1: \documentclass[prl, showpacs, superscriptaddress, twocolumn, floatfix]{revtex4}
2: \usepackage{color, graphicx, amssymb}
3:
4: \date{03/13/2009}
5:
6: \begin{document}
7:
8: \title{Persistent currents in normal metal rings}
9:
10: \author{Hendrik Bluhm}
11: \altaffiliation[Present address: ]{Department of Physics, Harvard University,
12: Cambridge, MA, USA}
13: \affiliation{Departments of Physics and Applied Physics, Stanford
14: University, Stanford, CA 94305}
15: \author{Nicholas C. Koshnick}
16: \affiliation{Departments of Physics and Applied Physics, Stanford
17: University, Stanford, CA 94305}
18: \author{Julie A. Bert}
19: \affiliation{Departments of Physics and Applied Physics, Stanford
20: University, Stanford, CA 94305}
21: \author{Martin E. Huber}
22: \affiliation{Departments of Physics and Electrical Engineering,
23: University of Colorado Denver, Denver, CO 80217}
24: \author{Kathryn A. Moler}
25: \email{kmoler@stanford.edu}
26: \affiliation{Departments of Physics and Applied Physics, Stanford
27: University, Stanford, CA 94305}
28:
29:
30:
31: \begin{abstract}
32: The authors have measured the magnetic response of 33 individual cold
33: mesoscopic gold rings, one ring at a time. The response of some
34: sufficiently small rings has a component that is periodic in the flux
35: through the ring and is attributed to a persistent current. Its period
36: is close to $h/e$, and its sign and amplitude vary between rings. The
37: amplitude distribution agrees well with predictions for the typical
38: $h/e$ current in diffusive rings. The temperature dependence of the
39: amplitude, measured for four rings, is also consistent with
40: theory. These results disagree with previous measurements of three
41: individual metal rings that showed a much larger periodic response
42: than expected. The use of a scanning SQUID microscope enabled in situ
43: measurements of the sensor background. A paramagnetic linear
44: susceptibility and a poorly understood anomaly around zero field are
45: attributed to defect spins.
46: \end{abstract}
47:
48: \pacs{73.23.Ra}
49: \maketitle
50: When a conducting ring is threaded by a magnetic flux $\Phi_a$, the
51: associated vector potential imposes a phase gradient on the electronic
52: wave functions, $\psi$, that can be transformed into a phase factor in
53: the boundary conditions: $\psi(L) = e^{i 2 \pi \Phi_a/\phi_0}
54: \psi(0)$, where $L$ is the circumference of the ring and $\phi_0
55: \equiv h/e$ the flux quantum \cite{ByersN:Theccq}. The $h/e$
56: periodicity of this phase factor is reflected in all properties of the
57: system. Here, we focus on the persistent current $I$ circulating the
58: ring, which is the first derivative of the free energy with respect to
59: $\Phi_a$, and thus a fundamental thermodynamical quantity. For a
60: perfect 1D ring without disorder populated by noninteracting
61: electrons, it is relatively straightforward to show that $I$ will be
62: of order $e v_F/L$ \cite{CheungHF:Percso}, the current carried by a
63: single electron circulating the ring at the Fermi velocity $v_F$.
64: Perhaps somewhat surprisingly, persistent currents are not destroyed
65: by elastic scattering \cite{ButtikerM:Josbsn}. In the diffusive limit,
66: i.e. for a mean free path $l_e < L$, $I \sim e/\tau_D$ is set by the
67: diffusive round trip time $\tau_D = L^2/D$, where $D = v_F l_e/3$ is
68: the diffusion constant \cite{CheungHF:Percmr, RiedelEK:Mespcs}.
69: Thermal averaging leads to a strong suppression of the persistent
70: current at temperatures above the correlation energy $E_c \equiv \hbar
71: \pi^2 D/L^2 \propto \hbar/\tau_D$.
72:
73: Like many mesoscopic effects in disordered systems, the persistent
74: current depends on the particular realization of disorder and thus
75: varies between nominally identical samples. In metal rings, the
76: dependence on disorder and $\cos (k_F L)$, which is random in
77: practice, leads to a zero ensemble average $\langle I_{h/e}\rangle$ of
78: the first, i.e. $h/e$-periodic, harmonic.
79: The magnitude of the fluctuations from sample to sample is given by the typical
80: value \cite{RiedelEK:Mespcs}
81: \begin{equation}\label{eq:theory}
82: \langle I_{h/e}^2 \rangle^{1/2} = \frac{E_c}{\phi_0} e^{-k_B T/E_c}.
83: \end{equation}
84: We have not included a factor 2 for spin because our Au rings are
85: in the strong spin-orbit scattering limit
86: \cite{MeirY:Uniess, Entin-WohlmanO:Effssm}.
87: An additional contribution that survives averaging over disorder
88: but oscillates with $k_F L$ \cite{BaryS:avge} is predicted to have a magnitude
89: $\overline{I_{h/e}} = (12/\pi^2) \sqrt{M L/l_e} (E_c/\phi_0) e^{-2L/l_e}$,
90: where $M$ is the number of channels
91: \cite{CheungHF:Percmr}.
92: Because of the exponential dependence on $L/l_e$, it is usually
93: negligible compared to Eq. \ref{eq:theory} for metallic rings.
94: Higher harmonics are generally smaller because they are more sensitive
95: to disorder and thermal averaging. However, due to interactions
96: \cite{AmbegaokarV:Cohapc, SchmidA:Percmr, SchechterM:Magrdm, Bary-SorokerH:Effpbm}
97: and differences between the canonical and grand canonical ensemble
98: \cite{AltshulerBL:Perdbc, VonOppenF:Avepcm, SchmidA:Percmr},
99: $\langle I_{h/2e}\rangle$ is expected to be nonzero.
100:
101: There are very few experimental results on persistent
102: currents, and most measured the total response of an ensemble of
103: rings \cite{LevyLP:Magmcr, ReuletB:DynriA, DeblockR:aceam,
104: DeblockR:Diaorm}. The experiments to date are all based on magnetic
105: detection and are considered challenging as they require a very high
106: sensitivity. The measurements of large ensembles are dominated by
107: $\langle I_{h/2e}\rangle$, whose contribution to the total current of
108: $N$ rings scales with $N$, whereas the $h/e$ periodic current scales
109: as $\sqrt{N}$ because of its random sign.
110: The measured values of $\langle I_{h/2e}\rangle$ are generally a factor
111: of a few larger than most theoretical predictions. A
112: plausible reconciliation was proposed
113: recently for metallic rings \cite{Bary-SorokerH:Effpbm}.
114:
115: Here, we address $\langle I_{h/e}^2 \rangle$ in diffusive rings by measuring
116: one ring at a time. The $h/e$ component has been measured in good
117: agreement with theory \cite{CheungHF:Percso} in a single ballistic
118: ring \cite{MaillyD:Expopc} and an ensemble of 16 nearly ballistic
119: rings \cite{RabaudW:Percmc} in semiconductor samples. Measurements of
120: three diffusive metal rings \cite{ChandrasekharV:Magrsi} on the other
121: hand showed periodic signals that were 10--200 times larger than
122: predicted \cite{RiedelEK:Mespcs}. Later results on the total current
123: of 30 diffusive rings \cite{JariwalaEMQ:Diapcd} showed a better
124: agreement with theory \cite{RiedelEK:Mespcs}, but did not allow one to
125: distinguish between the typical and average current, which would
126: require individual measurements of several rings or groups of
127: rings. Thus, there is an unresolved contradiction between experiment
128: and theory for the typical $h/e$ current, the investigation of which
129: is a major open challenge in mesoscopic physics.
130:
131: We report measurements of the individual magnetic responses of
132: 33 diffusive Au rings.
133: The use of a scanning SQUID
134: technique allowed us to measure many different rings, one by one, with
135: in situ background measurements \cite{HuberM:Susc,KoshnickN:Flusm}.
136: The response of some of the rings contains an $h/e$ periodic component
137: whose amplitude distribution -- including rings without a detectable
138: periodic signal -- is in good agreement with predictions for
139: $\langle I_{h/e}^2 \rangle^{1/2}$. Additional features in the total
140: nonlinear response most likely reflect a nonequilibrium response of
141: impurity spins. Different frequency and geometry dependencies allow
142: the distinction between those two components, and support the
143: interpretation of the periodic part as persistent currents.
144: Due to the necessity to subtract a mean background from our data and the small
145: number of rings, we are unable to extract any ensemble average from
146: our results.
147:
148: Our samples were fabricated using standard e-beam and optical
149: liftoff lithography and were e-beam evaporated from a 99.9999 \% pure Au
150: source onto a Si substrate with a native oxide.
151: The 140 nm thick rings were deposited
152: at a relatively high rate of 1.2 nm/s in order to achieve a large
153: $l_e$. The rings have an annulus width of 350 nm,
154: and radii $R$ from 0.57 to 1 $\mu$m. From resistance
155: measurements of wires fabricated together with the rings,
156: we obtain $D$ = 0.09 m$^2$/s, $l_e$ = 190 nm.
157: Weak localization measurements yield a dephasing length
158: $L_\phi$ = 16 $\mu$m at $T$ = 300 mK, so that $ L_\phi \approx 4 L$
159: for our most important $R$ = 0.67 $\mu$m rings.
160: Some rings were connected to large metallic banks
161: [See Fig. \ref{fig:scan}(b)] to absorb the inductively coupled
162: heat load from the sensor SQUID.
163:
164: The experiment was carried out using a dilution-refrigerator based
165: scanning SQUID microscope \cite{BjornssonPG:Scasqi}.
166: Our sensors \cite{HuberM:Susc} have an integrated field coil of
167: 13 $\mu$m mean diameter, which is used to apply a field to the sample.
168: The sample response is coupled into the SQUID via a 4.6 $\mu$m diameter
169: pickup loop.
170: A second, counter-wound pair of coils
171: cancels the response to the applied field
172: to within one part in $10^4$ \cite{HuberM:Susc}.
173: The sensor response to a current $I$ in a ring is $\Phi_{SQUID} = M I$, where
174: $M$ is the pickup-loop--ring inductance.
175: Independent estimates based on previous experiments
176: \cite{KoshnickN:Flusm,BluhmH:Magrms}
177: and modeling give $M = R^2 \cdot 0.3 \Phi_0/\mu$m$^2$mA, where
178: $\Phi_0 \equiv h/2e$ is the superconducting flux quantum.
179: Using the measured $D$, Eq. \ref{eq:theory} thus predicts a typical
180: $h/e$ response from persistent currents of
181: $M \langle I_{h/e}^2 \rangle^{1/2} = 0.17\; \mu\Phi_0 \cdot e^{-k_B T/E_c}$,
182: where $\Phi_0 \equiv h/2e$ is the superconducting flux quantum.
183: We neglect the contribution from $\overline{I_{h/e}}$, which is
184: 0.3 $E_c/\phi_0$ for our smallest rings and much less for larger rings.
185:
186: \begin{figure}
187: \includegraphics{fig1}
188: \caption{\label{fig:scan}(color online).
189: (a) Susceptibility scan of an isolated ring used to locate the ring and
190: to determine the indicated measurement positions. Background measurements
191: at positions ``o'' are subtracted from the data taken at positions ``+''
192: to obtain the ring response.
193: (b) Scanning electron micrograph of a heatsunk ring.
194: (c) Temperature dependence of the linear response of one heatsunk and three
195: isolated rings. The data in (a) and (c) reflect the total amplitude
196: of the linear response to a sinusoidal excitation of $\pm$45 G for (a) and
197: the 0.67 $\mu$m rings in (c), and $\pm$35 G for the 1 $\mu$m rings.
198: }
199: \end{figure}
200:
201: After coarse alignment by imaging a current-carrying
202: meander wire on the sample, accurately locating a ring is
203: facilitated by a paramagnetic susceptibility of our metal structures
204: that appears in scans of the linear response to an applied field
205: [Fig. \ref{fig:scan}(a)].
206: To measure the complete nonlinear response, we digitized the SQUID
207: signal at a sample rate of 333 kHz and averaged it over many sweeps of
208: the current through the field coil, which was varied sinusoidally
209: over the full field range at typically 111 Hz.
210: This raw signal of a few m$\Phi_0$ is
211: dominated by nonlinearities in the sensor background and a small phase
212: shift between the fluxes applied to the two pickup loop--field coil
213: pairs. To extract the response of a ring, we measured at the positions
214: indicated in Fig. \ref{fig:scan}(a), and subtracted datasets taken far
215: from the ring (o) from those near the ring (+). The total
216: averaging time for each ring was on the order of 12 hours.
217: The reduced coupling
218: to the SQUID at intermediate positions was accounted for through a
219: smaller prefactor. The symmetric measurement positions eliminate
220: linear variations of the sensor background, which in some cases are
221: larger and more irregular than the final signal. The reliability of
222: the final result can be assessed by checking if its features (typically
223: characterized by higher harmonics of the sensor response) show a
224: spatial dependence similar to that of the ring--pickup-loop coupling.
225: This check allowed us to identify and discard questionable datasets
226: with very irregular features of ~1 $\mu\Phi_0$ obtained in some sample
227: regions.
228:
229: The response of our rings is dominated by a paramagnetic linear
230: component of up to $\approx$150 $\mu\Phi_0$ at a field of 45 G
231: \cite{BluhmH:Susc}. Its temperature dependence is
232: shown in Fig. \ref{fig:scan}(c).
233: The linear response of heatsunk rings and heatsinks
234: \cite{BluhmH:Susc} (not shown) varies approximately as $1/T$. Thus, it
235: is likely due to spins. Its magnitude corresponds to a density of $4
236: \cdot 10^{17}$ spins/m$^2$ (assuming spin 1/2). If these spins were
237: identical to the metallic magnetic impurities that were shown
238: to cause excess dephasing \cite{PierreF:Depemm}, one would expect a
239: much larger spin flip dephasing rate than the upper bound obtained
240: from our $\tau_\phi$ measurements.
241:
242: The linear response of isolated rings varies little below $\approx$ 150 mK.
243: This indication of a saturating electron temperature agrees
244: with estimates of the heating effect
245: of the $~$10 $\mu$A, 10 GHz Josephson current in the SQUID pickup loop
246: \cite{aux}.
247: The different behavior of
248: heatsunk and isolated rings shows that the linear susceptibility
249: reflects the electron rather than phonon temperature.
250:
251: We now focus on the much smaller
252: nonlinear response, obtained after eliminating the linear response
253: (including a component that is out-of-phase with the sinusoidal applied field)
254: by subtracting a fitted ellipse.
255: This linear component varied by up to a factor 2 between
256: nominally identical rings.
257: Fig. \ref{fig:ensemble}(a) shows data from fifteen isolated
258: rings with $R$ = 0.67 $\mu$m.
259: While these raw data are not periodic in $\Phi_a$, most of them
260: can be described as the sum of a periodic component and a
261: step-like shape near $\Phi_a = 0$.
262: This unexpected,
263: poorly understood anomaly appeared in nearly all rings, and was
264: most pronounced in heatsunk rings \cite{aux}.
265: Its frequency dependence suggests that it is
266: due to nonequilibrium effects in the spin response, but it might
267: mask a persistent-current-like effect \cite{aux}.
268:
269: Since one might expect the same spin signal from each ring, whereas
270: persistent currents should fluctuate around a zero mean, we subtracted
271: the average of all fifteen datasets from each individual curve. The
272: results [Fig. \ref{fig:ensemble}(b)] show oscillations that can be
273: fitted with a sine curve of the expected period for most rings.
274: Datasets 4, 5 and 15 give better fits with a 30 \% larger period,
275: which corresponds to an effective radius close to the ring's inner
276: radius. This variation of the period may reflect an imperfect
277: background elimination, but could also be a mesoscopic fluctuation of
278: the effective ring radius.
279: The seemingly much larger period of datasets 13 and 14 appears to be
280: due to a different magnitude of the zero field anomaly. From the sine
281: curve fits to 13 datasets, we obtain an estimate for $M \langle
282: I_{h/e}^2 \rangle^{1/2}$ of 0.11 $\mu\Phi_0$ if fixing the period at
283: the value expected for the mean radius of the rings, or 0.12
284: $\mu\Phi_0$ if treating it as a free parameter. This value agrees
285: with the theoretical value of 0.12 $\mu\Phi_0$ from Eq.
286: \ref{eq:theory} for $T$ = 150 mK, which corresponds to $\langle
287: I_{h/e}^2 \rangle^{1/2} = 0.9$ nA for $R$ = 0.67 $\mu$m.
288:
289:
290: \begin{figure}
291: \includegraphics{fig2}
292: \caption{\label{fig:ensemble}(color online).
293: (a) Response of 15 nominally identical rings with $R$ = 0.67 $\mu$m after
294: subtracting the in- and out-of-phase component of the linear response.
295: The curve labeled ``mean'' is the average of datasets 1-15.
296: (b) Results of subtracting this mean from datasets 1-15 in (a).
297: The smooth lines are sinusoidal fits (including a linear background term)
298: with fixed (red/dark) and fitted period (green/light).
299: Datasets 13 and 14 were excluded from the analysis because of their stronger
300: zero field anomaly.
301: The rms amplitude estimated from the fixed and variable period fits
302: corresponds to a current of 0.8 and 0.9 nA, respectively, in agreement with
303: the expected value of $\langle I_{h/e}^2 \rangle^{1/2}$
304: from Eq. \ref{eq:theory}.
305: }
306: \end{figure}
307:
308: We checked the reproducibility of the response over several weeks
309: without warming up the sample for seven rings, and found good
310: consistency in five cases. Reducing the field sweep range from 45 to
311: 35 and 25 G or varying the frequency between 13 and 333 Hz changed the
312: step feature, but had little effect on the oscillatory component in
313: the difference between the responses of two rings \cite{aux}.
314:
315: Out of five measurements of rings with $R$ = 0.57 $\mu$m \cite{aux}, four gave
316: similar results after subtracting their mean response as the $R$ =
317: 0.67 $\mu$m rings. The rms value of the fitted sine amplitudes was
318: 0.06 and 0.07 $\mu\Phi_0$ for variable and fixed period, respectively.
319: A fifth ring was excluded from this analysis because it had a
320: significantly larger zero field anomaly. Data from additional three
321: rings were rejected because of a large variation of the sensor
322: background that was not connected with the rings.
323:
324: We also measured eight isolated rings with $R$ = 1 $\mu$m, which are
325: expected to give a smaller signal because of their smaller $E_c$ of
326: 170 mK and stronger heating from the SQUID \cite{aux}. Since the
327: magnitude of the zero-field anomaly varies significantly for these
328: rings, the mean subtraction procedure cannot fully remove it. One of
329: these rings shows a sinusoidal signal with a period of 1 to 1.15
330: $\phi_0$ and an amplitude of up to 0.1 $\mu\Phi_0$, but poor
331: reproducibility. Fitting sine curves, regardless of the absence of
332: clear oscillations for the other seven rings, gives $M \langle
333: I_{h/e}^2 \rangle^{1/2}$ = 0.03 $\mu\Phi_0$. None of those rings show
334: a signal at a period similar to those in Fig. \ref{fig:ensemble}. This
335: dependence of the signal on the ring size supports the interpretation
336: as persistent current, as opposed to an artifact of the spin response.
337:
338: The data discussed so far was taken at base temperature.
339: We have measured the temperature dependence of the responses of
340: four 0.67 $\mu$m rings with large oscillatory signals of opposite sign.
341: Taking the difference between their nonlinear responses,
342: which eliminates any common background signal,
343: leads to predominantly sinusoidal curves at most temperatures,
344: as shown in Fig. \ref{fig:tempdep}(a).
345: The period appears to be $T$-independent, and amplitudes from
346: fits with a fixed period are consistent with an $e^{-k_B T/E_c}$
347: dependence [Fig. \ref{fig:tempdep}(b)] with $E_c/k_B$ = 380 mK,
348: as obtained from the measured $D$.
349:
350: \begin{figure}
351: \includegraphics{fig3}
352: \caption{\label{fig:tempdep}(color online).
353: (a) Difference between the nonlinear responses of two rings with
354: a large oscillatory component (curves 1 and 2 in
355: Fig. \ref{fig:ensemble}) at $T$ = 0.035 K to 0.5 K.
356: (b) Temperature dependence of the amplitude of the sinusoidal fits
357: in panel (a). The exponential curve is a fit to
358: $\exp(-\mathrm{min}(T, 0.15 K)/0.38 K)$, taking the
359: the saturation of the electron temperature into account.
360: The error bars were obtained by analyzing the $x$ and $y$ scan across the
361: rings [cf. Fig. \ref{fig:scan}(a)] separately and averaging the
362: difference square of the respective results over all eight data points.
363: }
364: \end{figure}
365:
366: In this experiment, the $h/e$ persistent current in diffusive
367: rings is in good agreement with theory within the temperature range covered,
368: providing long-overdue experimental input to the questions raised
369: by an earlier experiment \cite{ChandrasekharV:Magrsi}.
370:
371: \acknowledgments{This work was supported by NSF Grants No.
372: DMR-0507931, DMR-0216470, ECS-0210877 and PHY-0425897 and by the
373: Packard Foundation. Work was performed in part at the Stanford
374: Nanofabrication Facility, supported by NSF Grant No. ECS-9731293.
375: We would like to thank Yoseph
376: Imry, Moshe Schechter and Hamutal Bary-Soroker for useful discussions.}
377:
378: \bibliography{pc_bibdata,susc_bibdata}
379: \end{document}
380: