1: \def\deg      {{^\circ \/}}

2: \def\H        {{$^1$H \/}}

3: \def\F        {{$^{19}$F \/}}

4: \def\P        {{$^{31}$P \/}}

5: \def\CaF      {{CaF$_2$ \/}}

6: \def\etal     {{\it et al. \/}}

7: \def\abinitio {{\it ab initio \/}}

8: \def\eg       {{\it e.g. \/}}

9: \def\ie       {{\it i.e. \/}}

10: \def\first    {{1$^{\rm st}$ \/}}

11: \def\second   {{2$^{\rm nd}$ \/}}

12: \def\third    {{3$^{\rm rd}$ \/}}

13: \def\th       {{$^{\rm th}$ \/}}

14: 

15: \newcommand{\ee}[1]{\cdot10^{#1}}

16: \newcommand{\mr}[1]{\mathrm{#1}}

17: \newcommand{\unit}[1]{\,\mathrm{#1}}

18: \newcommand{\um}{\,\mu{\rm m}}

19: \newcommand{\us}{\,\mu{\rm s}}

20: \newcommand{\kT}{k_{\rm B}T}

21: \newcommand{\kB}{k_{\rm B}}

22: 

23: %\newcommand{\wc}{\omega_c}

24: %\newcommand{\wci}{\omega_{c,i}}

25: \newcommand{\fci}{f_{i}}

26: \newcommand{\kci}{k_{i}}

27: \newcommand{\Qci}{Q_{i}}

28: \newcommand{\tm}{\tau_\mr{m}}

29: \newcommand{\itm}{\tau_\mr{m}^{-1}}

30: \newcommand{\wrf}{\omega_\mr{rf}}

31: \newcommand{\weff}{\omega_\mr{eff}}

32: \newcommand{\theff}{\theta_\mr{eff}}

33: \newcommand{\wo}{\omega_0}

34: \newcommand{\wone}{\omega_1}

35: \newcommand{\wmod}{\omega_\mr{mod}}

36: \newcommand{\Dw}{\Delta\omega}

37: \newcommand{\Gone}{\Gamma_1}

38: \newcommand{\Gr}{\Gamma_{1\rho}}

39: \newcommand{\SB}{S_B}

40: \newcommand{\DE}{\Delta E}

41: \newcommand{\Tr}{T_{1\rho}}

42: \newcommand{\dB}{\nabla|\bm B|}

43: \newcommand{\gradB}{\vec{\nabla}|\vec{B}|}

44: \newcommand{\dxB}{\partial \vec{B}/\partial x}

45: %\newcommand{\dxBx}{\partial B_x/\partial x}

46: %\newcommand{\dxBy}{\partial B_y/\partial x}

47: \newcommand{\dxBz}{\partial B_z/\partial x}

48: \newcommand{\dxBu}{\frac{\partial {\bm B}}{\partial x}}

49: %\newcommand{\dxBxu}{\frac{\partial B_x}{\partial x}}

50: %\newcommand{\dxByu}{\frac{\partial B_y}{\partial x}}

51: \newcommand{\dxBzu}{\frac{\partial B_z}{\partial x}}

52: \newcommand{\dxBzm}{\overline{\partial B_z/\partial x}}

53: \newcommand{\dxBzum}{\overline{\frac{\partial B_z}{\partial x}}}

54: %\newcommand{\dxBzum}{\overline{\left(\frac{\partial B_z}{\partial x}\right)}}

55: \newcommand{\SBz}{S_{Bz}(\wone)}

56: \newcommand{\Geff}{G_\mr{eff}}

57: \newcommand{\Btip}{{\bm B}_\mr{tip}}

58: \newcommand{\Btipz}{B_\mr{tip,z}}

59: \newcommand{\Bext}{B_\mr{ext}}

60: \newcommand{\vBtip}{\vec{B}_\mr{tip}}

61: \newcommand{\vBext}{\vec{B}_0}

62: \newcommand{\vecr}{{\bm r}}

63: \newcommand{\xrms}{x_\mr{rms}}

64: \newcommand{\Tmode}{T_\mr{mode}}

65: \newcommand{\Tmodet}{T_\mr{mode,3}}

66: \newcommand{\ez}{{\bm e}_z}

67: 

68: 

69: %\documentclass[preprint,prl,aps,showpacs]{revtex4}% Phys Rev Lett

70: \documentclass[prl,aps,twocolumn,showpacs]{revtex4}% Phys Rev Lett

71: 

72: %\usepackage[dvips]{graphicx}% Include figure files

73: \usepackage{graphicx}

74: %\usepackage{dcolumn}% Align table columns on decimal point

75: \usepackage{bm}% bold math

76: %\usepackage{epsfig}

77: \usepackage{amsmath}

78: \usepackage{amssymb}

79: %\usepackage{footnpag}

80: \usepackage[colorlinks=true, pdfstartview=FitV, linkcolor=blue, citecolor=blue, urlcolor=blue]{hyperref} % enable links

81: 

82: \bibliographystyle{apsrev.bst}

83: 

84: \begin{document}

85: 

86: % formatting

87: \global\emergencystretch = .1\hsize % adjust the line-breaking to avoid overfull \hbox (standard .15\hsize)

88: 

89: \title{Nuclear spin relaxation induced by a mechanical resonator}

90: 

91: \author{C. L. Degen$^{1}$}

92:   \email{degenc@gmail.com} 

93: \author{M. Poggio$^{1,2}$, H. J. Mamin$^1$, and D. Rugar$^1$}

94:   \affiliation{

95:    $^1$IBM Research Division, Almaden Research Center, 650 Harry Road, San Jose CA, 95120, USA. \\

96:    $^2$Center for Probing the Nanoscale, Stanford University, 476 Lomita Hall, Stanford CA, 94305, USA.}

97: \date{\today}

98: 

99: \begin{abstract}

100: We report on measurements of the spin lifetime of nuclear spins strongly coupled to a micromechanical cantilever as used in magnetic resonance force microscopy. We find that the rotating-frame correlation time of the statistical nuclear polarization is set by the magneto-mechanical noise originating from the thermal motion of the cantilever. Evidence is based on the effect of three parameters: (1) the magnetic field gradient (the coupling strength), (2) the Rabi frequency of the spins (the transition energy), and (3) the temperature of the low-frequency mechanical modes.

101: Experimental results are compared to relaxation rates calculated from the spectral density of the magneto-mechanical noise.

102: \end{abstract}

103: 

104: \pacs{76.60.-k, 85.85.+j, 05.40.Jc}

105: 

106: \maketitle

107: 

108: Sensitive detection of nuclear spin signals requires a sensor that can couple strongly to the weak nuclear magnetic moment.

109: For inductively detected nuclear magnetic resonance (NMR), where a resonant coil picks up the rf signal of the precessing spins,

110: this is achieved by scaling down the coil size so as to improve the current-per-flux ratio \cite{hoult76}.

111: Optimized inductive detectors are currently able to observe ensembles containing roughly $10^{12}$ proton spins,

112: equivalent to about $10^8$ net magnetic moments \cite{seeber01}.

113: Stronger couplings enabling higher spin sensitivity have been shown with magnetic force sensors,

114: such as used in magnetic resonance force microscopy (MRFM)  \cite{sidles95,rugar04}, which recently

115: measured the net moment of about $10^3$ nuclei \cite{mamin07}.

116: 

117: In order to detect spins with still higher sensitivity, the detector must be coupled even more tightly to the nuclear magnetic moment.

118: For a mechanical detector sensitive to forces in the $x$-direction, this is achieved by increasing the magnetic field gradient $\dxBu$,

119: which generates the magnetic force $F_x = {\bm \mu}\cdot\dxBu$,

120: where ${\bm \mu}$ is the total magnetic moment of the spin ensemble.

121: Present MRFM technology employs gradients up to $10^6\unit{T/m}$ \cite{mamin07}.

122: If the gradient can be successfully pushed into the range of $10^8\unit{T/m}$,

123: force detection of single nuclear spins may become feasible.

124: 

125: The strong interaction between spins and sensor, however, also increases the ``back action'' of the sensor on the spins and makes

126: them more susceptible to detector noise. Here we are concerned with magnetic noise generated by the thermally vibrating

127: cantilever, but similar effects can be expected in any real-world detector, as for example in an inductively-coupled rf circuit.

128: The interaction between a mechanical resonator and spins has been the subject of a number of theoretical studies,

129: and is predicted to lead to a host of intriguing effects. These range from shortening of spin lifetimes \cite{mozyrsky03,berman03},

130: to spin alignment by specific mechanical modes either at the Larmor frequency or in the rotating frame \cite{magusin00,butler05},

131: to resonant amplification of mechanical oscillations \cite{bargatin03}.

132: 

133: In this letter, we report direct experimental evidence for accelerated nuclear spin relaxation induced by a single, low-frequency mechanical mode.

134: The observed relaxation has its origin in the random magnetic field ${\bm B}(t) = \dxBu x(t)$

135: created by the thermal (Brownian) motion $x(t)$ of the cantilever in a large field gradient.

136: The nuclear spin and mechanical oscillator degrees of freedom are well decoupled in strong static fields,

137: since the spin precession frequency is orders of magnitude higher than typical cantilever frequencies.

138: However, when an on-resonance transverse rf magnetic field $B_1$ is present,

139: kHz frequency noise that overlaps with the Rabi frequency of the spin will induce spin relaxation.

140: The rate of nuclear spin transitions may be described by the rotating-frame relaxation time $\Tr$,

141: %

142: \begin{equation}

143: \Tr^{-1} \approx \frac{\gamma^2}{2} \SBz,

144: \label{eq_tr}

145: \end{equation}

146: %

147: where $\wone = \gamma B_1$ is the Rabi frequency of the spin, $\gamma$ is the gyromagnetic ratio,

148: and $\SBz$ is the (double-sided) power spectral density of the $z$-component of ${\bm B}(t)$ \cite{slichter_specdens,mozyrsky03}.

149: 

150: For a transversely oscillating cantilever as shown in Fig. \ref{fig_schematics},

151: $\SBz$ is related to the spectrum of cantilever tip motion by $\SBz = \left(\dxBzu\right)^2 S_x(\wone)$,

152: and will --- if $S_x(\wone)$ has thermal origin --- be proportional to $\kT$.

153: Hence, in a strong coupling regime where the spin lifetime is dominated by noise from the mechanical resonator,

154: we expect an explicit dependence on the field gradient, the Rabi frequency, the temperature,

155: and the mechanical mode spectrum.

156: 

157: %

158: \begin{figure}

159:       \begin{center}

160:       \includegraphics[width=0.45\textwidth]{fig1}

161:       \end{center}

162:       \caption{(a) Experimental arrangement (setup {\bf A}). A \CaF single crystal containing \F nuclei

163:       is attached to the end of the cantilever and placed in a fixed position $\sim100\unit{nm}$ above a nanoscale FeCo magnetic tip.

164:       An rf magnetic field induces magnetic resonance in a thin ``resonant slice'' of spins where the Larmor resonance condition is fulfilled.

165:       Raising or lowering the external field $\Bext$ shifts the resonant slice position up or down,

166:       allowing spins further from or closer to the tip to be selectively addressed, respectively.

167:       }

168:       \label{fig_schematics}

169: \end{figure}

170: %

171: We find evidence for mechanically induced spin relaxation while measuring nuclear spin correlation times for small ensembles

172: of statistically polarized \F spins in a \CaF single-crystal sample \cite{bloch46,degen07}.

173: The setup for these experiments, shown in Fig. \ref{fig_schematics}(a), combines three components: 

174: (1) an ultra-sensitive cantilever, (2) a nanoscale ferromagnetic tip, and (3) a micron-scale rf circuit for rf field generation.

175: Experiments are carried out with two different arrangements: setup {\bf A} uses a 90-$\um$-long single-crystal Si cantilever \cite{chui03}

176: together with a FeCo thin-film conical tip and an external rf microcoil \cite{mamin07}.

177: Setup {\bf B} employs a slightly longer cantilever ($120\unit{\um}$)

178: combined with a FeCo cylindrical pillar integrated onto a lithographically patterned rf microwire \cite{poggio07}.

179: The \CaF samples, a few $\um^3$ in size, are glued to the end of the cantilevers.

180: 

181: For spin detection we rely on the ``rf frequency sweep'' method to drive adiabatic spin inversions,

182: thereby modulating the $z$-component of the nuclear polarization at the fundamental cantilever frequency \cite{madsen04,poggio07}.

183: Here, the rf center frequency $\wrf/2\pi$ is $114\unit{MHz}$ with peak frequency deviation of the rf sweep $\Dw/2\pi$ in the range of $400$ to $1000\unit{kHz}$.

184: We measure the correlation time $\tm$ of the cantilever tip oscillation amplitude,

185: which reflects the correlation time of the nuclear polarization \cite{degen07}.

186: $\tm$ can be determined in several ways, for example by calculating the autocorrelation function \cite{degen07},

187: by measuring the linewidth of the associated power spectral density, or by using a bank of filters of different noise bandwidths \cite{rugar04}.

188: $\tm$ is very closely related to the rotating frame relaxation time $\Tr$, as will be discussed below.

189: 

190: In a first set of experiments we observe that spin relaxation depends strongly on the distance

191: between spins and magnetic tip. We attribute this observation

192: to an associated variation of the magnetic field gradient.

193: Figure \ref{fig_gradient}(a) plots the relaxation rate $\itm$ as a function of applied external field $\Bext$.

194: Since the external field sets the region of space where the resonance condition is met ---

195: indicated by the ``resonant slice'' in Fig. \ref{fig_schematics} ---

196: changing $\Bext$ is equivalent to changing the region in the sample where we probe the spins.

197: Spins located closer to the tip require less external field to satisfy the Larmor condition,

198: $\gamma |\Bext \hat{\bm z} +\Btip| = \wrf$.

199: For these spins, the resonance appears on the low end of the $\Bext$ scan [Fig. \ref{fig_gradient}(a)].

200: Likewise, spins far away from the tip experience the smallest $\Btip$ and require the highest $\Bext$.

201: 

202: In order to relate $\itm$ to the magnetic field gradient, we first calculate the tip field $\Btip(\vecr)$ as a function of position $\vecr$.

203: A good model for the tip can be obtained by inferring the tip size from a scanning electron micrograph,

204: and combining it with MRFM data for the tip moment \cite{mamin07}.

205: To estimate the effective lateral field gradient $\Geff(\Bext) \propto \dxBzu$ for each resonant slice,

206: we then derive the gradient from $\Btip(\vecr)$ and average it over the resonant slice volume \cite{Geff}.

207: As a result, we can relate the gradient to the external field and use this knowledge to plot $\itm$ as a function of $\Geff$,

208: shown in Fig. \ref{fig_gradient}(b) and (c).

209: 

210: %

211: \begin{figure}[t]

212:       \begin{center}

213:       \includegraphics[width=0.45\textwidth]{fig2}

214:       \end{center}

215:       \caption{

216:       (a) Spin correlation time $\tm$ as a function of external field $\Bext$.

217:       Data points (dots) represent experimental values, measured with setup {\bf A}.

218:       Spins are resonant at a total field of $|\Bext\hat{\bm z}+\Btip|=2.85\unit{T}$.

219:       (b) Effective field gradient $\Geff(\Bext)$ as a function of $\Bext$ calculated from tip model.

220:       (c) Relaxation rate $\itm$ as a function of field gradient $\Geff$ obtained by combining (a) and (b).

221:       Solid line is a best fit with the lowest point excluded. Dashed line is a guide to the eye.

222:       }

223:       \label{fig_gradient}

224: \end{figure}

225: %

226: We find that spin relaxation increases with the gradient as $\itm\propto \Geff^{1.23\pm0.16}$.

227: Since the gradient is the main parameter describing the coupling between spins and oscillator,

228: a stronger gradient is equivalent to a stronger coupling --- hence the same mechanical noise creates

229: more magnetic noise. While gradient-induced spin relaxation is the dominant relaxation mechanism

230: in Fig. \ref{fig_gradient}, other (intrinsic) processes will eventually dominate in the limit of small gradient.

231: In Fig. \ref{fig_gradient}, this may be the case for gradients $\Geff<0.3\unit{G/nm}$.

232: 

233: In a second set of experiments we investigate the spin relaxation rate as a function of rf field magnitude $\wone = \gamma B_1$.

234: Here we are able to measure $\itm$ for $\wone/2\pi$ between $60$ and $170\unit{kHz}$.

235: The rf field magnitude is calibrated by a spin nutation experiment \cite{poggio07}.

236: For small $\wone$, the spin correlation times are so short that

237: the measurement is dominated by the response time of the actively damped cantilever ($\sim 20\unit{ms}$).

238: The upper limit on $\wone$ is set by heating restrictions of the rf microwire.

239: 

240: The measured relaxation rates are shown in Fig. \ref{fig_driving}(a). 

241: We find that the spin relaxation rate decreases rapidly until reaching a distinct knee at $\wone/2\pi\approx125\unit{kHz}$.

242: We believe that this feature is directly related to the thermal vibration of the third cantilever mode,

243: having a resonance frequency of $f_3=122\unit{kHz}$ [see Fig. \ref{fig_modes}].

244: 

245: %

246: \begin{figure}[t]

247:       \begin{center}

248:       \includegraphics[width=0.45\textwidth]{fig3}

249:       \end{center}

250:       \caption{(a) Spin relaxation rate $\itm$ as a function of rf field magnitude $\wone=\gamma B_1$, measured with setup {\bf B}.

251:       Each time $\wone$ falls below a cantilever mode, indicated by dotted vertical lines, a new dissipation channel is added.

252:       At low fields $\tm$ is presumably limited by the adiabatic condition (dashed line, Ref. \cite{adiabaticcondition}).

253:       %(Data reproduced from Ref. \cite{poggio07}).

254:       (b) Spin relaxation rate $\itm$ as a function of mode temperature $\Tmode$, here for the third cantilever mode.

255:       The thermal motion of the mode is found to affect spin relaxation when $\wone/2\pi<f_3$ (upper figure),

256:       but not when $\wone/2\pi>f_3$ (lower figure). Solid line is a best fit.}

257:       \label{fig_driving}

258: \end{figure}

259: %

260: To further explore the role of the third mode we can modify its effective temperature, denoted by $\Tmode$.

261: For that purpose we excite the mode with bandwidth-limited noise using a piezoelectric actuator mechanically coupled to the cantilever.

262: At the same time as we measure $\tm$, we also monitor the mean-squared motion $\xrms^2$ of the cantilever tip in that mode.

263: We can assign an equivalent mode temperature $\Tmode=\kci\xrms^2/\kB$, where $\kci$ is the effective spring constant of mode $i$ and $\kB$ is Boltzmann's constant.

264: Note that while $\Tmode$ can become very large for strong mechanical actuation, the actual (or bath) temperature of the cantilever

265: remains at $T=4.5\unit{K}$, as the heat capacity of the single mode is much smaller than

266: the heat capacity of the cantilever's phonon bath and the modes are only very weakly coupled.

267: 

268: %

269: \begin{figure}[t]

270:       \begin{center}

271:       \includegraphics[width=0.45\textwidth]{fig4}

272:       \end{center}

273:       \caption{Power spectral density of the thermomechanical noise at $T=4.5\unit{K}$, showing the first three vibrational modes

274:       of cantilever {\bf B}. Sketches illustrate the bending of the cantilever beam.

275:       Frequency, spring constant and quality factor of the lowest mode are $f_{1}=2.57\unit{kHz}$, $k_{1}=86\unit{\mu N/m}$ and $Q_{1}=17,000$.

276:       }

277:       \label{fig_modes}

278: \end{figure}

279: %

280: We investigate the dependence of $\itm$ on $\Tmode$ at several different $B_1$ fields and find

281: the two different behaviors shown in Fig. \ref{fig_driving}(b):

282: If $\wone/2\pi$ lies above $f_3$, no change in spin relaxation is observed even for strong actuation.

283: On the other hand, for $\wone/2\pi$ below $f_3$, relaxation is greatly enhanced.

284: We can fit the dependence of the relaxation rate on the mode temperature and find $\itm \propto (\Tmode)^{0.43\pm0.07}$.

285: Because the slope does not level off even for low $\Tmode$, the spin relaxation rate is set by the thermal fluctuations of the third cantilever mode

286: over the entire investigated mode temperature range, in particular at thermal equilibrium where $\Tmode = T$.

287: At $\wone/2\pi=98\unit{kHz}$, we can also try to enhance spin relaxation by actuating the second mode.

288: No influence is observed, as expected, because $\wone/2\pi>f_2$.

289: 

290: In order to better understand the dependence of the spin correlation time on field gradient, rf field magnitude, and temperature, it is worthwhile to connect $\tm$ to the cantilever's mode spectrum. We calculate the transition rate following the analysis of Mozyrsky \etal \cite{mozyrsky03}.

291: To obtain an expression for $\tm$ similar to Eq. (\ref{eq_tr}),

292: taking into account the time-dependence of the rf field frequency $\wrf(t)$ during cyclic spin inversion,

293: we assume that $\tm$ can be described as the average relaxation rate over a frequency sweep,

294: %

295: \begin{equation}

296: \itm = \frac{1}{T_c} \int_{0}^{T_c/2} dt\, \frac{\wone^2}{\weff^2(t)}\, \gamma^2 S_{Bz}(\weff(t)).

297: \label{eq_itm_int}

298: \end{equation}

299: %

300: (See also Ref. \cite{mozyrsky03}, Eq. (7)). Here, $T_c=1/f_1$ is the oscillation period of the fundamental cantilever mode,

301: $\weff(t) = \{[\wrf(t)-\wo]^2+\wone^2\}^{1/2}$ is the effective magnetic field \cite{slichter_adiabaticsweeps},

302: and $\wrf(t)$ is modulated from $\wo-\Dw$ to $\wo+\Dw$. Note that because $\Dw\gg\wone$,

303: the Rabi frequency $\weff$ traverses a broad range of frequencies during the sweep.

304: A key difference between $\tm$ and $\Tr$ is therefore that $\tm$ is sensitive to noise

305: in a frequency band set by $\wone\leq\weff(t) \underset{\sim}{<} \Dw$, while $\Tr$ is influenced by noise in the vicinity of $\wone$ only.

306: 

307: The magnetic noise spectrum $S_{Bz}(\omega)$ is given by

308: %

309: \begin{equation}

310: S_{Bz}(\omega) = \Geff^2 \sum_{i=1}^n

311: \frac{\kT}{\pi \kci \fci \Qci} \ \frac{(2\pi\fci)^4}{((2\pi\fci)^2-\omega^2)^2+(2\pi\fci \omega/\Qci)^2},

312: \label{eq_modes}

313: \end{equation}

314: %

315: where $n$ is the number of modes with a significant noise contribution. Because of the high quality factors $Q_i$, $S_{Bz}(\omega)$ exhibits

316: a discrete set of sharp peaks at the mode frequencies $f_i$. We can facilitate the analysis of 

317: Eq. (\ref{eq_itm_int}) by treating the peaks as $\delta$-functions, and find that

318: %

319: \begin{equation}

320: \label{eq_itm}

321: \itm \approx \frac{\gamma^2\Geff^2\kT}{2\Dw} \sum_{i=n'}^{n} \frac{\wone^2}{2\pi\fci\kci \sqrt{(2\pi\fci)^2-\wone^2}} ,

322: \end{equation}

323: %

324: where $n'$ is the lowest mode whose resonance frequency $f_{n'}$ is above $\wone/2\pi$.

325: In other words, when $\wone$ is reduced a new relaxation channel is added each time

326: $\wone/2\pi$ becomes less than a cantilever resonance $\fci$. While $n$ can be large,

327: the lowest mode $n'$ will dominate (\ref{eq_itm}) because it has the lowest effective spring constant

328: and hence the largest thermal vibration amplitude.

329: 

330: As a result, we find that Eq. (\ref{eq_itm}) describes both the kink in Fig. \ref{fig_driving}(a) at the frequency of the third mode and

331: produces the correct trends in $\Geff$ and $\Tmode$. Eq. (\ref{eq_itm}) and Refs. \cite{sidles92,mozyrsky03}, however, predict

332: that $\itm \propto \Geff^2\Tmode$ --- a significantly stronger dependence on $\Geff$ and $\Tmode$ than our experimental observation,

333: $\itm \propto \Geff^{1.23\pm0.16}\Tmode^{0.43\pm0.07}$.

334: We are aware of two possible reasons for this discrepancy: First, the analysis of the spin transition rate [Eq. (\ref{eq_itm_int})]

335: is based on the Bloch-Redfield approximation, where the magnetic noise is assumed to be uncorrelated \cite{slichter_blochredfield}.

336: This assumption may be violated because the correlation time of the noise, $\tau_{c,i} = \Qci/\pi\fci$, can be significant, possibly even longer than $\tm$.

337: Furthermore, we assume that spins are non-interacting, which will not necessarily be valid, especially at low rf fields where local fields can easily exceed $B_1$.

338: 

339: In conclusion, we find that nuclear spin relaxation can be induced by a single, low-frequency mode of a micromechanical resonator.

340: The spin relaxation rate is observed to increase both with the gradient and the effective mode temperature.

341: We also find that long spin lifetimes are recovered when increasing the magnitude of the rf field to raise the Rabi frequency

342: above the lowest mechanical mode frequencies.

343: 

344: We thank C. Rettner, M. Hart and M. Farinelli for fabrication of microwire and magnetic tip,

345: B. W. Chui for cantilever fabrication, D. Pearson and

346: B. Melior for technical support, and M. Ernst for helpful discussions.

347: We acknowledge support from the DARPA QUIST

348: program administered through the US Army Research Office,

349: and the Stanford-IBM Center for Probing the Nanoscale, a NSF Nanoscale

350: Science and Engineering Center. C. L. D. acknowledges funding from

351: the Swiss National Science Foundation.

352: 

353: \noindent

354: \begin{thebibliography}{}

355: 

356: %%% conventional mri

357: \bibitem{hoult76}

358: D. I. Hoult, and R. E. Richards, J. Magn. Res. {\bf 24}, 71 (1976).

359: 

360: %%% conventional MRI

361: \bibitem{seeber01} %the 10^12 protons paper

362: D. A. Seeber, R. L. Cooper, L. Ciobanu, and C. H. Pennington,

363: %D. A. Seeber \etal,

364: %Rev. Sci. Instr. {\bf 72}, 2171 (2001).

365: %10.1063/1.1359190

366: \href{http://dx.doi.org/10.1063/1.1359190}{Rev. Sci. Instr. {\bf 72}, 2171 (2001)}.

367: 

368: %%% MRFM

369: \bibitem{sidles95}

370: J. A. Sidles, J. L. Garbini, K. J. Bruland, D. Rugar, O. Z\"uger, S. Hoen, and C. S. Yannoni,

371: %J. A. Sidles \etal,

372: \href{http://dx.doi.org/10.1103/RevModPhys.67.249}{Rev. Mod. Phys. {\bf 67}, 249 (1995)}.

373: 

374: \bibitem{rugar04}

375: %D. Rugar \etal,

376: D. Rugar, R. Budakian, H. J. Mamin, and B. W. Chui,

377: \href{http://dx.doi.org/10.1038/nature02658}{Nature {\bf 430}, 329 (2004)}.

378: 

379: %%% present day MRFM

380: \bibitem{mamin07}

381: %H. J. Mamin \etal,

382: H. J. Mamin, M. Poggio, C. L. Degen, and D. Rugar,

383: \href{http://dx.doi.org/10.1038/nnano.2007.105}{Nature Nanotechnology {\bf 2}, 301 (2007)}.

384: %\href{http://arxiv.org/abs/cond-mat/0702664}{cond-mat/0702664}.

385: 

386: %%% theory spin-mechanical oscillator interaction

387: \bibitem{mozyrsky03}

388: D. Mozyrsky, I. Martin, D. Pelekhov and P. C. Hammel,

389: %D. Mozyrsky \etal,

390: \href{http://dx.doi.org/10.1063/1.1554769}{Appl. Phys. Lett. {\bf 82}, 1278 (2003)}.

391: 

392: \bibitem{berman03}

393: G. P. Berman, V. N. Gorshkov, D. Rugar, and V. I. Tsifrinovich,

394: %10.1103/PhysRevB.68.094402

395: \href{http://dx.doi.org/10.1103/PhysRevB.68.094402}{Phys. Rev. B {\bf 68}, 094402 (2003)}.

396: 

397: %%% spin cooling

398: \bibitem{magusin00}

399: P. C. M. M. Magusin, and W. S. Veeman,

400: \href{http://dx.doi.org/10.1006/jmre.1999.1989}{J. Magn. Res. {\bf 143}, 243 (2000)}.

401: 

402: \bibitem{butler05}

403: M. C. Butler, V. A. Norton, and D. P. Weitekamp,

404: Abstracts of papers of the American Chemical Society {\bf 229}, 

405: U735-U735 217-PHYS, 2005; D. P. Weitekamp, Patent US 6,841,995 B2, 2005.

406: 

407: % cantilaser

408: \bibitem{bargatin03}

409: I. Bargatin, and M. L. Roukes,

410: %Phys. Rev. Lett. {\bf 91}, 138302 (2003).

411: %10.1103/PhysRevLett.91.138302 

412: \href{http://dx.doi.org/10.1103/PhysRevLett.91.138302}{Phys. Rev. Lett. {\bf 91}, 138302 (2003)}.

413: 

414: %%% spectral density / theory relaxation

415: 

416: \bibitem{slichter_specdens}

417: C. P. Slichter,

418: {\sl Principles of Magnetic Resonance}, $3^\mr{rd}$ edition, (Springer, Berlin, 1990), p. 206 ff and p. 242 ff.

419: 

420: %%% statistical polarization

421: 

422: \bibitem{bloch46}

423: F. Bloch, 

424: \href{http://dx.doi.org/10.1103/PhysRev.70.460}{Phys. Rev. {\bf 70}, 460 (1946)}.

425: 

426: \bibitem{degen07}

427: %C. L. Degen  \etal

428: C. L. Degen, M. Poggio, H. J. Mamin, and D. Rugar,

429: \href{http://dx.doi.org/10.1103/PhysRevLett.99.250601}{Phys. Rev. Lett. {\bf 99}, 250601 (2007)}.

430: %\href{http://arxiv.org/abs/0710.2323}{arXiv:0710.2323}.

431: 

432: %%% experimental setup, protocol

433: 

434: \bibitem{chui03}

435: B. W. Chui {\it et al.,}

436: {\sl Technical Digest of the 12$^{th}$ International Conference on Solid-State Sensors and Actuators (Transducers '03)},

437: (IEEE Boston MA, USA, 2003), p. 1120.

438: 

439: \bibitem{poggio07}

440: M. Poggio, C. L. Degen, H. J. Mamin, and D. Rugar,

441: %M. Poggio \etal,

442: \href{http://dx.doi.org/10.1063/1.2752536}{Appl. Phys. Lett. {\bf 90}, 263111 (2007)}.

443: 

444: \bibitem{madsen04}

445: L. A. Madsen, G. M. Leskowitz, and D. P. Weitekamp,

446: \href{http://dx.doi.org/10.1073/pnas.0405232101}{Proc. Nat. Acad. Sci. {\bf 101}, 12804 (2004)}.

447: 

448: %%% gradient

449: 

450: \bibitem{Geff}

451: The introduction of an effective gradient $\Geff$ is necessary because not all spins in the

452: resonant slice experience the same $\dxBzu$. We believe that an appropriate definition for $\Geff$ is

453: the weighted-rms value

454: %$\Geff(\Bext) = \left[ \sum_{k=1}^{n}  \dxBzu({\bm r}_k)^2 w_k \right]^{\frac{1}{2}}$,

455: $\Geff(\Bext) = \left\{ \sum_{k=1}^{n}  \left[\dxBzu({\bm r}_k)\right]^4 / \sum_{k=1}^{n} \left[\dxBzu({\bm r}_k)\right]^2 \right\}^{\frac{1}{2}}$,

456: because it gives more weight to spins contributing large magnetic forces $F_k^2 = \mu^2 \left[\dxBzu({\bm r}_k)\right]^2$ to the signal.

457: $\{\vecr_k\}_{k=1...n}$ are the positions of the spins for which $|\gamma\Btip(\vecr_k)-\wo|<\Dw$.

458: 

459: %%% analysis

460: 

461: \bibitem{slichter_adiabaticsweeps}%adiabatic condition AND rotating frame

462: Ref. \cite{slichter_specdens}, p. 20-24.

463: 

464: \bibitem{sidles92}

465: J. A. Sidles, J. L. Garbini, and G. P. Drobny,

466: %Rev. Sci. Instr. {\bf 63}, 3881 (1992).

467: %10.1063/1.1143837

468: \href{http://dx.doi.org/10.1063/1.1143837}{Rev. Sci. Instr. {\bf 63}, 3881 (1992)}.

469: 

470: \bibitem{slichter_blochredfield}%bloch-redfield approximation

471: Ref. \cite{slichter_specdens}, p. 199.

472: 

473: %%% captions

474: 

475: \bibitem{adiabaticcondition}

476: By simulations of the Bloch equations one can show that the adiabaticity parameter $A$

477: [see J. Baum, R. Tycko, and A. Pines,

478: \href{http://dx.doi.org/10.1103/PhysRevA.32.3435}{Phys. Rev. A {\bf 32}, 3435 (1985)}]

479: relates to the maximum possible $\tm$ as $\itm \approx 2f 10^{-0.68 A}$, where $f$ is the cantilever frequency.

480: 

481: \end{thebibliography}

482: 

483: \end{document}

484: