1: %Version last modified 10/28/2007

2: %\documentclass[preprint,prl,superscriptaddress,amsmath,amssymb,letterpaper]{revtex4}

3: \documentclass[twocolumn,prl,superscriptaddress,amsmath,amssymb,letterpaper]{revtex4}

4: \usepackage{graphicx}% Include figure files

5: \usepackage{bm}% bold math

6: 

7: \def\beq{\begin{equation}}

8: \def\eeq{\end{equation}}

9: \def\beqn{\begin{eqnarray}}

10: \def\eeqn{\end{eqnarray}}

11: \DeclareMathOperator{\Tr}{Tr} 

12: \DeclareMathOperator{\sinc}{sinc} 

13: \DeclareMathOperator{\Ai}{Ai}

14: \DeclareMathOperator{\Max}{Max}

15: \DeclareMathOperator{\Min}{Min}

16: \DeclareMathOperator{\diag}{diag}

17: \DeclareMathOperator{\Real}{Re}

18: \DeclareMathOperator{\si}{si}

19: \DeclareMathOperator{\ci}{ci}

20: 

21: \begin{document}

22: %\draft

23: %\tighten

24: %\onecolumn

25: %\twocolumn[\hsize\textwidth\columnwidth\hsize\csname

26: %@twocolumnfalse\endcsname

27: 

28: \title{Optical coupling to spin waves in the cycloidal multiferroic BiFeO$_3$}

29: \author{Rogerio de Sousa}

30: \affiliation{Department of Physics, University of California,

31: Berkeley, CA 94720}

32: \author{Joel E. Moore}

33: \affiliation{Department of Physics, University of California,

34: Berkeley, CA 94720} \affiliation{Materials Sciences Division,

35: Lawrence Berkeley National Laboratory, Berkeley, CA 94720}

36: \date{\today}

37: 

38: \begin{abstract}

39:   The magnon and optical phonon spectrum of an incommensurate

40:   multiferroic such as BiFeO$_3$ is considered in the framework of a

41:   phenomenological Landau theory. The resulting spin wave spectrum is

42:   quite distinct from commensurate substances due to soft mode

43:   anisotropy and magnon zone folding. The former allows electrical

44:   control of spin wave propagation via reorientation of the

45:   spontaneous ferroelectric moment. The latter gives rise to multiple

46:   magneto-dielectric resonances due to the coupling of optical phonons

47:   at zero wavevector to magnons at integer multiples of the cycloid

48:   wavevector.  These results show that the optical response of a

49:   multiferroic reveals much more about its magnetic excitations than

50:   previously anticipated on the basis of simpler models.

51: \end{abstract}

52: \pacs{

53: %Magnetomechanical and magnetoelectric effects, magnetostriction

54: 75.80.+q,

55: %Spin waves in magnetically ordered materials

56: 75.30.Ds,

57: %Magnetooptical effects

58: 78.20.Ls.}

59: %Infrared and Raman spectra (condensed matter) 

60: %78.30.-j.}

61: %

62: \maketitle

63: %\vskip2pc]

64: %\narrowtext

65: %

66: 

67: The coupling between different types of correlated electron order in

68: ``multiferroic'' materials possessing simultaneous magnetism and

69: ferroelectricity gives rise to several interesting phenomena

70: \cite{wang03,kimura03,pimenov06,sushkov07}.  This coupling leads to

71: rather complex broken-symmetry phases~\cite{harris06} and excitation

72: spectra.  Recent examples are the magnetically induced

73: ferroelectricity in TbMnO$_3$\cite{kimura03} and the switching of the

74: magnetic state by an electrical probe observed in

75: BiFeO$_3$~\cite{zhao06}.  Crystals of these two materials show

76: incommensurate magnetic order: in TbMnO$_3$, the incommensurate order

77: established at 41 K is believed to generate the spontaneous

78: polarization observed at 28 K, while in BiFeO$_3$, the cycloidal

79: antiferromagnetic order is established at 650 K, well below the

80: ferroelectric transition at 1120 K.  The cycloid disappears in

81: a strong (18 T) magnetic field~\cite{ruette04} or in thin films~\cite{bai05}.

82: 

83: This paper studies the excitations and electromagnetic response in a

84: cycloidal multiferroic such as bulk BiFeO$_3$.  The combination of

85: incommensurate order and magnetoelectric coupling is found to give a

86: strong coupling between a long-wavelength electric field and spin

87: waves at multiples of the cycloid wavevector.  This leads to a series

88: of resonances in the dielectric constant at integer and irrational

89: multiples of a fundamental frequency determined by the magnon energy

90: at the cycloid wavevector.  The lowest-lying resonances created by

91: this effect should be visible in standard transmission and

92: reflectivity measurements at room temperature in the far infrared

93: frequency range (below the lowest optical phonon at 2~THz).  Recent

94: Raman and optical reflectivity spectroscopy studies of bulk BiFeO$_3$

95: \cite{haumont06} focused on the optical phonon resonances, without any

96: interpretation of the far IR region.  The optical phonon spectra seem

97: to be well understood from first principles calculations

98: \cite{hermet07}.  Nevertheless, the latter completely ignores the

99: underlying cycloidal magnetic structure and its magnetoelectric

100: character. Our work reveals a rich sub-phonon dielectric response not

101: yet explored by optical experiments or first principle calculations.

102: Aside from allowing a direct observation of the basic physics of

103: multiferroic coupling in BiFeO$_3$ and related materials, this effect

104: could be used in devices based on electronic excitation of spin waves

105: \cite{khitun05}, or the development of fast magnetic probes for

106: magnetic domain switching.

107: 

108: It has been known for many years that magnetoelectric coupling in

109: uniformly ordered materials mixes spin waves (magnons) and

110: polarization waves (optical phonons) \cite{baryakhtar69,tilley82}.

111: This gives rise to low-frequency magneto-optical resonances in the

112: dielectric susceptibility, the so-called electromagnon excitations,

113: that may be visible in infrared experiments

114: \cite{pimenov06,sushkov07}.  However, the resonance frequencies are

115: primarily determined by the zero-wavevector magnon frequencies.  The

116: mixing with finite-wavevector magnons found here results specifically

117: from the {\it incommensurate} magnetic order; while such

118: incommensurate order complicates the calculation of excitation spectra

119: because a finite-wavevector analysis is now necessary, incommensurate

120: order is a common feature of many of the most studied multiferroic

121: materials.  Although recent theoretical work has clarified the

122: relationship between ferroelectricity and spiral ferromagnetic order

123: \cite{mostovoy06,katsura05}, the characteristic features of

124: electromagnon excitations in incommensurate magnets seem not to have

125: been studied before.

126: 

127: This paper assumes that the ground state consists of a uniform

128: polarization and an incommensurate magnetic structure, as in many

129: important multiferroic materials; an interesting effect when the

130: polarization and magnetization are both periodic and commensurate with

131: each other has recently been discussed \cite{betouras07}.

132: 

133: Our calculation is based on a dynamical Ginzburg-Landau treatment of

134: the coupled ferromagnetic, antiferromagnetic, and polarization orders,

135: in a cycloidal ground state and an applied AC electrical field.  It is

136: assumed that the system is far below its critical temperature so that

137: thermal fluctuations can be ignored in the dynamics.  The long-period

138: incommensurate cycloid is produced in our model by a Lifshitz term in

139: the effective Ginzburg-Landau free energy; the existence of such a

140: term was previously

141: argued~\cite{sparavigna94,sosnowska82,zalesski00,ruette04} as the

142: unique symmetry-allowed explanation for the observed order.  The model

143: free energy is

144: \begin{eqnarray}

145: F &=& {G L^4 \over 4} + {A L^2 \over 2} + {c \sum_i

146:   (\nabla L_i)^2 \over 2} \cr

147: && -\alpha \bm{P} \cdot \left[{\bf L} (\nabla \cdot {\bf L}) + {\bf L} \times

148: (\nabla \times {\bf L})\right] -\bm{P}\cdot \bm{E} \nonumber\\

149: &&+{r M^2\over 2}+ {a {P_z}^{2}\over 2}+{u{P_z}^{4}\over 4} + {a_\perp ({P_x}^2+{P_y}^2) \over 2} .

150: \end{eqnarray} 

151: Here $L=|\bm{M}_1-\bm{M}_2|$ is a N\'{e}el vector describing the

152: staggered sublattice magnetization, $M=|\bm{M}_1+\bm{M}_2|$ is the

153: total magnetization of the material, and $P_z$ is the magnitude of the

154: ferroelectric polarization along ${\bf \hat z}$ [the cubic (111) and

155: equivalent directions in BiFeO$_3$].  Many possible terms have been

156: omitted from the free energy as absent or unimportant in BiFeO$_3$.

157: Note that in the absence of the Lifshitz term, the AFM and FE orders

158: are decoupled, and the ground state is a commensurate, isotropic

159: Heisenberg antiferromagnet with $|{\bf L}_0|^{2}= {- A \over G}$ and

160: an easy-axis ferroelectric with uniform polarization given by

161: $\bm{P}_0=P_0 {\bf \hat{z}}$, $P_{0}^{2}= -a/u$.  Clearly $A < 0$ and

162: $r>0$ for an antiferromagnetic ground state, while $a<0$ and

163: $a_{\perp}>0$ for a ferroelectric ground state.  

164: %

165: %RdS: Removed.

166: %

167: %A $(\nabla {\bf

168: %  P})^2$ term has been omitted because the corresponding velocity is

169: %negligible compared to the spin-wave velocity, and the only source of

170: %antiferromagnetic anisotropy is the crystal distortion represented by

171: %${\bf P}$.

172: 

173: \begin{figure}

174:   \includegraphics[width=3in]{cycloid}

175: \caption{(color online). Antiferromagnetic lattice with harmonic cycloid order. 

176:   The cycloid plane is pinned by the polarization $\bm{P}_0$, with

177:   arbitrary propagation direction ${\bf \hat{q}}$ in the plane

178:   perpendicular to $\bm{P}_0$. We also depict the tangent vector ${\bf

179:     \hat{D}}$ related to phase fluctuations of the cycloid.}

180:   \label{fig1} \end{figure} 

181: 

182: The Lifshitz term induces an incommensurate cycloidal

183: order in which the antiferromagnetic moment rotates in an arbitrary

184: plane including ${\bf \hat z}$.  It also increases the magnitude of

185: $L_0$ and $P_0$.  Our reference ground state is a

186: cycloid with 

187: \begin{equation} 

188: {\bf L}_0(x) = L_0 \left[\cos(q x) {\bf \hat z}

189:   + \sin(qx) {\bf \hat x}\right].  

190: \label{l0}

191: \end{equation} 

192: Here the cycloid direction has been chosen along ${\bf \hat x}$.  The

193: pitch is $q = \alpha P_0 / c$ and the magnitudes of antiferromagnetic

194: and polar order are 

195: $L_{0}^{2} = (-A+\alpha^2 P_{0}^{2}/c)/G$,

196: $P_{0}^{2}=(-a+\alpha^2 L_{0}^{2}/c)/u$.  A check on the above is that

197: there are two zero-energy symmetry actions for fixed ${\bf P}$:

198: changing the cycloid phase and rotating the cycloid direction

199: in the $xy$ plane.  The displacement for the phase change from the

200: transformation $qx\rightarrow qx + \phi/L_0$ is

201: \begin{equation}

202: \delta \bm{L}=\phi \left[\cos(qx){\bf \hat{x}} -\sin(qx){\bf \hat{z}}\right].

203: \end{equation}

204: We can use the shorthand ${\bf \hat D}(x)$ for this spin vector

205: direction transverse to the local ground-state spin direction, and

206: write $\delta {\bf L} = \phi {\bf \hat D}(x)$ for the phase change.

207: The displacement for the cycloid direction change due to an

208: infinitesimal rotation $q{\bf \hat{x}}\rightarrow q({\bf \hat{x}}+\eta

209: {\bf \hat{y}})$ is 

210: $\delta {\bf L} = \eta L_0 q y {\bf

211:   \hat D}(x) + \eta L_0 \sin (q x) {\bf \hat y}$. This cycloid

212: direction symmetry is similar to the one found in smectic liquid

213: crystals in that it requires large displacements relative to the

214: original state at some point in the crystal.  

215: 

216: The linearized equations of motion for the ferroelectric and

217: antiferromagnetic order parameters are

218: \begin{subequations}

219: \begin{eqnarray}

220: \frac{\partial^{2} \bm{L}}{\partial t^{2}}

221: %&=& 

222: %r(\gamma L_0)^{2}{\bf \hat{L}}_{0}\times

223: %\left({\bf \hat{L}}_{0}\times \frac{\delta F}{\delta \bm{L}}

224: %\right)\cr

225: &=&-r(\gamma L_0)^2 \left[\frac{\delta F}{\delta \bm{L}}-\left(

226: {\bf \hat{L}}_{0}\cdot 

227: \frac{\delta F}{\delta \bm{L}}

228: \right)

229: {\bf\hat{L}}_{0}

230: \right],\label{ll}\\

231: \frac{\partial^{2} \bm{P}}{\partial t^{2}}&=&-f \frac{\delta F}{\delta \bm{P}}.\label{debye}

232: \end{eqnarray}

233: \end{subequations}

234: Here $\gamma$ is a gyromagnetic ratio with dimensions of (sG)$^{-1}$,

235: while $f$ has dimensions of s$^{-2}$ and plays a similar role in the

236: ferroelectric equation of motion. In the above polarization equation, we have

237: ignored damping and the possible Poisson-bracket term between ${\bf P}$ and

238: ${\bf M}$~\cite{bakerjarvis01}. The coupled spin and polarization equations are solved in terms of the parametrization

239: \begin{subequations}

240: \begin{eqnarray}

241: \delta \bm{L}&=& \phi(\bm{r}) \textrm{e}^{-i\omega t}{\bf \hat{D}}(x) 

242: +\psi(\bm{r}) \textrm{e}^{-i\omega t} {\bf \hat{y}},\\

243: \delta \bm{P}&=& \delta \bm{p}(\bm{r}) \textrm{e}^{-i\omega t}.

244: \end{eqnarray}

245: \end{subequations}

246: The field $\phi(\bm{r})$ denotes phase fluctuations of the cycloid

247: ground state, while $\psi(\bm{r})$ refers to spin fluctuations out of

248: the cycloid ($xz$) plane. The field

249: $\delta\bm{p}(\bm{r})=\bm{P}-\bm{P}_0$ denotes optical phonon

250: fluctuations related to longitudinal and transverse vibrations of the

251: ferroelectric moment $\bm{P}$ \footnote{While there are many optical

252:   phonons in a ferroelectric, this field $\bm{P}$ should be thought of

253:   as describing the particular polar phonon that goes soft at the

254:   ferroelectric transition.}.

255: 

256: %The required derivative of the free energy, to linear order in the

257: %displacements, is 

258: %\begin{eqnarray}

259: %\frac{\delta F}{\delta \bm{L}} &=& -c\nabla^2 \phi {\bf \hat{D}}-2cq(\partial_y \psi) {\bf \hat{z}}

260: %+\left[\left(cq^2 -c\nabla^2\right)\psi -2cq\sin(qx)(\partial_y\phi)\right]{\bf \hat{y}} \nonumber\\

261: %&&-\alpha L_0 \left[2q\cos(qx)(\delta \bm{p}_y) - {\bf \hat{L}}_0 \times \left(\nabla \times \delta \bm{p}\right)

262: %\right]+(\cdots){\bf \hat{L}}_0.\label{dfdl}

263: %\end{eqnarray}

264: \begin{widetext}

265: Calculating $\delta F/\delta \bm{L}$ to linear order in the

266: displacements, and inserting into Eq.~(\ref{ll}) leads to

267: \begin{subequations}

268: \begin{eqnarray}

269: \left[\omega'^2+c\nabla^2\right]\phi&=&2cq\sin(qx)(\partial_y \psi)

270: -\alpha L_0 \left(\nabla\times \delta\bm{p}\right)

271: \cdot{\bf \hat{y}},

272: \label{phieqn}\\

273: \left[\omega'^2-cq^2+c\nabla^2\right]\psi&=&

274: -2cq\sin(qx) (\partial_y\phi)

275: -\alpha L_0 \left[

276: 2q\cos(qx)(\delta p_y) -\left(\nabla\times \delta\bm{p}\right)\cdot {\bf \hat{D}}

277: \right],\label{psieqn}

278: \end{eqnarray}

279: with $\omega'=\omega /(\gamma L_0 \sqrt{r})$ defining the frequency in

280: units of $\sim 10^{12}$~rad/s.  The equation of motion for the

281: polarization becomes

282: \begin{equation}

283: \xi \omega'^2 (\delta\bm{p})=\frac{\delta F}{\delta \bm{P}}=

284: a_{\parallel}(\delta\bm{p}_z) +a_{\perp}(\delta\bm{p}_{\perp})

285: -\alpha L_0 \left[

286: q\cos(qx)(\psi) {\bf \hat{y}} +(\bm{\nabla} \phi) \times {\bf \hat{y}}-(\bm{\nabla}\psi) \times {\bf \hat{D}}

287: \right]-\bm{E},\label{dpeqn}

288: \end{equation}

289: \end{subequations}

290: where we define $\xi = r(\gamma L_0)^2/f$  ($\sim 10$ in BiFeO$_3$),

291: $a_{\parallel}=-2a+3\alpha^2L_{0}^{2}/c$, and the AC electric field is

292: $\bm{E}\textrm{e}^{-i\omega t}$.

293: \end{widetext}

294: 

295: Some intuition for Eqs.~(\ref{phieqn})-(\ref{dpeqn}) may be gained by

296: considering their invariance

297: under rotations of the cycloid in the $xy$ plane. Consider the

298: $\omega=0$ symmetry operation $\phi\rightarrow \eta q y $, $\psi

299: \rightarrow \eta \sin(qx)$, with $\delta\bm{p}$ left unchanged.  This

300: creates no terms in both sides of Eq.~(\ref{phieqn}). In

301: Eq.~(\ref{psieqn}), the left hand side becomes $-2cq^2\eta \sin(qx)$,

302: which cancels the term generated in the right hand side due to

303: the $y$-dependence in $\phi$.  For the square brackets in

304: Eq.~(\ref{dpeqn}), the first term adds $q\cos(qx)[\eta\sin(qx)]{\bf

305:   \hat{y}}$, with the third term $-\bm{\nabla}[\eta\sin(qx)]\times{\bf

306:   \hat{D}}$ giving the desired cancellation.

307: 

308: In the commensurate limit, $\alpha,q \rightarrow 0$ and we recover two

309: transverse AFM fluctuation modes with $\omega'^2=ck^2$.  There are two

310: dispersionless polarization wave modes transverse to $\bm{P}_0$ with

311: frequency $\omega'^2=a_{\perp}/\xi$, and one longitudinal mode with

312: frequency $\omega'^2=a_{\parallel}/\xi$.  The combination orthogonal

313: to $\phi{\bf \hat{D}}$, and $\psi {\bf \hat{y}}$, $S = {\bf \hat{L}}_0

314: \cdot {\delta \bm{L}}$, is constant in time in the equation of motion

315: above but decays dissipatively if non-Poisson bracket terms are kept.

316: It is not expected to modify the above propagating modes.  For $k

317: \rightarrow 0$ in the $\phi$ mode, we obtain a symmetry (the phase

318: shift) which should be nondissipative.  For $k \rightarrow 0$ in the

319: $\psi$ mode, there should be some finite dissipation.

320: 

321: We solve the system Eqs.~(\ref{phieqn})-(\ref{dpeqn}) using the ansatz

322: \begin{equation}

323: [\phi(\bm{r}),\psi(\bm{r}),\delta\bm{p}(\bm{r})]= 

324: \sum_{n} [\phi_n,\psi_n,(\bm{p})_n] \textrm{e}^{inqx}\textrm{e}^{i\bm{k}\cdot \rm{r}},

325: \end{equation}

326: %\begin{subequations}

327: %\begin{eqnarray}

328: %\phi(\bm{r})&=& \sum_{n} \phi_n \textrm{e}^{inqx}\textrm{e}^{i\bm{k}\cdot \rm{r}},\\

329: %\psi(\bm{r})&=& \sum_{n} \psi_n \textrm{e}^{inqx}\textrm{e}^{i\bm{k}\cdot \rm{r}},\\

330: %\delta \bm{p}(\bm{r})&=& \sum_{n} (\bm{p})_n \textrm{e}^{inqx}

331: %\textrm{e}^{i\bm{k}\cdot \rm{r}},

332: %\end{eqnarray}

333: %\end{subequations}

334: with $n$ an integer running from $-\infty$ to $\infty$ and the

335: restriction $|k_x|<q/2$. The complex numbers $\phi_n$, $\psi_n$, $(p_i)_n$ are

336: independent of $\bm{r}$ but have a $\bm{k}$ dependence.

337: Substitution of this ansatz into Eqs.~(\ref{phieqn})-(\ref{dpeqn}) yields a

338: linear system of equations whose characteristic polynomial determines

339: the resonance frequencies. 

340: 

341: Consider the unmixed spin waves in the limit

342: $(\bm{p})_n\rightarrow 0$. Substitution of the ansatz into

343: Eqs.~(\ref{phieqn}),~(\ref{psieqn}) gives

344: \begin{subequations}

345: \begin{eqnarray}

346: \left[ \omega'^2-c\tilde{k}^2\right]& \phi_n & - c k_y\left(\psi_{n-1}-\psi_{n+1}\right)=0,

347: \label{phin}\\

348: \left[\omega'^2 -c(\tilde{k}^2+q^2)\right] &\psi_n & + c k_y\left(\phi_{n-1}-\phi_{n+1}\right)=0,

349: \label{psin}

350: \end{eqnarray}

351: \end{subequations}

352: with $\tilde{k}^2=(k_x+nq)^2+k_{y}^{2}+k_{z}^{2}$.  We see that modes

353: propagating along the cycloid plane (with $k_y=0$) are simple plane

354: waves with cyclon ($\phi$) and out of plane ($\psi$) dispersions given

355: by $\omega'^2=c\tilde{k}^2$ and $\omega'^2=c(\tilde{k}^2+q^2)$

356: respectively. As expected from symmetry, the cyclon mode remains soft,

357: but the out of plane mode $\psi$ acquires a gap due to the pinning of

358: the cycloid plane by the ferroelectric moment. Note that the $\psi_0$

359: gap equals the cyclon energy at $n=\pm 1$. At $\bm{k}=0$, the AFM

360: resonance modes are simply $\omega'_{n}=\sqrt{c}q|n|$ and

361: $\omega'_{n}=\sqrt{c}q\sqrt{n^2+1}$ for the cyclon and out of plane

362: excitations respectively.  For BiFeO$_3$, $cq^2\sim 1$

363: \cite{sosnowska82,zalesski00}, and the modes are equally spaced by approximately

364: $10^{12}$~rad/s intervals ($0.16$~THz or $10$~K).

365: 

366: If the propagation vector has a small projection along the $k_y$

367: direction, the dispersion curves in the cycloid plane will repel each

368: other whenever they intersect. This leads to a series of small gaps

369: (anticrossings) in the propagation frequency as a function of the

370: reduced $k_x$, similar to the effect discussed by Bar'yakhtar in an

371: helical ferromagnet \cite{baryakhtar70}. Apart from these small gaps,

372: the high-frequency dispersion curves are nearly unaffected by the

373: incommensurate order.  A very different situation is found for the

374: low-frequency modes along the non-trivial $y$ direction. Consider for

375: example the lowest frequency mode $\phi_0$. Solving

376: Eqs.~(\ref{phin}),~(\ref{psin}) within ${\cal O}(k^{6})$ leads to

377: \begin{equation}

378: \omega'^2\approx c\left[k_{x}^{2}+k_{z}^{2}+\frac{3}{8}\frac{k_{y}^{4}}{q^2}

379: -\frac{k_{x}^{2}k_{y}^{2}}{q^2}\right].

380: \end{equation}

381: This soft mode dispersion is strongly anisotropic, and may be useful

382: for electrical control of the spin wave group velocity via

383: switching of the $\bm{P}_0$ direction. A similar effect is found for

384: phase fluctuations in smectic liquid crystals, and in an helical

385: ferromagnet such as MnSi \cite{kirkpatrick05}. The anisotropy is

386: related to the $\bm{q}$ rotation symmetry discussed above, which

387: forbids a $k_{y}^{2}$ term in the dispersion.

388: 

389: \begin{figure}

390:   \includegraphics[width=3in]{spectra} \caption{(color online). Spin

391:     wave spectra for a cycloidal multiferroic including one optical

392:     phonon mode (not shown).  The solid lines are for propagation along the

393:     non-trivial direction perpendicular to the cycloid plane ($y$ in

394:     Fig.~1). The dashed lines are the spectra for propagation along

395:     $z$, including zone folding.  The magnon modes at $nq$ are

396:     labelled by $\phi_n$ (cyclon) and $\psi_n$ (out of plane).}

397:   \label{fig2} \end{figure} 

398: 

399: A full numerical solution for propagation along $y$ is shown in

400: Fig.~\ref{fig2}. Note that in addition to the soft mode anisotropy

401: effect, $k_y\neq 0$ admixes the modes at $nq$ and $-nq$, splitting

402: their dispersions. Moreover, the second lowest frequency mode

403: (connected to $\phi_{\pm 1}$) acquires a negative group velocity.

404: 

405: The full solution of Eqs.~(\ref{phieqn})-(\ref{dpeqn}) with

406: $(\delta\bm{p})\neq 0$ shows avoided crossings between the optical phonon dispersion

407: and magnon branches at finite $k$. At these

408: anticrossings the spin wave excitations are highly mixed with the

409: polarization wave. However, an optical experiment only probes modes at

410: $\bm{k} \approx 0$ due to the large value of the velocity of light when

411: compared to the magnon velocity ($\sim 10^{6}$~cm/s). One can ask

412: whether any $n>0$ mode responds to optical excitation

413: with wavelength much larger than $2\pi/q\approx 600$~\AA. The answer

414: is affirmative as now shown by calculation of

415: the AC electric susceptibility.

416: 

417: An AC electric excitation $\bm{E}\textrm{e}^{-i\omega t}$ may excite

418: spin waves due to the appearence of magnetoelectric fields

419: such as $2q \cos{(qx)}(\delta p_y)$ 

420: on the right hand side

421: of Eq.~(\ref{psieqn}). This couples $\psi$ to an optical phonon at $k=0$.

422: %There are five coupled equations of motion for spin and polarization

423: %waves with an AC electric excitation $\bm{E}\textrm{e}^{-i\omega t}$.

424: %One equation couples $\psi$ to the polarization; at $\bm{k}=0$,

425: %\begin{eqnarray}

426: %%\left(\omega'^2-cq^2n^2\right)\phi_n&=&i\alpha qL_0n(p_z)_n,\\

427: %\left[\omega'^2-cq^2(n^2+1)\right]\psi_n&=&-\frac{\alpha q L_0}{2}\left[(n+1)(p_y)_{n-1}\right.\nonumber\\

428: %&&\left.-(n-1)(p_y)_{n+1}\right].

429: %%\left(\xi \omega'^2-a_{\perp}\right)(p_x)_n&=&-E_x \delta_{n0},\\

430: %%\left(\xi \omega'^2-a_{\perp}\right)(p_y)_n&=&-E_y \delta_{n0}-\frac{\alpha q L_0}{2}\left[

431: %%(n+2)\psi_{n+1}\right.\nonumber\\&&\left.-(n-2)\psi_{n-1}\right],\\

432: %%\left(\xi \omega'^2-a_{\parallel}\right)(p_z)_n&=&-E_z \delta_{n0}-i\alpha q L_o n\phi_n.

433: %\end{eqnarray}

434: %The AC electrical susceptibility tensor is defined as

435: %$(p_i)_0=\chi_{ij}E_{j}$. 

436: Solving the system of equations shows that the AC electrical susceptibility tensor 

437: $(p_i)_0=\chi_{ij}E_{j}$ is diagonal for our model, with $\chi_{xx}=-1/(\xi

438: \omega'^2-a_{\perp})$, $\chi_{zz}=-1/(\xi \omega'^2-a_{\parallel})$.

439: Therefore the optical response with light polarized within the cycloid

440: plane contains no magnetic anomaly but only simple poles at the bare

441: optical phonon frequencies. However, if the light is polarized along

442: the direction perpendicular to the cycloid plane, the situation is quite different,

443: \begin{equation}

444: \chi_{yy}(\omega)=

445: %\frac{-1}{(\xi \omega'^2-a_{\perp})-\frac{2(\alpha q L_0)^2}{(\omega'^2-2cq^2)}}=

446: \frac{-(\omega'^2-2cq^2)}{

447: (\omega'^{2}-\Omega'^{2}_{{\rm PH}})

448: (\omega'^{2}-\Omega'^{2}_{{\rm AFM}})}.

449: \end{equation}

450: This AC susceptibility has two poles at the following shifted phonon

451: and AFM ($\psi_{\pm1}$ mode) frequencies:

452: $\Omega'^{2}_{{\rm PH}}= a_{\perp}/\xi+2cq^2$,

453: $\Omega'^{2}_{{\rm AFM}}= 2cq^2-2(\alpha q L_0)^2/a_{\perp}$.

454: %where contributions of order ${\cal O}

455: %\left(\xi cq^2/a_{\perp}\right)^{2}$ were dropped.

456: 

457: Hence we conclude that the simplest model for BiFeO$_3$ already

458: contains a low-frequency magneto-optical resonance. Note the prominent

459: importance of magnon zone folding: Even though the AC electric field

460: is a zero wavevector excitation, it couples to a magnon at wavevector

461: $\pm q$. This is a special feature of electromagnetic excitations in

462: an incommensurate multiferroic material. 

463: 

464: The magnon excitations of a cycloid with wavevector $q$ may violate

465: momentum conservation by a multiple of $q$ (a reciprocal lattice

466: vector for the cycloid). In the case of a multiferroic material with

467: incommensurate magnetic order, this allows the possibility of

468: observing additional magneto-dielectric resonances due to the coupling

469: of a zero-wavevector phonon to magnon modes at other multiples of

470: $q$.    To see this effect, consider a contribution to the free energy in the

471: form of an easy-plane anisotropy

472: $\frac{\lambda}{2}\left(\bm{P}\cdot \bm{L}\right)^2$,

473: which is known to add anharmonicity to the cycloid order

474: \cite{sparavigna94}. To first order in $\lambda$, this term adds a third

475: harmonic to Eq.~(\ref{l0}). This gives an additional magnetoelectric

476: contribution to the right side of the equation determining the

477: cyclon field $\phi$ [Eq.~(\ref{phieqn})]:

478: $\lambda P_0 L_0/2 \left[\cos{(2qx)}(\delta p_x) - \sin{(2qx)}(\delta p_z)\right]$.

479: %\begin{equation}

480: %\frac{\lambda P_0 L_0}{2} \left[\cos{(2qx)}(\delta p_x) - \sin{(2qx)}(\delta p_z)\right].

481: %\end{equation}

482: As a result, the AC electric susceptibilities $\chi_{xx},\chi_{zz}$

483: develop poles at the magnon modes $\phi_{\pm 2}$. Similarly, the out

484: of plane field $\psi$ is excited by a term of the form $\cos{(3qx)}

485: (\delta p_y)$, showing that $\chi_{yy}$ has an additional

486: electromagnon pole at the mode $\psi_{\pm 3}$.  Considering additional

487: powers of $\lambda$ (higher harmonics of the cycloid) shows that

488: higher frequency resonances at $nq$ 

489: are present, with strength falling off as

490: higher powers of $\lambda$.

491: 

492: In conclusion, we determined the electromagnon spectra for a class of

493: cycloidal multiferroics including bulk BiFeO$_3$.  A simple

494: Landau-Ginzburg model shows that a low-frequency optical probe will

495: couple to magnons at multiples of the cycloid wave vector, leading to

496: a series of magneto-dielectric resonances at integer $n$ and

497: irrational $\sqrt{1+n^2}$ multiples of a fundamental cyclon frequency

498: ($\phi_n$ and $\psi_n$ magnons respectively).  Interestingly, only one

499: mode ($\psi_{\pm 1}$) becomes electric dipole active due to the strong

500: linear magnetoelectric effect driving the harmonic cycloid order. The

501: remaining modes acquire a small electric dipole character due to

502: weaker quadratic magnetoelectric effects and anharmonicity in the

503: cycloid order. As a result, optical experiments such as

504: transmittivity, reflectivity, and Raman spectroscopy will reveal much

505: more about the magnetic order of multiferroics than previously

506: expected on the basis of homogeneous models.  For example, the

507: direction dependence and intensity of higher resonances is directly

508: related to anharmonicity in the magnetic order.  Other incommensurate

509: multiferroics may show similar phenomena even if the details of the

510: multiferroic coupling are different.

511: 

512: The authors acknowledge useful conversations with S. Mukerjee, J.

513: Orenstein, R. Ramesh, I. Souza, N. Spaldin, and A. Vishwanath. This

514: work was supported by WIN (RdS) and by NSF DMR-0238760 (JEM).

515: 

516: \begin{thebibliography}{99}

517: 

518: %Giant magnetoelectric effect in BiFeO3 heterostructures

519: \bibitem{wang03} J. Wang {\it et al.}, Science {\bf 299}, 1719 (2003).

520: 

521: %Experiment claiming magnetic control of ferroelectricity in RMnO_3

522: \bibitem{kimura03} T. Kimura {\it et al.}, Nature {\bf 426}, 55 (2003).

523: 

524: %Optical experiments in RMnO_3 (low temperature electromagnon)

525: \bibitem{pimenov06} A. Pimenov {\it et al.}, Nature Physics {\bf 2}, 97 (2006).

526: 

527: %Optical experiments in RMnO_5 (low temperature electromagnon)

528: \bibitem{sushkov07} A.B. Sushkov {\it et al.}, \prl {\bf 98}, 027202 (2007). 

529: 

530: %Landau-Ginzburg for multiferroics

531: \bibitem{harris06} A. B. Harris, \prb {\bf 76}, 054447 (2007).

532: 

533: %Electrical control of magnetic order in BiFeO3

534: \bibitem{zhao06}  T. Zhao {\it et al.}, Nature Materials  {\bf 5}, 823 (2006).

535: 

536: %ESR experiment confirming the cycloid to homogeneous phase transition at high B

537: \bibitem{ruette04} B. Ruette {\it et al.}, \prb {\bf 69}, 064114 (2004).

538: 

539: %RdS: Added additional reference for film, very useful paper. 

540: %Destruction of Cycloid in (111) thin films.

541: \bibitem{bai05} F. Bai {\it et al.}, Appl. Phys. Lett. {\bf 86},

542:   032511 (2005);  H. B\'{e}a {\it et al.}, Phil. Mag. Lett. {\bf 87},

543:   165 (2007).

544: 

545: % Raman and reflectivity experiments in THz region for bulk BFO

546: \bibitem{haumont06} R. Haumont, J. Kreisel, P. Bouvier, and F.

547:   Hippert, \prb {\bf 73}, 132101 (2006); S. Kamba {\it et al.}, \prb

548:   {\bf 75}, 024403 (2007).

549: 

550: % spin-wave devices

551: \bibitem{khitun05} A. Khitun and K. L. Wang, Superlattices and

552:   Microstructures {\bf 38}, 184 (2005).

553: 

554: %Theory of optical phonons in bulk BFO

555: \bibitem{hermet07} P. Hermet, M. Goffinet, J. Kreisel, and Ph. Ghosez,

556:   \prb {\bf 75}, 220102(R) (2007).

557: 

558: %Original work on electromagnons

559: \bibitem{baryakhtar69} V.G. Bar'yakhtar and I.E. Chupis, Sov. Phys.

560:   Solid State {\bf 10}, 2818 (1969); {\it ibid} {\bf 11}, 2628 (1970).

561: 

562: %AC susceptibilities in magnetoelectric BaMnF_4

563: \bibitem{tilley82} D.R. Tilley and J.F. Scott, \prb {\bf 25}, 3251

564:   (1982); G.A. Maugin, \prb {\bf 23}, 4608 (1981).

565: 

566: %Landau theory for spiral induced ferroelectricity

567: \bibitem{mostovoy06} M. Mostovoy, \prl {\bf 96}, 067601 (2006).

568: 

569: %Microscopic mechanism for spiral induced ferroelectricity; 

570: %Spin waves inducing ferroelectricity in RMnO_3

571: \bibitem{katsura05} H. Katsura, N. Nagaosa, A.V. Balatsky, \prl {\bf 95}, 057205 (2005);

572: {\it ibid.} {\bf 98}, 027203 (2007).

573: 

574: \bibitem{betouras07} J.J. Betouras, G. Giovannetti, and J. van den

575:   Brink, \prl {\bf 98}, 257602 (2007).

576: 

577: %Theory for the cycloid

578: \bibitem{sparavigna94} A. Sparavigna, A. Strigazzi, and A. Zvezdin, \prb {\bf 50}, 2953 (1994).

579: 

580: %Original neutron scattering experiment reporting a cycloid order in BiFeO3 

581: \bibitem{sosnowska82} I. Sosnowska, T. Peterlin-Neumaier, and E.

582:   Steichele, J. Phys. C: Solid State Phys. {\bf 15}, 4835 (1982).

583: 

584: %NMR reporting cycloid anharmonicity as a function of temperature

585: \bibitem{zalesski00} A.V. Zalesski\u{i}, A. K. Zvezdin, A.A. Frolov, and A.A. Bush, JETP Lett.

586:   {\bf 71}, 465 (2000).

587: 

588: %Poisson brackets between P and M

589: \bibitem{bakerjarvis01} J. Baker-Jarvis and P. Kabos, \pre {\bf 64}, 056127 (2001).

590: 

591: %Original spin wave work on a spiral ferromagnet (no magnetoelectric coupling!)

592: \bibitem{baryakhtar70} V.G. Bar'yakhtar and E.P. Stefanovskii, Sov. Phys. Solid State {\bf 11}, 1566 (1970).

593: 

594: %Soft mode anisotropy in a spiral ferromagnet in the context of observed non-Fermi liquid behavior in MnSi

595: \bibitem{kirkpatrick05} T. R. Kirkpatrick and D. Belitz, \prb {\bf 72}, 180402(R) (2005).

596: 

597: \end{thebibliography} 

598: 

599: \end{document}