1: %5/15/02---Feedback from PBL and KR

2: %

3: %Third draft 5/13/02 with new figure and feedback from JS and MD

4: %

5: %Second draft 5/10/02 with figures and feedback from JS and MD.

6: %

7: %Working draft 5/10/02 by PC.

8: %

9: %\documentstyle{article}

10: \documentstyle[preprint,floats,aps,epsf,epsfig,prb]{revtex}

11: \title{Thickness-Dependence of the Coercive Field in Ferroelectrics}

12: %

13: \author{P. Chandra,$^1$ M. Dawber,$^2$ P.B. Littlewood$^3$ and J.F. Scott$^2$}

14: \address{$^{1}$NEC, 4 Independence Way, Princeton, NJ 08540, U.S.A.}

15: \address{$^{2}$Symetrix Centre for Ferroics, Earth Sciences

16: Department, Downing Street, University of Cambridge, Cambridge U.K.

17: CB2 3EQ}

18: \address{$^{3}$Cavendish Laboratory, University of Cambridge,

19: Madingley Road, Cambridge U.K. 

20: CB3 OHE} 

21: %

22: \renewcommand\floatpagefraction{.7}

23: \renewcommand\dblfloatpagefraction{.7}

24: \newcommand\joinreld{\mathrel{\mkern-10mu}}

25: \newcommand \joinrely{\mathrel{\mkern-7mu}}

26: \newcommand \relbd{\mathrel{{\bf\smash{{\phantom- \above1pt \phantom-

27: }}}}}

28: \newcommand \ltdash{\raise-1.8pt\hbox{$\scriptscriptstyle |$}}

29: \newcommand \dsp {\displaystyle}

30: \newcommand \bea {\begin{eqnarray} }

31: \newcommand \eea {\end{eqnarray}}

32: \newcommand \vsk {\vskip 1.5 truein}

33: \newcommand \dg {^{\dagger}}

34: \newcommand \rarrow{\rightarrow}

35: \newlength{\bxwidth}\bxwidth=0.8\textwidth

36: \newcommand\frm[1]{\epsfig{file=#1,width=\bxwidth}}

37: \newcommand\prm[2]{$ $\vskip 2.4 truein

38: \centerline{\epsfig{file=#1,width=\bxwidth} }\vskip 0.5truein

39: \centerline{{\bf Fig.} #2}}

40: \newcommand\prk[2]{$ $\vskip 2.4 truein\vskip -2 truein

41: \centerline {\epsfig{file=#1,width=\bxwidth}} \vskip 0.5truein

42: \centerline{{\bf Fig.} #2}}

43: %

44: \begin{document}

45: \maketitle

46: \begin{abstract}

47: For forty years researchers on ferroelectric switching have used the 

48: Kay-Dunn theory to model the thickness-dependence of the coercive 

49: field; it works surprisingly well, despite the fact that it is 

50: based upon homogeneous nucleation and a small-field expansion, 

51: neither of which is realized in thin films. Here we 

52: demonstrate that this result can be obtained from a more general 

53: Kolmogorov-Avrami model of (inhomogeneous) nucleation and growth.  By 

54: including a correction to the switching field across the 

55: dielectric that includes Thomas-Fermi screening in the metal 

56: electrode, we show that our theory quantitatively describes the coercive 

57: fields versus thickness in several different families of 

58: ferroelectric (lead zirconate-titanate [PZT], potassium nitrate, and 

59: polyvinylidenefluoride [PVDF]) over a wide range of 

60: thickness (5 decades).  This agreement is particularly satisfying

61: in the case of PVDF, as it indicates that the switching kinetics

62: are domain-wall limited down to 1 nanometer and thus

63: require no new effects.

64: \end{abstract}

65: \vskip 0.2truein

66: \pacs{77.55.+f, 77.80.Fm, 68.15.+e, 85.50.+k}

67: 

68: 

69: 

70: \newpage

71: 

72: Demand for integrated microelectronics with increased densities and decreasing

73: voltage standards requires detailed characterization of the coercive field,

74: $E_c$, for the design and fabrication of

75: competitive, reliable ferroelectric

76: devices.  More specifically, recent measurements on nanometer 

77: ferroelectric films\cite{Bune98}

78: have revealed a weakness in our current understanding 

79: of the thickness-dependence

80: of $E_c$.  

81: Motivated by these experiments, in this Letter we present a new

82: treatment of the coercive field as a function of thickness, $E_c(d)$, that incorporates

83: both inhomogenous nucleation in a finite-size film and field penetration in

84: the electrodes.  The $E_c(d)$ that emerges

85: from our approach is in good

86: agreement with observed behavior 

87: for films that range from

88: 100 microns to 1 nanometer.  Furthermore our approach indicates 

89: that the minimum

90: film thickness, determined by the magnitude of the 

91: depolarization field,\cite{Batra72} can be tuned by varying

92: the spontaneous polarization of the ferroelectric and the screening length of

93: the electrodes.  Other implications of our results for the processing windows

94: of next-generation FeRAMS and DRAMS are also discussed.

95: 

96: For the last forty years the Kay-Dunn scaling law\cite{Kay62} has been

97: successfully

98: used to 

99: describe the thickness-dependence of the coercive field in 

100: ferroelectric films ranging from 100 microns to 200 nanometers.\cite{Scott00} 

101: However the Kay-Dunn model is 

102: based on assumptions, particularly that of homogenous nucleation,

103: which are inappropriate for thin films.

104: As discussed in the original

105: Kay-Dunn paper,\cite{Kay62} the calculated energy barrier associated with such 

106: nucleation is several orders of magnitude 

107: larger than any experimentally determined

108: value; furthermore the presence of long-range elastic interactions

109: increases

110: this barrier still further.\cite{Littlewood86}  Imaging measurements confirm the

111: site-specific nature

112: of nucleation in thin ferroelectric films, indicating that new domains nucleate

113: at electrodes and at twin boundaries.\cite{Lajzerowicz81,Shur91,Houchmandzadeh92,Ganpule00}  

114: The good agreement between measured values and the Kay-Dunn treatment of

115: $E_c(d)$ is thus quite

116: fortuitous.

117: 

118: Here we show that the Kay-Dunn scaling result can be recovered from an adapted

119: Kolmogorov-Avrami model\cite{Ishibashi95} of inhomogeneous

120: nucleation 

121: in a confined geometry.

122: In this approach

123: the transformation of a sample from  paraelectric to ferroelectric

124: is treated by considering the nucleation and growth of a 

125: single domain.  The coercive field 

126: is determined by the condition that

127: \begin{equation}

128: P(E,f) \left \vert_{E = E_c} = \frac{1}{2} \right.

129: \end{equation}

130: where $P(E,f)$ is the untransformed  sample fraction as a function of

131: applied field $E$ and frequency $f$.  For

132: inhomogeneous nucleation 

133: $P(E,f) \sim \exp -N(E) (v(E) \frac{2\pi}{f})^{D}$ 

134: where

135: $N(E)$ and $v(E)$ are the field-dependent 

136: number of nucleation sites per unit

137: area and domain growth velocity respectively and $D$ is

138: the dimension of the nucleus.

139: The factor $(v(E) \frac{2\pi}{f})^{D}$ enters the formula

140: for $P(E,f)$ as the characteristic volume of a growing nucleus

141: over the cycle time of the electric field.  However if the nuclei

142: elongate through the thickness of the film before colliding

143: with another domain, this factor should be replaced

144: by $(\kappa d)$ where $\kappa$ is the (small) transverse

145: area of the needle-shaped domain and $d$ is the film

146: thickness.  Such a morphology is supported by experimental

147: observation.\cite{Fatuzzo67}

148: Then 

149: we obtain

150: $P(E) \sim \exp -  N(E)  (\kappa d)$ where $N \sim E^\alpha$.

151: $\alpha$ is known to equal $3/2$ over a fairly

152: broad range of fields both experimentally\cite{Merz54}

153: and theoretically\cite{Stadler58} from switching kinetics; this then

154: leads to the desired result,

155: $E_c \sim d^{-2/3}$.

156: Here $0 < \kappa < 1$

157: indicates the size of the nucleating grain as a fraction of the

158: film thickness. We note that the scaling of $E_c$ with thickness follows directly from

159: the field-dependence of the number of nucleation sites per unit

160: area, $N(E)$; this result could be checked independently by

161: tuning

162: $N(E)$ and observing the predicted change in $E_c(d)$.

163: 

164: Recently it has been possible to measure the coercive field of

165: ferroelectric PVDF films with thicknesses less than 100 nanometers; 

166: on these length-scales significant deviation from the scaling

167: (cf. Figure 1) described

168: above is observed.\cite{Bune98}  In retrospect,

169: there were already hints of this new behavior in  

170: $E_c(d)$ for films between 100 and 200 

171: nanometers,\cite{Nagarajan99}

172: but the recent measurements down to 1 nanometer

173: are quite conclusive.\cite{Bune98}

174: The scaling treatment of $E_(d)$ assumes

175: perfect conducting plates, and here we have modified it

176: to include field-penetration in the electrodes.

177: We show that the latter effect becomes important

178: for small film thicknessses and is responsible for the observed

179: deviation from the expected scaling behavior (cf. Figure 1).  Furthermore we

180: demonstrate that, using these ideas, we can include a correction to the

181: measured coercive fields and recover the scaling result for films

182: ranging from 100 microns down to 1 nanometer (cf. Figure 2).

183: 

184: In an idealized ferroelectric capacitor, the plates are assumed to be

185: perfect conductors and charge resides on a plane

186: of negligible thickness at the electrode-ferroelectric interface,

187: compensating for the spontaneous polarization in the

188: ferroelectric

189: film.  Realistically this charge

190: distributed over a small but finite

191: length-scale in the metal, $\lambda$.  

192: The resulting electric field leads to an associated voltage 

193: drop in the metal electrodes.  

194: A compensating depolarization

195: potential, $\Phi_{dp}$,

196: must exist across the film 

197: to ensure that it is an equipotential 

198: in the absence of an externally applied voltage.

199: The associated field, $E_{dp} = \frac{\Phi_{dp}}{d}$,

200: assists in the switching

201: process so that the measured coercive field, $E_c^{meas}$, is

202: \begin{equation}

203: E_c^{meas} = E_c^{i} -  E_{dp}

204: \label{ec}

205: \end{equation}

206: where $E_c^{i}$ refers to a perfect capacitor with

207: negligible screening and is described by the Kay-Dunn scaling.

208: 

209: In order to determine $\Phi_{dp}$,

210: and its associated field,

211: we assume that the charge density takes the form

212: \begin{equation}

213: \rho(x) = 

214: \frac{Q}{\lambda} e^{ \left( - \frac{x}{\lambda}\right)}

215: \end{equation} 

216: and, 

217: furthermore we

218: assume symmetric capacitor plates; then, using the fact

219: that the potential in the electrodes is related to the induced

220: charge density,  we find 

221: \begin{equation}

222: \left \vert  \Phi_{dp} \right\vert = \left \vert \left(\frac{4

223: \pi}{\epsilon_m}\right) \frac{ 2\lambda Qd}{A}\right \vert

224: \end{equation}

225: where $\epsilon_m$ refers to the metallic dielectric constant,

226: and $Q = \frac{P_s A}{d}$ in a ferroelectric where $P_s$ is the

227: spontaneous polarization.

228: Here we estimate $\lambda$ using

229: the assumption of a slowly varying potential,\cite{Kittel76}

230: resulting in $\lambda^2 = \epsilon_m \lambda_{TF}^2$ where $\lambda_{TF}$

231: is the Thomas-Fermi screening length.   

232: Experimentally $P_s$ is observed to be independent of film

233: thickness in the nanometer PVDF samples studied\cite{Bune98},

234: indicating that $\Phi_{dp}$ is a constant; note that

235: in this case the voltage drop across the electrodes, $V_{electrodes}$,

236: compensates for the depolarization potential $\Phi_{dp}$.

237: However $E_{dp}$

238: scales inversely with $d$, so that it assumes increasing

239: importance in the expression for $E_c^{meas}$, (\ref{ec}),

240: with decreasing film thickness.  

241: 

242: Field-penetration of the electrodes has been studied in the

243: context of very thin ($\sim 30$ angstroms) film capacitors

244: by several authors, \cite{Mead61,Ku64,Simmons65,Black99}

245: and here we argue that it is relevant for thicker

246: films whose dielectric constant, $\epsilon_f$ is

247: large.

248: The effective capacitance of the metal-film-metal sandwich 

249: can be determined by contributions from two capacitors

250: in series\cite{Ku64}

251: \begin{equation}

252: \frac{1}{C} = \frac{1}{C_{film}} + \frac{1}{C_{electrodes}}

253: \end{equation}

254: where $C_{film}$ and $C_{electrodes}$ are the film

255: and the electrode capacitances respectively;

256: the first term is geometric and the second provides

257: an upper bound to the capacitance with decreasing film thickness.

258: For a dielectric film, the voltage drop is

259: $V_{film} =  \frac{4\pi}{\epsilon_{f}} \left( \frac{Qd}{A}\right)d$,

260: so that the fraction of the total potential difference, $V = V_{film}

261: + V_{electrode}$, across

262: the electrodes is

263: \begin{equation}

264: \frac{V_{electrodes}}{V} 

265: = \frac{2\lambda \left( \frac{\epsilon_f}{\epsilon_m}\right)}

266: { d + 2\lambda 

267: \left(\frac{\epsilon_f}{\epsilon_m}\right)}

268: \end{equation}

269: which can be substantial for large $\epsilon_f$ and small $d$.

270: Therefore we note that field-penetration effects will be important

271: for both (non-switching) high-dielectric DRAM capacitors and in 

272: switching FeRAMS.

273: 

274: Let us now turn to a more quantitative analysis.

275: In Figure 1, we display the measured coercive field data\cite{Bune98} for

276: PVDF and the Kay-Dunn scaling prediction.  We note that there is

277: deviation from the scaling behavior at approximately 100 nanometers.

278: Next, using (\ref{ec}), we plot the coercive field including

279: the effect of field-penetration.  We have used the value

280: of $\epsilon_f = 14$

281: for PVDF from experiment,\cite{Furukawa81}

282: and used  

283: $\epsilon_m = 1$ and $\lambda_{TF} = 0.45$ angstroms for the aluminum

284: electrodes.\cite{Kittel76}

285: For thicknesses greater than 100 nm, the modified and the scaling

286: curves are identical, indicating that the depolarization contribution

287: is not important in this regime.  However

288: the modified $E_c$ continues to display a good fit to the data for

289: nanometer

290: thicknesses, particularly as it

291: deviates from the scaling curve.   We note that the observed rollover

292: in $E_c(d)$ will correspond to the thickness where the depolarization

293: and the idealized coercive fields are comparable; here we expect a

294: polarization

295: instability\cite{Batra72}

296: that will determine the minimum allowed film thickness, $d_{min}$.

297: We note that our results suggest that this instability

298: can be tuned by varying the spontaneous polarization of

299: the ferroelectric and the screening length of the electrodes.

300: This is consistent with the results of Ghosez and Rabe\cite{Ghosez00}

301: who found that $d_{min}$ for a ferroelectric

302: film with perfect electrodes was less than that found with

303: semiconducting ones.\cite{Batra72}

304: 

305: Alternatively we can use (\ref{ec}) to add depolarization corrections

306: to the measured coercive field; the resulting idealized $E_c^i$

307: displays Kay-Dunn scaling for five decades in thicknesses ranging

308: from 100 microns to 1 nanometers (cf. Figure 2).  We present the

309: modified data for three different materials in the main part of

310: the figure; since $E_c = C \times d^{-2/3}$ where $C$

311: is material-specific, we expect universal behavior for 

312: $\log E_c - \log C$.  Indeed this is the case, as indicated

313: in the inset of Figure 2.  It is unusual to see universal scaling

314: over such a large number of decades.

315: The agreement between theory and experiment is particularly

316: satisfying for the case of PVDF, since it demonstrates

317: that the switching kinetics of films on nanometer scales are

318: domain-wall

319: limited and are qualitatively similar to those of their 100 micron

320: counterparts.  No unusual effects special to ultrathin films

321: are required; this has been a topic of discussion

322: in the literature.\cite{Fridkin00}

323: 

324: Our results have several implications for the design and fabrication

325: of ferroelectric devices.  Ideally FeRAMS should be designed at

326: film thicknesses where the measured coercive field is less than

327: the scaling value, but before it is equal to the depolarization

328: field; this gives a processing window for PVDF, the polymeric ferroelectric

329: used for this study, for 10 - 100 nm.  Furthermore the minimum

330: film thickness, associated with a polarization instability,

331: can be tuned by varying $P_s$ and $\lambda$.  Oxide electrodes,

332: known for reducing ferroelectric fatigue\cite{Scott00}, would

333: {\sl not} be optimal for very thin ($< 10$ nm) devices due to

334: large field-penetration in the capacitor plates.  Indeed the

335: optimization of very thin devices should emphasize the

336: screening properties of the electrodes, favoring

337: Pt or Au over $SrTiO_3$.  For the DRAM market, the primary objective

338: is maximizing total capacitance; our results indicate that it is

339: limited by the electrodes.  Thus use of larger dielectric materials

340: and thinner films will {\sl not} improve it significantly.  Instead

341: one should optimize the electrical resistivity of the ferroelectric

342: film and the screening properties of the electrodes.

343: 

344: In summary, in this Letter we have brought together the ideas of Kay and

345: Dunn,\cite{Kay62} Batra and Silverman,\cite{Batra72}

346: and Ghosez and Rabe\cite{Ghosez00} to develop a model for

347: the coercive field as a function of thickness.  For thin films,

348: we have shown the importance of field-penetration in the electrodes;

349: the resulting depolarization field is a function of the spontaneous

350: polarization of the ferroelectric and the Thomas-Fermi screening

351: length of the metal electrode.  Naturally when $E_{dp}$ becomes equal

352: to the idealized coercive field the polarization is unstable;\cite{Batra72}

353: however when it is comparable but less than $E_c^i$, it contributes

354: to the switching.  In Figure 1, with no adjustable parameters, we

355: display good agreement between our modified coercive field and

356: that measured in PVDF, particularly for thicknesses $<$ 100 nm

357: where there is deviation from scaling behavior.  In Figure 2

358: the measured data are corrected to accommodate field-penetration

359: in the electrodes; the result obeys Kay-Dunn scaling for

360: five decades in length-scale.  Our results indicate that the choice

361: of electrode is very important for optimizing performance for

362: very thin film devices.  Given increasing clock speeds, an

363: improved understanding of the coercive fields at finite

364: frequencies would also be of great interest for the development 

365: of future ferroelectric devices.

366: 

367: We are grateful to S. Bhattacharya, D.J. Jung, K. Rabe, A. Ruediger

368: and M.M.J. Treacy for discussions.

369: 

370: \begin{references}

371: 

372: 

373: \bibitem{Bune98}

374: A.V. Bune et al., {\sl Nature} {\bf 391}, 874 (1998);

375: S. Ducharme et al., {\sl Phys. Rev. Lett.} {\bf 84}, 175 (2000).

376: 

377: 

378: \bibitem{Batra72}

379: I.P. Batra and B.D. Silverman, {\sl Sol. State Comm.} {\bf 11}, 291 (1972). 

380: 

381: \bibitem{Kay62}

382: H.F. Kay and J.W. Dunn, {\sl Phil. Mag.} {\bf 7}, 2027 (1962).

383: 

384: \bibitem{Scott00}

385: J.F. Scott, {\sl Ferroelectric Memories} (Springer-Verlag, Berlin,

386: 2000).

387: 

388: \bibitem{Littlewood86}

389: P.B. Littlewood and P. Chandra, {\sl Phys. Rev. Lett.} {\bf 57}, 2415

390: (1986);

391: P. Chandra, {\sl Phys. Rev. A} {\bf 39}, 3672 (1989).

392: 

393: \bibitem{Lajzerowicz81}

394: J. Lajzerowicz, {\sl Ferroelectrics} {\bf 35}, 219 (1981);

395: 

396: \bibitem{Shur91}

397: V. Shur et al, {\sl JETP Lett.} {\bf 53}, 615 (1991).

398: 

399: \bibitem{Houchmandzadeh92}

400: B. Houcmandzadeh, J. Lajzerowicz and E.K.H. Salje, {\sl

401: J. Phys. Cond. Mat.} {\bf 4}, 9779 (1992).

402: 

403: \bibitem{Ganpule00}

404: C.S. Ganpule et al., {\sl Appl. Phys. Lett.} {\bf 77}, 292 (2000);

405: C.S. Ganpule et al., {\sl Phys. Rev. B} {\bf 65}, 14101, (2001).

406: 

407: \bibitem{Ishibashi95}

408: Y. Ishibashi and H. Orihara, {\sl Integ. Ferroelec.} {\bf 9}, 57

409: (1995). 

410: 

411: \bibitem{Fatuzzo67}

412: E. Fatuzzo and W.J. Merz, {\sl Ferroelectricity} (North-Holland,

413: Amsterdam, 1967).

414: 

415: \bibitem{Merz54}

416: W. Merz, {\sl J. Appl. Phys.} {\bf 27}, 938 (1954).

417: 

418: \bibitem{Stadler58}

419: E.H.L. Stadler and P.J. Zachminidis, {\sl J. Appl. Phys.} {\bf 29},

420: 1285 (1958); {\sl ibid} {\bf 33}, 3487 (1962);

421: {\sl ibid} {\bf 34}, 3255 (1963). 

422: 

423: \bibitem{Nagarajan99}

424: V. Nagarajan et al., {\sl J. Appl. Phys.} {\bf 86}, 595 (1999).

425: 

426: \bibitem{Mead61}

427: C.A. Mead, {\sl Phys. Rev. Lett.} {\bf 6}, 545 (1961).

428: 

429: \bibitem{Ku64}

430: H.Y. Ku and F.G. Ullman, {\sl J. Appl. Phys.} {\bf 35}, 265 (1964).

431: 

432: \bibitem{Simmons65}

433: J.G. Simmons, {\sl Appl. Phys. Lett} {\bf 6} 54 (1965);

434: J.G. Simmons, {\sl Brit. J. Appl. Phys.} {\bf 18}, 269 (1967).

435: 

436: \bibitem{Black99}

437: C.T. Black and J.J. Welser, {\sl IEEE Trans. on Electron Devices}

438: {\bf 46}, 776 (1999).

439: 

440: \bibitem{Furukawa81}

441: T. Furukawa and G.E. Johnson, {\sl Ferroelectrics} {\bf 39}, 1241

442: (1981); {\sl Appl. Phys. Lett.} {\bf 38}, 1027 (1981).

443: 

444: \bibitem{Kittel76}

445: C. Kittel, {\sl Introduction to Solid State Physics}, (John Wiley, New

446: York, 1976).

447: 

448: 

449: \bibitem{Ghosez00}

450: Ph. Ghosez and K.M. Rabe, {\sl Appl. Phys. Lett.} {\bf 76}, 2767 (2000).

451: 

452: \bibitem{Fridkin00}

453: V. M. Fridkin et al., {\sl Phys. Rev. Lett.}{ bf 84}, 175 (2000),

454: A. M. Bratkowsky and A. P. Levanyuk, {\sl Phys. Rev. Lett.}{\bf 87},

455: 19701 (2001). 

456: 

457: 

458: 

459: 

460: 

461: 

462: 

463: 

464: 

465: 

466: 

467: 

468: \end{references}

469: 

470: \newpage

471: 

472: \noindent{\bf Figure Captions}

473: 

474: \renewcommand{\labelenumi}{{\bf Fig.} \theenumi .}

475: \begin{enumerate}

476: 

477: \item 

478: Measured coercive field data$^1$ for PVDF,

479: the Kay-Dunn scaling and the current theory vs. thickness; as

480: explained in the text, the latter incorporates field-penetration effects via a

481: depolarization contribution to the field as a function of thickness.

482: \label{1}

483: 

484: 

485: \item 

486: {\bf (Main Figure)} The corrected log coercive field vs. log thickness

487: where the depolarization contribution has been subtracted from the

488: measured data {\bf (Inset)}

489: Normalized corrected 

490: log coercive field vs. thickness which indicates Kay-Dunn scaling

491: over five decades of lengths; here universal behavior is shown

492: for $\log E_c - \log C$ where $C$ is defined by the expression

493: $E_c = C\times d^{-2/3}$ and is material- and sample-specific.  

494: \label{2}

495: 

496: 

497: \end{enumerate}

498: 

499: \newpage

500: 

501: 

502: %\begin{figure}

503: 

504: %\bxwidth=0.6\textwidth

505: \prk{fig1.eps}{1}

506: 

507: 

508: %\caption{}

509: 

510: %\begin{figure}

511: %\includegraphics[height=5cm]{fig1.eps}

512: %\end{figure}

513: 

514: %\newpage

515: %\bxwidth=0.9\textwidth

516: \prm{fig2.eps}{2}

517: 

518: 

519: 

520: 

521: 

522: \end{document}

523: 

524: 

525: 

526: 

527: 

528: 

529: 

530: 

531: 

532: 

533: 

534: 

535: 

536: 

537: 

538: 

539: