1: \documentclass[prl,aps,floatfix,twocolumn,superscriptaddress]{revtex4}

2: 

3: %\documentclass[prl,aps,floatfix,11pt,superscriptaddress]{revtex4}

4: %%%%%%%%%%%%

5: \usepackage{amsmath}

6: \usepackage{graphicx}

7: 

8: \begin{document}

9: 

10: \title{THz emission from ultrafast optical orientation}

11: \author{F. Nastos}

12: \affiliation{Department of Physics and Institute for Optical

13: Sciences, University of Toronto, 60 St. George Street, Toronto,

14: Ontario, Canada M5S 1A7}

15: \author{R. W. Newson}

16: \affiliation{Department of Physics and Institute for Optical

17: Sciences, University of Toronto, 60 St. George Street, Toronto,

18: Ontario, Canada M5S 1A7}

19: \author{J. H{\"u}bner}

20: \affiliation{Institute for Solid State Physics, University of

21: Hannover, Appelstr. 2, 30167 Hannover, Hannover, Germany}

22: \author{J. E. Sipe}

23: \affiliation{Department of Physics and Institute for Optical

24: Sciences, University of Toronto, 60 St. George Street, Toronto,

25: Ontario, Canada M5S 1A7}

26: \author{H. M. van Driel}

27: \affiliation{Department of Physics and Institute for Optical

28: Sciences, University of Toronto, 60 St. George Street, Toronto,

29: Ontario, Canada M5S 1A7}

30: 

31: \begin{abstract}

32: We show both theoretically and experimentally that the magnetization

33: density accompanying ultrafast excitation of a semiconductor with

34: circular polarized light varies rapidly enough to produce a

35: detectable THz field.

36: \end{abstract}

37: 

38: \maketitle

39: 

40: A host of recent fundamental studies on the behavior of spins in

41: semiconductors have been motivated by the idea of using the spin

42: degree of freedom in information processing, the goal of a field

43: that has come to be known as \textit{spintronics}.  Central to many

44: of these investigations has been the optical injection of spin

45: polarized electrons and holes into the band structure of a

46: semiconductor.  These spin polarized carriers are then available to

47: be dragged by a bias voltage, for example, to regions of interest in

48: any proposed device structure.  Such optical injection of spin

49: polarized carriers, also called \textit{optical orientation}, can be

50: effected by the absorption of circularly polarized light

51: \cite{DyakanovPerel,ZuticRMP04}, which takes advantage of the

52: selection rules to populate particular angular momentum states of

53: the band structure.

54: 

55: In the analysis of most of these experiments it is assumed that the

56: holes can be neglected, because in bulk semiconductors the hole spin

57: lifetimes are typically much shorter than the electron spin

58: lifetimes. This makes it difficult, if not impossible, to measure

59: the optical orientation of hole spins in bulk semiconductors using

60: usual techniques employed to study electron spin orientation.

61: However, Hilton and Tang \cite{HiltonPRL02} recently showed that

62: pump-probe techniques could be used to probe the hole spin dynamics

63: in bulk GaAs.  They found a time of $\tau_h = 110\,\text{fs}$ for

64: the hole spin relaxation time, which has spurred theoretical

65: investigations \cite{YuPRB05}.  All such techniques measure optical

66: orientation {\em indirectly} by monitoring, after excitation, the

67: population in various bands, or its effect on luminescence or

68: Faraday rotation \cite{OestrichSST17,AwschalomPT52}.

69: 

70: In this manuscript we present an approach for the \textit{direct}

71: measurement and study of ultrafast optical orientation.  While we

72: focus on GaAs, the technique is applicable to a wide range of

73: semiconductors.  It relies on the magnetization that accompanies the

74: injected spin density in optical orientation: Under ultrafast

75: excitation this magnetization varies rapidly over a sub-picosecond

76: timescale, and radiates in the terahertz (THz) regime.  Such

77: radiation has already been used to study a host of other ultrafast

78: processes

79: \cite{BeaurepaireAPL04,SchmuttenmaerCR04,JohnstonPRB02,CoteAPL02},

80: and we show here that detailed measurement and analysis of the

81: emitted THz electric field trace can be used to study the electron

82: and hole spin dynamics.

83: 

84: To estimate the THz field strength radiated from ultrafast optical

85: orientation, we compare the injected magnetization source $M(t)$ in

86: optical orientation to the injected polarization source $P(t)$ in

87: the shift current (also called the intrinsic photovoltaic effect

88: \cite{vonBaltzPRB81}).  When a semiconductor is excited by photons

89: with energies above the band gap, electrons and holes are injected

90: at a rate that closely follows the temporal pulse profile

91: $\mathcal{I}(\mathbf{r},t)$, which is the intensity profile of the

92: laser pulse in the sample.  The carrier density injection rate

93: $\dot{n}$ is given by $\dot{n}(\mathbf{r},t)=\xi_\omega

94: \mathcal{I}(\mathbf{r},t)$, where $\omega$ refers to the carrier

95: frequency of the laser, and $\xi_\omega$ is proportional to the

96: absorption coefficient. The spatial coordinate $\mathbf{r}$ accounts

97: for the variation from the exponential decay into the surface, and

98: the variation in intensity across the spot-size.  The polarization

99: and magnetization sources we consider both arise from the carrier

100: injection, although they have different symmetry characteristics and

101: underlying microscopic origins.

102: 

103: In III-V semiconductors, the shift current appears for excitation by

104: certain {\em linear} polarizations of the field

105: \cite{noteAboutPGEffect}.  In this effect the center of charge of

106: the carriers moves as the carriers are promoted from the valence

107: band to the conduction band, and a net current directly results,

108: \begin{equation}

109: \dot{P}_{\text{inj}}(t) = p\dot{n}(\mathbf{r},t), \label{pdot}

110: \end{equation}

111: where $p=ed$ is the average injected dipole moment per carrier.

112: Here, $e$ is the electron charge and the displacement $d$ is on the

113: order of a Bohr radius $a_\text{B}$.  Detailed calculations show

114: that for GaAs under excitation with linear polarized light along

115: [110] this distance $d$ is very close to the GaAs bond length,

116: $d=2.54\,a_\text{B}$ \cite{NastosPRB06}. Since the polarization

117: $P_{\text{inj}}(t)$ rises on a subpicosecond time-scale, it radiates

118: in the THz regime \cite{neglectOR}. However, the polarization

119: survives for much longer than a picosecond, and so its decay is not

120: expected to radiate in the THz regime. Thus for the polarization

121: that contributes to the far field radiation we have $\ddot{P}(t) =

122: \ddot{P}_{\text{inj}}(t)$.

123: 

124: Optical orientation arises from {\em circularly} polarized

125: excitation of carriers near the band edge.  Because of spin-orbit

126: splitting these carriers can have a net spin polarization, and

127: accompanying this spin density is a magnetization density

128: $M(\mathbf{r},t)$.  For optically injected electrons, close to the

129: conduction band edge, the injected magnetization density is given by

130: \begin{equation}

131: \dot{M}_{e,\text{inj}}(\mathbf{r},t) = \bar\mu_e

132: \dot{n}(\mathbf{r},t), \label{mdot}

133: \end{equation}

134: where $\bar{\mu}_e$ is the mean magnetic moment of an injected

135: electron.  Unlike for free electrons -- where the magnetic moment of

136: an electron is simply $\mu = g\mu_{\text B} s/\hbar$, where

137: $\mu_{\text B}$ is the Bohr magneton, $s=\hbar/2$, and $g$ is the

138: free electron $g$-factor, $g=2.0023$  -- the average magnetic moment

139: of an injected electrons can be written as $\bar\mu_e = g_e^\ast

140: \mu_{\text B} S / \hbar$, where $g_e^\ast$ is an effective

141: $g$-factor of an electron, and $S$ is the injected degree of spin

142: polarization in units of $\hbar/2$.  The effective $g$-factor can be

143: calculated from Roth's expression for the magnetic moment of a Bloch

144: state in the $n$-th band~\cite{RothPR60},

145: \begin{equation}

146: \mu_n^z = - \frac{g \mu_\text{B} \sigma^z_n}{2}

147: -\frac{2}{mi}\sum_{s\ne n}

148: \frac{p^x_{ns}p^y_{sn}-p^y_{ns}p^x_{sn}}{\epsilon_{n}-\epsilon_{s}},

149: \label{Roth}

150: \end{equation}

151: and defining the effective $g$-factor {\it via} $g_e^\ast =

152: -2\mu_n/\mu_\text{B}$.  In Eq.~\eqref{Roth}, $\sigma^z_n$ is the

153: expectation value of the Pauli spin matrix in the $n$-th band,

154: $p^x_{ns}$ is the $x$ component of the momentum matrix element

155: between the bands $n$ and $s$, and $\epsilon_n$ is the energy of the

156: $n$-th band.  The $z$ direction is the direction along which the

157: magnetization is aligned. The appearance of $\sigma^z_n$ in the

158: first term is required to generalize the more common $g$-factor

159: expression \cite{ChadiPRB76} beyond $j=1/2$ states.  For optically

160: injected electrons in GaAs, close to the conduction band edge,

161: $g_e^\ast$ is $-0.44$ at low temperatures~\cite{HermannPRB77}.

162: 

163: This magnetization density $M_e(t)$ will vary on an ultrafast

164: time-scale with the laser pulse, and it will radiate in the THz

165: regime.  The radiation will be magnetic-dipole like, in contrast to

166: the electric-dipole like radiation from the shift current.  Since

167: the lifetime of the injected spins is known to be on the order of

168: many picoseconds~\cite{ZuticRMP04}, their decay will not contribute

169: to the THz and we have

170: $\ddot{M}_e(\mathbf{r},t)=\ddot{M}_{e,\text{inj}}(\mathbf{r},t).$

171: 

172: The far-field radiation from both magnetic and polarization sources

173: is dependent on the second derivatives of the sources, $\ddot{M}(t)$

174: and $\ddot{P}(t)$.  A simple estimate of the radiation from just the

175: electron magnetization ${M}_{e}(t)$ can be made by comparing this

176: source to the polarization induced by the shift current.  Since both

177: of these sources survive long enough that their decay profiles are

178: not expected to radiate in the THz regime, it is mainly the

179: generation (or turn on) of these sources that provides the THz

180: radiation.  The ratio of the peak strengths for the magnetization

181: and polarization sources are directly comparable, since the

182: different Fresnel coefficients governing the transmission of

183: radiation for each type of source only determine the radiation

184: patterns.  So to first order we can estimate the strength of the

185: magnetic radiation from the ratio $\ddot{M}_e(t) / \ddot{P}(t)$.

186: Using Eq.~(\ref{mdot}) and~(\ref{pdot}) gives

187: \begin{equation}

188: \frac{\ddot{M}_e}{\ddot{P}}= \frac{\bar\mu_e}{p} =

189: \frac{Sg_e^\ast\mu_{\text B}}{\hbar ed}=

190: \frac{(S/\frac{\hbar}{2})g_e^\ast}{4(d/a_\text{B})}\alpha,

191: \label{eq1}

192: \end{equation}

193: where $\alpha$ is the fine structure constant.  In GaAs optical

194: orientation gives 50\% spin polarization ($S=0.5 \hbar/2$), so the

195: ratio $\ddot M_e/\ddot P$ is roughly $1.6\times 10^{-4}$.   The THz

196: fields from the shift current have been measured to be about 1500

197: V/m, so an absolute sensitivity of about 0.2 V/m is needed to

198: measure the THz radiation from optical orientation of bulk GaAs

199: electrons alone.  This requirement is experimentally challenging

200: \cite{LloydHughesPRB04} but would not be impossible.  Note that this

201: is because the THz field amplitude, rather than the intensity, is

202: measured experimentally.  Were the intensity measured, the signal

203: from the magnetization effect would be down from that of the shift

204: current by a factor of $(\ddot M_e/\ddot P)^{2}$.

205: 

206: The relatively short lifetime of the holes, however, suggests that

207: the hole spins may radiate in the THz regime.  Including the decay

208: of the hole spins, the magnetization of the holes follows

209: \begin{equation}

210: \dot{M}_h(\mathbf{r},t) = \dot{M}_{h,\text{inj}}(\mathbf{r},t) -

211: \frac{{M}_h(\mathbf{r},t)}{\tau_h},

212: \end{equation}

213: where $\dot{M}_{h,\text{inj}}$ is the hole injection rate

214: \begin{equation}

215: \dot{M}_{h,\text{inj}}(\mathbf{r},t)=\bar\mu_h\dot{n}(\mathbf{r},t),

216: \end{equation}

217: analogous to the electron injection rate, but here $\bar\mu_h$ is

218: the mean magnetic moment of the injected holes.

219: 

220: On the one hand, the decay of the holes adds more temporal variation

221: to the magnetization, and enhances the THz signal;  on the other

222: hand, since the decay lowers the overall magnitude of the

223: magnetization, it is expected to decrease the THz signal. To include

224: both effects, we have solved for the far-field radiation from an

225: injected magnetization using a Green function formalism tailored to

226: planar interfaces \cite{SipeJOSAB87}.  We find that for pulse widths

227: of 100\,fs the effect of the hole spin decay

228: ($\tau_h=110\,\text{fs}$) is to decrease the peak THz radiation from

229: this source by roughly a factor of two.  That is, the hole spins

230: decay fast enough so that the decay process will contribute the THz

231: radiation, but slow enough that the overall signal from it is not

232: washed out.

233: 

234: Yet the relatively short lifetimes of holes, in bulk materials such

235: as GaAs, prevents the study of their dynamics by many other

236: techniques. Thus the hole $g$-factors, and their magnetic moments in

237: general, have been largely neglected in bulk materials.  In GaAs

238: based nanostructures however, hole $g$-factors have been the subject

239: of much recent attention \cite{PryorPRL06}. It is typically found

240: that in those systems that the holes have a $g$-factor many times

241: larger than the electron $g$-factor.  In this work however, we are

242: interested in the optically injected magnetic moments $\bar\mu_h$,

243: rather than the $g$-factors themselves.  The average magnetic moment

244: $\bar\mu_h$ will depend on the details of the populations injected

245: into the heavy-hole and light-hole bands. We calculate directly the

246: injected hole magnetization into these states, but for easy

247: comparison to the electron magnetization, we quote our result as a

248: hole $g$-factor $\tilde{g}_h^\ast$ relative to the injected electron

249: spin density, so that

250: \begin{equation}

251: \bar\mu_h = \frac{ \tilde{g}_h^{\ast} \mu_{\text B} S}{\hbar}.

252: \end{equation}

253: The tilde is to emphasize that we are not defining the $g$-factor

254: relative to the hole spins.  Following a similar approach as

255: in~\cite{BhatPRL05} we use Fermi's golden rule together with Roth's

256: formula Eq.~\eqref{Roth} for the magnetization, and use a 30-band

257: $\mathbf{k}\cdot\mathbf{p}$ band structure~\cite{RichardPRB04} to

258: evaluate $\dot{M}_{h,\text{inj}}$ directly.  We find that

259: $\tilde{g}^{\ast}_h = 14.1$, which implies that the hole

260: magnetization injection is approximately 32 times larger than the

261: electron magnetization injection and clearly dominates

262: it~\cite{LDAgFactor}.

263: 

264: Combining this with our calculations for the radiation emission

265: (following \cite{SipeJOSAB87}) we find that for GaAs the hole

266: radiation should be of the order $3\times 10^{-3}$, or about $1/300$

267: times the shift current radiation.

268: 

269: Despite the relatively large expected signal from the the holes, the

270: ratio of THz emission from magnetization to shift current is still

271: small, and one must choose an experimental geometry which minimizes

272: the shift current. Differences in the spatial symmetry

273: characteristics of shift current and magnetization sources are

274: exploited to distinguish the weak magnetization source from the much

275: stronger shift current source.

276: 

277: \begin{figure}

278: \includegraphics[scale=0.75]{expsetupa.eps} \\

279: \includegraphics[scale=0.75]{expsetupb.eps} \\

280: \caption{(color online) (a) Experimental setup. (b) Top view of

281: sample with epoxy hemisphere.  PEM: photoelastic modulator, HWP:

282: half waveplate, QWP: quarter waveplate, PD: photodiode

283: \label{expsetup}}

284: \end{figure}

285: 

286: Fig. \ref{expsetup} illustrates the experimental setup for observing

287: the transient THz radiation.  An 80\,MHz Ti:sapphire oscillator

288: delivers 1\,nJ, 100\,fs pulses at 800\,nm.  For this wavelength, the

289: injected kinetic energy of heavy (light) holes is 3.6\,meV

290: (9.5\,meV), much less than the 300\,meV needed to excite holes in

291: the spin-orbit split-off band.  The pulses are split into a pump

292: beam for sample excitation and a probe beam to sample THz radiation

293: via electro-optic sampling.  The pump beam is incident onto a

294: 100\,$\mu$m thick (110) GaAs sample with a spot diameter of

295: 25\,$\mu$m, producing a peak incident intensity of 500\,MW/cm$^2$

296: and a carrier density of $2\times 10^{18}\,\text{cm}^{-3}$ over an

297: absorption depth of about 1\,$\mu$m. Since the total magnetization

298: or electric polarization is determined by total number of carriers

299: excited, and not their density, there is no premium associated with

300: beam focusing.  The emitted THz radiation emitted along the surface

301: normal direction from the sample back side is collected with a f =

302: 50\,mm, $90^\circ$ off-axis parabolic and focused with the same type

303: of mirror onto a $500\,\mu$m thick (110) ZnTe electro-optic sampling

304: crystal. The $800\,\text{nm}$ probe pulse is spatially overlapped

305: with the THz field inside the ZnTe crystal and temporally scanned.

306: The pump beam polarization is controlled by half- and quarter-wave

307: plates (QWP) and a photo-elastic modulator (PEM). The half-wave

308: plate (HWP) is oriented to provide light linearly polarized

309: 45$^\circ$ from the $\mathbf{\hat{s}}$ polarization vector (the

310: vertical). The PEM operates in quarter-wave mode and modulates the

311: polarization state at a frequency of 42\,kHz, enabling lock-in

312: detection techniques.

313: 

314: A major difficulty in measuring the emitted THz is that the injected

315: magnetic dipoles are parallel to the laser beam propagation

316: direction.  For normal incidence excitation, this injects magnetic

317: dipoles that are perpendicular to the surface.  The component of the

318: emitted radiation {nearly} parallel to the surface is small, and so

319: there is little transmission.  In the shift current effect however,

320: the polarization is parallel to the surface, and this allows the

321: radiated THz field to easily transmit through the surface.  To

322: enhance the emission of THz from the ultrafast optical orientation

323: it is thus necessary to irradiate the sample at {a high angle of

324: incidence; we use} an angle of incidence of $55^\circ$, exciting

325: through an epoxy hemisphere (see Fig. \ref{expsetup}b), to give an

326: angle of refraction inside the GaAs, $\theta_R$, of about

327: $20^\circ$. Even for this small angle it follows from the Maxwell

328: {\it saltus} (boundary) conditions that the emitted THz radiation

329: mainly propagates in the direction of the surface normal, because

330: the induced electron-hole plasma is $< 1\,\mu\text{m}$ thick,

331: 

332: We define $\beta$ as the angle between the QWP's fast axis and

333: $\mathbf{\hat{s}}$.  The shift current and magnetization signals

334: both depend on $\beta$ in a characteristic and unique way, so we

335: focus on that variable to distinguish between the two THz sources.

336: If we define $\gamma$ as the angle between the [001] crystal axis

337: (lying in the plane of the sample) and $\mathbf{\hat{s}}$, the

338: internal THz field emitted by the two types of sources are given by

339: \begin{eqnarray}

340: \Delta E_{\text{THz}}^{\text{shift}} & = & C e d \cos{(\theta_R)}

341: (2 \sin{\gamma} - 3 \sin^3{\gamma}) \cos{(2 \beta)}, \\

342: \Delta E_{\text{THz}}^{\text{mag}} & = & C \mu_{\text{eff}}

343: \sin{(\theta_R)} \sin^2{(2 \beta)},

344: \end{eqnarray}

345: where the proportionality constant $C$ includes universal constants

346: and Fresnel reflection and transmission coefficients.  The effective

347: magnetic moment $\mu_\text{eff}$ describes the sum of electron and

348: hole magnetic moments, although we expect it to be dominated by hole

349: term.  It is evident that the shift current (magnetization) signal

350: is maximized (minimized) when $\beta = 0^\circ$ and minimized

351: (maximized) when $\beta = 45^\circ$. This is because when

352: $\cos(2\beta)=0$ ($\sin(2\beta)=0$), the light polarization is

353: modulating between $+45^\circ$ and $-45^\circ$ linear (right and

354: left circular) states.  We choose $\gamma = 0.02\,\text{rad}$

355: ($1.2^\circ$), for which $2\sin\gamma -3\sin^3\gamma\approx

356: 2\gamma$, in order to provide a small shift current source with

357: which to compare the magnetization signal.

358: 

359: Fig. \ref{expresults}a illustrates the expected variation of the THz

360: field strength from the shift and magnetization sources, taking the

361: ratio of the source strengths to be 0.015.  Other possible sources

362: of THz radiation were neglected.  Optical rectification is small

363: relative to the shift current effect above the band gap, and shares

364: the same symmetry dependence.  Sources of THz radiation related to

365: Dember fields have no dependence on crystal orientation or pump

366: polarization.  Fig. \ref{expresults}b shows results from

367: experimental measurements of the THz field strength as a function of

368: $\beta$. The THz field strength is defined as the peak-to-peak value

369: of the observed THz trace, although several other definitions of

370: signal strength were used and each resulted in a similar pattern as

371: a function of $\beta$.

372: 

373: \begin{figure}

374: \includegraphics[scale=0.5]{expresults.eps}

375: \caption{(color online) (a) Simulation of THz signal vs QWP angle

376: $\beta$. red dashed: shift current signal, blue thin: magnetization

377: signal, black thick: total signal (b) Experimental results (points)

378: with Fourier series fit (curve) \label{expresults}}

379: \end{figure}

380: 

381: The deviation of the expected sinusoidal pattern created by the

382: shift current alone is attributed to the signal from the

383: magnetization source.  The “zero crossings” of the Fourier series

384: fit to the data of Fig. \ref{expresults}b are separated by more than

385: $90^\circ$. This behavior indicates the presence of an additional

386: THz source with a $\beta$-dependence as shown in Fig.

387: \ref{expresults}a. The difference between maximum and minimum signal

388: values recorded at $\beta=90^\circ$ and $0^\circ$ respectively was

389: $0.05\pm 0.14\,\mu$V. This difference would have had to be on the

390: order of $1\,\mu\text{V}$ to attribute the data's deviation to a

391: uniform vertical shift. Experimental data indicates that the

392: relative strengths of the magnetization and shift current sources

393: have a ratio 0.017, which is within a factor of six of the

394: theoretical value of the hole contribution.  More work is required

395: to confirm that the offset in Fig.~\ref{expresults}b is mainly due

396: to the magnetization source. Given all the possible sources of

397: systematic error, we consider this extremely good agreement in the

398: first study of this effect.

399: 

400: Increasing the THz emission from the spin injection can be achieved

401: in a few ways: The THz generation depends on the injection rate of

402: spin polarized carriers, and so using a shorter pulse will give a

403: larger injection rate and increase the THz signal.  One could also

404: use materials with a larger $g$-factor. Materials with $g$-factors

405: near 50, such as InAs, have been reported \cite{RothPR59}. Finally,

406: some materials also allow for a higher degree of spin polarization

407: $S$ to be achieved; in strained GaAs or the II-VI wurtzite

408: materials, for example, optical orientation can give an almost 100\%

409: polarized electron spin population.

410: 

411: In conclusion, we have investigated the possibility of measuring THz

412: radiation from the ultrafast optical orientation of electrons and

413: holes. This THz radiation is due to a rapidly-varying magnetic

414: source, and we have shown that it is large enough to be detected

415: against the more usual THz signal from electric current sources.

416: Measurements of the THz radiation from spin-injection can complement

417: the wide range of experiments on spin polarization decay already

418: existing in the literature, providing a direct measurement of the

419: injected spin dynamics of both electrons and holes in

420: semiconductors.

421: 

422: We thank J. Prineas, C. Pryor, A. Bristow, M. Betz and J. Rioux for

423: useful discussions.  Preliminary experimental work by Q. H. Wang is

424: also acknowledged.  This work is supported by NSERC, Photonics

425: Research Ontario, and DARPA SpinS.

426: 

427: \bibliographystyle{apsrev}

428: \bibliography{THz_biblio}

429: 

430: \end{document}

431: