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

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

3: \def\bull{\vrule height .9ex width .8ex depth -.1ex}

4: \newcommand{\ba}{\begin{eqnarray}}                                             

5: \newcommand{\ea}{\end{eqnarray}}     

6: \newcommand{\de}{$\Delta E$ }

7: 

8: \newcommand{\xx}{$\gamma^{(3)}$}

9: \newcommand{\oso}{oscillator strength }

10: \newcommand{\w}{$\omega$}

11: \newcommand{\bu}{$B_u$}

12: 

13: 

14: 

15: \newcommand{\hp}{hyperpolarizability~}

16: \newcommand{\hpe}{hyperpolarizability}

17: \newcommand{\gm}{$\gamma~$}

18: \newcommand{\gme}{$\gamma$}

19: \newcommand{\gx}{$\gamma_{max}~$}

20: 

21: \newcommand{\uni}{$U_{\rm Ni}~$}

22: \newcommand{\ux}{$U_{\rm X}~$}

23: 

24: \newcommand{\gee}{$\gamma$}

25: \newcommand{\go}{\gamma_0}

26: \newcommand{\U}{\uparrow}

27: \newcommand{\D}{\downarrow}

28: \newcommand{\R}{\rightarrow}

29: \newcommand{\z}{\leftarrow}

30: 

31: \newcommand{\xam}{$|\gamma|~$}

32: \newcommand{\xae}{$|\gamma|$}

33: 

34: \newcommand{\dg}{$^{\circ}$ }

35: \newcommand{\dge}{$^{\circ}$}

36: \newcommand{\na}{$\rm NaV_2O_5$ }  

37: \newcommand{\nae}{$\rm NaV_2O_5$}  

38: \newcommand{\dxy}{$d_{xy}$ }  

39: \newcommand{\dxys}{$d_{xy}^*$ }  

40: \newcommand{\be}{\begin{equation}}

41: \newcommand{\ee}{\end{equation}}

42: \newcommand{\la}{\langle}

43: \newcommand{\no}{\noindent}

44: \newcommand{\ra}{\rangle}

45: \newcommand{\et}{{\it et al. }}

46: \def\oa{$\rm O_1$}

47: \def\ob{$\rm O_2$}

48: \def\oc{$\rm O_3$}

49: \def\cpc{Comp.\ Phys.\ Commun.\ }

50: \def\jpsj{J. Phys. Soc. Jpn.\ }

51: 

52: 

53: 

54: 

55: 

56: \newcommand{\raw}{\rightarrow}

57: \newcommand{\wo}{\omega}

58: 

59: \newcommand{\figone}

60: {

61: \vspace{5cm}

62: \centerline {FIGURE 1, {Zhang and George, PRL }}

63: \newpage}

64: 

65: \newcommand{\figtwo}

66: {

67: \vspace{5cm}

68: \centerline {FIGURE 2, {Zhang and George, PRL }}

69: \newpage}

70: 

71: 

72: \newcommand{\figthree}

73: {

74: \vspace{5cm}

75: \centerline {FIGURE 3, {Zhang and George, PRL }}

76: \newpage}

77: 

78: 

79: \newcommand{\al}{\eta}

80: 

81: 

82: \newcommand{\ws}{\omega_{\sigma}}

83: 

84: \newcommand{\xawm}{$|\gamma(-3\wo;\wo,\wo,\wo)|~$}

85: \newcommand{\xawe}{$|\gamma (-3\wo;\wo,\wo,\wo)|$}

86: 

87: \newcommand{\xazm}{$|\gamma(0)|~$}

88: \newcommand{\xaze}{$|\gamma(0)|$}

89: \newcommand{\mgm}{MgB$_2$ }

90: \newcommand{\mge}{MgB$_2$}

91: 

92: 

93: 

94: \begin{document}

95: \draft

96: 

97: \title{An angle-resolved soft x-ray spectroscopy study of the

98: electronic states of single crystal \mge}

99: 

100: \author{G. P. Zhang,$^{1,2,*}$ G. S. Chang,$^1$ T. A. Callcott$^1$}

101: 

102: \address{$^1$Department of Physics and Astronomy, The University of

103: Tennessee at Knoxville, TN 37996} 

104: 

105: \address{$^2$Department of Physics, Indiana State University, Terre

106: Haute, IN 47809$^*$}

107: 

108: 

109: 

110: \author{D. L. Ederer}

111: \address{Physics Department, Tulane University, New Orleans, LA 70118}

112: 

113: 

114: \author{W. N. Kang, Eun-Mi Choi, Hyeong-Jin Kim, Sung-Ik Lee}

115: 

116: \address{National Creative Research Initiative Center for

117:         Superconductivity and Department of Physics, Pohang University

118:         of Science and Technology, Pohang 790-784, Korea}

119: 

120: \date{\today}

121: 

122: \maketitle

123: 

124: 

125: \begin{abstract}

126: 

127: Angle-resolved soft x-ray measurements made at the boron K-edge in

128: single crystal \mgm provide new insights into the B-2p local partial

129: density of both unoccupied and occupied band states. The strong

130: variation of absorption with incident angle of exciting x-rays permits

131: the clear separation of contributions from $\sigma$ states in the

132: boron plane and $\pi$ states normal to the plane. A careful comparison

133: with theory accurately determines the energy of selected critical $k$

134: points in the conduction band. Resonant inelastic x-ray emission at an

135: incident angle of 15\dg shows a large enhancement of the emission

136: spectra within about 0.5 eV of the Fermi level that is absent at 45\dg

137: and is much reduced at 60\dge.  We conclude that momentum transferred

138: from the resonant inelastic x-ray scattering (RIXS) process couples

139: empty and filled states across the Fermi level.

140: 

141: 

142: 

143: \end{abstract}

144: 

145: \pacs{74.25.Jb, 71.20.-b, 78.70.Dm, 78.70.En}

146: 

147: \begin{multicols}{2}

148: 

149: \section{Introduction}

150: 

151: 

152: The discovery of superconductivity in \mgm came as a

153: surprise.\cite{mgb2} Its unusually high transition-temperature

154: immediately motivated intensive investigations.  Although the

155: conventional BCS mechanism most likely underlies this peculiar

156: superconductivity, the precise mechanisms are less clear. Different

157: sample qualities and different experimental techniques show

158: inconsistent gap values ranging from 2 meV to 7 meV.\cite{gap}

159: However, in a single crystal sample, results are less ambiguous.  A

160: recent angle-resolved photoemission (PE) study on a small \mgm single

161: crystal showed detailed dispersion of the occupied band along the

162: $\Gamma-K$ and $\Gamma-M$ directions.\cite{pe} However, due to the

163: large background and low intensity for normal emission, the dispersion

164: along the $k_z$-direction was not resolved.

165: 

166: Soft x-ray absorption (SXA) spectroscopy \cite{xray} complements PE by

167: providing an element and angular momentum selected local, partial

168: density of states (LPDOS) of the unoccupied bands.  Similarly, soft

169: x-ray fluorescence (SXF) spectra provide a measure of the LPDOS for

170: the occupied valence bands.  Although these spectroscopies are not

171: generally selective for momentum, they can often provide an accurate

172: measure of density of states features associated with critical points

173: of the band structure and thus provide a detailed check on theoretical

174: band calculations. These features of x-ray spectroscopies are clearly

175: illustrated in a recent paper (the first paper of Ref. [5]) on

176: \mge,\cite{prb_mgb2} which used SXF and SXA studies to map the boron

177: p-LPDOS for a polycrystalline sample.

178: 

179: We have recently shown that in an anisotropic material such as

180: NaV$_2$O$_5$,\cite{prl_nav2o5} additional selectivity can be obtained

181: by varying the incident angle of the exciting

182: x-rays. \cite{prb_nav2o5} By varying the incident angle of exciting

183: x-rays so that the polarization vector is varied with respect to a

184: bond direction, it is possible to selectively excite electrons to

185: electronic orbitals having particular spatial orientation. In this

186: paper, we report angle resolved soft x-ray absorption and emission

187: measurements at the boron $K$-edge of a single crystal film of

188: \mge. The soft x-ray absorption spectra provide a measure of the B-2p

189: LPDOS for unoccupied states above the Fermi level. The angular

190: resolved spectra allow us to separate excitation to band states

191: derived from B $\sigma$ states directed within the boron plane and

192: $\pi$ states directed perpendicular to the plane. The results enable

193: us to identify critical points of unoccupied band states along several

194: crystal-momentum directions, which can be directly compared to our

195: theoretical calculations.

196: 

197: 

198: For excitation near a band edge, the excitation and emission processes

199: may be coupled so that the emission spectra must be interpreted in

200: terms of a resonant inelastic x-ray scattering (RIXS) process.  In the

201: RIXS process incident x-rays excite electronic excitations and are

202: scattered with reduced energy.  Previous RIXS experiments in

203: graphite,\cite{graphite} diamond, silicon, cBN\cite{cbn} and

204: hBN\cite{hbn} show that in well screened and delocalized electronic

205: systems the momentum conservation inherent in the RIXS process can

206: provide direct information about band dispersion.  These analyses

207: generally assume that the momentum transferred in the RIXS process is

208: small on the scale of the Brillouin zone so that the electronic

209: excitations induced by RIXS are nearly vertical in the reduced

210: zone. In \mge, however, we observe a strong enhancement of a narrow

211: band of states near the top of the occupied band that requires an

212: electronic excitation across the Fermi level and thus a transfer of

213: momentum.  The variation of this enhancement with incident angle

214: demonstrates that the enhancement is a result of momentum transfer

215: from the RIXS process, and not primarily associated with other

216: processes such as phonon-assisted transitions that might allow

217: momentum to be conserved.

218: 

219: The paper is arranged as follows. In Section II, we present the

220: angular-resolved absorption spectra. The angular-resolved RIXS spectra

221: are presented in Section III. 

222: 

223: \section{Angular resolved x-ray absorption spectra}

224: 

225: A single crystal of \mgm was grown as a thin film of 400 nm thickness

226: on the $R$-plane ($1\bar{1}02$) of an Al$_2$O$_3$

227: substrate.\cite{sample} An amorphous B thin film was deposited by

228: pulsed laser deposition and a high-quality and single crystal \mgm

229: film was obtained by sintering the thin film at 900$^{\circ}$C for 20

230: minutes in Mg vapor. The crystal quality and orientation was confirmed

231: by x-ray diffraction.  The process is described in detail

232: elsewhere.\cite{sample} The dc magnetization measured with a

233: superconducting quantum interference device (SQUID) magnetometer

234: showed that the superconducting transition temperature (Tc) is 39 K,

235: and the 10-90\% transition width is 0.7 K.  Our soft x-ray

236: measurements were made at room temperature at Beamline 8.0 of the

237: Advanced Light Source at Lawrence Berkeley National Laboratory.  The

238: slits of the beamline monochromator were set to provide a resolution

239: of about 0.1 eV for the absorption spectra, while the resolution of

240: the emission spectrometer is about 0.2 eV at the 190 eV energy of the

241: Boron K spectra.

242: 

243: The experimental geometry is indicated in Fig. 1.  \mgm is a layered

244: compound with alternating layers of boron and Mg stacked along the $c$

245: axis. Graphite-like hexagonal layers of boron lie in the $a-b$

246: plane. Linearly polarized x-ray light from an undulator and

247: monochromator was incident on the sample. By rotating the sample

248: holder, we can scan the polarization vector from the $a-b$ plane

249: toward the $c$ axis with $\theta$ changed from 15\dg to 45\dge.

250: $\theta $ measures the angle between the incident x-ray beam and the

251: sample normal as well as the angle between the polarization vector of

252: incident photons and the $a-b$ plane.

253: 

254: 

255: 

256: Figure 2(a) shows the dramatic change of the absorption spectra

257: measured in total fluorescent yield mode with $\theta$.  The spectra

258: shown are taken at 7.5\dg intervals from 15\dg to 45\dg with angles

259: measured from the $a-b$ plane.  The experimental spectra of Fig. 2(a)

260: are normalized with respect to the total areas from 183.55 to 208.55

261: eV (boron $K$-edge).  Experimental and theoretical spectra are aligned

262: with peak labeled $a$.  The Fermi level is set to zero (long dashed

263: line).  The zero energy corresponds to an energy of 187.28 eV with

264: uncertainty of about 1 eV in the absolute energy, consistent with our

265: previous measurement.\cite{prb_mgb2} The relative energies of emission

266: and absorption spectra are calibrated to be accurate to about 0.1 eV.

267: Figure 2 shows that the Fermi level cuts through the pre-edge of peak

268: $a$, a result which agrees with the photoemission data \cite{pe} and

269: shows normal metallic behavior. We identify six spectral features

270: labeled from $a$ to $f$.  At $\theta=15$\dge, only two peaks $a$ and

271: $f$ are observable leaving a 4 eV-wide valley from 1 eV to 5 eV.  As

272: the polarization rotates from the $a-b$ plane toward the $c$ axis,

273: peak $a$ grows slightly while peak $f$ drops. The most remarkable

274: change is in the valley where the whole valley is gradually filled. We

275: will concentrate on the two broad peaks at 2.2 and 2.8 eV above the

276: Fermi level, which increase in amplitude to a point comparable to peak

277: $a$.  We note here that in a polycrystalline or powder sample these

278: angular dependent changes in the absorption spectra are not observed

279: because they are intrinsically angle averaged.\cite{refereea}

280: 

281: Our calculated {\it ab initio} results are shown in Fig. 2(b), where

282:  we also present two additional spectra at $\theta=0$\dg and 90\dge.

283:  For an easy comparison, we plot spectra calculated for the exact same

284:  experimental conditions.  The angular variation results from explicit

285:  calculation of the dipole matrix elements taking into account the

286:  projection of the polarization vector on the orbital amplitude in

287:  real space. The theoretical calculations have been described in

288:  greater detail in a recent paper on NaV$_2$O$_5$.\cite{prb_nav2o5} In

289:  brief, we calculate the eigenvalues and wavefunctions by solving the

290:  Kohn-Sham equation, \be \hat{H}\Psi_{n{\bf k}}({\bf r})=\{

291:  -\frac{{\hbar}^2}{2m}{\nabla}^2+ \hat{V}_{\rm eff}\} \Psi_{n{\bf

292:  k}}({\bf r})=E_{n{\bf k}}\Psi_{n{\bf k}}({\bf r})~. \nonumber \ee

293:  Then, we compute transition matrix elements between the core level

294:  and unoccupied bands.  It is quite impressive that the theoretical

295:  spectra reproduce the experimental ones very accurately. Both the

296:  overall shape and peaks and valleys are consistent with the

297:  experimental counterparts.  A quantitative comparison is given in

298:  Table I. One notices that except for peak $f$, theoretical results

299:  for peaks and valleys from $b$ to $e$ are about 0.2 eV larger than

300:  the experimental ones, which may suggest band narrowing possibly due

301:  to correlation effects.  This band narrowing, if real, is much

302:  smaller than the simple metals such as Na, Li and Mg,\cite{na}

303:  especially considering the experimental uncertainty of about 0.2 eV.

304:  In any case, the effect is small.

305: 

306: \subsection{Incident at $\theta=15$\dge}

307: 

308: The good agreement between theory and experiment paves the way for a

309: detailed investigation of the origin of the peaks.  To help our

310: analysis, we draw several vertical short-dashed lines across both the

311: absorption spectra [Figs. 2(a) and 2(b)] and the band structure

312: [Fig. 2(c)] to highlight important connections. We label those

313: crossing points of peaks $a$, $c$, $d$ and $f$ by $a_1$, $a_2$,

314: etc. With an incident angle of 15\dge, the polarization vector is

315: 15\dg from the $a-b$ plane, so that we mostly probe the

316: $\sigma$-orbitals directed between B atoms lying in the a-b plane.

317: For excitation to states within 0.5 eV of the absorption edge, we see

318: from the spectra that only peak $a$ is excited.  Now comparing with

319: the band structure, we can see six different points from $a_1$ to

320: $a_6$ which potentially contribute to this peak and distribute almost

321: evenly along all momentum directions. For instance, points $a_1$ and

322: $a_4$ in Fig. 2(c) fall near the $\Gamma$ point along the $\Gamma-M$

323: and $\Gamma-K$ lines respectively, while $a_2$ is close to the $M_1$

324: point. $a_3$ is almost in the middle of the $\Gamma-K$ line, and

325: $a_5$, representing two nearby points along the $A-L$ line, is close

326: to the $A$ point, whereas $a_6$ along the $A-L$ line is close to the

327: $L$ point.

328: 

329: If we solely depended on the energy position, we would not be able to

330: resolve these different band contributions to the $a$ peak, a common

331: complication in soft x-ray spectroscopy. Fortunately, \mgm is highly

332: anisotropic so that there are strong variations in the angular

333: dependent transition matrix elements that allow us to resolve the

334: contributions from different bands. We can use the transition matrix

335: elements $\la \psi_{nk}(r)|\vec{E}\cdot \vec{r}|\psi_{B1s}(r)\ra$ to

336: resolve these features, where $|\psi_{nk}(r)\ra$ is the band state and

337: $|\psi_{B1s}(r)\ra$ is the boron 1s core level. Transition matrix

338: elements contain both the direct space and reciprocal space information,

339: which helps us to probe the spatial orientation of orbitals and reveal

340: the dispersive band in momentum space.  

341: 

342: In Fig. 3, in the upper part, we show the BZ in the three dimension,

343: while in the lower part, we plot the calculated transition matrix

344: elements at various positions in the BZ and for different angles of

345: the incident light. The four columns represent the transition matrix

346: elements for four peaks from $a$ to $f$ in the momentum space. Rows 1

347: and 2 refer to $k_z=0$ plane at the bottom of the BZ which contains

348: the $\Gamma$ point while 3 and 4 refer to $k_z=2\pi/c$ plane at the

349: top of the BZ which contains the $A$ point (see the top graph in

350: Fig. 3).  The first and third rows refer to the results at

351: $\theta=15$\dge, while the second and fourth ones the results at

352: $\theta=45$\dge.  The high symmetry $k$ points are also indicated in

353: Figs. 3(a) and 3(c) for the top and bottom planes. Numbers near the

354: contours in Fig. 3 indicate the magnitude of transition matrix

355: elements (in arbitrary units).  First focus on Figs. 3(a) and 3(c),

356: which show the matrix elements contribution to peak $a$ for

357: $\theta=15$\dge. From those two figures, one can see clearly that the

358: largest transition matrix elements are well localized at the $\Gamma$

359: point and in the vicinity of the $A$ point while contributions from

360: other directions are very small.  In three dimensions, it is found

361: that the matrix elements remain equally strong for the region

362: surrounding the $\Gamma-A$ line that lies within the tubular Fermi

363: surface \cite{band} that surrounds this line (see the top graph of

364: Fig. 3).  Returning to Fig. 2(c), now we can identify that band states

365: at $a_1(a_4)$ and $a_5$ contribute most strongly to peak

366: $a$. Therefore, the position of peak $a$ can be used to measure the

367: position for $\Gamma_1$, which is 0.27 eV above the Fermi surface. Our

368: assignment is also consistent with the nature of the $\sigma$ band

369: along the $\Gamma_1-A_1$ line. It is well known that this famous

370: $\sigma$ band is derived from electronic orbitals in the $a-b$

371: plane. Since our present electric field polarization is nearly in the

372: $a-b$ plane, this band is strongly excited.

373: 

374: If we increase the incoming photon energy to 0.69 eV (with respect to

375: the Fermi level), the spectra start to fall sharply since there is no

376: density of states above $A_1$ and bands along $M-K-\Gamma$ can not be

377: effectively excited (see below). When we increase the photon energy

378: further up to the $\Gamma_2$ point, we begin to access a new band

379: ($c_4$ and $c_5$, or $d_4$ and $d_5$), thus we would expect a strong

380: increase of the spectral weight. Surprisingly, we see a 4 eV-wide

381: valley instead. The valley also appears in the theoretical spectra

382: (see Fig. 2(b)).  The transition-matrix element contours in Figs. 3(e)

383: and 3(g) are for peak $c$, and in Figs. 3(i) and 3(k) for peak $d$,

384: but the matrix elements are small.

385: 

386: In order to obtain a qualitative understanding of these variations in

387: the matrix elements, we calculated the spatial charge density

388: distribution. For the bands at point $c_6$ (similarly $d_6$, $c_2$ and

389: $c_3$), we find that they are derived from $\pi$-orbitals and are

390: highly oriented along the $c$ axis (see Fig. 4(a)), which explains why

391: they can not be effectively excited. However, this does not explain

392: bands at points $c_4$, $c_5$, $d_4$ and $d_5$ since they do lie in the

393: $a-b$ plane (see Fig. 4(b)). A very careful investigation reveals the

394: secret: those orbitals in Fig. 4(b) are strongly localized around Mg

395: sites, not B sites. Since our measurement is element-specific and is

396: done at the boron $K$ edge, we are not sensitive to orbitals around

397: the Mg sites, which explains the appearance of the 4-eV wide valley at

398: $\theta=15$\dge.

399: 

400: 

401: Above the valley labeled $e$ at 3.66 eV, the spectra increase in

402: intensity.  Further increasing the photon energy to 6.0 eV above the

403: Fermi level, we pick up high-lying $\sigma$ bands along the

404: $\Gamma_3-A_2$ direction and the $\Gamma-K$ (A-L) line. This can be

405: seen directly from the transition-matrix contour plot in Figs. 3(m)

406: and 3(o). Therefore, the energy position of peak $f$ is a good measure

407: of the position of the $\Gamma_3$ and $A_2$ points. Our experiment

408: gives the value of 6.65 eV, which can be compared with the theoretical

409: value of 6.55 eV.

410: 

411: \subsection{Incident at $\theta=45$\dg}

412: 

413: 

414: When we rotate the polarization vector from the $a-b$ plane toward the

415: $c$ axis ($\theta=45$\dge), we begin to probe orbitals which are

416: oriented along the $c$ axis.  We start from peak $a$. Comparing

417: Figs. 3(a) with 3(b) (Figs. 3(c) with 3(d)), one notices that the

418: spectral weight moves away from the $\Gamma$ (A) point to the M-K

419: (H-L) line. Qualitatively we can say that the contribution to this

420: peak from the $\sigma$-bands is decreasing with angle while the

421: contribution from the $\pi$-bands bands is strongly increasing.

422: Clearly, these angular measurements give us the ability to cleanly

423: separate the contribution of bands with different spatial orientation

424: to the measured density of states.

425: 

426: As the angle is increased, two broad peaks $c$ and $d$ appear at 2.2

427: and 2.8 eV above the Fermi surface (see Figs. 2(a) and 2(b)),

428: respectively.  Although there are six crossing points contributing to

429: peak $c$, we can safely exclude $c_1$, $c_4$ and $c_5$ since they are

430: from Mg sites.  The transition matrix contour in Fig. 3(f) shows that

431: states with large elements are near the K point, while at

432: $k_z=2\pi/c$, the matrix elements are only half that of the states at

433: $k_z=0$ (see Fig. 3(h)).  Therefore, peak $c$ predominantly comes from

434: states at $c_2$ and $c_3$ and determines their energy positions.  The

435: momenta for these states are (0.38,0.25,0) and (0.3,0.3,0),

436: respectively.  They form a hot spot which leads to peak $c$.  For peak

437: $d$, the transition matrix contour plot (see Fig. 3(j)) shows that the

438: dominant contribution comes from states along the $M-K$ and $\Gamma-K$

439: lines.  States at $d_4$ and $d_5$ (see Fig. 2(c)) do not contribute

440: because those states originate from Mg sites. The main contribution is

441: from bands at $d_2$ and $d_3$ which are highly oriented along the $c$

442: axis (see Fig. 4(a)).  They form another hot spot which ultimately

443: leads to peak $d$. By carefully analyzing the data, we find that point

444: $d_3$ is at (0.28, 0.28, 0) in the reciprocal space while point $d_2$

445: is at (0.46, 0.09,0), very close to the $M$ point. The position of

446: peak $d$ of 2.8 eV is a good measure of the energy position for the

447: $M_2$ point.

448: 

449: Summarizing all the data, we tabulate in the last column of Table I

450: the measured values of unoccupied band states at selected points of

451: both high and low symmetry in k-space. These measured values of

452: particular points in k-space provide values that may be compared

453: directly to calculated values.

454: 

455: 

456: \section{Angle resolved resonant inelastic scattering}

457: 

458: We now turn our attention to the soft x-ray (SX) emission spectra

459: produced when electrons from occupied valence states refill the

460: excited core state.  Well above the excitation threshold, the

461: excitation and emission processes are always decoupled so that the

462: normal SX fluorescence (SXF) spectra represent the boron p-LPDOS of

463: the filled valence band.  Near threshold, however, the absorption and

464: emission processes are often strongly coupled and must be described in

465: terms of a resonant inelastic x-ray scattering (RIXS) process. An

466: important defining feature of RIXS spectra is that the energy

467: difference between incident and scattered electrons is equal to the

468: energy of the electronic excitations so that RIXS features move in

469: energy with the excitation energy, while normal SXF features remain

470: fixed in energy.  For near threshold excitation, normal SXF and RIXS

471: are often superimposed. In the present data, resonant features in the

472: SXF spectra are observed near threshold for excitation to the $\sigma$

473: states that contribute to peak $a$ in Fig. 2. For energies more than 1

474: eV above the absorption threshold, the spectra do not change with

475: further increases in energy and are representative of SXF spectra.

476: 

477: The emission spectra are presented in Fig. 5 for excitation energies

478: within 1 eV of the boron K edge. Note that the Fermi level is at about

479: 187.28 eV.  Spectra are shown for incident angles of 15\dg and

480: 60\dge. Figure 5(a) shows the results at 15\dg where we have seen that

481: excitation is mainly into the $\sigma$ states. When we excite at

482: 187.25 eV, only an elastic peak A can be seen. Increasing the

483: excitation energy to the threshold of 187.50 eV, a shoulder appears on

484: the low energy edge of A and a broad peak B appears around 185.05 eV.

485: Its energy position stays the same even with higher excitation

486: energies of 188.00 and 188.25 eV, which identifies it as a normal

487: fluorescence peak, where the excitation and de-excitation processes

488: are decoupled.

489: 

490: 

491: From Fig. 5(a), we notice that at the excitation energy of 187.50 eV,

492: the elastic peak A becomes asymmetric and broader. The shape of its

493: left shoulder indicates a possible peak hidden there.  This is indeed

494: the case.  When we further increase the excitation energy to 187.75

495: eV, peak C is revealed at 187.30 eV. Two significant features of this

496: peak may be noted. The strong resonant enhancement of the peak remains

497: visible so long as absorption is into the $\sigma$-band and ends when

498: the excitation energy exceeds the $A_1$ point of this band (see

499: Fig. 2). The resonance region extends only to about 1 eV below the

500: Fermi edge and thus does not include the lower energy $\sigma$ states.

501: With excitation at an incident angle of 60\dg shown in Fig 5(b), the

502: resonant enhancement is greatly reduced, but is still present. For

503: excitation at 45\dg (not shown), no enhancement at all is observed in

504: the SXF spectra. Comparing Figs. 5(a) with 5(b), we find that they are

505: very similar, and within our experimental resolution of 0.2 eV the

506: peak positions are identical (see Table II). The only significant

507: difference is the reduction in the magnitude of the resonant

508: enhancement near the Fermi level when the angle is increased from

509: 15\dg to greater angles.

510:  

511: 

512: It is difficult to account for the resonant enhancement in terms of an

513: ordinary RIXS process in a delocalized system, because k-conservation

514: is expected for the two photon scattering process.  K-conservation

515: follows from the Kramers-Heisenberg formula that describes the RIXS

516: process \cite{prl_nav2o5} \ba S(\omega,\omega')\propto\sum_f\left |

517: \sum_m\frac{\la f| p\cdot A |m\ra\la m|p\cdot

518: A|gs\ra}{\omega+E_{gs}-E_m-i\Gamma}\right |^2 \nonumber\\

519: \times \delta(E_{gs}+\omega-E_f-\omega')~, \ea where $|gs\ra$, $|m\ra$, and

520: $|f\ra$ are initial, intermediate, and final states, and $E_{gs}$,

521: $E_m$, and $E_f$ are their energies, respectively; $\omega$ and

522: $\omega'$ are the incident and emitted photon energies; $\Gamma$ is

523: the spectral broadening due to the core lifetime in the intermediate

524: state. $p\cdot A$ is the transition operator.  If $|\psi_{nk_1}\ra$ is

525: the intermediate state $|m\ra$ with momentum $k_1$ and

526: $|\psi_{nk_2}\ra$ is the final state $|f\ra$ with momentum $k_2$,

527: besides the energy conservation, the total momentum should be

528: conserved, i.e., $q+k_1=q'+k_2$, where $q$ and $q'$ are momenta for

529: the incoming and outgoing photons, respectively.  If the photon

530: momenta $q$ and $q'$ were negligible as is the case in visible

531: spectroscopy, we would have $k_1\approx k_2$ or zero momentum

532: transfer.

533: 

534: 

535: Previous calculations\cite{band} show that the empty states from the

536: $\sigma$ band lie in a small tubular region in k-space surrounding the

537: $\Gamma$-A axis,\cite{band} while the filled states from this band are

538: located outside of this tube.  This behavior can also be seen for the

539: $\Gamma$-K, $\Gamma$-M and A-L axes in Fig. 2(c) and are similar for

540: all other directions normal to the $\Gamma$-A axis.  Thus the resonant

541: enhancement requires a mechanism that provides momentum normal to this

542: axis and couples the empty states inside the tube to the filled states

543: outside the tube.

544: 

545: First, we will estimate the magnitude of the energy loss and momentum

546: transfer for the enhanced emission. When excited at 187.75 eV, the

547: resonant inelastic peak C in Fig. 5(a) has the energy loss that is

548: centered at about 0.5 eV and covers a range of about 1 eV. To excite

549: this range of energies requires a momentum transfer in the range of

550: 0.1 on the scale of 2$\pi/a$ which sets the scale of the BZ. All

551: subsequent momentum values are quoted on this scale.

552: 

553: 

554: We have considered two possible processes that might provide the

555: required momentum.  They are phonon assisted transitions and momentum

556: transfer from the RIXS scattering process.\cite{refereeb} A previous

557: calculation of the phonon spectrum\cite{phonon} shows that the

558: in-plane phonon mode $E_{2g}$ has a very strong electron-phonon

559: coupling (EPC) along the $\Gamma$-A, $\Gamma$-M, and A-L

560: directions. Such a strong but selective coupling might provide way to

561: transfer momentum from the phonon subsystem to the electronic

562: subsystem.  Since there is no valence state available along the

563: $\Gamma$-A direction, its strong coupling is irrelevant to the

564: anomalous enhancement.  The most important coupling is along the

565: $\Gamma$-M and A-L directions.  To be more specific, the in-plane

566: $E_{2g}$ phonon mode with the strongest EPC is the most

567: important\cite{phonon}. Along the $\Gamma$-M direction, phonon can

568: transfer momentum ${\bf q}$ to maximum values of (0.15,0,0), while

569: along the A-L direction, transfers occur to maximum values of ${\bf

570: q}=(0.2,0,0.5)$.  These momentum transfers are of the correct order of

571: magnitude to efficiently couple the excitation and de-excitation

572: processes.

573:  

574: 

575: Since momentum is transferred in the SX scattering process, we

576: also considered the possibility that this accounts for the observed

577: enhancement. As indicated in Fig. 6(a), the scattering of X-ray photons

578: through 90\dg requires a transfer of momentum in the interaction with

579: the surface along a line 45\dg from the incident direction.  For 15\dg

580: angle of incidence on the sample, the momentum transfer normal to the

581: $\Gamma$-A line is calculated to be about 0.03 (2$\pi/a$), while for a

582: 45\dg angle of incidence, there is no momentum transfer normal to this

583: line. The mechanism is illustrated schematically in Fig. 6(b).

584: 

585: The variation of the magnitude of the resonant enhancement as the

586: incident angle is varied allows us to choose between these two

587: mechanisms.  A phonon-assisted process should not depend on angle so

588: that the enhancement in the spectra should only be reduced moderately

589: by the change in the projection of the polarization on the $\sigma$

590: orbitals in the $a-b$ plane. If the momentum is provided by the RIXS

591: scattering process, the enhancement should be absent at 45\dg and

592: greatly reduced at 60\dge.  The measured enhancement is indeed found to

593: be unobservable at 45\dg and greatly reduced at 60\dge. We conclude

594: that the enhancement is enabled by the momentum provided by the RIXS

595: process, allowing non-vertical transitions within the BZ.

596: 

597: 

598: 

599: 

600: 

601: We note finally that two separate angular dependent effects are

602: present in the RIXS spectra. First the projection of the polarization

603: vector of the incident x-rays on the orbitals determines the intensity

604: of the absorption into the derived intermediate states.  This factor

605: falls off as a function of $\cos\theta$ with increasing incident

606: angle.  And second, the resonant enhancement near threshold depends on

607: the momentum transfer perpendicular to the $\Gamma-A$ axis provided by

608: the RIXS process, which varies as $\cos(\theta-45^{\circ})$ with a

609: 90\dg scattering angle.  A new experimental endstation will soon be

610: available that will permit the scattering angle in RIXS measurements

611: to be varied over a wide range, which will greatly facilitate the

612: study of momentum transfer effects.

613: 

614: 

615: We acknowledge many helpful discussions with Eric L. Shirley.  This

616: research is supported by NSF Grant No. DMR-9801804. Measurements were

617: carried out at the Advanced Light Source at LBNL supported by DOE

618: contract DE-A003-76SF00098.  This work is also supported in part by

619: the Ministry of Science and Technology of Korea through the Creative

620: Research Initiative Program.

621: 

622: 

623: $^*$Mailing address.

624: 

625: \begin{references}

626: 

627: \bibitem{mgb2} J. Nagamatsu, N. Nakagawa, T. Muranaka, Y. Zenitani, and

628: J. Akimitsu, Nature (London) {\bf 63}, 410 (2001).

629:     

630: \bibitem{gap}G. Karapetrov, M. Iavarone, W. K. Kwok, G. W. Crabtree, and

631: D. G. Hinks, Phys. Rev.  Lett. {\bf 86}, 4374 (2001);

632: G. Rubio-Bollinger, H.  Suderow, and S. Vieira, Phys. Rev.  Lett. {\bf

633: 86}, 5582 (2001).

634: 

635: \bibitem{pe}H. Uchiyama, K. M. Shen, S. Lee, A. Damascelli, D.  H. Lu,

636: D. L. Feng, Z.-X.  Shen, and S. Tajima, Phys. Rev.  Lett. {\bf 88},

637: 157002 (2002).

638: 

639: \bibitem{xray}T. A. Callcott, in Vacuum Ultraviolet Spectroscopy II,

640: edited by J. A. Samson and D. L. Ederer, Experimental Methods in the

641: Physical Sciences, Vol. 32 (Academic Press, New York, 1998),

642: pp. 279-300.

643: 

644: 

645: 

646: \bibitem{prb_mgb2}T. A. Callcott, L. Lin, G. T. Woods, G. P. Zhang,

647: J. R. Thompson, M. Paranthaman, and D. L. Ederer, \prb {\bf 64}, 132504

648: (2001); E. Z. Kurmaev, I. I. Lyakhovskaya, J. Kortus, A. Moewes,

649: N. Miyata, M. Demeter, M. Neumann, M. Yanagihara, M. Watanabe, T.

650: Muranaka, and J. Akimitsu, Phys. Rev. B {\bf 65}, 134509 (2002).

651: 

652: 

653: \bibitem{prl_nav2o5}G. P. Zhang, T. A. Callcott, G. T. Woods, L. Lin,

654: Brian Sales, D. Mandrus and J. He, \prl {\bf 88}, 077401 (2002); {\bf

655: 88}, 189902 (2002, Erratum).

656: 

657:  

658: \bibitem{prb_nav2o5}G. P. Zhang, G. T. Woods, Eric L. Shirley,

659: T. A. Callcott, L. Lin, G.  S. Chang, B. C.  Sales, D. Mandrus, and

660: J. He, \prb {\bf 65}, 165107 (2002); G. T. Woods, G. P. Zhang,

661: T. A. Callcott, L.  Lin, G. S. Chang, B.  Sales, D. Mandrus, and

662: J. He, \prb {\bf 65} 165108 (2002).

663: 

664: 

665: \bibitem{graphite} J. A. Carlisle, Eric L.  Shirley, E. A. Hudson,

666: L. J. Terminello, T. A.  Callcott, J. J. Jia, D. L.  Ederer, R. C. C.

667: Perera, and F. J.  Himpsel, Phys. Rev.  Lett. {\bf 74}, 1234 (1995).

668: 

669: \bibitem{cbn}E. L. Shirley, J. A. Soininen, G. P. Zhang,

670: J. A. Carlisle, T. A. Callcott, D. L. Ederer, L. J. Terminello and

671: R. C. C. Perera, J. El. Spectr. \& Rel. phenomena {\bf 114-116}, 939

672: (2001).

673: 

674: \bibitem{hbn}J. A. Carlisle, Eric L.  Shirley, L. J. Terminello, J.

675: J. Jia, T. A. Callcott, D. L.  Ederer, R. C. C. Perera, and

676: F. J. Himpsel, \prb {\bf 59}, 7433 (1999); J. Jia, T. A.  Callcott,

677: Eric L.  Shirley, J. A.  Carlisle, L. J.  Terminello, A.  Asfaw, D. L.

678: Ederer, F. J.  Himpsel, and R. C.  C. Perera, Phys. Rev.  Lett. {\bf

679: 76}, 4054 (1996).

680: 

681: \bibitem{sample} W. N. Kang, Hyeong-Jin Kim, Eun-Mi Choi, C. U. Jung,

682: and Sung-Ik Lee, Science {\bf 292}, 1521 (2001).

683: 

684: \bibitem{refereea} Since x-ray spectrum data depend on not only the

685: polar angle $\theta$ but also the azimuthal angle $\phi$ (which is in

686: the a-b plane), in order to compare the data in the polycrystalline

687: sample, one should average the data over both $\theta$ {\it and}

688: $\phi$.  This explains why the spectrum at 45 degrees is different

689: from the data in the polycrystalline sample. The averaged absorption

690: spectrum represents the boron partial density of states.  This is also

691: the reason why all three set data in the polycrystalline or powder

692: samples from different groups [Y. Zhu {\it et al}. \prl {\bf 88},

693: 247002 (2002), T. A. Callcott {\it et al}. \prb {\bf 64}, 132504

694: (2001), and McGuinness {\it et al}, Europhys. Lett. {\bf 56}, 112

695: (2001)] are consistent. In the sintered sample [McGuinness et al,

696: Europhys. Lett. {\bf 56}, 112 (2001)], as noted by those authors,

697: surface boron oxides are more substantial and may significantly change

698: the spectrum. Consequently their sinistered MgB2 emission spectrum

699: looks more similar to the B$_2$O$_3$ spectrum shown in Fig. 4 of the

700: first paper of Ref. 5.

701: 

702: 

703: 

704: 

705: 

706: 

707: \bibitem{na}T. A. Callcott, E. T.  Arakawa, and D. L.  Ederer,

708: Phys. Rev. B {\bf 18}, 6622 (1978); T. A. Callcott and D. L.  Ederer,

709: Phys. Rev.  Lett. {\bf 62}, 1322 (1989).

710: 

711: 

712: 

713: \bibitem{band}J. Kortus {\it et al.}, Phys. Rev.  Lett. {\bf 86}, 4366

714: (2001).

715: 

716: 

717: \bibitem{refereeb} 

718: The self-absorption in our RIXS data is very weak.  The

719: reason is as follows.  First, the self-absorption effect becomes

720: stronger with increase in the incident angle as seen from the

721: absorption spectra. Consequently, if the self-absorption had a strong

722: contribution to our RIXS data, we would expect a monotonic decrease in

723: the intensity of peak C. This does not agree with our experimental

724: observation: the intensity of peak C decreases and after 45 degrees,

725: it increases.  Therefore, we conclude that the self-absorption effect

726: less likely plays a dominant role in our RIXS spectra.

727: 

728: 

729: \bibitem{phonon} Y. Kong, O. V. Dolgov, O. Jepsen and O. K. Anderson,

730: Phys. Rev. B {\bf 64}, 020501 (2001)

731: 

732: \end{references}

733: 

734: \end{multicols}

735: 

736: \begin{table}

737: \begin{caption}

738: {A comparison between experimental and theoretical energy positions

739: for six peaks and valleys. The last column shows their corresponding

740: crystal momenta and assigned special symmetry points. All the energies

741: are referenced to the Fermi level. }

742: \end{caption}

743: 

744: \vspace{1cm}

745: 

746: \begin{tabular}{ccccl}

747: Peak \& valley &~~~Experiment (eV)~~~&~~ Theory (eV)~~&~~Deviation (eV)~~&$k-$point\\

748: \hline

749: a& 0.27&           0.27&0.00&$\Gamma_1$ \\

750: b& 1.25  &         1.49&0.24&$\Gamma_2$\\

751: c& 2.24    &       2.43&0.19& (0.38, 0.25, 0) \\ 

752:  &         &           &    & (0.3, 0.3, 0)\\       

753: d& 2.83     &      2.96&0.13& (0.46, 0.09, 0)($\rm M_2$)\\

754:  &         &           &    & (0.28, 0.28, 0)\\       

755: e& 3.66      &     3.79&0.13& $\rm M_3$\\

756: f& 6.65       &    6.55&-0.10& $\Gamma$, ($\rm A_2$) \\

757: \end{tabular}

758: 

759: \vspace{2cm}

760: 

761: \begin{caption}

762: { The resonant inelastic peak positions for two different angles

763: $\theta=15$\dg and $\theta=60$\dge.  At $\theta=15$\dg the strong

764: electron-phonon coupling in the flat $\sigma$ band leads to a surprising

765: enhancement of peak C at 187.30 eV. The energy units are in eV.  The

766: experimental uncertainty is about 0.2 eV.}

767: \end{caption}

768: 

769: \vspace{1cm}

770: 

771: \begin{tabular}{lllll}

772: Excitation energy  &\multicolumn{2}{c}

773: {$\theta=15$\dge} &\multicolumn{2}{c} {$\theta=60$\dge} \\

774: \hline

775: A & B  &   C             & B  &  C      \\

776: \hline

777: 187.50  & 185.05  &                & 185.05  &        \\

778: 187.75  & 185.05  & 187.30         & 185.05  & 187.07 \\

779: 188.00  & 185.05  & 187.21         & 185.05  & 186.95 \\

780: 188.25  & 185.05  & 186.98         & 185.05  & 186.95 

781: \end{tabular}

782: 

783: 

784: \end{table} 

785: 

786: \begin{figure}

787: \caption{Experimental geometry.  A linearly polarized photon with

788: energy $\omega_{in}$ is incident on the sample, and the outgoing

789: photon energy becomes $\omega_{out}$.  ${\bf n}$ is surface normal,

790: ${\bf E}$ is polarization of incident photon, $(a-b)$ is plane of

791: boron atoms, and ${\bf c}$ is normal to boron plane.  }

792: \end{figure}

793: 

794: 

795: 

796: 

797: \begin{figure}

798: \caption{ (a) The angular-resolved experimental absorption spectra at

799: the boron $K-$edge. $\theta$ represents the angle between the electric

800: polarization $\rm \vec{E}$ and the $a-b$ plane. We scan from 15\dg to

801: 45\dg with incremental angle of 7.5\dge. Six peaks and valleys are

802: labeled by $a-f$.  The experimental geometry is shown in the

803: inset. (b) Theoretical spectra with the exact same condition. Two

804: dashed curves represent results at 0 and 90\dge, respectively.  (c)

805: Band structure. Vertical short-dashed lines highlight the connection

806: between the band structure and absorption spectra. Crossing points are

807: denoted with $a_1$, $a_2$, $a_3$, etc. The long vertical line

808: represents the Fermi level which is set to zero. }

809: \end{figure}

810: 

811: 

812: 

813: 

814: \begin{figure}

815: \caption{ Transition-matrix element contour in the reciprocal

816: space. The $x$ axis denotes the $k_x$ while the $y$ axis the $k_y$

817: axis. Four columns represent results for peaks $a$, $c$, $d$ and $f$,

818: respectively. The first and third rows are calculated at

819: $\theta=15$\dge, while the second and fourth rows are at

820: $\theta=45$\dge. The first and second rows are for the crystal

821: momentum $k_z=0$ while the third and fourth rows are for

822: $k_z=2\pi/c$. Numbers (in arbitrary units) near contours show the

823: magnitude of transition matrix elements. The high symmetry points are

824: shown in (a) and (c), and are also shown in 3-dimensional BZ.  }

825: \end{figure}

826: 

827: 

828: \begin{figure}

829: \caption{ Unoccupied spatial charge density plot in the $y-z$ plane

830: (the $z$ axis is same as the $c$ axis).  (a) represents charge

831: distribution for bands at point $c_3$ (see Fig. 1(c)) and other points

832: $c_2$, $d_2$ and $d_3$. The charge is mainly located around boron

833: sites. (b) represents those bands at point $c_4$ (see Fig. 1(c)) as

834: well as other points $c_1$, $c_5$, $d_1$, $d_4$, and $d_5$. The

835: density is located around magnesium sites, which makes little

836: contributions to our absorption spectra. }

837: \end{figure}

838: 

839: \begin{figure}

840: \caption{ Resonant inelastic x-ray scattering spectra measured at (a)

841: $\theta=15$\dg and (b) $\theta=60$\dge. The main difference is that at

842: $\theta=15$\dge, there is an enhancement of the inelastic peak C at

843: excitation energy of 187.75 eV. The Fermi energy is at about 187.28

844: eV. }

845: \end{figure}

846: 

847: 

848: \begin{figure}

849: \caption{(a) Momentum transfer $\Delta \bf k$ in RIXS process.  The

850: projection of the momentum transfer along the a/b plane changes with

851: the incident angle, with the minimum at 45\dg in our experimental

852: geometry.  $\bf k_{in}$ and $\bf k_{out}$ are incident and outgoing

853: momenta, respectively.  ${\bf n}$ denotes the surface normal or the

854: $c$ axis.  (b) With momentum transfer $\Delta k$, valence electrons

855: from different bands can recombine with the core hole, showing a

856: strong enhancement in the RIXS spectra (see Fig. 5(a)).  }

857: \end{figure}

858: 

859: 

860: 

861: 

862: 

863: \newpage

864: 

865: \hspace{5cm}\psfig{figure=fig1.eps,width=8cm,angle=0}

866: 

867: \vspace{4cm}

868: 

869: \centerline{FIGURE 1}

870: 

871: \hspace{1cm}\psfig{figure=fig2.eps,width=12cm,angle=180}

872: 

873: \vspace{1cm}

874: 

875: \centerline{FIGURE 2}

876: 

877: \newpage

878: 

879: 

880: \hspace{7.5cm}\psfig{figure=fig3son.eps,width=3cm,angle=0}

881: 

882: \vspace{-0.5cm}

883: 

884: \hspace{1.5cm}\psfig{figure=fig3L.eps,width=18cm,angle=270}

885: 

886: 

887: \vspace{1cm}

888: 

889: \centerline{FIGURE 3}

890: 

891: 

892: 

893: 

894: 

895: \hspace{0.8cm}\psfig{figure=fig4.eps,width=12cm,angle=270}

896: 

897: 

898: \vspace{2cm}

899: 

900: \centerline{FIGURE 4}

901: 

902: 

903: 

904: \hspace{-1.5cm}\psfig{figure=fig5.eps,width=17cm,angle=270}

905: 

906: 

907: \vspace{2cm}

908: 

909: \centerline{FIGURE 5}

910: 

911: 

912: \hspace{-.5cm}\psfig{figure=fig6.eps,width=17cm,angle=0}

913: 

914: 

915: \vspace{2cm}

916: 

917: \centerline{FIGURE 6}

918: 

919: 

920: \end{document}

921: 

922: 

923: 

924: 

925: