1: \documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
2: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
3: \usepackage{graphicx}% Include figure files
4: \usepackage{dcolumn}% Align table columns on decimal point
5: \usepackage{bm}% bold math
6: \begin{document}
7: \title{Magnetization of a strongly interacting two-dimensional electron system in perpendicular magnetic fields}
8: \author{S. Anissimova, A. Venkatesan, A.~A. Shashkin$^*$, M.~R. Sakr$^\dag$, and S.~V. Kravchenko}
9: \affiliation{Physics Department, Northeastern University, Boston, Massachusetts 02115, U.S.A.}
10: \author{T.~M. Klapwijk}
11: \affiliation{Kavli Institute of Nanoscience, Delft University of Technology, 2628 CJ Delft, The Netherlands}
12: \begin{abstract}
13: We measure the thermodynamic magnetization of a low-disordered,
14: strongly correlated two-dimensional electron system in silicon in
15: perpendicular magnetic fields. A new, parameter-free method is used
16: to directly determine the spectrum characteristics (Land\'e
17: $g$-factor and the cyclotron mass) when the Fermi level lies outside
18: the spectral gaps and the inter-level interactions between
19: quasiparticles are avoided. Intra-level interactions are found to
20: strongly modify the magnetization, without affecting the determined
21: $g^*$ and $m^*$.
22: \end{abstract}
23: \pacs{71.30.+h, 73.40.Qv}
24: \maketitle
25:
26: Magnetization is one of the least studied properties of
27: two-dimensional (2D) electron systems: signals associated with the
28: magnetization of 2D electrons are weak, and measuring them is a
29: challenging experiment. Few experimental observations of the
30: de~Haas–-van~Alphen effect in 2D electron systems were made using
31: SQUID magnetometers \cite{stormer83}, pick up coils lithographed
32: above the gate \cite{fang83}, or torque magnetometers
33: \cite{eisenstein85}. A novel method has recently been used by Prus
34: {\it et al}.\ \cite{prus03} and Shashkin {\it et al}.\
35: \cite{shashkin04a} to measure the spin magnetization of 2D electrons
36: in silicon metal-oxide-semiconductor field-effect transistors
37: (MOSFETs). This method entails modulating the magnetic field with an
38: auxiliary coil and measuring the imaginary (out-of-phase) component
39: of the AC current induced between the gate and the 2D electron
40: system, which is proportional to $\partial\mu/\partial B$ (where
41: $\mu$ is the chemical potential). Using the Maxwell relation,
42: $\partial\mu/\partial B=-\partial M/\partial n_s$, one can then
43: obtain the magnetization $M$ by integrating the induced current over
44: the electron density, $n_s$. Pauli spin susceptibility has been
45: observed to behave critically near the 2D metal-insulator transition,
46: in agreement with previous transport measurements
47: \cite{kravchenko04,shashkin04b}.
48:
49: Here we apply a similar method to study the thermodynamic
50: magnetization of a low-disordered, strongly correlated 2D electron
51: system in silicon MOSFETs in perpendicular and tilted magnetic
52: fields. By measuring $\partial\mu/\partial B$ at non-integer filling
53: factors, we directly determine the spectrum characteristics without
54: any fitting procedures or parameters. As compared to previously used
55: measuring techniques, the remarkable advantage of the novel method is
56: that it probes the spectrum of the 2D electron system with the Fermi
57: level lying outside the spectral gaps so that the effects of
58: interactions between quasiparticles belonging to different energy
59: levels (inter-level interactions) are avoided. Although
60: intra-level interactions are found to strongly affect the
61: magnetization, the extracted Land\'e $g$-factor and the cyclotron
62: mass are insensitive to them. Therefore, measured spectrum
63: characteristics are likely to be identical with those of a continuous
64: spectrum. The so-obtained $g$-factor has been found to be weakly
65: enhanced and practically independent of the electron density down to
66: the lowest densities reached ($\approx1.5\times10^{11}$~cm$^{-2}$),
67: while the cyclotron mass becomes strongly enhanced at low $n_s$.
68:
69: Measurements were made in an Oxford dilution refrigerator on clean
70: (100)-silicon samples with peak electron mobilities of 3~m$^2$/Vs at
71: 0.1~K and oxide thickness of 149~nm. Magnetic field $B$ was modulated
72: with a small AC field $B_{\text{mod}}$ in the range of 0.005 --
73: 0.03~T at a low frequency $f=0.05-0.45$~Hz to minimize possible
74: mechanical resonances and avoid overheating the sample. The latter
75: was verified by monitoring the temperature-dependent sample
76: resistance at $B=0$ and in the range of filling factors 1 to 6 with
77: and without modulation. Noticeably higher amplitudes $B_{\text{mod}}$
78: and/or frequencies $f$ caused overheating of the mixing chamber and
79: were avoided. The in-phase and out-of-phase components of the current
80: between the gate and the 2D electron system were measured with high
81: precision ($\sim10^{-16}$~A) using a current-voltage converter and a
82: lock-in amplifier. The imaginary (out-of-phase) current component is
83: equal to $\mbox{Im }i=(2\pi fCB_{\text{mod}}/e)\,d\mu/dB$, where $C$
84: is the capacitance of the sample. For measurements of the
85: capacitance, a similar circuit was used with a distinction that the
86: gate voltage was modulated and thus the imaginary current component
87: was proportional to the capacitance. The electron density was
88: determined from the capacitance oscillations.
89:
90: Typical experimental traces of the gate current in a perpendicular
91: magnetic field of 5~T are displayed in Fig.~\ref{fig1}. Sharp dips in
92: the out-of-phase component, seen at integer filling factors
93: $\nu\equiv n_shc/eB_\perp$, reflect gaps in the density of states:
94: dips at odd filling factors correspond to the valley splitting, the
95: ones at $\nu=2$ and 6 are due to the spin splitting, and the dip at
96: $\nu=4$ is due to the cyclotron splitting. However, there are no
97: corresponding features in the in-phase current component, which
98: ensures that we reach the low-frequency limit and the measured
99: $\partial\mu/\partial B$ is not distorted by lateral transport
100: effects. This is further confirmed by the fact that the out-of-phase
101: current is proportional to the excitation frequency as displayed in
102: the right-hand inset to Fig.~\ref{fig1}. Magnetization per electron
103: can be extracted by integrating the measured out-of-phase signal with
104: respect to $n_s$, as shown in the left-hand inset to Fig.~\ref{fig1}
105: for illustration. The magnetization exhibits the expected sawtooth
106: oscillations, with sharp jumps at integer filling factors (note that
107: the height of the jumps yields values that are smaller than the level
108: splitting by the level width).
109:
110: \begin{figure}
111: \scalebox{0.48}{\includegraphics[clip]{fig1.eps}}
112: \caption{\label{fig1} Out-of-phase (solid line) and in-phase (dotted
113: line) current components as a function of the electron density in a
114: perpendicular magnetic field of 5~T and $T=0.8$~K.
115: $B_{\text{mod}}=0.022$~T and $f=0.45$~Hz. The value $d\mu/dB$ is
116: indicated in units of the Bohr magneton $\mu_B$. In the right-hand
117: inset, we demonstrate proportionality of $\mbox{Im }i$ to frequency:
118: the solid and dashed lines (vertically shifted for clarity)
119: correspond to $0.45$ and $0.1$~Hz, respectively; the $y$-component of
120: the latter is multiplied by 4.5. The left-hand inset illustrates
121: magnetization per electron.}
122: \end{figure}
123:
124: If the disorder and interactions are disregarded, in quantizing
125: magnetic fields (except at integer filling factors) the derivative
126: $\partial\mu/\partial B=-\partial M/\partial n_s$ is equal to
127: \begin{equation}
128: \frac{\partial\mu}{\partial B}=\mu_B\left[\left(\frac{1}{2}+N\right)\frac{2m_e}{m_b}\pm\frac{1}{2}g_0\right],
129: \end{equation}
130: where $\mu_B$ is the Bohr magneton, $N$ is the Landau level number,
131: $m_e$ and $m_b=0.19\,m_e$ are the free electron mass and band mass,
132: respectively, and $g_0=2$ is the $g$-factor in bulk silicon. Disorder
133: smears out the dependences which otherwise would consist of a series
134: of delta-functions. Interactions modify this picture in two ways:
135: (i)~by renormalizing the values of the cyclotron mass and $g$-factor
136: and (ii)~by providing a negative contribution of order
137: $-(e^2/\varepsilon l_B)\{\nu\}^{1/2}$ to the chemical potential
138: \cite{macdonald86,efros88} (here $\varepsilon$ is the dielectric
139: constant, $l_B$ is the magnetic length, and $\{\nu\}$ is the
140: deviation of the filling factor from the nearest integer). The latter
141: effect, which is caused by the intra-level interactions between
142: quasiparticles, leads to the so-called negative thermodynamic
143: compressibility near integer filling factors predicted by Efros
144: \cite{efros88} and experimentally observed in
145: Refs.~\cite{kravchenko89,eisenstein92}.
146:
147: \begin{figure}
148: \scalebox{0.52}{\includegraphics[clip]{fig2.eps}}
149: \caption{\label{fig2} (a)~Capacitance in $B=8$~tesla and in $B=0$ as
150: indicated. The (noise averaged) geometric capacitance is depicted by
151: a dashed line. (b)~$\mbox{Im }i\propto d\mu/dB$ in a perpendicular
152: magnetic field of 8~tesla. The maximum values possible in a
153: non-interacting system (see text) are depicted by dashed lines.}
154: \end{figure}
155:
156: In Fig.~\ref{fig2}, we compare capacitance $C$ with
157: $\partial\mu/\partial B$, measured at the same magnetic field value
158: and plotted versus $n_s$ around the filling factor $\nu=2$. The
159: capacitance consists of two contributions:
160: $1/C=1/C_{\text{geo}}+1/Ae^2(dn_s/d\mu)$, where $C_{\text{geo}}$ is
161: the geometric capacitance \cite{rem} depicted by the dashed line in
162: Fig.~\ref{fig2}~(a), and $A$ is the sample area. (Note that the
163: geometric capacitance slightly increases with $n_s$ since the
164: thickness of the 2D electron layer --- and, therefore, the average
165: distance between the 2D layer and the gate --- decreases with the
166: gate voltage.) The second term is responsible for the dip centered at
167: $n_s=3.87\times10^{11}$~cm$^{-2}$, corresponding to $\nu=2$, and
168: sharp maxima on both sides of it. Note that at these maxima, the
169: capacitance exceeds $C_{\text{geo}}$, which corresponds to the
170: negative thermodynamic compressibility discussed above. Farther from
171: integer filling factors, the intra-level interaction corrections
172: become weak, being proportional to $\{\nu\}^{-1/2}$, and the measured
173: capacitance approaches $C_{\text{geo}}$ (as long as the broadening of
174: Landau levels is negligible, {\it i.e.},
175: $dn_s/d\mu\gg\left.dn_s/d\mu\right|_{B=0}$).
176:
177: Similar maxima on both sides of $\nu=2$ are seen in the magnetization
178: data shown in Fig.~\ref{fig2}~(b). At the maxima, the derivative
179: $\partial\mu/\partial B$ exceeds maximum values possible in a
180: non-interacting 2D electron gas, which are determined by Eq.~(1) and
181: are depicted in the figure by dashed lines. The possibility that
182: $\partial\mu/\partial B$ might exceed its maximum non-interacting
183: values due to intra-level Coulomb interactions between quasiparticles
184: was predicted by MacDonald {\it et al}.\ \cite{macdonald86}; in fact,
185: this is how negative compressibility \cite{efros88} manifests itself
186: in magnetization measurements. Sharp spike just above $\nu=4$ and
187: maxima on both sides of $\nu=2$ in the dependence shown in
188: Fig.~\ref{fig1} are of the same nature.
189:
190: \begin{figure*}
191: \scalebox{1.09}{\includegraphics[clip]{fig3.eps}}
192: \caption{\label{fig3} Illustration of how the effective $g$-factor
193: (a, b) and the cyclotron mass (c, d) have been measured. The
194: imaginary current component is plotted as a function of the deviation
195: of the filling factor from $\nu=2$. In perpendicular magnetic fields,
196: the difference between $\partial\mu/\partial B$ for spin-down
197: ($\downarrow$) and spin-up ($\uparrow$) electrons yields $g^*$ in
198: units of the Bohr magneton. In tilted magnetic fields, the difference
199: between $\partial\mu/\partial B$ for electrons with $N=1$ and $N=0$
200: is equal to $2\mu_B\,(m_e/m^*)\cos\,\phi$. The dashed lines show
201: noise-averaged values. $B_{\text{mod}}=0.022$~T (a, b) and 0.0055~T
202: (c, d).}
203: \end{figure*}
204:
205: It is straightforward to obtain the effective $g$-factor from the
206: data for $\partial\mu/\partial B$. In accordance with Eq.~(1), it is
207: equal (in units of the Bohr magneton) to the difference between
208: $\partial\mu/\partial B$ for spin-down ($\downarrow$) and spin-up
209: ($\uparrow$) electrons belonging to the same Landau level:
210: $\mu_B\,g^*= (\partial\mu/\partial
211: B)_\downarrow-(\partial\mu/\partial B)_\uparrow$. It is important
212: that this method of determining the $g$-factor does not require the
213: use of any fitting procedures or parameters. Figure~\ref{fig3}~(a, b)
214: shows measured $\partial\mu/\partial B$ as a function of the
215: deviation of the filling factor from 2 at two values of magnetic
216: field. Near $\nu=2$, there are sharp intra-level interaction-induced
217: structures discussed above; these regions have been excluded from the
218: analysis. However, farther from $\nu=2$, the dependences for $\nu<2$
219: and $\nu>2$ become parallel to each other. This ensures that the
220: so-determined $g^*$ is not affected by the valley splitting
221: \cite{hrapai03,valley} and intra-level interaction effects
222: \cite{macdonald86,efros88} discussed above. The latter contribute
223: equally to both spin-up and spin-down dependences and cancel each
224: other out. Disorder also contributes equally to $\partial\mu/\partial
225: B$ on both sides of $\nu=2$: we have found that at magnetic fields
226: down to approximately 3~T, there are wide regions of filling factors
227: where capacitance ({\it i.e.}, the density of states) is symmetric
228: around $\nu=2$ (see, {\it e.g.}, Fig.~\ref{fig2}~(a)); furthermore,
229: closeness of the capacitance to $C_{\text{geo}}$ attests that the
230: disorder-induced corrections are small. At lower magnetic fields,
231: however, the electron-hole symmetry around $\nu=2$ breaks down, which
232: sets the lower boundary for the range of magnetic fields (and,
233: consequently, electron densities). Note that temperature smears out
234: the dependences in a way similar to disorder: at higher temperatures,
235: the capacitance at half-integer filling factors decreases, which
236: leads to a worsening of the method accuracy.
237:
238: In Fig.~\ref{fig4} we plot the measured $g$-factor along with the one
239: previously obtained from transport measurements (solid line). One can
240: see that there is no systematic dependence of the $g$-factor on
241: $n_s$: it remains approximately constant and close to its value in
242: bulk silicon even at the lowest electron densities, which is in good
243: agreement with the transport \cite{shashkin02} and magnetocapacitance
244: \cite{hrapai03} results.
245:
246: \begin{figure}
247: \scalebox{0.52}{\includegraphics[clip]{fig4.eps}}
248: \caption{\label{fig4} The effective $g$-factor (circles) and the
249: cyclotron mass (squares) as a function of the electron density. The
250: solid and long-dashed lines represent, respectively, the $g$-factor
251: and effective mass, previously obtained from transport measurements
252: \cite{shashkin02}, and the dotted line is the Pauli spin
253: susceptibility obtained by magnetization measurements in parallel
254: magnetic fields \cite{shashkin04a}. The critical density $n_c$ for
255: the metal-insulator transition is indicated.}
256: \end{figure}
257:
258: The same method can be used for determination of the cyclotron mass
259: in tilted magnetic fields strong enough to completely polarize the
260: electron spins \cite{remark}. If (and only if) the spin splitting
261: exceeds the cyclotron splitting, the gap at $\nu=2$ lies between
262: Landau levels 0$\uparrow$ and 1$\uparrow$, and the difference
263: $(\partial\mu/\partial B)_{N=1}-(\partial\mu/\partial B)_{N=0}$ is
264: equal to $2\mu_B\,(m_e/m^*)\cos\,\phi$, where $\phi$ is the tilt
265: angle. Once the electron spins are fully polarized at filling factors
266: above $\nu=2$, the tilt angle is automatically large enough for the
267: level crossing to have occurred. The region of explorable electron
268: densities is restricted from above by the condition that the
269: electrons must be fully spin-polarized, while with our current
270: set-up, the maximum magnetic field at which we can apply the
271: modulation is only 8~tesla capable of polarizing the electron spins
272: up to $n_s^*\approx2\times10^{11}$~cm$^{-2}$
273: \cite{shashkin04a,vitkalov00}. Figure~\ref{fig3}~(c, d) shows
274: $\partial\mu/\partial B$ as a function of $|\nu-2|$ under the
275: condition $n_s<n_s^*$ at two tilt angles \cite{angle}. The extracted
276: cyclotron mass at electron densities $1.55$ and
277: $1.35\times10^{11}$~cm$^{-2}$ is significantly enhanced. At densities
278: below $1.35\times10^{11}$~cm$^{-2}$, the symmetry of capacitance on
279: both sides of the $\nu=2$ gap breaks down, making the determination
280: of $m^*$ impossible. As a result, we were only able to obtain two
281: data points. Nevertheless, good agreement with the effective mass
282: previously obtained by transport measurements (Fig.~\ref{fig4})
283: demonstrates the applicability of the new method and adds credibility
284: to both transport and magnetization results.
285:
286: We stress once again that the advantage of the new method we use here
287: is that it allows determination of the spectrum of the 2D electron
288: system under the condition that the Fermi level lies outside the
289: spectral gaps, and the inter-level interactions are avoided. Being
290: symmetric about $\nu=2$, the intra-level interactions are canceled
291: out in the data analysis and do not influence the extracted
292: $g$-factor and cyclotron mass. Therefore, the obtained values $g^*$
293: and $m^*$ are likely to be identical with those for a continuous
294: spectrum, and the comparison with previously found values of the
295: $g$-factor and the effective mass is valid.
296:
297: To summarize, thermodynamic magnetization measurements in
298: perpendicular and tilted magnetic fields allow determination of the
299: spectrum characteristics of 2D electron systems and show that
300: enhancement of the $g$-factor is weak and practically independent of
301: the electron density, while the cyclotron mass becomes strongly
302: enhanced as the density is decreased. The obtained data agree well
303: with the $g$-factor and effective mass obtained by transport
304: measurements, as well as with the Pauli spin susceptibility obtained
305: by magnetization measurements in parallel magnetic fields, even
306: though the lowest electron densities reached in the experiment are
307: somewhat higher. Thus, we arrive at the conclusion that, unlike in
308: the Stoner scenario, it is indeed the effective mass that is
309: responsible for the dramatically enhanced spin susceptibility at low
310: electron densities.
311:
312: We gratefully acknowledge discussions with V.~T. Dolgopolov, B.~I.
313: Halperin, and M.~P. Sarachik. This work was supported by the National
314: Science Foundation grant DMR-0403026, the ACS Petroleum Research Fund
315: grant 41867-AC10, the RFBR, RAS, and the Programme ``The State
316: Support of Leading Scientific Schools''.
317:
318: {\it Note added in proof}.---After this work had been completed, we
319: learned that Punnoose and Finkelstein \cite{punnoose05} made a
320: renormalization group analysis for multi-valley 2D systems. Their
321: conclusion that the effective mass dramatically increases at the
322: metal-insulator transition while the $g$-factor remains nearly intact
323: is consistent with our experimental results.
324:
325: \begin{thebibliography}{apssamp}
326: \bibitem[*]{1} Permanent address: Institute of Solid State Physics, Chernogolovka, Moscow District 142432, Russia.
327: \bibitem[\dag]{2} Present address: Department of Physics and Astronomy, UCLA, Los Angeles, CA 90095, U.S.A.
328: \bibitem{stormer83} H. Stormer, T. Haavasoja, V. Narayanamurti, A.~C. Gossard, and W. Wiegmann, J.\ Vac.\ Sci.\ Tech.\ B {\bf 1}, 423 (1983); I. Meinel, D. Grundler, S. Bargstadt-Franke, C. Heyn, and D. Heitmann, Appl.\ Phys.\ Lett.\ {\bf 70}, 3305 (1997).
329: \bibitem{fang83} F.~F. Fang and P.~J. Stiles, Phys.\ Rev.\ B {\bf 28}, 6992 (1983).
330: \bibitem{eisenstein85} J.~P. Eisenstein, Appl.\ Phys.\ Lett.\ {\bf 46}, 695 (1985); J.~P. Eisenstein, H.~L. Stormer, V. Narayanamurti, A.~Y. Cho, A.~C. Gossard, and C.~W. Tu, Phys.\ Rev.\ Lett.\ {\bf 55}, 875 (1985); S.~A.~J. Wiegers, M. Specht, L.~P. Levy, M.~Y. Simmons, D.~A. Ritchie, A. Cavanna, B. Etienne, G. Martinez, and P. Wyder, Phys.\ Rev.\ Lett.\ {\bf 79}, 3238 (1997); M. Zhu, A. Usher, A.~J. Matthews, A. Potts, M. Elliott, W.~G. Herrenden-Harker, D.~A. Ritchie, and M.~Y. Simmons, Phys.\ Rev.\ B {\bf 67}, 155329 (2003).
331: \bibitem{prus03} O. Prus, Y. Yaish, M. Reznikov, U. Sivan, and V. Pudalov, Phys.\ Rev.\ B\ {\bf 67}, 205407 (2003).
332: \bibitem{shashkin04a} A.~A. Shashkin, S. Anissimova, M.~R. Sakr, S.~V. Kravchenko, V.~T. Dolgopolov, and T.~M. Klapwijk, Phys.\ Rev.\ Lett.\ {\bf 96}, 036403 (2006).
333: \bibitem{kravchenko04} S.~V. Kravchenko and M.~P. Sarachik, Rep.\ Prog.\ Phys.\ {\bf 67}, 1 (2004).
334: \bibitem{shashkin04b} A.~A. Shashkin, Physics-Uspekhi {\bf 48}, 129 (2005).
335: \bibitem{efros88} A. L. Efros, Solid State Commun. {\bf 65}, 1281 (1988).
336: \bibitem{macdonald86} A.~H. MacDonald, H.~C.~A. Oji, and K.~L. Liu, Phys.\ Rev.\ B {\bf 34}, 2681 (1986).
337: \bibitem{kravchenko89} S.~V. Kravchenko, V.~M. Pudalov, and S.~G. Semenchinsky, Phys.\ Lett.\ A {\bf 141}, 71 (1989).
338: \bibitem{eisenstein92} J.~P. Eisenstein, L.~N. Pfeiffer, and K.~W. West, Phys.\ Rev.\ Lett.\ {\bf 68}, 674 (1992).
339: \bibitem{rem} The geometric capacitance has been determined according
340: to $1/C_{\text{geo}}=1/C|_{B=0}-1/Ae^2\left.(dn_s/d\mu)\right|_{B=0}$
341: (for more on this procedure, see Ref.~\cite{hrapai03}).
342: \bibitem{hrapai03} V.~S. Khrapai, A.~A. Shashkin, and V.~T. Dolgopolov, Phys.\ Rev.\ Lett.\ {\bf 91}, 126404 (2003); Phys.\ Rev.\ B\ {\bf 67}, 113305 (2003).
343: \bibitem{valley} The bare valley splitting is independent of the
344: magnetic field and does not contribute to $\partial\mu/\partial B$.
345: However, it may be enhanced by inter-level interactions
346: \cite{hrapai03}. The fact that $\partial\mu/\partial B$ for $\nu<2$
347: and $\nu>2$ are parallel to each other ensures that a possible
348: influence of the enhanced valley splitting is negligible in our
349: experiment.
350: \bibitem{shashkin02} A.~A. Shashkin, S.~V. Kravchenko, V.~T. Dolgopolov, and T.~M. Klapwijk, Phys.\ Rev.\ B {\bf 66}, 073303 (2002).
351: \bibitem{remark} The effects of finite layer thickness, which lead to
352: an increase of the effective mass with parallel magnetic field, are
353: negligible in silicon MOSFETs \cite{kravchenko04,shashkin04b}.
354: \bibitem{vitkalov00} S.~A. Vitkalov, H. Zheng, K.~M. Mertes, M.~P. Sarachik, and T.~M. Klapwijk, Phys.\ Rev.\ Lett.\ {\bf 85}, 2164 (2000); A.~A. Shashkin, M. Rahimi, S. Anissimova, S.~V. Kravchenko, V.~T. Dolgopolov, and T.~M. Klapwijk, Phys.\ Rev.\ Lett.\ {\bf 91}, 046403 (2003).
355: \bibitem{angle} Using $n_s^*(8~\mbox{T})\approx2\times
356: 10^{11}$~cm$^{-2}$, one can estimate that the coincidence of spin and cyclotron splittings for $\nu=2$ occurs at
357: $\phi^*=\cos^{-1}(n_s^*hc/2eB)\approx59^\circ$.
358: \bibitem{punnoose05} A. Punnoose and A.~M. Finkelstein, Science {\bf 310}, 289 (2005).
359: \end{thebibliography}
360:
361: \end{document}
362: