg_velichko@lycos.com}

37: \affiliation{Department of Physics, Belarusian State University on

38: Informatics and Radioelectronics, 6 P. Brovka Str., Minsk 220013,

39: Belarus

40: \\}

41: 

42: 

43: \author{V. A. Dobrushkin} \email[]{dobrush@dam.brown.edu}

44: \affiliation{Brown University, Division of Applied Mathematics,

45: Providence, Rhode Island 02912, USA

46: \\}

47: 

48: \author{A. M. Mironov}

49: \affiliation{Institute of Applied Physics Problems, 7 Kurchatov

50: Str., Minsk 220064, Belarus

51: \\}

52: % e-mail: MironovA@bsu.by

53: 

54: \begin{abstract}

55: 

56: A model of arsenic clustering in silicon is proposed and analyzed.

57: The main feature of the proposed model is the assumption that

58: negatively charged arsenic complexes play a dominant role in the

59: clustering process. To confirm this assumption, electron density

60: and concentration of impurity atoms incorporated into the clusters

61: are calculated as functions of the total arsenic concentration at

62: a temperature of 1050\,$^{\circ}$C. A number of the negatively

63: charged clusters incorporating a point defect and one or more

64: arsenic atoms $(\text{DAs}_{1})^{-}$, $(\text{DAs}_{1})^{2-}$,

65: $(\text{DAs}_{2})^{-}$, $(\text{DAs}_{2})^{2-}$,

66: $(\text{DAs}_{3})^{-}$, $(\text{DAs}_{3})^{2-}$,

67: $(\text{DAs}_{4})^{-}$, and $(\text{DAs}_{4})^{2-}$ are

68: investigated. It is shown that for the doubly negatively charged

69: clusters or for clusters incorporating more than one arsenic atom the

70: electron density reaches a maximum value and then monotonically

71: and slowly decreases as total arsenic

72: concentration increases. In the case of cluster $(\text{DAs}_{2})^{2-}$, the

73: calculated electron density agrees well with the experimental

74: data. Agreement with the experiment confirms the conclusion that

75: two arsenic atoms participate in the cluster formation. Among all

76: present models, the proposed model of clustering by formation of

77: $(\text{DAs}_{2})^{2-}$ gives the best fit to the experimental

78: data and can be used in simulation of high concentration arsenic

79: diffusion.

80: \end{abstract}

81: 

82: \pacs{61.72.Ji, 61.72.Tt, 61.72.Ss, 85.40.Ry}% PACS

83: %Doping and impurity implantation in germanium and silicon                             % Classification Scheme.

84: 

85: \keywords{clusters; annealing; doping effects; arsenic; silicon}

86: 

87: %Use showkeys class option if keyword

88:                               %display desired

89: 

90: \maketitle

91: 

92: \section{Introduction}

93: 

94: Using low energy high dose arsenic ion implantation, one can

95: produce the active regions of modern integrated microcircuits

96: characterized by very shallow junctions and high dopant

97: concentrations. Thermal annealing is applied after implantation

98: for the arsenic activation and damage reduction. During the

99: initial stage of annealing, the arsenic atoms become electrically

100: active, occupying the substitutional positions. Then a fraction of

101: impurity atoms incorporates into the clusters, thereby decreasing

102: the layer conductivity.

103: \cite{Tsukamoto_80,Tsai_80,Fahey_89,Nobili_94,

104: Solmi_98,Krishnamoorthy_98,Nobili_99,Uematsu_00,

105: Nobili_01,Solmi_01,Solmi_03} Due to clustering, the concentration

106: of charge carriers is less than the total impurity concentration

107: because the clustered dopant atoms do not serve as a source of

108: free charge carriers. We can observe the clustering phenomenon in

109: many cases of the formation of highly doped semiconductor

110: structures; for example, in the region with high impurity

111: concentration during thermal diffusion of arsenic

112: atoms.\cite{Fair_73,Fair_Weber_73,Murota_79} Deactivation of the

113: electrically active dopant atoms also indicates that clustering

114: can occur during thermal treatment of the supersaturated arsenic

115: layers created by ion implantation with subsequent thermal or

116: laser annealing.

117: \cite{Nobili_94,Nobili_99,Parisini_90,Luning_92,Rousseau_94,

118: Rousseau_98,Solmi_02} Moreover, a reverse annealing or transient

119: increase of the carrier density can be observed if a doped

120: specimen, previously annealed at temperature $T_{1}$, is further

121: annealed at higher temperature $T_{2}$. \cite{Solmi_00} It is

122: supposed that transient dissolution of a part of the clusters

123: occurs during such a reverse annealing. \cite{Solmi_00}

124: 

125: 

126: At very high dopant concentrations exceeding the solid solubility

127: of As in Si, a significant fraction of arsenic atoms can form

128: precipitates.

129: \cite{Fahey_89,Nobili_94,Solmi_98,Krishnamoorthy_98,Uematsu_00,Solmi_01,

130: Parisini_90, Nobili_83,La Via_91} The typical distributions of the

131: total dopant concentration and electron density after thermal

132: annealing of silicon substrates heavily implanted by As were

133: presented by Nobili et al. \cite{Nobili_94} (Fig. 7) and by Solmi

134: \cite{Solmi_01} (Fig. 1c). As can be seen from Fig. 1c,

135: \cite{Solmi_01} the carrier concentration reaches its saturation

136: value $n_{e}=3.57\times10^{8}$ $\mu $m$^{{\rm - }{\rm 3}}$,

137: whereas the total arsenic concentration increases to the

138: solubility limit $C_{sol}=3.27\times10^{9}$ $\mu $m$^{{\rm - }{\rm

139: 3}}$ \cite{Nobili_01} and continues to increase in the region

140: adjoint to the surface. It is supposed that the cluster formation

141: occurs at high dopant concentrations,

142: \cite{Tsukamoto_80,Tsai_80,Fahey_89,Nobili_99,Nobili_01,Solmi_01,Fair_73,

143: Parisini_90,Luning_92} more precisely, at the dopant

144: concentrations approximately ranging from $n_{e}$ to $10\,n_{e}$.

145: If dopant concentration exceeds the solubility limit $C_{sol}$,

146: precipitation occurs. \cite{Nobili_94,Solmi_01} Arsenic clustering

147: and precipitation have attracted evergrowing attention of the

148: researchers, as they are greatly important for the silicon

149: integrated circuits technology. Both clusters and precipitates do

150: not serve as a source of charge carriers and the total arsenic

151: concentration may be higher than the electron density by about two

152: orders of magnitude. The clusters and precipitates are metastable

153: and can dissociate under subsequent thermal treatments. This means

154: that stringent requirements should be placed upon the accuracy of

155: the clustering models used to calculate the carrier concentrations

156: in heavily doped silicon layers.

157: 

158: \section{Analysis of clustering phenomena}

159: 

160: The development of a model for As clustering necessitates a

161: consistent analysis of the experimental data and theoretical

162: substantiation of the processes proceeding in heavily doped

163: silicon layers.

164: 

165: \subsection{Clustering during thermal diffusion

166: and annealing of  ion-implanted layers}

167: 

168: 

169: A great number of clustering models with different types of

170: clusters have been developed and proposed since 1973.

171: \cite{Tsukamoto_80,Tsai_80,Fahey_89,Fair_73,Parisini_90,Luning_92,

172: Guerrero_82,Berding_98,Mueller_03} The variety of the proposed

173: models reflects discrepancies in the experimental results

174: \cite{Nobili_83} and a great uncertainty in the nature of the

175: clusters discussed. For example, inactive cluster $\text{VAs}_{2}$

176: was considered in \cite{Fair_73} to explain the difference between

177: the total and electrically active dopant concentrations for

178: arsenic thermally diffused from a constant source. For the best

179: fit to the diffusion profiles of ion-implanted arsenic, it was

180: suggested that the electrically inactive arsenic atoms were

181: incorporated into the clusters with four arsenic atoms per

182: cluster. \cite{Tsukamoto_80} According to Tsai et al.

183: \cite{Tsai_80} who have also investigated diffusion in the

184: ion-implanted layers, clustering of arsenic may be expressed as

185: 

186: \begin{equation} \label{Cluster_Reaction_Tsai}

187: 3\text{As}^{+}+e^{-}{\stackrel{\mbox{Annealing}}{\qquad

188: \longleftrightarrow \qquad }}

189: \text{As}_{3}^{2+}{\stackrel{\mbox{25\,$^{\circ}$C}}{\quad

190: \longrightarrow \quad }} \text{As}_{3},

191: \end{equation}

192: 

193: 

194: \noindent where $\text{As}^{ +} $  is the substitutionally

195: dissolved arsenic atom participating in the cluster formation;

196: $e^{ - }$ is the electron.

197: 

198: As follows from reaction (\ref{Cluster_Reaction_Tsai}), the

199: clustered As atoms are electrically active at annealing

200: temperatures, but are neutral at room temperature.

201: 

202: It was found in \cite{Uematsu_00} that the formation of clusters

203: $(\text{As}_{2}\text{V})^{\times}$ occurs due to the following

204: reaction, which

205: 

206: \begin{equation} \label{Cluster_Reaction_Uematsu}

207: 2\text{As}^{+}+\text{V}^{2-}\longleftrightarrow

208: (\text{As}_{2}\text{V})^{\times}

209: \end{equation}

210: 

211: \noindent is most liable to cause deactivation because the

212: concentration of inactive As species increases to the fourth power

213: of the active As concentration. \cite{Tsukamoto_80,Murota_79} Here

214: $\text{V}^{2-}$ is the doubly negatively charged vacancy

215: participating in the clustering.

216: 

217: Indeed, if a local equilibrium is assumed, it follows from the mass action

218: law for reaction (\ref{Cluster_Reaction_Uematsu}) that

219: 

220: \begin{equation} \label{MAL_Uematsu}

221: C^{A}=A\tilde{C}^{V\times}\chi^{2}C^{2},

222: \end{equation}

223: 

224: \begin{equation} \label{Constant_Uematsu}

225: A=H^{A} C_{i}^{V\times} n_{i}^{2},

226: \end{equation}

227: 

228: \begin{equation} \label{Rel_vacancy_concentration}

229: \tilde{C}^{V\times}=\frac{C^{V\times}}{C_{i}^{V\times}},

230: \end{equation}

231: 

232: \noindent where $C^{A}$ and $C$ are the concentrations of the

233: impurity atoms incorporated into clusters and substitutionally

234: dissolved arsenic atoms, respectively; $H^{A}$ is the constant of

235: the local equilibrium for reaction

236: (\ref{Cluster_Reaction_Uematsu}); $C^{V\times}$ and

237: $C_{i}^{V\times}$ are the actual and the equilibrium

238: concentrations of neutral vacancies, respectively;

239: $\chi=\displaystyle \frac{n}{n_{i}}$ is the electron density

240: normalized to the concentration of intrinsic charge carriers in a

241: semiconductor during thermal treatment $n_{i}$.

242: 

243: One would expect from (\ref{MAL_Uematsu}) that reaction

244: (\ref{Cluster_Reaction_Uematsu}) yields the fourth power

245: dependence $C^{A}\propto C^{4}$, since for high concentrations

246: $C>>n_{i}$ it might be reasonable to assume $\displaystyle

247: \chi=\frac{n}{n_{i}}\thickapprox \frac{C}{n_{i}}$. However, as

248: seen from Fig. 7 (Ref. \cite{Nobili_94}) and Fig. 1c (Ref.

249: \cite{Solmi_01}), the electron density $n\thickapprox n_{e}=const$

250: in the region of concentrations associated with active clustering.

251: Consequently, reaction (\ref{Cluster_Reaction_Uematsu}) yields the

252: second power dependence $C^{A}\backsim C^{2}$, which  contradicts

253: the experimental data. \cite{Tsukamoto_80,Murota_79}

254: 

255: 

256: Because of the vacancy participating in cluster formation, silicon

257: self-interstitials were in turn ejected or left behind during

258: clustering. \cite{Krishnamoorthy_98} This assumption had been made

259: much earlier in the study \cite{Parisini_90,Rousseau_94} of

260: impurity deactivation within the laser-annealed layers. However,

261: the results obtained in \cite{Krishnamoorthy_98} show that the

262: point defects induced during clustering and/or precipitation make

263: no contribution to the enhanced transient diffusion of arsenic

264: implanted at low energy, while these defects did coalesce to form

265: extended defects at the projected range. Analysis of the defect

266: microstructure has revealed that the transition between arsenic

267: clustering and SiAs precipitation is not abrupt, pointing to

268: possible coexistence of arsenic clusters and SiAs precipitates

269: (although precipitates are not directly observed).

270: \cite{Krishnamoorthy_98}

271: 

272: 

273: The model \cite{Tsai_80} was generalized by Solmi and Nobili

274: \cite{Solmi_98}

275: 

276: \begin{equation} \label{Cluster_Reaction_Solmi}

277: m\text{As}^{+}+e^{-}{\stackrel{\mbox{Annealing}}{\qquad

278: \longleftrightarrow \qquad }}

279: \text{As}_{m}^{(m-1)+}{\stackrel{\mbox{25\,$^{\circ}$C}}{\quad

280: \longrightarrow \quad }} \text{As}_{m}

281: \end{equation}

282: \noindent to take into account a saturation behavior of the

283: carrier density with increasing dopant concentration. For

284: diffusion at a temperature of 1050\,$^{\circ}$C, the best agreement

285: between the modeled and experimental curves describing diffusivity

286: against total arsenic concentration was achieved for $m=4$. And at

287: 900\,$^{\circ}$C, the neutral cluster model \cite{Fair_73} provided

288: a better agreement with the experimental data. The reaction

289: (\ref{Cluster_Reaction_Solmi}) was also used in \cite{Nobili_01}

290: to describe deactivation in the arsenic-doped layers, which were

291: formed by ion implantation and high temperature annealing of

292: silicon-on-insulator specimens. However, Solmi et al.

293: \cite{Solmi_03} have recently used the following reaction

294: 

295: \begin{equation} \label{Cluster_Reaction_Solmi_03}

296: m\text{As}^{+}+\text{V}\longleftrightarrow \text{As}_{m}+\text{I}

297: \end{equation}

298: 

299: \noindent to model the transient arsenic diffusion. Here

300: $m$ is assumed to have the values between 2 and 4 to take into

301: account the fact that the As cluster was formed around a vacancy with

302: the subsequent injection of the self-interstitial.

303: 

304: 

305: It is important to note that reaction

306: (\ref{Cluster_Reaction_Solmi}) was first theoretically studied by

307: Guerrero et al. in paper, \cite{Guerrero_82} which will be

308: considered below.

309: 

310: \subsection{Clustering during deactivation of supersaturated

311: arsenic layers formed by laser annealing}

312: 

313: Let us consider the clustering models used for the explanation of

314: deactivation in the supersaturated arsenic layers formed by ion

315: implantation and subsequent laser annealing. In \cite{Luning_92}

316: it was found that a cluster consisting of four positively charged

317: arsenic atoms and two additional negative charges gave the best

318: least-squares fit with respect to the experimental data obtained

319: in the case of deactivation within the supersaturated arsenic

320: layers created by ion implantation and subsequent laser annealing.

321: If two negative charges are associated with a doubly charged

322: vacancy, this model gains support from the theoretical

323: calculations of Pandey et al. \cite{Pandey_88} Such a cluster has

324: an intuitive geometrical and structural appeal --- four arsenic

325: atoms tetrahedrally arranged around a vacancy may relax inward,

326: thus relieving lattice strains due to the As size effect.

327: \cite{Luning_92} Using this cluster species, a single-step

328: clustering reaction was postulated in \cite{Luning_92}

329: 

330: \begin{equation} \label{Cluster_Reaction_Luning}

331: 4\text{As}^{+}+\text{V}^{2-}\longleftrightarrow

332: (\text{As}_{4}\text{V})^{2+}.

333: \end{equation}

334: 

335: 

336: The assumption of self-interstitial ejection during clustering was

337: supported by the experimental data obtained by Parisini et al.

338: \cite{Parisini_90} who have observed that extended defects at the

339: projected range were extrinsic or interstitial in nature. The reaction

340: 

341: \begin{equation} \label{Cluster_Reaction_Parisini}

342: (\text{As}_{2})^{2+}+2e^{-}\longrightarrow

343: (\text{As}_{2}\text{V})^{+}+\text{I}^{-}

344: \end{equation}

345: 

346: \noindent was proposed in \cite{Parisini_90} only for the initial

347: step of the deactivation process in arsenic implanted silicon

348: specimens, first laser-annealed and then thermally annealed. Here

349: $\text{I}^{-}$ is the negatively charged self-interstitial. The

350: results obtained by Rousseau et al. \cite{Rousseau_94} confirm the

351: conclusion that arsenic is indeed deactivated by vacancies, with a

352: concurrent injection of self-interstitials. The latter follows

353: from the observed significant enhancement of the diffusion in the

354: buried boron layer underneath the As structure subjected to

355: deactivation. Investigations of the enhanced diffusion in buried

356: boron layers were continued in, \cite{Rousseau_98} where it was

357: proposed that small arsenic clusters of various sizes were formed

358: around a vacancy during deactivation with injection of the

359: associated interstitial into the bulk.

360: 

361: Considering that the majority of the models described in this

362: section are based essentially on the experimental investigations

363: of the defect subsystem, we first concern ourselves with these

364: experimental results.

365: 

366: 

367: \subsection{Experimental investigations of defect subsystem}

368: 

369: The methods of transmission electron spectroscopy (TEM),

370: \cite{Nobili_94,Krishnamoorthy_98,Parisini_90,La

371: Via_91,Dokumachi_95} extended x-ray-absorption fine-structure

372: (EXAFS), \cite{Parisini_90,Erbil_86,Allain_92,d'Acapito_04}

373: Rutherford backscattering (RBS), \cite{La Via_91,Brizard_93}

374: positron lifetime measurements

375: \cite{Makinen_89,Lawther_95,Myler_96,Szpala_96} combined with the

376: measurements of electron momentum distributions,

377: \cite{Polity_98,Saarinen_99,Ranki_03,Ranki_Saarinen_03,Ranki_04}

378: and other methods \cite{Nobili_83,Herrera-Gomez_96} are commonly

379: used to investigate silicon specimens doped by As. For example,

380: TEM observations show that As-lased samples are completely free of

381: any visible defects. \cite{Dokumachi_95} On the other hand, TEM

382: analysis reveals the presence of extended defects after subsequent

383: thermal treatment. \cite{Parisini_90,Dokumachi_95} For example,

384: small \{113\} interstitial loops begin to appear beyond the

385: plateau region of the As profile after low-temperature treatment.

386: \cite{Parisini_90} After high-temperature thermal treatment

387: ($>$850\,$^{\circ}$C), a low density of very small As-related

388: precipitates (about 2 nm in diameter) is observed. The

389: precipitated As fraction cannot be responsible for the total

390: amount of inactive As. \cite{Parisini_90}

391: 

392: It was established that EXAFS measurements provided particular

393: details about the local atomic structure around the dopant and

394: confirmed the fact that As atoms are substitutional after laser

395: annealing. \cite{Erbil_86,Allain_92} Note that, according to the

396: x-ray standing-wave spectroscopy measurements, the As atoms

397: remained at the substitutional positions even after 85\% of the

398: electrical activity has been lost due to thermal annealing of the

399: laser melted layers. \cite{Herrera-Gomez_96} On the other hand, a

400: large amount of nonsubstitutional arsenic was detected by

401: Rutherford back-scattering after diffusion from $\text{TiSi}_{2}$

402: source. \cite{La Via_91} These defects are interpreted as

403: precipitates, probably formed due to the stress induced by

404: $\text{TiSi}_{2}$ layer. Indeed, when thermal annealing is carried

405: out after the complete removal of titanium silicide (i.e., without

406: stress), one can observe the activation of all the arsenic atoms

407: diffused in the silicon substrate. \cite{La Via_91} Thus, such

408: full activation indicates the significant influence of stress on

409: the formation of precipitates.

410: %

411: % very important for future analysis

412: %

413: 

414: According to EXAFS measurements, \cite{Allain_92} subsequent

415: thermal annealing (350\,\,--\,750\,$^\circ $C) of the As-lased

416: layers leads to the formation of $VAs_{m}$ clusters including up

417: to $7\pm4$ As atoms around the vacancy.  However, at high

418: temperatures it was observed that the number of the first

419: neighbors of Si has tended back to 4 atoms.

420: %

421: % very important for this analysis

422: %

423: This is probably due to the precipitates observed by TEM over

424: this temperature range. Quantitative analysis of EXAFS data for

425: the ultra-low energy implanted layers after rapid or spike

426: annealing \cite{d'Acapito_04} has revealed a site for As that was

427: different from the pure substitutional one, suggesting the presence

428: of clusters of As atoms coupled to vacancies. All the observed

429: phenomena, namely, a low value of the total coordination number in

430: the annealed samples, small $\text{As--Si}$ bond length, and the

431: presence of $\text{As--As}$ coordinations, may  be explained by

432: the assumption that clustering of As ions with vacancies takes

433: place to form $\text{VAs}_{m}$ structures.

434: 

435: Ion channeling and RBS were used in \cite{Brizard_93} to

436: complement the EXAFS and electron microscopy results. In has been

437: found that the deactivated As atom is displaced with respect to

438: the lattice sites and, moreover, the average displacement is

439: constant over the temperature range 450\,\,--\,900\,$^{\circ}$C,

440: being equal to $0.23 \pm 0.06$ \AA. Also, in \cite{Brizard_93} it

441: was proposed  that the formation of larger clusters

442: $\text{V}_{n}\text{As}_{m}$ ($n\thickapprox 5$ and $m\thickapprox

443: 10$) occurred starting from $\text{VAs}_{2}$ or $\text{VAs}_{4}$.

444: The formation of these larger clusters is in agreement with a

445: decrease in the first nearest neighbors of As. At higher

446: temperatures, an increase in the number of the first nearest

447: neighbors of As to 4 was observed by EXAFS measurements.

448: \cite{Allain_92} This means that the cluster species may differ at

449: intermediate (350\,\,--\,750\,$^{\circ}$C) and higher

450: temperatures.

451: 

452: Very interesting information about clustering was obtained during

453: the experimental investigations of the doped layers by  positron

454: annihilation spectroscopy

455: \cite{Lawther_95,Myler_96,Saarinen_99,Ranki_03,Ranki_Saarinen_03,Ranki_04}

456: and by positron annihilation spectroscopy combined with the

457: Hall-effect/resistivity measurements. \cite{Ranki_03} The defect

458: subsystem was generally investigated before and after thermal

459: treatment. To support these measurements, the positron lifetimes

460: and core electron momentum distributions were calculated for

461: different vacancy-donor complexes. \cite{Ranki_04} Lawther et al.

462: \cite{Lawther_95} have investigated the doped layers melted by an

463: excimer laser to obtain the profile with constant As concentration

464: and a sharp fall-off at a depth of 0.2 $\mu$m. Arsenic

465: deactivation was initiated by annealing at 750\,$^{\circ}$C for 15

466: s or by conventional thermal treatment for 2 h. It was found that

467: $\text{VAs}_{m}$ complexes with the average values of $m$ greater

468: than 2 caused arsenic  deactivation in heavily doped Si. Myler et

469: al. \cite{Myler_96} have studied the silicon layers (with arsenic

470: concentration 4$\times 10^{8}$ $\mu \text{m}^{3}$) fully activated

471: by laser melting and subsequently annealed for 15 s at 500 and

472: 750\,$^{\circ}$C. The changes in the positron annihilation spectra

473: after the thermal treatment at 500 or 750\,$^{\circ}$C were also

474: attributed to the formation of

475: $\text{As}_{m}\text{Si}_{4-m}\text{V}_{ac}$ complexes. In this

476: experiment, the impossibility of determining the number of

477: impurity atoms incorporated in the impurity-vacancy complex was

478: established. Ranki at al. \cite{Ranki_03} have studied the samples

479: implanted by As during MBE growth at 450\,$^{\circ}$C. The samples

480: were annealed in $\text{N}_{2}$, either by RTA at 900\,$^{\circ}$C

481: for 10\,\,--\,170 s or furnace annealing at

482: 800\,\,--\,900\,$^{\circ}$C for two minutes. Based on the positron

483: annihilation and Hall-effect/resistivity experiments, Ranki et al.

484: \cite{Ranki_03} have concluded that the dominant defect in

485: As-grown samples was $\text{VAs}_{3}$. The measurements

486: demonstrated high concentrations (above $10^{5}$ $\mu

487: \text{m}^{-3}$) of $\text{VAs}_{3}$ complexes in heavily-doped

488: silicon. Moreover, larger V-As complexes, probably

489: $\text{V}_{2}\text{As}_{5}$, may occur together with

490: $\text{VAs}_{3}$ at high As concentrations. A relative amount of

491: $\text{V}_{2}\text{As}_{5}$ increases with annealing. This cluster

492: is even dominant after annealing at 800\,$^{\circ}$C. The

493: $\text{VAs}_{3}$ and $\text{V}_{2}\text{As}_{5}$ complexes become

494: unstable at 800\,$^{\circ}$C and 900\,$^{\circ}$C, respectively,

495: then their concentrations decrease. Cluster reconstruction may

496: occur during cooling , starting from $\text{VAs}_{1}$ to

497: $\text{VAs}_{2}$ and then to $\text{VAs}_{3}$. The clusters

498: $\text{VAs}_{2}$ and $\text{VAs}_{3}$ may be in turn transformed

499: to the $\text{V}_{2}\text{As}_{5}$ complexes. Moreover, some

500: $\text{VAs}_{4}$ and other vacancy complexes may be also formed,

501: but with much lower concentrations.

502: 

503: 

504: High energy electron irradiation is commonly used in positron

505: experiments for the generation of nonequilibrium vacancies, mobile

506: even at room temperature and liable to interact with the dopant

507: atoms. \cite{Saarinen_99,Ranki_02,Ranki_Saarinen_03,Ranki_04} For

508: example, positron experiments \cite{Saarinen_99} were performed

509: both with the As-grown samples doped by arsenic in concentrations

510: of $10^{7}$ and $10^{8}$ $\mu \text{m}^{-3}$ and samples subjected

511: to 2 MeV electron irradiation at 300 K. It was established that

512: heavily As-doped silicon contained $\text{VAs}_{3}$ complex as a

513: native defect. Before irradiation, the concentration of

514: $\text{VAs}_{3}$ was equal to $\sim 10^{7}$ $\mu \text{m}^{-3}$ at

515: a doping level of $10^{8}$ $\mu \text{m}^{-3}$. After electron

516: irradiation, one can observe the pairs ($\text{VAs}_{1}$) formed

517: by a vacancy and a single impurity atom.  It was demonstrated in

518: \cite{Ranki_02} that the migration of these $\text{VAs}_{1}$

519: defects started at around 450 K, leading to formation of

520: $\text{VAs}_{2}$ defects. And, in turn, these defects were

521: transformed to $\text{VAs}_{3}$ defects at 700 K. The

522: $\text{VAs}_{3}$ defects were stable at 700 K, representing the

523: dominant vacancy-impurity cluster in heavily doped n-type Si at

524: this temperature. The formation of larger $\text{VAs}_{2}$ and

525: $\text{VAs}_{3}$ complexes was significantly dependent on the As

526: concentration. This interpretation of the experimental data was

527: confirmed by the results obtained in. \cite{Ranki_Saarinen_03} In

528: these experiments, the electron irradiated samples were annealed

529: isochronally (30 min) at 300\,\,--\,1220 K. Irradiation of the

530: heavily doped Si samples has produced mainly the vacancy-donor

531: pairs ($\text{VAs}_{1}$) with a small concentration of

532: divacancies. Considering that, after irradiation, $\text{VAs}_{1}$

533: concentration was much higher than the initial concentration of

534: $\text{VAs}_{3}$, no signals from $\text{VAs}_{3}$ were observed

535: for the As-irradiated As-doped sample. From the core-region

536: electron momentum distribution measurements, it was found

537: \cite{Ranki_Saarinen_03} that defects in As-irradiated samples may

538: be identified as $\text{VAs}_{1}$, in samples annealed at 600 K

539: --- as $\text{VAs}_{2}$, and in those annealed at 775 K --- as

540: $\text{VAs}_{3}$. It is well known that a drastic drop in the

541: conductivity occurs when heavily doped Si is annealed at

542: temperatures between 400\,$^{\circ}$C and 500\,$^{\circ}$C. This

543: deactivation of the dopants is partially reversible by annealing

544: at 800\,\,--\,1000\,$^{\circ}$C. Taking into account these

545: experimental data, Ranki et al. \cite{Ranki_Saarinen_03} have

546: suggested that the formation and annealing of $\text{VAs}_{3}$ at

547: 700 and 1100 K, respectively, are responsible for the observed

548: behavior of the conductivity. The experiments conducted in

549: \cite{Ranki_02} were completed by studies in \cite{Ranki_04}

550: associated with thermal treatment up to 1220 K. It has been found

551: that dissociation of $\text{VAs}_{3}$ began at 1100 K, and

552: at 1220 K these defects were annealed away.

553: 

554: 

555: Thus, the analyzed data show that complexing of As atoms with

556: vacancies occurs in the layers heavily doped by arsenic. Based on

557: the experimental data, various $\text{V}_{m}\text{As}_{n}$

558: clusters are possible but $\text{VAs}_{2}$ and $\text{VAs}_{3}$

559: are the most likely. Taking clustering into account, one can

560: explain the phenomenon of compensation at high doping levels.

561: Nevertheless, some aspects are not clearly understood. For

562: example, $\text{VAs}_{3}$ starts to dissociate at 1100 K. This

563: annealing temperature agrees well with the observation that the

564: number of Si first neighbors tends back to 4 atoms

565: \cite{Allain_92} at temperatures higher than 750\,$^{\circ}$C. At

566: the same time, clustering occurs at temperatures higher than

567: 750\,$^{\circ}$C as well. \cite{Solmi_01} This means that such a

568: high temperature clustering is hardly explained by

569: $\text{VAs}_{3}$ complexes.

570: 

571: % grown by MBE and doped layers irradiated by electrons

572: 

573: \subsection{First principles studies}

574: 

575: Along with various experimental investigations, a number of the

576: theoretical {\it ab initio} calculations have been performed to

577: explain deactivation of arsenic atoms in silicon. From the

578: calculations of Pandey et al., \cite{Pandey_88} it follows that

579: $\text{VAs}_{4}$ complex including a vacancy surrounded by four

580: arsenic atoms is energetically favored over both substitutional,

581: isolated As in Si and substitutional $\text{Si\,-As}_{4}$

582: configurations. This cluster is electrically inactive, being

583: responsible for arsenic deactivation and structural changes in

584: heavily doped silicon. Larger defect clusters (e.g.,

585: $\text{V}_{2}\text{As}_{4}$) should also form during annealing,

586: whereas $\text{VAs}_{4}$ is only the first step in the clustering

587: process. These investigations were continued in,

588: \cite{Ramamoorthy_96} where general $\text{V}_{n}\text{As}_{m}$

589: complexes have been considered. The formation energies of the

590: vacancy-donor complexes $\text{VAs}_{1}$, $\text{VAs}_{2}$,

591: $\text{VAs}_{3}$, and $\text{VAs}_{4}$ were found to be 2.47,

592: 0.82, -0.53, and -2.39 eV, respectively. Moreover, it was proposed

593: that the complex $\text{VAs}_{2}$ was mobile, as were the

594: $\text{VAs}_{1}$ pairs. As these pairs moved, they reacted with

595: other defects to form larger, immobile complexes. When considering

596: the mobile $\text{VAs}_{2}$ complexes, Ramamoorthy and Pantelides

597: \cite{Ramamoorthy_96} have tried to explain the coupled diffusion

598: phenomenona and clustering. It was supposed that $\text{VAs}_{2}$

599: and $\text{VAs}_{3}$ were the dominant complexes in the

600: deactivated specimens near the enhanced-diffusion threshold.

601: \cite{Larsen_86,Larsen_93} A high rate of As diffusion observed

602: during rapid thermal annealing \cite{Larsen_86} was due to

603: $\text{VAs}_{2}$ migration over a very short time period

604: (6\,\,-\,60 s), when no extensive clustering could occur. During

605: their motion, $\text{VAs}_{2}$ complexes reacted with other

606: defects and formed larger, immobile complexes, which decreased the

607: As diffusivity.

608: 

609: Nevertheless,  further investigation is necessary because some

610: theoretical results disagree with the experimental data. For

611: example, in \cite{Chadi_97,Citrin_03} it was stated that the

612: formation of $\text{VAs}_{3}$ or $\text{VAs}_{4}$ defects was

613: exothermic, but, according to the first-principles calculations,

614: \cite{Ramamoorthy_96} buildup of $\text{VAs}_{1}$ or

615: $\text{VAs}_{2}$ structures was endothermic. This means that even

616: at low arsenic concentrations all impurity atoms must form

617: $\text{VAs}_{3}$ or $\text{VAs}_{4}$ clusters after long-time

618: treatment. In point of fact, at low dopant concentrations, one can

619: observe intensive diffusion by means of $\text{VAs}_{1}$ pairs

620: rather than clustering.

621: 

622: Berding et al., \cite{Berding_98} and Berding and Sher

623: \cite{Berding_Sher_98} have used the electronic quasichemical

624: formalism to calculate a free energy of various clusters,

625: including $\text{As\,Si}_{4}$, $\text{VAs}_{4}$,

626: $\text{VAs}_{3}\text{Si}_{1}$, $\text{VAs}_{2}\text{Si}_{2}$. Here

627: $\text{As\,Si}_{4}$ is the arsenic atom in the substitutional

628: position. The neutral cluster composed of threefold-coordinated

629: second-neighbor arsenic atoms DP(2) was proposed by Chadi et al.

630: \cite{Chadi_97} as an alternative to exothermic $\text{VAs}_{4}$.

631: The clusters DP(2), $\text{V}_{2}\text{As}_{6}$, and

632: $\text{Si}_{8}$ were also included in the calculations performed

633: in. \cite{Berding_Sher_98} In contrast to, \cite{Ramamoorthy_96}

634: Berding et al., \cite{Berding_98} Berding and Sher

635: \cite{Berding_Sher_98} take into account the ionized states and

636: entropy of the cluster formation. The entropy is unfavourable for

637: the formation of a large defect complex such as $\text{VAs}_{4}$.

638: Therefore, the complete free-energy calculation is needed to

639: determine the role of $\text{VAs}_{4}$ in deactivation. Based on

640: the complete free-energy calculation, it was found that cluster

641: $\text{VAs}_{4}$ was neutral, in agreement with the previous

642: findings. \cite{Pandey_88} Also, the energy of this complex (-1.62

643: eV) was in a rough agreement with the value given in.

644: \cite{Pandey_88} The $\text{VAs}_{3}$ and $\text{VAs}_{2}$

645: clusters were found to have one and two acceptor levels within the

646: energy gap. Consequently, their formation energy may be

647: effectively decreased when the Fermi energy is near the

648: conduction-band edge. The equilibrium cluster concentrations

649: depending on the temperature and concentration of dopant atoms are

650: obtained using the minimized free energy of the system. For all

651: the dopant concentrations and temperatures considered

652: (400\,\,--\,1000\,$^{\circ}$C), three classes of clusters are

653: dominant under equilibrium conditions: $\text{Si}_{8}$,

654: $\text{AsSi}_{4}$, and $\text{VAs}_{4}$. In all cases, the Fermi

655: energy was near the conduction-band edge. Predominantly,

656: $\text{VAs}_{3}$ and $\text{VAs}_{2}$ were singly and doubly

657: ionized, respectively. At low arsenic concentrations (up to $\sim

658: 5\times10^{6}$ $\mu$\text{m}$^{-3}$), noticeable deactivation was

659: absent for temperatures higher than 500\,$^{\circ}$C. With the

660: arsenic concentration raised to $5\times10^{7}$ $\mu

661: $\text{m}$^{-3}$ and higher, significant deactivation was

662: predicted, mainly due to the formation of $\text{VAs}_{4}$

663: clusters. Both concentration and temperature influenced the

664: contributions of different defects. For example, the concentration

665: of $\text{VAs}_{4}$ clusters has reached the concentration of

666: isolated arsenic atoms in the lattice $\text{As\,Si}_{4}$ for

667: $\thicksim 5\times10^{8}$ $\mu $\text{m}$^{-3}$ at a temperature

668: of 700\,$^{\circ}$C and for $\thicksim 2\times10^{9}$ $\mu

669: $\text{m}$^{-3}$ at a temperature of 1000\,$^{\circ}$C.

670: 

671: In \cite{Berding_98,Berding_Sher_98} the authors have attempted to

672: explain the effect of the electron concentration saturation at

673: high doping levels. However, the calculations performed in these

674: papers show that saturation is not reachable even at very high

675: arsenic concentrations. For example, saturation was not reached

676: with the arsenic concentration of $2\times10^{9}$ $\mu $m$^{-3}$

677: at a temperature of 700\,$^{\circ}$C, which is in disagreement

678: with the experimental data (see Fig. 2 from \cite{Berding_98}). It

679: is important that, according to, \cite{Berding_Sher_98} DP(2)

680: clusters proposed by Chadi et al. in \cite{Chadi_97} may be

681: present as a deactivating species, insignificantly contributing to

682: the deactivation under conditions of full equilibration.

683: 

684: These new defects, first mentioned in \cite{Chadi_97} and called

685: DP, represent a pair of two three-fold coordinated donor atoms.

686: The lowest energy DP structures are DP(2) and DP(4), where the

687: donor atoms occupy either second- or fourth-neighbor Si sites

688: along the $<110>$ direction. It was shown that DP(2) exist in the

689: stable electrically active 2+ charge state that is donating

690: electrons to the conduction band or in the metastable neutral

691: charge state capturing two electrons from the Fermi sea. A very

692: important conclusion made in \cite{Chadi_97} is a threefold

693: coordination of each dopant atom in DP(2) in the neutral trap

694: state. Thus, relaxation of the neighboring Si atoms creates an

695: increased volume, which is consistent with the experimental

696: observations. Therefore, such an increased volume around the donor

697: atoms is possible without vacancies. The main ideas of

698: \cite{Chadi_97} were further developed in. \cite{Citrin_03} Citrin

699: et al. \cite{Citrin_03} have proposed a new class of defects,

700: called the donor-pair-vacancy-interstitials and composed of two

701: dopant donor atoms near the displaced Si atom that is forming a

702: vacancy-interstitial pair. This defect, which is like a hybrid of

703: the donor-pair and Frenkel-pair defects, is denoted DP(i)V-I. For

704: the case of Sb, the data of annular dark-field (``Z-contrast'')

705: scanning transmission electron microscopy (ADF-STEM) clearly

706: demonstrate that, in heavily Sb-doped silicon grown at low

707: temperatures, the primary deactivating defect contains only two Sb

708: atoms. \cite{Voyles_02} Thus, energetically favorable

709: $\text{VSb}_{3}$ and $\text{VSb}_{4}$ may be ruled out from the

710: deactivation process. Only $\text{VSb}_{2}$, DP(2), and DP(4) may

711: be considered as liable candidates. Using additional STEM

712: measurements, x-ray absorption data, and first-principles

713: calculations, Voyles et al. \cite{Voyles_02} have shown that

714: neither $\text{VSb}_{2}$ nor DP(2) and DP(4) defects are important

715: in heavily Sb-doped Si. At that time, DP(2)V-I and DP(4)V-I were

716: in line with the available experimental results, including

717: positron annihilation spectroscopy data. \cite{Citrin_03} Both

718: DP(2)V-I and DP(4)V-I are independent out of the pre-existing

719: vacancy population, which is in conformity with the observations

720: of. \cite{Solmi_02}

721: 

722: Nevertheless, as follows from the experimental data,

723: \cite{Solmi_01} a deactivation mechanism can be different in the

724: As- and Sb-doped silicon. The mechanism of arsenic clustering

725: calls for further investigation. In, \cite{Mueller_03} Mueller et

726: al. have investigated the electronic structure and charge states

727: of various vacancy-impurity clusters using the first-principles

728: density-functional theory (DFT). It was found that the

729: $\text{VAs}_{1}$ complex can trap up to two conduction electrons

730: at a high n-doping level. Both $\text{VAs}_{2}$ and

731: $\text{VAs}_{3}$ act as a single-electron trap center. In

732: agreement with,

733: \cite{Pandey_88,Ramamoorthy_96,Berding_98,Berding_Sher_98} Mueller

734: et al. \cite{Mueller_03} have found that in $\text{VAs}_{4}$ the

735: fifth valence electrons of all four As atoms are strongly bound

736: and the complex is expected to remain neutral for any position of

737: the Fermi level. Note that according to the calculations of,

738: \cite{Berding_98} the cluster $\text{VAs}_{2}$ has two acceptor

739: levels, whereas Solmi et al. \cite{Solmi_00} have found from the

740: carrier mobility measurements that the complex to be electrically

741: neutral at room temperature. In \cite{Xie_99} the first-principles

742: calculations were performed to explain the coupled diffusion

743: phenomenona and clustering in heavily arsenic-doped silicon. As

744: contrast to, \cite{Ramamoorthy_96} it was found that the

745: $\text{VAs}_{2}$ cluster is less mobile due to its high migration

746: barrier. However, these clusters can contribute to the diffusion

747: of As at elevated temperatures.

748: 

749: Proceeding from the presented analysis, our understanding of the

750: clustering mechanisms is neither complete nor firmly established.

751: This concerns the number of arsenic atoms in the cluster as well as

752: its structure. However, a study based on the first-principles

753: shows that clusters of a certain kind have an acceptor level and can

754: be negatively charged. Let us consider the models for the clusters

755: with different charge states.

756: 

757: \subsection{Models based on the mass action law}

758: 

759: The models based on thermodynamic formalism are usually used for

760: simulation of high concentration dopant diffusion taking place

761: during semiconductor processing. Assuming a local thermodynamic

762: equilibrium among substitutionally dissolved arsenic, the dopant

763: atoms incorporated into clusters, and electrons, one can use the

764: mass action law to calculate the concentration of clustered dopant

765: atoms. \cite{Tsai_80,Fair_73,Fair_Weber_73,Guerrero_82} The charge

766: conservation law for the reaction of cluster formation under the

767: assumption of local charge neutrality is also useful for

768: describing the clustering phenomenon. A widespread model of Tsai

769: et al. \cite{Tsai_80} was the first to account for the saturation

770: of the charge carriers at increasing doping levels.

771: \cite{Guerrero_82} It is interesting to compare this model with

772: the latest experimental data. The mass action law for the reaction

773: (\ref{Cluster_Reaction_Tsai}) at the diffusion temperature yields

774: \cite{Tsai_80}

775: 

776: \begin{equation} \label{MAL_Tsai}

777: C^{Cl}=A^{Cl}nC^{3},

778: \end{equation}

779: 

780: \noindent where $C^{Cl}$ and $n$ are the concentrations of

781: clusters and electrons at the annealing temperature, respectively;

782: $A^{Cl}$ is the constant of local thermodynamic equilibrium. After

783: cooling, all clusters become neutral and $C\approx n_{R}$ if $C\gg

784: n_{i}$. Here $n_{R}$ is the concentration of electrons at room

785: temperature. It follows from the assumption of local charge

786: neutrality that the electron concentration at the annealing

787: temperature for $C\gg n_{i}$ is

788: 

789: \begin{equation} \label{Electron_equation_Tsai}

790: n\approx C+2C^{Cl}=C+2A^{Cl}nC^{3}.

791: \end{equation}

792: 

793: Solving this equation, Tsai et al. \cite{Tsai_80} obtained

794: 

795: \begin{equation} \label{Electron_Tsai}

796: n=\frac{C}{1-2A^{Cl}C^{3}}

797: \end{equation}

798: 

799: \noindent and

800: 

801: \begin{equation} \label{Total_Tsai}

802: C^{T}=C+\frac{3A^{Cl}C^{4}}{1-2A^{Cl}C^{3}}=\frac{C+A^{Cl}C^{4}}{1-2A^{Cl}C^{3}}.

803: \end{equation}

804: 

805: From Eq.(\ref{Total_Tsai}) it is inferred that under condition

806: 

807: \begin{equation} \label{Csat_Tsai}

808: C=C_{sat}=\frac{1}{\sqrt[3]{2A^{Cl}}}

809: \end{equation}

810: 

811: \noindent both the concentration of arsenic atoms incorporated

812: into clusters and the total concentration of arsenic atoms

813: approach infinity. The value $n_{Re}=C_{sat}$ is interpreted in

814: \cite{Tsai_80} as a maximum level of electron concentration. From

815: the experimental data it is seen that a maximum equilibrium

816: carrier concentration at the annealing temperature in the arsenic

817: implanted layers may be described by the following relation

818: \cite{Solmi_01}

819: 

820: \begin{equation} \label{Electron_Solmi}

821: n_{e}=1.3\times10^{11}\exp\left(-\frac{0.42 eV}{k_{B}T}\right).

822: \end{equation}

823: 

824: Using Eqs.\;(\ref{Electron_Solmi}) and (\ref{Csat_Tsai}), the

825: value of $A^{Cl}$ may be derived to be

826: 

827: \begin{equation} \label{Constant_Tsai}

828: A^{Cl}=\frac{1}{2n_{e}^{3}}.

829: \end{equation}

830: 

831: To illustrate, $n_{e}=3.57\times10^{8}$ $\mu $m$^{-3}$ at a

832: temperature of 1050\,$^{\circ}$C and $A^{Cl}=1.10\times10^{-26}$

833: $\mu $m$^{9}$.

834: 

835: The values of $n_{R}$ as a function of the total arsenic

836: concentration calculated by the model of  Tsai et al.

837: \cite{Tsai_80} are shown in Fig. 1. The experimental values of

838: $n_{R}$ based on the data from \cite{Solmi_98} is also given for

839: comparison. As seen from Fig. 1, saturation of the electron

840: concentration for the model \cite{Tsai_80} is observed at the

841: extremely high total As concentrations, whereas the experimental

842: data show that the carrier concentration reaches its maximum at

843: $C^{T}\approx 8\times10^{8}$ $\mu $m$^{-3}$. Thus, the model

844: \cite{Tsai_80} is inconsistent with the experimental data.

845: Moreover, as follows from the experimental data of Solmi and

846: Nobili, \cite{Solmi_98} the value of $n_{e}$ obtained at room

847: temperature corresponds to the peculiarity of the curve for

848: diffusivity as a function of the total arsenic concentration.

849: Because this feature characterizes impurity diffusion at the

850: annealing temperature, it was concluded in \cite{Solmi_98} that

851: $n_{e}$ had a physical meaning. Thus, the model \cite{Tsai_80}

852: that includes cluster neutralization during cooling is also

853: inconsistent with the experimental data for arsenic diffusivity.

854: Indeed, the electron concentration at the annealing temperature

855: calculated by this model and associated with the peculiarity of

856: the diffusivity is three times greater than $n_{e}$.

857: 

858: 

859: \begin{figure}[ht]

860: \centering {

861: \begin{minipage}[ht]{8.6 cm}

862: {\includegraphics[ scale=0.74]{Fig1.eps}}

863: 

864: \end{minipage}

865: }\caption{Calculated electron concentration against the total

866: arsenic concentration for different models of clustering: solid

867: line --- model of Tsai et al. (Ref. \cite{Tsai_80}), dash-dotted

868: line --- neutral clusters $\text{VAs}_{2}$, and dashed line ---

869: neutral clusters $\text{VAs}_{4}$. The experimental data (circles)

870: are taken from Solmi and Nobili \cite{Solmi_98} for the diffusion

871: at a temperature of 1050\,$^{\circ}$C. \label{As_Cluster1}}

872: \end{figure}

873: 

874: 

875: 

876: Fig. 1 presents the functions $n=n(C^{T})$ calculated for the

877: cases when neutral clusters $VAs_{2}$ and $VAs_{4}$ are formed. In

878: these cases, saturation of the electron density is not observed and

879: the calculations do not agree with the experimental data. The

880: absence of saturation for the electron concentration corresponds

881: to the first-principles calculations in \cite{Berding_98} (Fig. 2)

882: and \cite{Berding_Sher_98} (Fig. 3). As seen from these figures,

883: no saturation is observed at arsenic concentrations when $VAs_{4}$

884: clustering is prevailing.

885: 

886: The reactions

887: 

888: \begin{equation} \label{Reaction_Guerrero}

889: m\text{As}^{+}+k e^{-}\rightleftarrows \text{As}_{m}^{r}

890: \end{equation}

891: 

892: \noindent were used in the basic model for the clustering of As in

893: silicon proposed by Guerrero et al. \cite{Guerrero_82} Here $m$

894: and $k$ are the numbers of arsenic atoms and electrons

895: participating in the clustering process; $r$ is the electric

896: charge of a cluster. The function $C=C(C^{T})$ was investigated by

897: means of the mass action law combined with the charge conservation

898: law for the cluster formation reaction on the assumption of local

899: charge neutrality

900: 

901: \begin{equation} \label{LCN_Guerrero}

902: n=C+rC^{Cl}=C+rA^{Cl}C^{m}n^{k}.

903: \end{equation}

904: 

905: It was supposed in \cite{Guerrero_82} that the charge of clusters

906:  becomes zero when the sample is cooled down to room

907: temperature. Therefore, $C$ is interpreted as the concentration of

908: electrically active arsenic at room temperature, which equals

909: the electron density $n$ in the sample at room temperature. Four

910: main conclusions have been drawn from the basic model:

911: 

912: 1. When the electrons do not participate in the reaction

913: (\ref{Reaction_Guerrero}), i.e., $k=0$, or the clusters are

914: electrically neutral at high temperature, i.e., $r=0$, $C$ is a

915: monotonically increasing function of $C^{T}$ and also there is no

916: finite limit to $C$ and electron density $n$ at room temperature.

917: 

918: 2. If the clusters are positively charged at a high temperature

919: ($r\geq 1$) and exactly one electron takes part in the reaction

920: (\ref{LCN_Guerrero}), i.e., $k=1$, $C$ is a monotonically

921: increasing function of $C^{T}$, which approaches a finite

922: ``saturation'' value $C_{max}$ with increasing $C^{T}$.

923: 

924: As can be see from the reaction (\ref{Cluster_Reaction_Tsai}), the

925: model of Tsai et al. \cite{Tsai_80} satisfies these conditions. On

926: the other hand, from our calculations for this model (see Fig. 1),

927: it follows that saturation of the electron concentration is

928: observed at extremely high total As concentrations. It means that

929: the saturation value obtained in such a manner is inconsistent

930: with the experimental data.

931: 

932: 

933: 3. Provided that the clusters are positively charged at high

934: temperatures ($r\geq 1$) and there are two or more electrons

935: taking part in the reaction (\ref{Reaction_Guerrero}), i.e.,

936: $k\geq 2$, $C$ increases with increasing $C^{T}$ only up to a

937: finite value $C^{T}_{max}$. A maximum concentration value

938: $C=C_{max}$ is observed for $C^{T}=C^{T}_{max}$. With further

939: increase in the total arsenic concentration $C^{T}>C^{T}_{max}$

940: the concentration of the substitutionally dissolved arsenic atoms

941: $C$ and hence electron density $n$ at room temperature, are

942: decreased.

943: 

944: This behavior of $n$ is at variance with the experimental data,

945: because experimentally measured $n$ is approximately constant

946:  with increasing $C^{T}$ (see Fig. 1).

947: 

948: 4. If the clusters are negatively charged at high temperatures

949: ($r\leq -1$), then $C=C(C^{T})$  behaves as in case 1, i.e.,

950: $C$, and hence electron density $n$ at room temperature,

951: monotonically increase without limit for increasing  $C^{T}$. This

952: behavior of $n$ also conflicts with the experimental data.

953: 

954: In what follows, the basic model is extended to take into account

955: the different charge states of vacancies participating in the

956: cluster formation. \cite{Guerrero_82} It is shown that the main

957: conclusions from the basic model are valid for the case when

958: vacancy is involved in the formation of clusters. Thus, the

959: generalized model of arsenic clustering proposed in

960: \cite{Guerrero_82} is not in agreement with the experimental data

961: for all possible cases of the cluster formation considered in this

962: model. In our opinion, this contradiction is due to the assumption

963: that charged clusters become neutral during cooling. As follows

964: from, \cite{Solmi_98} the hypothesis that deactivation is taking

965: place during cooling of the substrates contradicts the

966: experimental data.

967: 

968: Evidently, the models of arsenic clustering should give an

969: adequate explanation not only for saturation of the electron

970: density, but also for the plateau formed on the carrier

971: concentration profile after annealing of the ion implanted layers.

972: \cite{Solmi_98,Solmi_01} As follows from the above analysis, there

973: is no model satisfying these requirements.

974: 

975: The performed analysis enables us to formulate the purpose of this

976: study: the development of a more adequate model %for arsenic clustering

977: to explain the plateau phenomenon of the charge carrier profile at

978: high concentration arsenic diffusion.

979: 

980: %For example, at higher temperatures concentration of isolated

981: %arsenic atom in the lattice $AsSi_{4}$ prevails on the

982: %concentration of $VAs_{4}$.

983: 

984: \section{\textbf{Model}}

985: 

986: To develop a model for arsenic clustering that can describe the

987: saturation phenomenon of the electron density, we assume that not

988: only neutral, but also negatively charged, arsenic clusters are

989: formed. The possibility for the formation of the negatively

990: charged arsenic clusters is confirmed by the calculations in.

991: \cite{Berding_98,Mueller_03} It has been shown by Berding et al.

992: \cite{Berding_98} that clusters $VAs_{3}$ and $VAs_{2}$ have one

993: and two acceptor levels, respectively. The calculations carried

994: out by Mueller et al. \cite{Mueller_03} also show that for the

995: electron densities $n>10^{4}$ $\mu \text{m}^{-3}$ the majority of

996: complexes $\text{VAs}_{2}$ will be negatively charged. According

997: to, \cite{Mueller_03} the complex $\text{VAs}_{3}$ exhibits an

998: acceptor level. This cluster is singly ionized at elevated Fermi

999: levels, resulting in a loss of four mobile carriers for the

1000: formation of a $\text{VAs}_{3}$ aggregate.

1001: 

1002: It is interesting that saturation of the electron density is

1003: observed in the layers of GaAs heavily doped by silicon. To

1004: explain this phenomenon, it was supposed (see, for example

1005: \cite{Greiner_84,Velichko_93}) that silicon atoms in heavily doped

1006: GaAs are dissolved substitutionally not only in the Ga sublattice

1007: (donor $\text{Si}_{\text{Ga}}^{+}$), but also in the As sublattice

1008: (acceptor $\text{Si}_{\text{As}}^{-}$).  It was supposed in

1009: \cite{Adler_95} that a maximum electrical conductivity in Sb-doped

1010: silicon is limited by the defect formation at high dopant

1011: concentrations. The Sb-vacancy pairs were considered as defect

1012: dominating at high doping levels. Also, it was supposed that this

1013: defect acts as an electron acceptor and is responsible for

1014: saturation of charge carriers at high doping levels. Although a

1015: mechanism of such a saturation of the electron density in silicon

1016: heavily doped by Sb is different from that in As-doped silicon,

1017: \cite{Solmi_01} the idea of the limited electron density due to

1018: compensation by the acceptors is a very fruitful one. For example,

1019: it has been supposed in \cite{Velichko_Dobrushkin_Pakula_05} that

1020: phosphorus clusters are negatively charged. Due to this

1021: assumption, saturation of the electron density during high

1022: concentration phosphorus diffusion was explained.

1023: 

1024: 

1025: As distinct from the models of arsenic clustering considered in,

1026: \cite{Tsai_80,Guerrero_82} we assume that arsenic clusters have

1027: the same negative charge both at high diffusion temperatures and

1028: at room temperature. As follows from the analysis of the

1029: experimental data and theoretical investigations, the structure of

1030: arsenic clusters under diffusion temperature is still not

1031: conclusively established. Therefore, it is assumed that a point

1032: defect $\text{D}_{1}$ can be involved in the cluster formation and

1033: another defect $\text{D}_{2}$ can be generated during clustering.

1034: In this case, a reaction for the formation and dissolution of

1035: clusters can be written as

1036: 

1037: \begin{equation}\label{Reaction_Basic}

1038: m\text{As}^{+}  + m_{1}\text{D}^{r_{1}}+ ke^{ -}

1039: {\longleftrightarrow} \Phi^{r_{Cl} } + m_{2}\text{D}^{r_{2}} ,

1040: \end{equation}

1041: 

1042: \noindent where A is the substitutionally dissolved arsenic atom

1043: $\text{As}^{+}$; $e^{-}$ is the electron; $\Phi$ is the cluster

1044: formed; $m_{1}$ and $m_{2}$ are respectively the numbers of

1045: defects $\text{D}_{1}$ and $\text{D}_{2}$ participating in the

1046: cluster formation; $r_{1}$, $r_{2}$, and $r_{Cl}$  are the charge

1047: states of defect $\text{D}_{1}$, defect $\text{D}_{2}$, and

1048: cluster $\Phi^{r_{Cl} }$, respectively.

1049: 

1050: The charge conservation law for the chemical reaction

1051: (\ref{Reaction_Basic}) is of the form

1052: 

1053: \begin{equation} \label{CCL_Basic}

1054: m + m_{1} z_{1} - k = z_{Cl}+ m_{2} z_{2}  ,

1055: \end{equation}

1056: 

1057: \noindent where $z_{1}$ and  $z_{2}$ are the charges of defects

1058: $\text{D}_{1}$ and $\text{D}_{2}$, respectively; $z_{Cl}$ is the

1059: charge of cluster $\Phi^{r_{Cl} }$ in terms of the elementary

1060: charge.

1061: 

1062: The mass action law for reaction (\ref{Reaction_Basic}) yields

1063: 

1064: \begin{equation} \label{MAL_Basic}

1065: K_{L} \;C^{\DS{m}} \;(C_{D1}^{r_{1}} )^{\DS{m_{1}}}

1066: \;\chi^{\DS{k}}n_{i}^{\DS{k}}  = K_{R} \;C_{\Phi}

1067: \;(C_{D2}^{r_{2}} )^{\DS{m_{2}}}  ,

1068: \end{equation}

1069: 

1070: \noindent where $K_{L}$ is the rate of chemical reaction

1071: (\ref{Reaction_Basic}) in the forward direction; $K_{R}$ is the

1072: rate of this reaction in the opposite direction; $C$ is the

1073: concentration of impurity atoms participating in reaction

1074: (\ref{Reaction_Basic}) ; $C_{D1}^{r_{1}}$  is the concentration of

1075: defects in the charge state $r_{1}$ participating in the cluster

1076: formation; $C_{D2}^{r_{2}}$  is the concentration of defects in

1077: the charge state $r_{2}$ generated during clustering; $C_{\Phi}$

1078: is the concentration of the clusters in the charge state $r_{Cl}$;

1079: $\chi$ is the electron density normalized to the concentration of

1080: the intrinsic charge carriers in a semiconductor during diffusion

1081: $n_{i}$.

1082: 

1083: Due to a high dopant concentration, the approximation of local

1084: charge neutrality can be used for the evaluation of $\chi$

1085: 

1086: 

1087: \begin{eqnarray} \label{chi}

1088: \chi = \DS{\frac{n}{n_{i}}} &=& \frac{1}{2n_{i}} \biggl[ C+ z_{Cl}

1089: \; C_{\Phi} - N_{B} \nonumber \\

1090:     &+& \left. \sqrt {\left( C +z_{Cl} \; C_{\Phi} - N_{B}

1091: \right)^{2} + 4n_{i}^{2}}\, \right]\, ,

1092: \end{eqnarray}

1093: 

1094: 

1095: \noindent where $N_{B}$ is the summarized concentration of

1096: acceptors; $n_{ie}$ is the effective concentration of the

1097: intrinsic charge carriers at a diffusion temperature calculated

1098: for a high doping level.

1099: 

1100: It follows from equation (\ref{Reaction_Basic})  that the

1101: concentration of impurity atoms incorporated into clusters $C^{A}

1102: = mC_{\Phi}$. Moreover, considering highly mobile electrons, the

1103: mass action law for the reaction of defect conversion between the

1104: charge states is valid

1105: 

1106: \begin{equation} \label{MAL_Electron_1}

1107: C_{D1}^{r_{1}}  = h_{D1}^{r_{1}} \;n^{\DS{-z_{1}}} \,

1108: C_{D1}^{\times} = h_{D1}^{r_{1}} \, n_{i}^{\DS{ - z_{1}}} \;\chi

1109: ^{\DS{ - z_{1}}} \, C_{D1}^{\times} \quad {  ,}

1110: \end{equation}

1111: 

1112: \begin{equation} \label{MAL_Electron_2}

1113: C_{D1}^{r_{2}}  = h_{D2}^{r_{2}} \;n^{\DS{-z_{2}}} \,

1114: C_{D2}^{\times} = h_{D2}^{r_{2}} \, n_{i}^{\DS{ - z_{2}}} \;\chi

1115: ^{\DS{ - z_{2}}} \, C_{D2}^{\times} \quad {  ,}

1116: \end{equation}

1117: 

1118: \noindent where $h_{D1}^{r_{1}}$ and $h_{D2}^{r_{2}}$ are the

1119: constants for the local thermodynamic equilibrium in the reactions

1120: when neutral defects $\text{D}_{1}^{\times}$ and

1121: $\text{D}_{2}^{\times}$ are converted into charge states $r_{1}$

1122: and $r_{2}$, respectively.

1123: 

1124: 

1125: Substituting (\ref{MAL_Electron_1}) and (\ref{MAL_Electron_2})

1126: into (\ref{MAL_Basic}) and taking into consideration that $C_{AC}

1127: = mC_{Cl}$, we obtain

1128: 

1129: \begin{equation} \label{CA}

1130: C_{AC} = K\; \tilde {C}_{D} \;\chi ^{\displaystyle{(

1131: \;k\;+r_{2}m_{2}-r_{1}m_{1})}} \;C^{\displaystyle{m}} ,

1132: \end{equation}

1133: 

1134: \noindent where

1135: 

1136: \begin{equation} \label{Constant}

1137: K = \frac{m\;K_{L} \,(h_{D1 \DS{r_{1}}})^{\DS{m_{1}}} \,

1138: n_{i}^{\DS{(k+r_{2}m_{2}-r_{1}m_{1})}} \,(C_{D1

1139: \DS{\times}}^{eq})^{\DS{m_{1}}}} {K_{R} \,(h_{D2

1140: \DS{r_{2}}})^{\DS{m_{2}}} \,(C_{D2 \DS{\times}}^{eq})^{\DS{m_{2}}}

1141: \, },

1142: \end{equation}

1143: 

1144: 

1145: \begin{equation} \label{Rel_Defect}

1146: \tilde {C}_{D}=\frac{(\tilde {C}_{D1})^{\DS{m_{1}}}}{(\tilde

1147: {C}_{D2})^{\DS{m_{2}}}} \, , \quad \tilde {C}_{D1}= \frac{{C}_{D1

1148: \DS{\times}}}{{C}_{D1 \DS{\times}}^{eq}} \, ,\quad \tilde

1149: {C}_{D2}= \frac{{C}_{D2 \DS{\times}}}{{C}_{D2 \DS{\times}}^{eq}}

1150: \, .

1151: \end{equation}

1152: 

1153: \noindent Here $C_{AC}$ is the concentration of clustered arsenic

1154: atoms and $C_{T} = C + C_{AC}$ is the total arsenic concentration.

1155: The parameter $K$ has a constant value depending on the

1156: temperature of diffusion. This value can be extracted from the

1157: best fit to the experimental data. The quantities ${C}_{D1

1158: \DS{\times}}^{eq}$ and ${C}_{D2 \DS{\times}}^{eq}$ are the

1159: equilibrium concentrations of corresponding defects in the neutral

1160: charge state.

1161: 

1162: 

1163: 

1164: \section{\textbf{Results of calculations}}

1165: 

1166: Analysis of Eqs. (\ref{chi}) and (\ref{CA}) shows that saturation

1167: of the electron density will be observed for the formation of

1168: singly negatively charged clusters incorporating one arsenic atom

1169: and a lattice defect. The calculations of the functions $C =

1170: C(C^{T})$, $C^{A} = C^{A}(C^{T})$, and $n = n(C^{T} )$ for this

1171: case are presented in Fig. 2. For comparison with the experimental

1172: data, \cite{Solmi_98} the processing temperature was chosen to be

1173: 1050\,$^{\circ}$C ($n_{i} $ = 1.1665$\times10^{7}$ $\mu$m$^{-3}$).

1174: To calculate the functions $C = C\left( C^{T} \right)$,

1175: $C^{A}(C^{T})$, and $n(C^{T})$, a system of nonlinear equations

1176: (\ref{chi}), (\ref{CA}) was solved numerically by Newton's method.

1177: In the case of singly negatively charged clusters

1178: $(\text{DAs})^{-}$ incorporating a lattice defect and one arsenic

1179: atom we adopt $r_{Cl}$ = -1 and $m_{1} $ = 1. Eq.

1180: (\ref{CCL_Basic}) for $m$ = 1 yields that $- z_{1}m_{1} -

1181: z_{2}m_{2} + k$ = 2, and the system (\ref{CA}), (\ref{chi}) is of

1182: the form:

1183: 

1184: \begin{equation} \label{CA1}

1185: C^{A} = K\;\tilde {C}_{D} \;\chi ^{2}\;C \; ,

1186: \end{equation}

1187: 

1188: \begin{eqnarray} \label{chi1}

1189: \chi &=& \frac{1}{2n_{i}} \biggl[ C - K \tilde{C}_{D} \chi^{2}C -

1190: N_{B}+ \nonumber \\

1191:      &+& \left.  \sqrt {\left( C - K \tilde {C}_{D} \chi ^{2}C - N_{B}

1192: \right)^{2} + 4n_{ie}^{2}}\quad\right] {  .}

1193: \end{eqnarray}

1194: 

1195: As can be seen from Fig. 1, the concentration of charge carriers

1196: in the saturation region is approximately equal to the value

1197: $n_{e}=$ 3.5665$\times $10$^{8}$ $\mu$m$^{-3}$. In Fig. 2 for

1198: $C^{T}\rightarrow+\infty$, the electron concentration reaches

1199: saturation at the value $n \sim $ 3.765$\times $10$^{8}$

1200: $\mu$m$^{-3}$ associated with the plateau of the charge carrier

1201: profile given in Fig. 1. The value of parameter $K\tilde{C_{D}}$

1202: for this case was chosen as $ 1.0698\times

1203:  10^{{ -4}}$ a. u. to provide a fit to the experimental value of

1204: $n_{e}$. As seen from Fig. 2, the function $n=n(C^{T})$ is similar

1205: to that predicted by the model of Tsai et al. \cite{Tsai_80}

1206: Saturation of the electron density occurs at extremely high values

1207: of the total impurity concentration $C^{T}$, which is in conflict

1208: with the experimental data.

1209: 

1210: %\begin{figure}

1211: %\includegraphics{}

1212: %\caption{\label{}}

1213: %\end{figure}

1214: 

1215: \begin{figure}[ht]

1216: \begin{minipage}[ht]{8.6 cm}

1217: {\includegraphics[ scale=0.74]{Fig2.eps}}

1218: \end{minipage}

1219: 

1220: \caption{Calculated concentrations of substitutionally dissolved

1221: arsenic atoms (solid line), clustered arsenic atoms (dashed line),

1222: and electron density (dotted line) against the total dopant

1223: concentration for the formation of singly negatively charged

1224: clusters incorporating one arsenic atom. The measured electron

1225: density (circles) is taken from Solmi and Nobili \cite{Solmi_98}

1226: for the diffusion at a temperature of 1050\,$^{\circ}$C.

1227: \label{As_Cluster2}}

1228: \end{figure}

1229: 

1230: 

1231: In Fig. 3a the system of nonlinear equations (\ref{chi}),

1232: (\ref{CA}) is solved numerically for the formation of singly

1233: negatively charged  clusters $(DAs_{2})^{-}$ incorporating a

1234: lattice defect and two arsenic atoms ($r_{Cl} $ = -1, $m_{1} $ =

1235: 1, and $m$ = 2). Then, from (\ref{CCL_Basic}), (\ref{chi}), and

1236: (\ref{CA}) it follows that:

1237: 

1238: \begin{equation} \label{CA2}

1239: C^{A} = K\;\tilde {C}_{D} \;\chi ^{3}\;C^{2} \, ,

1240: \end{equation}

1241: 

1242: \begin{eqnarray} \label{chi2}

1243: \chi &=& \frac{1}{2n_{i}} \Biggl[ C - \DS{\frac{1}{2}}K

1244: \tilde{C}_{D} \chi^{3}C^{2} - N_{B}+ \nonumber \\

1245:      &+& \left.  \sqrt {\left( C - \DS{\frac{1}{2}}K \tilde {C}_{D} \chi ^{3}C^{2} - N_{B}

1246: \right)^{2} + 4n_{ie}^{2}}\, \right] .

1247: \end{eqnarray}

1248: 

1249: 

1250: The calculations for $n=n(C^{T})$ show that, with the increasing

1251: total concentration of impurity atoms $C^{T}$, the concentration of

1252: charge carriers $n$  reaches its maximum value $n_{max}$

1253: =3.750$\times 10^{8}$ $\mu$m$^{-3}$ at  $C^{T}_{\max} \sim

1254: 1.634\times$10$^{9}$ $\mu$m$^{-3}$ and then decreases

1255: monotonically. At maximum the concentration of the

1256: substitutionally dissolved arsenic atoms $C_{\max}$ is

1257: $7.943\times$10$^{8}$ $\mu$m$^{-3}$. The parameter

1258: $K\tilde{C_{D}}$ used in this calculation is equal to

1259: 4.0$\times$10$^{-14}$ $\mu$m$^{3}$. The calculated curve agrees

1260: with the experimental function $n=n(C^{T})$ much better than those

1261: obtained by the model of Tsai et al. \cite{Tsai_80} and the models

1262: of neutral $(\text{DAs}_{2})$ and neutral $(\text{DAs}_{4})$

1263: clusters.

1264: 

1265: \begin{figure}[ht]

1266: \begin{minipage}[ht]{8.6cm}

1267: {\includegraphics[ scale=0.74]{Fig3a.eps}} \centerline{(a)}

1268: \end{minipage}

1269: 

1270: \begin{minipage}[t]{8.6cm}

1271: {\includegraphics[ scale=0.74]{Fig3b.eps}}

1272: 

1273: \centerline{(b)}

1274: \end{minipage}

1275: 

1276: \caption{Calculated concentrations of substitutionally dissolved

1277: arsenic atoms (solid line), clustered arsenic atoms (dashed line),

1278: and electron density (dotted line) against the total dopant

1279: concentration for the formation of  singly (a) and doubly (b)

1280: negatively charged clusters incorporating two arsenic atoms. The

1281: measured electron density (circles) is taken from Solmi and Nobili

1282: \cite{Solmi_98} for the diffusion at a temperature of

1283: 1050\,$^{\circ}$C. \label{As_Cluster3a-b}}

1284: \end{figure}

1285: 

1286: The calculations of $n=n(C^{T})$ and  $C^{A}=C^{A}(C^{T})$ were

1287: also performed for the clusters $(\text{DAs}_{2})^{2-}$,

1288: $(\text{DAs}_{3})^{-}$, $(\text{DAs}_{3})^{2-}$,

1289: $(\text{DAs}_{4})^{-}$, and $(\text{DAs}_{4})^{2-}$. And similar

1290: behavior of  $n=n(C^{T})$ was observed for all these cases. Also,

1291: the concentration of charge carriers increases with increasing

1292: $C^{T}$, reaches a maximum, and then decreases monotonically. This

1293: decrease is more pronounced with an increased number of arsenic atoms

1294: in the cluster. Besides, the position of maximum carrier

1295: concentration is shifted to smaller dopant concentrations with

1296: increases in the negative charge of the cluster.

1297: 

1298: It follows from comparison of the calculated distributions

1299: $n(C^{T})$ with the experimental data that the best fit is

1300: observed for the doubly negatively charged cluster

1301: $(\text{DAs}_{2})^{2-}$ (see Fig. 3b). To obtain the curves shown

1302: in Fig. 4, we solve the system (\ref{chi}), (\ref{CA}) having the

1303: following form for the clusters $(\text{DAs}_{2})^{2-}$ \, :

1304: 

1305: \begin{equation} \label{CA22}

1306: C^{A} = K\;\tilde {C}_{D} \;\chi ^{4}\;C^{2} \; ,

1307: \end{equation}

1308: 

1309: \begin{eqnarray} \label{chi22}

1310: \chi &=& \frac{1}{2n_{i}} \Biggl[ C - K \tilde{C}_{D}

1311: \chi^{4}C^{2} - N_{B}+ \nonumber \\

1312:      &+& \left.  \sqrt {\left( C - K \tilde {C}_{D}

1313:       \chi ^{4}C^{2} - N_{B}

1314: \right)^{2} + 4n_{ie}^{2}} \, \right] \, .

1315: \end{eqnarray}

1316: 

1317: At maximum value of $n_{max}=3.694\times$10$^{8}$ $\mu$m$^{-3}$,

1318: the concentration of the substitutionally dissolved arsenic atoms

1319: $C_{\max}$ is equal to $7.943\times$10$^{8}$ $\mu$m$^{-3}$. The

1320: calculated curve $n=n(C^{T})$ agrees well with the experimental

1321: data, although only one fitting parameter $K\tilde{C_{D}}=$

1322: 0.67$\times$10$^{-15}$ $\mu$m$^{3}$ has been used. Full agreement

1323: is reached in the region of the transition to saturation of the

1324: electron density $n$. This means that two arsenic atoms are

1325: incorporated in the cluster, and this arsenic cluster is doubly

1326: negatively charged, at least at the total dopant concentrations

1327: $C^{T}\leq$ 1.0$\times$10$^{9}$ $\mu$m$^{-3}$. Minor differences

1328: in values of the theoretical curve and experimental data for

1329: $C^{T}>$ 2.3$\times$10$^{9}$ $\mu$m$^{-3}$ can arise due to heavy

1330: doping effects (change of the silicon zone structure, changes in

1331: the constants of forward $K_{L}$ and backward $K_{R}$ reactions,

1332: etc). It is clear from expressions (\ref{chi}), (\ref{CA}) that

1333: variation of the defect concentration $\tilde {C}_{D}$ in the

1334: region of high impurity concentrations also influences the

1335: concentration of impurity atoms incorporated in the clusters and,

1336: hence, the concentration of charge carriers. A change of $\tilde

1337: {C}_{D}$ due to generation (absorption) of the defects in the

1338: region with high arsenic concentration is quite possible. For

1339: example, such a generation can occur due to the lattice expansion.

1340: Besides, it is possible that a high doping level leads to error in

1341: the determination of the experimental values for electron density,

1342: and the data for $C^{T}>$ 2.3$\times$10$^{9}$ $\mu$m$^{-3}$ may be

1343: incorrect.

1344: 

1345: Figs. 4 and b show the functions $C = C(C^{T})$, $C^{A} =

1346: C^{A}(C^{T})$, and $n = n(C^{T})$ calculated for clusters

1347: $(\text{DAs}_{3})^{-}$ and $(\text{DAs}_{4})^{-}$, respectively.

1348: As can be seen from these figures, fitting to the experimental

1349: data is worse in comparison with Fig. 3b, especially for the total

1350: concentrations greater than 2.0$\times$10$^{9}$ $\mu$m$^{-3}$.

1351: 

1352: 

1353: \begin{figure}[ht]

1354: \begin{minipage}[ht]{8.6cm}

1355: {\includegraphics[ scale=0.74]{Fig4a.eps}} \centerline{(a)}

1356: \end{minipage}

1357: 

1358: \begin{minipage}[ht]{8.6 cm}

1359: {\includegraphics[ scale=0.74]{Fig4b.eps}} \centerline{(b)}

1360: \end{minipage}

1361: 

1362: 

1363: \caption{Calculated concentrations of substitutionally dissolved

1364: arsenic atoms (solid line), clustered arsenic atoms (dashed line),

1365: and electron density (dotted line) against the total dopant

1366: concentration for the formation of singly negatively charged

1367: clusters incorporating three (a) and four (b) arsenic atoms. The

1368: measured electron density (circles) are taken from Solmi and

1369: Nobili \cite{Solmi_98} for the diffusion at a temperature of

1370: 1050\,$^{\circ}$C. \label{As_Cluster4a-b}}

1371: \end{figure}

1372: 

1373: 

1374: Note that multiple clustering can occur, too, and different clusters

1375: can coexist in the equilibrium state. For example, the clusters

1376: incorporating different numbers of arsenic atoms

1377: $(\text{DAs}_{1})^{-}$ and $(\text{DAs}_{2})^{-}$ or clusters in

1378: different charge states $(\text{DAs}_{2})^{\times}$,

1379: $(\text{DAs}_{2})^{-}$, and $(\text{DAs}_{2})^{2-}$ can coexist.

1380: As follows from the calculated curves for the electron density,

1381: when clusters $(\text{VAs}_{1})^{-}$ (Fig. 2) and

1382: $(\text{VAs}_{2})^{\times}$ (Fig. 1) are formed, it may be

1383: expected that the fall of the electron density in the plateau region

1384: will be less pronounced in the case of multiple clustering.

1385: However, this problem requires further investigation.

1386: 

1387: \section{\textbf{Conclusions}}

1388: 

1389: Based on analysis of the experimental results and theoretical

1390: calculations for clustering of arsenic atoms in silicon, it was

1391: shown that the developed clustering models fail to explain the

1392: available experimental data, especially the effect of electron

1393: density saturation at high arsenic concentrations. Therefore, a

1394: new model of arsenic clustering was proposed and analyzed. The

1395: main feature of the proposed model is the assumption that

1396: negatively charged arsenic complexes play a dominant role in the

1397: clustering process. To confirm this assumption, the concentration

1398: of the impurity atoms incorporated in clusters and electron

1399: density were calculated as a function of the total arsenic

1400: concentration at a temperature of 1050\,$^{\circ}$C. Different

1401: cases of the formation of negatively charged clusters

1402: incorporating a point defect and one arsenic atom

1403: $(\text{DAs}_{1})^{-}$, $(\text{DAs}_{1})^{2-}$ or more arsenic

1404: atoms $(\text{DAs}_{2})^{-}$, $(\text{DAs}_{2})^{2-}$,

1405: $(\text{DAs}_{3})^{-}$, $(\text{DAs}_{3})^{2-}$,

1406: $(\text{DAs}_{4})^{-}$, and $(\text{DAs}_{4})^{2-}$ were

1407: investigated. It was shown that in the case of

1408: $(\text{DAs}_{1})^{-}$ the concentration of charge carriers

1409: reaches saturation with increase in the total arsenic

1410: concentration and concentration of the substitutionally dissolved

1411: impurity atoms. However, saturation was observed for very high

1412: values of the total arsenic concentration, conflicting with the

1413: experimental data. In the cases of doubly negatively charged

1414: cluster $(\text{DAs}_{1})^{2-}$ and clusters incorporating more

1415: than one arsenic atom, the electron density reached its maximum

1416: value, then  decreased monotonically and slowly with an increase

1417: in the total arsenic concentration. The electron density

1418: calculated for the formation of $(\text{DAs}_{2})^{2-}$ clusters

1419: agrees well with the experimental data and confirms the conclusion

1420: that two arsenic atoms participate in the cluster formation. Minor

1421: difference between the theoretical curve and experimental values

1422: for $C^{T}>$ 2.3$\times$10$^{9}$ $\mu$m$^{-3}$ may be explained by

1423: various heavy doping effects or uncertainty due to the

1424: experimental errors.

1425: 

1426: Thus, the plateau on the experimental distribution of the charge

1427: carriers, characterized by a practically constant value of the

1428: electron density in the region of high doping level, may be

1429: attributed to the negatively charged clusters

1430: $(\text{DAs}_{2})^{2-}$. The parameters describing arsenic

1431: clustering at a temperature of 1050\,$^{\circ}$C were determined

1432: by fitting the calculated values of the electron density to the

1433: experimental data. The proposed model of clustering by the

1434: formation of the negatively charged $(\text{DAs}_{2})^{2-}$ system

1435: gives the best fit to the experimental data among all the

1436: present-day models and may be used in simulation of high

1437: concentration arsenic diffusion.

1438: 

1439: \newpage

1440: 

1441: \begin{thebibliography}{99}

1442: 

1443: %Furnace annealing

1444: 

1445: %Ion-implanted layers

1446: \bibitem{Tsukamoto_80} K. Tsukamoto, Y. Akasaka, and K. Kuima,

1447: Jpn. J. Appl. Phys. \textbf{19}, 87 (1980).

1448: 

1449: \bibitem{Tsai_80} M. Y. Tsai, F. F. Morehead, J. E. E. Baglin,

1450: and A. E. Michel, J. Appl. Phys. \textbf{51}, 3230 (1980).

1451: 

1452: \bibitem{Fahey_89} P. M. Fahey, P. B. Griffin, and

1453: J. D. Plummer, Rev. Mod. Phys. \textbf{61}, 289 (1989).

1454: 

1455: \bibitem{Nobili_94} D. Nobili, S. Solmi, A. Parisini, M. Derdour,

1456:   A. Armigliato, and  L. Moro, Phys. Rew. B

1457: \textbf{49}, 2477 (1994).

1458: 

1459: \bibitem{Solmi_98} S. Solmi and D. Nobili, J. Appl. Phys.

1460: \textbf{83}, 2484 (1998).

1461: 

1462: \bibitem{Krishnamoorthy_98} V. Krishnamoorthy, K. Moller, K. S. Jones,

1463:   D. Venables, J. Jackson, and L. Rubin, J. Appl. Phys. \textbf{84},

1464:   5997 (1998).

1465: 

1466: \bibitem{Nobili_99} D. Nobili, S. Solmi, M. Merli, and J. Shao,

1467:   J. Electrochem. Soc., \textbf{146}, 4246 (1999).

1468: 

1469: \bibitem{Uematsu_00} M. Uematsu, Jpn. J. Appl. Phys., Part. 1

1470: \textbf{39},  1006 (2000).

1471: 

1472: \bibitem{Nobili_01} D. Nobili, S. Solmi, and J. Shao, J.

1473: Appl. Phys. \textbf{90}, 101 (2001).

1474: 

1475: \bibitem{Solmi_01} S. Solmi, in {\itshape Encyclopedia of Materials:

1476:   Science and Technology}, edited by K. H. J. Buschow, R. W. Cahn, M. C.

1477: Flemings, B. Ilschner, E. J. Kramer, S. Mahajan, and P.

1478: Veyssi\`{e}re (Elsevier Science Ltd., 2001) p. 2331.

1479: 

1480: \bibitem{Solmi_03} S. Solmi, M. Ferri, M. Bersani, D. Giubertoni, and

1481:   V. Soncini, J. Appl. Phys. \textbf{94}, 4950 (2003).

1482: 

1483: %Thermal diffusion

1484: 

1485: \bibitem{Fair_73} R. B Fair and G. R. Weber, J. Appl. Phys.

1486: \textbf{44}, 273 (1973).

1487: 

1488: \bibitem{Fair_Weber_73} R. B Fair and G. R. Weber, J. Appl. Phys.

1489: \textbf{44}, 280 (1973).

1490: 

1491: \bibitem{Murota_79} J. Murota, E. Arai, K. Kobayashi, and

1492:   K. Kubo. J. Appl. Phys. \textbf{50}, 804 (1979).

1493: 

1494: %Deactivation after laser anneling

1495: 

1496: \bibitem{Parisini_90} A. Parisini, A. Bourret, A. Armigliato,

1497:   M. Servidori, S. Solmi, R. Fabbri, J. R. Regnard, and J. L. Allain,

1498:   J. Appl. Phys. \textbf{67}, 2320 (1990).

1499: 

1500: \bibitem{Luning_92} S. Luning, P. M. Rousseau, P. B. Griffin, P. G. Carey,

1501: and J. D. Plummer, Tech. Dig. Int. Electron. Devices Meet. 457 (1992).

1502: 

1503: \bibitem{Rousseau_94} P. M. Rousseau, P. B. Griffin, and

1504: J. D. Plummer, Appl. Phys. Lett. \textbf{65}, 578 (1994).

1505: 

1506: \bibitem{Rousseau_98} P. M. Rousseau, P. B. Griffin, W. T. Fang,

1507: and J. D. Plummer, J. Appl. Phys. \textbf{84}, 3593 (1998).

1508: 

1509: \bibitem{Solmi_02} S. Solmi, M. Attari, and D. Nobili, Appl.

1510: Phys. Lett. \textbf{80}, 4774 (2002).

1511: 

1512: %Reverse annealing

1513: 

1514: \bibitem{Solmi_00} S. Solmi, D. Nobili, and J. Shao, J. Appl.

1515: Phys. \textbf{87}, 658 (2000).

1516: 

1517: %Precipitation

1518: 

1519: \bibitem{Nobili_83} D. Nobili, A. Carabelas, G. Celotti, and

1520: S. Solmi, J. Electrochem. Soc. \textbf{130}, 922 (1983).

1521: 

1522: \bibitem{La Via_91}F. La Via, V. Privitera, S. Lombardo, C. Spinella,

1523:   V. Raineri, E. Rimini, P. Baeri, and G. Ferla, J. Appl. Phys.

1524: \textbf{69}, 726 (1991).

1525: 

1526: \bibitem{Guerrero_82}

1527: E. Guerrero, H. P\"{o}tzl, R. Tielert, M. Grasserbauer, and G.

1528: Stingeder, J. Electrochem. Soc. \textbf{129},1826 (1982).

1529: 

1530: \bibitem{Pandey_88} K. C. Pandey, A. Erbil, G. S. Cargill, III,

1531:   R. F. Boehme, and David Vanderbilt, Phys. Rev. Lett. \textbf{61},

1532:   1282 (1988).

1533: 

1534: \bibitem{Berding_98}

1535: M.A. Berding, A. Sher, and M. van Schilfgaarde, Appl. Phys. Let.

1536: \textbf{72}, 1492 (1998).

1537: 

1538: \bibitem{Mueller_03} D. C. Mueller, E. Alonso, and W. Fichtner,

1539: Phys. Rev. B \textbf{68}, 045208 (2003).

1540: 

1541: % Low-temperature treatments

1542: 

1543: % Experimental investigations of defect subsystem

1544: 

1545: % TEM

1546: 

1547: \bibitem{Dokumachi_95} 0. Dokumaci, P. Rousseau, S. Luning,

1548:   V. Krishnamoorthy, K. S. Jones, and M. E. Law,

1549:   J. Appl. Phys. \textbf{78}, 828 (1995).

1550: 

1551: % EXAFS

1552: 

1553: \bibitem{Erbil_86} A. Erbil, W. Weber, G. S. Cargill III, and R. F.

1554: Boehme, Phys. Rev. B \textbf{34}, 1392 (1986).

1555: 

1556: \bibitem{Allain_92} J. L. Allain,  J. R. Regnard, A. Bourret,

1557:   A. Parisini, A. Armigliato, G. Tourillon, and S. Pizzini, Phys. Rev. B

1558: \textbf{46}, 9434 (1992).

1559: 

1560: \bibitem{d'Acapito_04} F. d'Acapito, C. Maurizio, and M. Malvestuto,

1561: Mater. Sci. Eng., B \textbf{114-115}, 386 (2004).

1562: 

1563: % RBS

1564: 

1565: \bibitem{Brizard_93} C. Brizard, J. R. Regnard, J. L. Allain,

1566:   A. Bourret, M. Dubus, A. Armigliato, and A. Parisini,

1567:   J. Appl. Phys. \textbf{75}, 126 (1994).

1568: 

1569: % positron

1570: 

1571: \bibitem{Makinen_89} J. M\"{a}kinen, C. Corbel, P. Hautoj\"{a}rvi,

1572:   P. Moser, and F. Pierre, Phys. Rev. B \textbf{39}, 10162 (1992).

1573: 

1574: \bibitem{Lawther_95} D.W. Lawther, U. Myler, P. J. Simpson,

1575:   P. M. Rousseau, P. B. Griffin, and J. D. Plummer,

1576:   Appl. Phys. Lett. \textbf{67}, 3575 (1995).

1577: 

1578: \bibitem{Myler_96} U. Myler, R. D. Goldberg, A. P. Knights,

1579:   D. W. Lawther, and P. J. Simpson, Appl. Phys. Lett. \textbf{69},

1580:   3333 (1996).

1581: 

1582: \bibitem{Szpala_96} S. Szpala, P. Asoka-Kumar, B. Nielsen, J. P. Peng,

1583: S. Hayakawa, and K. G. Lynn, Phys. Rev. B \textbf{54}, 4722

1584: (1996).

1585: 

1586: % positron + electron moments

1587: 

1588: \bibitem{Polity_98} A. Polity, F. B\H{o}rner, S. Huth, S. Eichler, and

1589: R. Krause-Rehberg, Phys. Rev. B \textbf{58}, 10363 (1998).

1590: 

1591: \bibitem{Saarinen_99} K. Saarinen, J. Nissil\H{a}, H. Kauppinen,

1592: M. Hakala, M. J. Puska, P. Hautoj\"{a}rvi, and C. Corbel, Phys.

1593: Rev. Lett. \textbf{82}, 1883 (1999).

1594: 

1595: \bibitem{Ranki_02} V. Ranki, J. Nissil\"{a}, and K. Saarinen,

1596: Phys. Rev. Lett. \textbf{88}, 105506 (2002).

1597: 

1598: \bibitem{Ranki_03} V. Ranki, K. Saarinen, J. Fage-Pedersen,

1599: J. Lundsgaard Hansen, and A. Nylandsted Larsen, Phys. Rev. B

1600: \textbf{67}, 041201(R) (2003).

1601: 

1602: \bibitem{Ranki_Saarinen_03} V. Ranki and K. Saarinen, Physica B

1603: \textbf{340-342}, 765 (2003).

1604: 

1605: \bibitem{Ranki_04} V. Ranki, A. Pelli, and K. Saarinen,

1606: Phys. Rev. B \textbf{69}, 115205 (2004).

1607: 

1608: % x-ray standing-wave spectroscopy

1609: 

1610: \bibitem{Herrera-Gomez_96} A. Herrera-G\"{o}mez, P. M. Rousseau,

1611:   G. Materlik, T. Kendelewicz, J. C. Woicik, P. B. Griffin,

1612:   J. Plummer, and W. E. Spicer, Appl. Phys. Lett. \textbf{68}, 3090 (1996).

1613: 

1614: % First principles

1615: 

1616: \bibitem{Ramamoorthy_96} M. Ramamoorthy and S. T. Pantelides,

1617: Phys. Rev. Lett. \textbf{76}, 4753 (1996).

1618: 

1619: \bibitem{Chadi_97} D. J. Chadi, P. H. Citrin, C. H. Park, D. L. Adler,

1620:   M. A. Marcus, and H.-J. Gossmann, Phys. Rev. Lett. \textbf{79}, 4834 (1997).

1621: 

1622: \bibitem{Citrin_03} P.H. Citrin, P.M. Voyles, D.J. Chadi, and D.A. Muller,

1623: Physica B. \textbf{340-342}, 784 (2003).

1624: 

1625: \bibitem{Larsen_86} A. Nylandsted Larsen, S. Yu. Shiryaev, E. S. S\o

1626:   rensen, and P. Tidemand-Peterson, Appl. Phys. Lett. \textbf{48}, 1805 (1986).

1627: 

1628: \bibitem{Larsen_93} A. Nylandsted Larsen, K. Kyllesbech Larsen,

1629: P. E. Andersen, and B. G. Svensson, J. Appl. Phys. \textbf{73},

1630: 691 (1993).

1631: 

1632: \bibitem{Berding_Sher_98} M. A. Berding and A. Sher, Phys. Rev.

1633: B. \textbf{58}, 3853 (1998).

1634: 

1635: % ADF-STEM experiments

1636: \bibitem{Voyles_02} P. M. Voyles,  D. A. Muller, J. L. Grazul, P. H. Citrin,

1637: and H.-J. L. Gossmann, Nature, \textbf{416}, 826 (2002).

1638: 

1639: % First principles

1640: 

1641: \bibitem{Xie_99} J. Xie and S. P. Chen, Phys. Rev. Lett.

1642: \textbf{83}, 1795 (1999).

1643: 

1644: \bibitem{Greiner_84} M. E. Greiner and J. F. Gibbons. Appl.

1645: Phys. Lett. \textbf{44}, 750 (1984).

1646: 

1647: \bibitem{Velichko_93} O. I. Velichko, A. A. Egorov, and S. K. Fedoruk,

1648: Inzh.-Fiz. Zhurnal  {\bf65}, 567 (1993) [J. Eng. Phys. Thermophys.

1649: {\bf65}, 1091 (1993)]. [Russian journal reference with English

1650: translation]

1651: 

1652: \bibitem{Adler_95}D. L. Adler,  J. D. Chadi, M. A. Marcus, H.-J. Gossmann,

1653: and P. H. Citrin, Bull. Amer. Phys. Soc. {\bf40}, 396 (1995).

1654: 

1655: \bibitem{Velichko_Dobrushkin_Pakula_05}

1656: O. I. Velichko, V. A. Dobrushkin, L. Pakula, Mater. Sci. Eng. B

1657: {\bf123}, 176 (2005).

1658: 

1659: \end{thebibliography}

1660: 

1661: \end{document}

1662: