1: \documentclass[12pt]{article}

2: 

3: \usepackage{graphicx}

4: \usepackage{amssymb,latexsym}

5: \usepackage{amsmath}

6: \usepackage[dvips]{epsfig}

7: \setlength{\hoffset}{-0.5in}

8: \setlength{\voffset}{-0.5in}

9: \setlength{\textheight}{8.0in}

10: \setlength{\textwidth}{6.5in}

11: 

12: \begin{document}

13: 

14: \title{\bf{Surface effects on the statistics of the local density of states in metallic

15: nanoparticles: manifestation on the NMR spectra.}}

16: \author{Jos\'{e} A. Gasc\'{o}n$^1$  \\

17: Department of Chemistry, Yale University\\

18: P.O. Box  208107, New Haven, CT 06520-8107, USA\\

19: \\

20: Horacio M. Pastawski\\

21: Facultad de Matem\'{a}tica Astronom\'{\i}a y F\'{\i}sica, Universidad\\

22: Nacional de C\'{o}rdoba, Ciudad Universitaria, 5000 C\'{o}rdoba, Argentina}

23: 

24: \maketitle

25: 

26: \footnotetext[1]{e-mail: jose.gascon@yale.edu (author to whom correspondence should

27: be addressed)}

28: 

29: 

30: \begin{abstract}

31: 

32: \setlength{\baselineskip}{0.3in}

33: 

34: In metallic nanoparticles, shifts in the ionization energy of surface atoms with 

35: respect to bulk atoms can lead to surface bands. Within a simple Tight Binding model 

36: we find that the projection of the electronic density of states on these sites 

37: presents two overlapping structures. One of them is characterized by the level 

38: spacing coming from bulk states and the other arises from the surface states. In very 

39: small particles, this effect contributes to an over-broadening of the NMR absorption 

40: spectra, determined by the Knight shift distribution of magnetic nuclei. We 

41: compare our calculated Knight shifts with experiments on aluminum nanoparticles, 

42: and show that the deviation of the scaling law as a function of temperature and 

43: particle size can be explained in terms of surface states.

44:   

45: \end{abstract}

46: 

47: \setlength{\baselineskip}{0.3in}

48: 

49: 

50: \section{Introduction}

51: 

52: Several NMR experiments on metallic nanoparticles have shown \cite{Halp-Brom,Slichter,V.D.Klink} 

53: quantum size effects: the absorption spectra becomes quite broad and asymmetric when either 

54: temperature or particle size diminishes. Studies of relaxation time \cite{V.D.Klink} 

55: $T_{1\text{ \ }}$ indicate that the spectra is inhomogeneous, implying phenomena of a local

56: nature. This inhomogeneity has been interpreted \cite{Efetov} considering a strictly local 

57: description for the electron polarization in the presence of a magnetic field $H$. Given the 

58: Bohr magneton, $\mu _{\text{B}}$, and $N(\varepsilon _{{\rm F}},${\bf r}$)$, the local density 

59: of {\it s} states at the Fermi level $\varepsilon _{{\rm F}}$, the {\it local} Pauli

60: susceptibility is written as: 

61: %

62: \begin{equation}

63: \chi _{\text{p}}({\bf r)}=\mu _{\text{B}}^{2}N(\varepsilon _{{\rm F}},{\bf r} 

64: ).  \label{Eq. 1}

65: \end{equation}

66: %

67: This equation expresses that, due to the finite size effects and

68: interference phenomena proper of the mesoscopic systems, the electronic spin

69: polarization is {\it inhomogeneous}. The hyperfine coupling between nuclear

70: spins and electronic spins produces different shifts in the resonance

71: frequency for {\it each} nucleus in the nanoparticle:

72: %

73: \begin{equation}

74: \Delta \omega ({\bf r)\propto }\chi _{\text{p}}({\bf r)}H.  \label{Eq. 2}

75: \end{equation}

76: %

77: Therefore, an inhomogeneously broadened NMR\ absorption line is the key to

78: indirectly measure the fluctuations of the local density of states (LDOS).

79: 

80: In previous works \cite{Pastaw-Gasc} we have calculated the NMR line shape

81: in metallic nanoparticles within a Tight Binding model with $M$ orbitals

82: accounting for the fluctuations of the LDOS. A remarkable outcome,

83: consistent with many experimental results \cite{Halp-Brom,Volokitin}, is a

84: universal scaling behavior of the line position and shape with respect to

85: the variation of the thermal energy $k_{{\rm B}}T$ and the mean spacing level 

86: $\Delta \cong 1/\left[ M\times N_{0}(\varepsilon _{{\rm F}})\right] $. This

87: results in a universal scaling law for the Knight shifts and the Pauli susceptibility 

88: \cite{Volokitin} whose relevant parameter is 

89: %

90: \begin{equation}

91:  \alpha = k_{{\rm B}}T/\Delta. 

92: \label{alpha}

93: \end{equation}

94: %

95: However, in very small crystalline particles, the experimental line shape

96: presents an anomalous over-broadening at intermediate temperatures \cite{Brom}

97: . In addition, a.c. conductivity measurements show that tunneling among

98: particles plays a relevant role in the electronic properties \cite{Brom2}. This suggests

99: the need to explore for additional sources of fluctuations on the statistics of energy levels, 

100: particularly the effects of surface states. It is known that self

101: consistent calculations of metal surfaces \cite{Guevara,Ganduglia} give

102: different ionization energies for surface and bulk sites. This fact is

103: expected both in metallic particles with clean surfaces and as well as those

104: with chemisorbed atoms. In our previous calculation we made no attempt to

105: consider surface states. However the natural connectivity of the surface

106: orbitals caused slight departures from the universal scaling law \cite{Pastaw-Gasc}.

107: 

108: In the present work we use a Tight Binding Hamiltonian with an {\it ad hoc} 

109: shift in the surface site energies \cite{Kalkstein,Ganduglia}. This

110: produces surface states, and as a consequence, large fluctuations on

111: the LDOS distribution. According to the Eqs. \ref{Eq. 1} and \ref{Eq. 2},

112: this leads to a modified distribution of Knight shifts detected as a

113: wider NMR absorption lines. We compare our calculations of the Knight shift

114: as a function of the scaling parameter with experimental values on

115: aluminum nanoparticles and show that the observed deviation of the scaling

116: law is a manifestation of surface effects.

117: 

118: \section{Model Hamiltonian}

119: 

120: A metal particle with $M$ atoms can be modeled with the Hamiltonian

121: 

122: \begin{equation}

123: {\cal H}=\sum_{i=1}^{M}E_{i}c_{i}^{+}c_{i}+\sum_{j>i}^{M} 

124: \sum_{i=1}^{M}(V_{ji}c_{j}^{+}c_{i}+V_{ij}c_{i}^{+}c_{j}),  \label{H}

125: \end{equation}

126: where $E_{i}$ is the energy of an $s$ state centered at site $i$ of a cubic

127: lattice.{\it \ }$V_{ij}\equiv V$ is the kinetic energy involved in hopping

128: between nearest neighbors sites $i$ and $j$. To represent shape and crystal

129: inhomogeneities, the sites energies are taken in the range $-W/2$ and $W/2$

130: (Anderson's disorder)$.$ We define \emph{surfaces sites} as the sites which

131: being at the surface of the cube, have an additional energy shift $U$.

132: The rest of the sites are \emph{bulk sites}, even though they may belong to the

133: faces of the cube with un-shifted site energies. This allows to model different surface to bulk

134: ratios maintaining $M$ constant. Tunneling among particles produces an

135: inhomogeneous broadening \cite{GLBE} of the atomic energy levels ($ 

136: E_{i}\rightarrow E_{i}-$i$\Gamma _{i}$), taking its higher values $\Gamma

137: _{s}$ at particular sites (\emph{contacts}) at the surface. Assuming that $ 

138: \Gamma _{i}\ll V,$ and neglecting localization effects we can assume $\Gamma

139: _{i}\equiv \eta _{0}\approx \Gamma _{s}/M$ for all sites in the

140: nanoparticle. As shown in the Ref. 4 finite temperature effects are

141: included by an additional broadening on the energy levels, $\eta =\eta

142: _{0}+k_{{\rm B}}T$.  Properties of the single particle excitation spectrum are contained in the retarded (advanced) Green's function

143: %

144: \begin{equation}

145: G_{i,j}^{R(A)}(\varepsilon _{}^{})=\sum_k\frac{a_k(\mathbf{r_i})a_k^{*}(\mathbf{r_j})}{\varepsilon + [(-){\rm i}\eta -E_k]},  \label{GR}

146: \end{equation}

147: %

148: where $\psi _k({\bf r})=\sum_ia_k(\mathbf{r_i})\varphi _i({\bf r})$ and $E_k$ are the

149: exact eigenfunctions (molecular orbitals) and eigenenergies for the isolated

150: particle, respectively. $\eta $ is a natural broadening of the electronic states and the $%

151: (-)$ sign corresponds to the retarded Green's function. The local density

152: of states per atom (LDOS) at the $i$-th site is evaluated as:

153: %

154: \begin{equation}

155: N(\varepsilon,\mathbf{r_i} )=-(2\pi {\rm i})^{-1}[G_{i,i}^R(\varepsilon

156: )-G_{i,i}^A(\varepsilon )],  \label{Ni}

157: \end{equation}

158: %

159: from which the relevant contribution to the density of states per unit

160: volume at the $i$-th nucleus, $N(\varepsilon ,{\bf r}_i),\,$is obtained. 

161: The evaluation of the local Green's function in Eq.~\ref{GR} is obtained via the 

162: Matrix Continued Fraction method \cite{Pastaw}. Its basic idea is to exploit the short range

163: interactions in the Hamiltonian (\ref{H}) by indexing states in a way that

164: subspaces representing layers interact through nearest neighbor subspaces.

165: In matrix form:

166: 

167: \begin{equation}

168: {\cal H=}\left[

169: \begin{array}{ccccc}

170: \ddots & \ddots & {\bf 0} & {\bf 0} & {\bf 0} \\

171: \ddots & {\bf E}_{n-1,n-1} & {\bf V}_{n-1,n} & {\bf 0} & {\bf 0} \\

172: {\bf 0} & {\bf V}_{n,n-1} & {\bf E}_{n,n} & {\bf V}_{n,n+1} & {\bf 0} \\

173: {\bf 0} & {\bf 0} & {\bf V}_{n+1,n} & {\bf E}_{n+1,n+1} & \ddots \\

174: {\bf 0} & {\bf 0} & {\bf 0} & \ddots & \ddots

175: \end{array}

176: \right] ,  \label{HM}

177: \end{equation}

178: 

179: \noindent where ${\bf 0}$'s are null matrices, ${\bf E}$'s in the diagonal represent

180: intra-layer interactions while the only non-zero off-diagonal matrices ${\bf V}_{n,n\pm 1}$ 

181: connect nearest neighbor layers. Detailed structure of

182: the sub-matrices depends on the lattice, for the cubic structure ${\bf V}%

183: _{n,n\pm 1}=V{\bf 1}$, with ${\bf 1}$ the identity matrix. The local

184: retarded Green's functions connecting sites $i$ and $j$ within the $n$-th

185: layers are arranged in a matrix

186: %

187: \begin{equation}

188: {\bf G}_{n,n}^R\left( \varepsilon \right) =\left[ \left( \varepsilon +{\rm i}%

189: \eta \right) {\bf 1}{\Bbb -}{\bf E}_{n,n}-{\bf \Sigma }_n^{R\,+}\left(

190: \varepsilon \right) -{\bf \Sigma }_n^{R\,-}\left( \varepsilon \right)

191: \right] ^{-1}

192: \end{equation}

193: %

194: where the matrix self energies ${\bf \Sigma }_n^{R\,+}$ and ${\bf \Sigma }%

195: _n^{R\,-}$ are calculated in terms of Matrix Continued Fractions (MCF)

196: defined through the recurrence relations:

197: \begin{equation}

198: {\bf \Sigma }_n^{R\,\,\pm }={\bf V}_{n,n\pm 1}\frac{{\bf 1}}{\left(

199: \varepsilon +{\rm i}\eta \right) {\bf 1}{\Bbb -}{\bf E}_{n\pm 1,n\pm 1}-{\bf %

200: \Sigma }_{n\pm 1}^{R\,\,\pm }}{\bf V}_{n\pm 1,n},

201: \end{equation}

202: which are calculated with the boundary conditions: ${\bf \Sigma }%

203: _{L}^{+}\equiv {\bf \Sigma }_1^{-}\equiv {\bf 0}$, where $L$ denotes the number of layers.

204: 

205: 

206: \textbf{Model parameters.} Parameters that model an $s$-band are \cite{sband}: 

207: $V\cong 0.9{\rm eV}$, which is

208: consistent with a bandwidth of $B=12V\cong 0.8$Ry $=11{\rm eV;}$ a shift $ 

209: U=2V$, consistent with a shift of $0.12$Ry estimated for metal clusters and

210: a Fermi energy of $\varepsilon _{{\rm F}}=3V+B/2$, which measured from the

211: band bottom gives a ratio $\varepsilon _{{\rm F}}/B=0.75$, close to typical

212: values of metals \cite{sband}. In this work we use $\eta =0.05V$, which assumes that

213: the main contribution to the level broadening comes from tunneling i.e. $ 

214: \eta _{0}>k_{{\rm B}}T$ which is consistent with the conductivity

215: measurements \cite{Brom2}.

216: 

217: The statistical distribution of the LDOS is calculated taking an ensemble of

218: ten Fermi energies in a range $\delta \varepsilon _{{\rm F}}=0.5V$ around $ 

219: \varepsilon _{{\rm F}}.$ For each energy ten disorder configurations are

220: considered. Each LDOS $N(\varepsilon _{{\rm F}},{\bf r})$ is normalized to

221: the bulk value $N_{0}(\varepsilon _{{\rm F}})$, which is evaluated at the

222: central site of a particle with $15\times 15\times 15$ orbitals. The

223: normalized LDOS is defined as $x=N(\varepsilon _{{\rm F}},{\bf r} 

224: )/N_{0}(\varepsilon _{{\rm F}}),$ and occurs with a probability $I(x)$.

225: According with Eqs. \ref{Eq. 1} and \ref{Eq. 2}, $I(x)$ is also the absorption at

226: normalized frequency shifts: $x=(\omega -\omega _{0})/(\omega _{K}-\omega

227: _{0}).\,$Here $\omega _{K}$ is the bulk metal frequency and $\omega _{0}$

228: the frequency of metal nuclei in dielectric materials (i.e. salt). 

229: Due to this correspondence, from now on, $I(x)$ will refer indistinctly to either

230: the distribution of the normalized LDOS at the Fermi energy or the Knight shift NMR spectrum.

231: Quantum size effects manifest as shifts of the NMR line maximum:

232: as the particle size or temperature decreases, the shift goes from $x=1$ (metal

233: bulk) to $x=0$ (salt) while the line broadens and turns asymmetrical 

234: \cite{Efetov,Pastaw-Gasc}. It is important to emphasize that the simple model used in this work

235: only attempts to provide physical insight into the quantum size effects as

236: manifested in the LDOS distribution, as well as a qualitative description of surface

237: effects.

238: 

239: \section{Results}

240: 

241: To model small surface to volume ratios proper of a big particle, the

242: surface sites (with mean site energy $U=2V$) were chosen on a single

243: face of a $7\times 7\times 7$ cube, i.e. $M_{{\rm surf}}=7\times 7=49$ are

244: surface sites{\it \ }and $M_{{\rm bulk}}=6\times 7\times 7=294$ are bulk

245: sites. In this case the ratio of level spacing is $\Delta _{{\rm surf.}}/\Delta _{{\rm bulk}}=4$. 

246: Results for this particle can be extrapolated

247: to an actual particle with $40\times 40\times 40$ orbitals having

248: roughly the same surface to bulk ratio. Figure 1 shows the Knight shift

249: spectrum obtained from the LDOS occurrence distribution at 

250: surface sites (thick line). 

251: %

252: \begin{figure}[ht]

253: \begin{center}

254: \includegraphics*[scale=1.0]{figure1.eps}

255: \caption{Knight shift spectrum of surface sites when $U=2V$ (thick line),

256: compared with the case with $U=0$ (thin line). $M=7\times 7\times 7,$

257: $M_{{\rm surf.}}=49$ (corresponding to one side of the cube) and

258: $M_{{\rm bulk}}=294.$ $W=1V$ and $\eta =0.05V.$}

259: \end{center}

260: \end{figure}

261: %

262: It shows that the shift in the site energies produces an over-broadening with

263: respect to the Knight shift distribution from the same sites for the homogeneous

264: configuration with $U=0$ (thin line). The over-broadening suggests the

265: existence of multiple fluctuation scales in the NMR spectrum. To clarify this, 

266: Figure 2(a) shows the projection of the density of states (DOS) at 

267: surface sites, $N_{\text{surf.}}(\varepsilon)$. It reveals two structures, one 

268: corresponding to surface sites with a typical spacing $\Delta _{{\rm surf.}}$ and 

269: another one with

270: typical spacing $\Delta _{{\rm bulk}}$ corresponding to bulk sites. Notice

271: that the structure of bulk states (Fig. 2.b), characterized by a mean level spacing of

272: $\Delta _{{\rm bulk}}$, enters into the surface DOS as a substructure. These bulk states

273: have a small weight on the surface band because $U$ provides a

274: barrier that prevents the mixing between states with the same kinetic energy

275: ($E_{k_{\Vert }}$) parallel to the surface. 

276: 

277: \begin{figure}[ht]

278: \begin{center}

279: \includegraphics*[scale=1.0]{figure2.eps}

280: \caption{DOS projection on a) surface sites{\it \ }with energies $U=2V$ and

281: b) {\it bulk sites }with energies $U=0.$ $M=7\times 7\times 7,$ $M_{{\rm 

282: surf.}}=49$ and $M_{{\rm bulk}}=294.$ $W=1V$ and $\eta =0.05V.$}

283: \end{center}

284: \end{figure}

285: 

286: The isolated surface and bulk

287: bands $N_{{\rm surf.}}^{0}(\varepsilon )$ and $N_{{\rm bulk}

288: }^{0}(\varepsilon ),$ provide a rationale to understand the structure of the

289: surface DOS, and consequently the multiple fluctuation scales in the NMR spectrum. 

290: Typically $N_{{\rm surf.}}(\varepsilon )\cong aN_{{\rm surf.} 

291: }^{0}(\varepsilon )+bN_{{\rm bulk}}^{0}(\varepsilon )$. Within a

292: perturbative calculation, $a\leq 1-(V^{2}/U)^{2}=3/4$ and $a+b=1$. 

293: The main peaks in $N_{{\rm surf.}}(\varepsilon )$ (Fig. 2.a) arise from those in 

294: $N_{{\rm surf.}}^{0}(\varepsilon ).$ They produce a wide range of high LDOS

295: values with low probability (tail in the $I_{{\rm surf.}}(x)$, Fig. 1). The

296: valleys of the global structure will produce lower density values with high

297: probability (peak in the $I_{{\rm surf.}}(x)$). In the same way, the

298: denser spectrum of the bulk band $N_{{\rm bulk}}^{0}(\varepsilon )$, which

299: contributes with small weight, provides a lower scale fluctuations to $N_{ 

300: {\rm surf.}}(\varepsilon )$.

301: % 

302: \begin{figure}[ht]

303: \begin{center}

304: \includegraphics*[scale=1.0,keepaspectratio=true]{figure3.eps}

305: \caption{Knight shift spectrum of bulk sites when $U=2V$ (thick line),

306: compared with the case with $U=0$ (thin line). $M=7\times 7\times 7,$ $M_{ 

307: {\rm surf.}}=49$ and $M_{{\rm bulk}}=294.$ $W=1V$ and $\eta =0.05V.$}

308: \end{center}

309: \end{figure}

310: %

311: Again, the peaks and valleys of $N_{{\rm bulk} 

312: }^{0}(\varepsilon )$ produce an asymmetric distribution $I_{{\rm bulk}}(x)$ 

313: with a maximum. Therefore, the superposition of these two LDOS will cause the

314: over-broadening in the distribution of Knight shifts for the surface sites roughly

315: represented by:

316: %

317: \begin{equation}

318: I_{{\rm surf.}}(x)\approx \frac{1}{ab}\int I_{{\rm surf.}}^{0}

319: (\frac{x-x^{\prime }}{a})I_{{\rm bulk}}^{0}(\frac{x^{\prime}}{b})dx^{\prime} 

320: \end{equation}

321: %

322: Even without attempting a fitting, this shows that the predictions of

323: numerical results are consistent with the spectral analysis.

324: 

325:  

326: \begin{figure}[ht]

327: \begin{center}

328: \includegraphics*[scale=1.0,keepaspectratio=true]{figure4.eps}

329: \caption{Knight shift spectrum of bulk sites setting $U=2V$ (thick line),

330: compared with the case with $U=0$ (thin line). $M=6\times 6\times 6,$ $M_{ 

331: {\rm surf.}}=152$ and (corresponding to the six sides of the cube) $M_{{\rm  

332: bulk}}=64.$ $W=1V$ and $\eta =0.05V.$}

333: \end{center}

334: \end{figure}

335:  

336: \begin{figure}[ht]

337: \begin{center}

338: \includegraphics*[scale=1.0,keepaspectratio=true]{figure5.eps}

339: \caption{Knight shift spectrum of surface sites{\it \ }when $U=2V$ (thick line),

340: compared with the case with $U=0$ (thin line). $M=6\times 6\times 6,$ $M_{ 

341: {\rm surf.}}=152$ and $M_{{\rm bulk}}=64.$ $W=1V$ and $\eta =0.05V.$}

342: \end{center}

343: \end{figure}

344:  

345: \begin{figure}[ht]

346: \begin{center}

347: \includegraphics*[scale=1.0]{figure6.eps}

348: \caption{Relative Knight shift $x_{\text{max.}}$ \emph{vs.} the scaling parameter $\alpha$.

349: The solid lines correspond to simulations of various particle sizes with

350: surface energy $U=2V$ for all faces of the cube.

351: Experimental values correspond to aluminum nanoparticles \cite{Kobayashi}. In both

352: simulations and experiments a deviation of the universal scaling law is observed.} 

353: \end{center}

354: \end{figure}

355: 

356: In order to evaluate the effect of surface states on bulk NMR signal of big

357: particles, Figure 3 shows the Knight shift spectrum (thick line) for the

358: bulk sites together with a case with unperturbed surfaces $(U=0)$

359: represented by the thin line. $I_{{\rm bulk}}(x)$ does not present much

360: over-broadening with respect to the homogeneous configuration$.$ The reason

361: for this can be understood by analyzing the DOS projection on bulk sites.

362: The relevant scale in $N_{\text{bulk}}(\varepsilon)$ (Fig. 2.b) is $\Delta _{{\rm bulk}}$.

363: Since surface states enter into the bulk with small weight, and since 

364: $\Delta _{{\rm surf.}}>\Delta _{{\rm bulk}}$ ($\Delta _{{\rm surf.}}=4\Delta _{{\rm bulk}}$), 

365: the relevant scale is dictated by spacing of bulk levels. Therefore, the surface band does 

366: not affect the statistics of the LDOS of bulk sites. 

367: 

368: The previous analysis suggests that if the surface to volume ratio is big,

369: as in very small particles where $\Delta _{{\rm bulk}}>\Delta _{{\rm surf.} 

370: }, $ the over-broadening should occur on the LDOS distribution of bulk

371: sites{\it . }The cluster configuration must be such that the surface sites 

372: {\it \ }are in higher proportion with respect to the bulk sites{\it . }This

373: situation can be obtained for a structure with $6\times 6\times 6$ orbitals,

374: taking all faces of the cube with shifted site energies $U=2V$ ($M_{{\rm surf. 

375: }}=152$ and $M_{{\rm bulk}}=64$). In this case $\Delta _{{\rm bulk} 

376: }=3.5\Delta _{{\rm surf.}}$. Figure 4 shows the Knight shift spectrum of 

377: bulk sites for this new configuration. The distribution (thick line) is

378: broader than the corresponding homogeneous case (thin line). The reason is

379: the same as the one discussed in connection with Fig. 1 provided that the

380: relation $\Delta _{{\rm surf.}}/\Delta _{{\rm bulk}}$ is now inverted.

381: However, the NMR spectrum of surface

382: sites (Fig. 5) exhibits a considerable over-broadening, unlike the case presented 

383: in Fig. 3. This is due to the fact that, because of the topology of surface sites, completely

384: surrounding bulk sites, there is a strong mixing between surface states and

385: interior states, which also produces two fluctuation scales on the LDOS distribution

386: at the surface.

387: 

388: We now discuss the surface effects on the scaling law. Fig. 6 shows the relative

389: Knight shift as a function of the scaling parameter $\alpha=\eta/\Delta$. The

390: solid lines correspond to our results for particles of various sizes

391: with surface energies on all faces of the cube. 

392: These curves correspond to particles of $7\times 7\times 7$,

393: $10\times 10\times 10$ and $12\times 12\times 12$ sites.  

394: Experimental results by Kobayashi \cite{Kobayashi} \emph{et al} on aluminum

395: nanoparticles are also shown. The qualitative dependency of the relative knight

396: shift with $\alpha$ agrees very well with our result. In addition, a deviation of

397: the scaling law is clearly observed in both, simulations and experiment. 

398: Large particles lie further to the right than small particles. This can be 

399: interpreted as follow: due to the existence

400: of surface states, the mean spacing level at the Fermi energy is no longer $\Delta$,

401: since this last spacing is defined according to the density of states of the bulk. 

402: Therefore, $\alpha$ is no longer the scaling parameter. If we define $\Delta'$ as

403: the mean spacing level at the Fermi energy in presence of surface states and

404: $\alpha'$ the corresponding scaling parameter, then, according to Eq.~\ref{alpha}, 

405: $\alpha=\alpha'\frac{\Delta'}{\Delta}$. 

406: Since bulk states and surface states become uncorrelated and consequently the 

407: repulsion of levels around the Fermi energy decreases, $\Delta'<\Delta$. Therefore,

408: the factor $\frac{\Delta'}{\Delta}$ counts for the deviation of the scaling law. 

409: It is smaller for smaller particles than for larger particles which is consistent

410: with the experimental and predicted results in Fig. 6. 

411: 

412: \section{Conclusions}

413: 

414: According to our results, we expect that the NMR spectra of either surface

415: or bulk nuclei in small metallic nanoparticles, manifest the existence of

416: surface states as an over-broadened line which could also be interpreted as a

417: strong \emph{disorder}. In fact, the deviation of the scaling law shown in  

418: the experiment on aluminum nanoparticles \cite{Kobayashi} is an indication of

419: different distributions of energy spacings coming from bulk and surface states.

420: Additionally to the over-broadening of crystalline Ni

421: particles \cite{Brom} which is consistent with this view, there are other

422: ways in which these effects might be observed and tested: the 

423: NMR technique SEDOR (spin echo double resonance) consists of

424: simultaneous irradiation of resonances of nuclei of two different species

425: which are coupled either by direct dipolar interaction or through the

426: conduction electrons. With this method the contribution to the resonance

427: spectra of the nuclei from the surface can be inferred. A very interesting

428: application \cite{SEDOR} is the test of Platinum nanoparticles using Carbon

429: as a local probe. The last is chemisorbed on the Pt surface as CO. There,

430: the presence of a second peak in the Pt SEDOR\ line might be interpreted as

431: the coexistence of surface and bulk states.

432: 

433: Also based on the CO chemisorption at the surface of a metallic

434: nanoparticle, it has been observed \cite{Becerra,Shore,Zilm} that Carbon

435: NMR presents a shift toward a metallic nature. Not only the Korringa

436: equation for the relaxation time $T_{1}$ is verified ($1/T_{1}\propto

437: (N(\varepsilon _{{\rm F}},{\bf r}))^{2}/k_{B}T$) but the line shape presents

438: the inhomogeneity proper of quantum size effects. That is, due to the

439: mixing between the conduction band of the metal and the CO molecular

440: orbitals, the frequency shift of the C nuclei contains information of the

441: LDOS on the surface nuclei . Eventually an NMR study of chemisorbed molecules

442: also would give evidence of anomalous fluctuations in surface LDOS.

443: 

444: \section{Acknowledgments}

445: This work was performed in part at LANAIS de RMN (UNC-CONICET) with financial

446: support from SeCyT-UNC, CONICOR, CONICET and FUNDACION\ ANTORCHAS. The

447: authors thank discussions with V. N. Prigodin, H. B. Brom, K. Efetov, A.

448: Bonivardi, M.V. Ganduglia-Pirovano and J. Guevara as well as correspondence

449: with J. J. van der Klink and L. R. Becerra.

450: 

451: \begin{thebibliography}{99}

452: \bibitem{Halp-Brom}  W. P Halperin, Rev. Mod. Phys. {\bf 58}, (1986) 533; H.

453: B. Brom, D. van der Putten and J. L. de Jongh, in {\it Physics and Chemistry

454: of Metal Clusters, }edited by L. J. de Jongh (Kluwer Academic Publishers,

455: Dordrecht, Netherlands, 1994).

456: 

457: \bibitem{Slichter}  H. E. Rhodes, P. K. Wang, H. T. Stokes, C. H. Slichter

458: and J. H. Sinfelt, Phys. Rev. B {\bf 26}, (1982) 3559.

459: 

460: \bibitem{V.D.Klink}  J. P. Bucher and J. J. van der Klink, Phys. Rev. B {\bf  

461: 38}, (1988) 11038.

462: 

463: \bibitem{Efetov}  K. B. Efetov and V. N. Prigodin, Phys. Rev. Lett. {\bf 70} 

464: , (1993) 1315.

465: 

466: \bibitem{Pastaw-Gasc}  H. M. Pastawski and J. A. Gasc\'{o}n, Phys. Rev. B 

467: {\bf 56}, (1997) 4887; H. M. Pastawski, J. A. Gasc\'{o}n and E. Medina,

468: Anal. Asoc. Qu\'{\i }m. Arg. {\bf 84}, (1996) 397.

469: 

470: \bibitem{Volokitin}  Y. Volokitin, J. Sinzig, L. J. de Jongh, G. Schmid, M.

471: N. Vargaftik and I. I. Moiseev, Nature {\bf 384}, (1996) 621.

472: 

473: \bibitem{Brom}  H. B. Brom, J. J. van der Klink, F. C. Fritschij, L. J. de

474: Jongh, R. Della Pergola and A. Ceriotti, Z. Phys. D {\bf 40}, (1976) 559.

475: 

476: \bibitem{Brom2}  F. C. Fritschij, H. B. Brom, L. J. de Jongh, ans G. Schmid, 

477: Phys. Rev. Lett. {\bf 82} (1999) 2167.

478: 

479: \bibitem{Guevara}  J. Guevara, E. Parisi, A. M. Llois, M. Weissmann,Phys.

480: Rev. B {\bf 55}, (1997) 13283; J. Guevara (private comm.)

481: 

482: \bibitem{Kalkstein}  D. Kalkstein and P. Soven, Surface Science 26, (1971) 85;

483: J. R. Schrieffer and P. Soven, Phys. Today, {\bf 29} (4) (1975) 24.

484: 

485: \bibitem{Ganduglia}  M. V. Ganduglia-Pirovano and M. H. Cohen, Phys. Rev. B 

486: {\bf 50}, (1994) 11142.

487: 

488: \bibitem{GLBE}  H. M. Pastawski, Phys. Rev. B {\bf 46}, (1992) 4053; J.L.

489: D'Amato and H. M. Pastawski, Phys. Rev. B. {\bf 41}, (1990) 7411.

490: 

491: \bibitem{Pastaw}  H. M. Pastawski, J. F. Weisz and S. Albornoz, Phys. Rev. B 

492: {\bf 28}, (1993) 6896.

493: 

494: \bibitem{sband} S. H. Yang, M. J. Mehl, and D. A. Papaconstantopoulos, Phys. Rev. B.

495: {\bf 57}, (1998) 2013, and references therein.

496: 

497: \bibitem{Kobayashi} S. Kobayashi, T. Takahashi, and W. Sasaki, J. Phys. Soc.

498: Japan {\bf 31}, (1971) 1442. 

499: 

500: \bibitem{SEDOR}  C. D. Makowka, C. P. Slichter and J. H. Sinfelt, Phys. Rev

501: B {\bf 31}, (1985) 5663.

502: 

503: \bibitem{Becerra}  L. R. Becerra, C. P. Slichter and J. H. Sinfelt, Phys.

504: Rev B {\bf 52}, (1995) 11457.

505: 

506: \bibitem{Shore}  S. E. Shore, J. P. Ansermet, C. P. Slichter and J. H.

507: Sinfelt, Phys. Rev Lett. {\bf 58}, (1987) 953.

508: 

509: \bibitem{Zilm}  K. W. Zilm, L. Bonneviot, G. L.Haller, O. H. Han and M.

510: Kermarec, J. Phys. Chem. {\bf 94}, (1990) 8495.

511: 

512: \end{thebibliography}

513: 

514: \end{document}

515: