1: \documentstyle[prl,aps,multicol,psfig,epsfig]{revtex}

2: \special{papersize=8.5in,11in}

3: 

4: \begin{document}

5: \draft

6: 

7: \author{V.\ M.\ Pudalov$^{a,b}$, M.\ E.\ Gershenson$^{a}$, and H.\ Kojima$^{a}$}

8: \address{$^a$ Department of Physics and Astronomy, Rutgers University,

9: New Jersey 08854, USA}

10: \address{$^b$ P.\ N.\ Lebedev Physics

11: Institute/Lebedev Research Centre in Physics, 119991 Moscow, Russia}

12: 

13: \date{\today}

14: \title{Absence of Ferromagnetic Instability

15: at the Metal-Insulator Transition in Si-inversion Layers}

16: \maketitle

17: 

18: \begin{abstract}

19: We have measured the

20: Shubnikov-de Haas oscillations in high-mobility Si MOS structures

21: over a wide range of the carrier densities $n\geq 0.77\times

22: 10^{11}$cm$^{-2}$. This range includes the critical density $n_c$

23: of the metal-insulator transition for two samples studied. The

24: periodicity of oscillations clearly demonstrates that the electron

25: states remain fourfold degenerate down to and at the 2D MIT. Both

26: the effective  spin susceptibility $\chi^*$ and mass $m^*$

27: remain finite and show no signatures of divergency

28: at the critical density for both samples  studied. To

29: test possible divergency of  $\chi^*(n)$ and

30:  $m^*(n)$ at even lower densities, we have analyzed the

31: data on $\chi^*(n)$ and $m^*(n)$ in terms of a critical dependence

32: $\chi^*, m^* \propto (n/n_0 -1)^{-\alpha}$. Our data suggest that

33: $\chi^*$ and $m^*$ may diverge at $n_0 \lesssim 0.5\times 10^{11}$cm$^{-2}$

34: ($r_s\geq 12$), which is significantly  smaller than $n_c$.

35: 

36: 

37: \end{abstract}

38: 

39: 

40: \pacs{71.30.+h, 73.40.Qv}

41: 

42: 

43: \vspace{-0.1in}

44: \begin{multicols}{2}

45: 

46: Despite intensive experimental and theoretical efforts (see, e.g.,

47: Ref.~\cite{aks} for a  bibliography),

48: the origin  of the apparent

49: ``metal-insulator transition in two dimensions'' (2D MIT) remains to

50: be the subject of ongoing discussion. This phenomenon addresses a

51: fundamental problem of the ground state of strongly correlated and

52: disordered electron systems. In (100) Si inversion layers, the 2D

53: MIT is observed at a sample-dependent  critical electron density

54: $n_c \sim 1\times 10^{11}$\,cm$^{-2}$ \cite{JETPL_98a}.

55: 

56: 

57: One of the important unsolved problems is a possible magnetic

58: instability \cite{finkelstein} in spin  or valley systems. The

59: electron-electron interactions drive a 2D

60:  system towards magnetic transition;

61: numerical calculations for the critical value of $r_s$

62: \cite{note1} at the instability vary from 13 to 20

63: \cite{isihara_82}. According to these calculations, the

64: ferromagnetic transition is likely to be of first order with a

65: complete rather than a partial (ferrimagnetic) spin polarization.

66: An interesting interpretation

67: \cite{shashkin_0007402,vitkalov_0009454} of the parallel-field

68: magnetoresistance in Si inversion layers suggested a

69: ferromagnetic instability at or very close to $n_c$ [where $n_c

70: \approx (0.8-0.85)\times10^{11}$cm$^{-2}$, which corresponds to

71: $r_s \approx 9$]. The idea of magnetic instability and  its

72: possible link to the 2D MIT is important and requires a careful

73: examination.

74: 

75: 

76: In this Letter, we report on our experimental test of two possible

77: scenarios for the magnetic

78: instability: (i) complete spin and/or valley polarization occurs

79: spontaneously at a sample-dependent critical density of the MIT,

80: and (ii) $\chi^*$ diverges at a universal (sample-independent)

81: value of $n=n_0$. To test the first scenario, we measured the

82: Shubnikov-de Haas (SdH) oscillations and determined the degeneracy of

83: the electron system across the 2D MIT in two samples (down to

84: $n=0.77\times 10^{11}$cm$^{-2}$). We found that the period of

85: oscillations corresponds to the double-degenerate spin and valley

86: states even in the presence of external magnetic field $B\approx 0.5$\,T.

87: This rules out the possibility of a complete

88: spin/valley polarization at the 2D MIT. To test the

89: second scenario, we analyzed independent measurements of

90:  $\chi^*(n)$ and  $m^*(n)$, and found

91: that each could  be described by the same critical dependence

92: $(n/n_0-1)^{-\alpha}$, if we impose an upper limit on the density

93: of the instability, $n_0 \lesssim 0.5\times

94: 10^{11}$cm$^{-2}$ ($r_s\geq 12$), and a lower limit on the

95: critical index $\alpha \gtrsim 0.6$. The density $n_0$ is

96: significantly lower than $n_c$ for the samples studied.

97: 

98: The measurements were performed on two Si-MOS samples: Si6-14

99: (peak mobility $\mu^{\rm peak}\simeq 2.2$m$^2$/Vs) and Si5

100: ($4.3$m$^2$/Vs), with the critical density $n_c$ of

101: the apparent MIT $1.0\times10^{11}$cm$^{-2}$ and

102: $0.77\times10^{11}$cm$^{-2}$, respectively.

103: The MOS structures are made on a (001)-Si wafer

104: with [100] source-drain orientation; the gate oxide

105: thickness was $\approx 190$\,nm. The density of  electrons

106: was controlled by the gate voltage $V_g$ and determined from the

107: period of SdH oscillations. An in-plane field $B_{\parallel}\geq

108: 0.02$\,T was applied for quenching the superconductivity of the Al

109: contact pads and gate electrode. Details of the experimental technique

110: can be found in Refs.~\cite{gm,crossed}

111: 

112: Typical SdH oscillations of the resisitivity $\rho_{xx}\equiv \rho$

113: are shown in Fig.~1\,a as a

114: function of $B_\perp$. Due to a high electron mobility, oscillations

115: were detected  in fields down to 0.14\,T. To examine directly the

116: first scenario, we focus on the  period of

117: the SdH oscillations, the quantity which is not renormalized by

118: interactions.

119: 

120: For such low densities as presented in Fig.~1, the oscillations

121: $\rho(B_\perp)$ are ``shaped'' mostly by the spin energy gaps

122: \cite{okamoto,gm,termination94,kravSSC2000}. Figure~1\,b shows

123: that the magnitude of oscillations increases with an in-plane

124: magnetic field $B_{\parallel}$. This confirms that the ratio of

125: the Zeeman energy $E_Z=g^*\mu_B B_\perp$ to the cyclotron energy

126: $\hbar \omega_c$ is within the interval $1/2 <

127: E_Z/\hbar\omega_c < 1$, in good agreement with the measured

128: values of $\chi^*(n)$ \cite{gm}, which control the calculated

129: energy spectrum (the upper inset to Fig.~1\,b). Curve {\em 5} in

130: Fig.~1\,a corresponds to the density $n=n_c$ for sample Si6-14.

131: The latter has been

132: determined in the insulating regime, $\rho

133: \propto \exp(-\Delta/k_B T)$,  by extrapolating the density dependence of the

134: activation energy $\Delta(n)$  to zero  \cite{prl93} (see the inset to Fig.~1\,b).

135: 

136: 

137: 

138: 

139: 

140: \vspace{0.05in}

141: \begin{figure}

142: \centerline{\psfig{figure=Fig1.eps,width=235pt,height=190pt

143: }}

144: \vspace{0.1in}

145: \begin{minipage}{3.2in}

146: \caption{a) SdH oscillations for  the sample Si6-14 at six

147: densities near $n_c$,  $T=0.2$\,K. Curves $3-6$ are terminated

148: at the onset of a large insulating peak in $\rho$

149: \protect\cite{pud_92}. b) Enhancement of oscillations with

150: $B_{\parallel}$. The upper inset shows the energy spectrum for

151: $B_\perp=0.5$\,T, $n=1.0\times 10^{11}$cm$^{-2}$, $g^*m^*/2m_b=4.35$,

152: $m^*=0.5m_e$ \protect\cite{gm}. Vertical arrows depict spin

153: polarization and the direction of the corresponding level shift

154: with $B_{\parallel}$. The lower inset illustrates

155: determination of the critical density

156: for Si6-14.

157: Densities are given in units of

158: $10^{11}$\,cm$^{-2}$.}

159: \end{minipage}

160: \label{fig1}

161: \end{figure}

162: 

163: 

164: In order to emphasize the low-field region, and to clearly

165: illustrate the SdH periodicity, we present the experimental data

166: normalized by the amplitude of the first SdH harmonic $A_1(B_\perp)$ \cite{SdH,gm}.

167: In evaluating $A_1$, we used the values of $\chi^*(n) \propto

168: g^*m^*$ and $m^*(n)$ measured in Ref.~\cite{gm}; the Dingle

169: temperature was adjusted to match damping of the measured oscillations.

170: Figure~2 shows oscillations of the resistivity $\delta

171: \rho/ \rho_{0}A_1$ as a function of the Landau level filling,

172: $\nu=nh/eB_\perp$. The $\delta \rho(\nu)/\rho_{0}A_1$ data (dots)

173: in Figs.~2\,a--c correspond to the $\rho(B_\perp)$ data {\em 1, 5} and {\em

174: 6} in Fig.~1\,a.

175: It is important to limit the field range $B_\perp \leq

176: 1T$ in the

177: analysis of the SdH oscillations  in order to diminish the magnetic-field-induced spin

178: polarization and reentrant quantum Hall effect-to-insulator

179: transitions \cite{pud_92}. The former limitation  was violated at

180: $\nu <5$: doubling of the period for $\nu=4$ in Fig.~2\,a

181: illustrates lifting  the spin degeneracy  by the perpendicular

182: field $B_\perp \sim 1.3\,T$. The latter limitation was violated for

183: $\nu<10$ in Figs.~2\,b-e and  may account for the large oscillation

184: amplitude.

185: 

186: 

187: For all the densities studied, including $n = n_c$ (Figs. 2\,b\,e), the

188: low-field oscillations $\delta \rho/ \rho_{0}$ have the

189: period $\Delta\nu = 4$ that corresponds to a double-degenerate

190: {(\em i.e. unpolarized)} spin and valley system. The minima of

191: $\delta \rho$ in Figs.~2 are located at $\nu=6, 10, 14, 18$,

192: in contrast to  $\nu =4, 8, 12, 16$, as observed for higher

193: densities. This is in agreement with earlier results

194: \cite{termination94,kravSSC2000,pud_92} and with the measured

195: $\chi^*$ values \cite{gm}. The sign of oscillations changes

196:  due to the Zeeman factor

197: $cos(\pi E_Z/\hbar\omega_c)$,  when $E_Z$ exceeds $\hbar\omega_c/2$

198: (at $r_s > 6.3$) \cite{gm,SdH}.

199: 

200: 

201: \vspace{0.05in}

202: \begin{figure}

203: \centerline{\psfig{figure=Fig2.eps,width=230pt}}

204: \vspace{0.1in}

205: \begin{minipage}{3.2in}

206: \caption{Oscillatory component of

207: the resistivity

208: measured for samples Si6-14

209: (panels a-c) and Si5

210: (panels d, e). The data are shown as dots, the

211: fits as lines \protect\cite{gm}; both are normalized by $A_1(B_\perp)$.

212: The temperature is 0.2\,K for traces (a - c) and 0.03\,K for

213: traces d and e.

214: The values of $n$ (in units of $10^{11}$\,cm$^{-2}$),

215:  $r_s$, and $g^*m^*$ are shown in each panel.}

216: \end{minipage}

217: \label{Fig.2}

218: \end{figure}

219: 

220: 

221: 

222: From these  data, we can estimate the spontaneous spin

223: polarization $P_0$  near  $n_c$.

224: The total spin polarization of the interacting

225: 2D electron system is:

226: \begin{equation}

227: P  \equiv \frac{n_{\uparrow}-n_{\downarrow}}{n} = P_Z +P_0=

228: \frac{\chi^* B_\perp}{2 \mu_B n} +P_0.

229: \end{equation}

230: Figures~2\,c and 2\,e show

231: that the complete spin (or valley) polarization does not occur at least down to

232: $\nu=6$ ($P_Z\approx 0.3$), for both $B_\parallel =0$ and

233: $B_\parallel=0.3$T (Fig.~1b); thus, the spontaneous

234: component $P_0 <1-0.3=0.7$.  We can impose even  more restrictive

235: upper limit on the spontaneous spin polarization across the 2D

236: MIT, $P_0 \lesssim 0.18$, by noting that no nodes of beating

237: are seen in SdH oscillations in

238: Figs.~2\,c, 2\,e over the  interval of $\nu = 5 - 20$.

239: For example, a larger value of $P_0= 0.19$ would induce a beating node

240: in the SdH oscillations at $\nu=15$.

241: For even lower densities, the period

242: of the $\rho$ oscillations \cite{termination94,pud_92} demonstrates that

243: both the spin and valley states remain double-degenerate across

244: the 2D MIT.

245: These {\em direct} data provide very strong evidence against

246:  a complete spontaneous spin/valley polarization

247: at the sample-dependent critical density for $n \geq 0.77\times 10^{11}$cm$^{-2}$.

248: 

249: 

250: 

251: Below, we  explore the second scenario, a divergence of

252:  $\chi^*$ and  $m^*$

253: at a sample-independent density

254: $n_0$. Indeed, it has been found in  recent measurements

255: \cite{okamoto,gm} that  both quantities

256: increase with decreasing carrier concentration and are sample-independent

257: ($\chi^*$ - to within $\pm 2\%$, and $m^*$ - $\pm 4$\%).

258: We examine our data on $\chi^*(n)$ and $m^*(n)$ \cite{gm,twomasses}

259: for sample Si6-14

260:  for possible critical density dependence in the form

261: $(n/n_0-1)^{-\alpha}$.

262: For a given $n_0$, the exponents $\alpha_\chi$ and $\alpha_m$

263: have been obtained from fitting

264:  $\chi^*(n)$ and $m^*(n)$. An example with

265: $n_0=0.53\times 10^{11}$cm$^{-2}$ and $\alpha=0.63$

266: is shown in  Figs.~3\,a,\,b. The exponents $\alpha_\chi$ and $\alpha_m$

267: versus $n_0$ (hash marks indicate the error bars) are

268: shown  in the inset in Fig.~3a.

269: The standard deviation of $\alpha(n_0)$  has such a shallow minimum

270: that the optimal $n_0$ value

271: could not be determined reliably;

272: the large uncertainty was mainly caused by

273: unknown critical range of densities.

274: Similar uncertain situation was encountered in the critical analysis

275: of the $m^*$ data for sample Si6-14 (see Figs.~3b and 3c).

276: 

277: We now include  into consideration

278: the data for sample Si5 which has

279: substantially lower $n_c$.  Three additional $m^*(n)$ points from Si5

280: are shown as diamonds in Fig.~3\,c. We find that the critical dependence cannot fit the data

281: when $n_0 $ is taken greater than  $0.65\times 10^{11}$cm$^{-2}$; this value sets

282:  the upper limit for possible $n_0$. Although only the $m^*(n)$

283:  data were available for Si5, $\chi^*(n)$ can be

284: estimated independently from the lineshape and phase of the SdH oscillations.

285: The SdH pattern is very sensitive to the ratio

286: $E_Z/\hbar\omega_c$: when this ratio becomes greater than 3/2 ($g^*m^*\geq 7.89$)

287: or smaller than 1/2 ($g^*m^*\leq 2.63$), the phase of oscillations

288: changes by $\pi$. Theoretical curves in Figs.~2\,d and 2\,e show that, even

289: before the phase reverses for all oscillations, the highest-field oscillation ($\nu=6$)

290: splits starting from $g^*m^* \approx 7$.

291: The absence of  such behavior in the measured $\delta\rho(B_\perp)$

292: traces enables us to obtain,

293: correspondingly, the upper and lower estimates for $g^*m^*$  at six densities in the range

294: from $n=0.768$ to $0.884\times 10^{11}$cm$^{-2}$.

295: Three of these estimates are shown by

296: vertical bars in Fig.~3a (three others, located between the shown bars,

297: are omitted for clarity).

298: 

299: 

300: 

301: 

302: Small values of $n_0 \ll n_c$ can be certainly accommodated by the critical dependence;

303: however, we searched for the upper limit on $n_0$, in order to determine

304: how close  $n_0$ could be to $n_c$. For this reason, we have used the {\em upper}

305: limits for $\chi^*$ (the top of bars in Fig.~3\,a) when we plotted the critical

306: dependence in Fig.~3\,a.

307: The $\chi^*(n)$ data  for both Si5 and Si6-14 obey a

308: common critical dependence only if

309:  we choose $n_0 \lesssim 0.53\times 10^{11}$cm$^{-2}$.

310:  This choice of $n_0$ also provides the lower

311: limit for the critical index $\alpha_\chi \geq 0.63$.

312: This procedure defines the range of densities

313: where the critical behavior holds:

314: e.g., for $n_0=0.53\times 10^{11}$cm$^{-2}$, this range corresponds to

315: $n <1.5\times 10^{11}$cm$^{-2}$.

316:  A similar but less restrictive conclusion follows  from the

317: critical analysis for $m^*$ (Figs.~3b,c): $n_0 \lesssim 0.65\times

318: 10^{11}$cm$^{-2}$ and $\alpha_m \geq 0.4$.

319: It is important to note that the upper limit

320: on $n_0$ is  a factor of 1.5--2 lower than the critical density

321: $n_c$ for the samples studied.

322: 

323: 

324: 

325: \vspace{0.05in}

326: \begin{figure}

327: \centerline{\psfig{figure=Fig3.eps,width=230pt}}

328: \vspace{0.1in}

329: \vspace{0.05in}

330: \begin{minipage}{3.2in}

331: \caption{ Log-log plots of (a) the spin susceptibility

332: $\chi^*/\chi_b$  and (b,c) the  mass $m^*/m_b$  vs $(n/n_0

333: -1)$: dots for sample Si6-14,  diamonds and bars for  Si5. The $m^*(n)$ data

334: are plotted for two values of $n_0 = 0.53$  and $0.65\times

335: 10^{11}$\,cm$^{-2}$. On panels (a):

336: the vertical bars extend from the upper to lower   limits

337: for $\chi^*$, as discussed in the text.

338: On panels (b,c): open and closed symbols depict the upper

339: and lower estimate for $m^*$, obtained from the $T$-dependence

340: of SdH amplitude \protect\cite{gm,twomasses}. The

341: dashed lines show the critical behavior $\chi^*, m^* \propto

342: (n/n_0-1)^{-\alpha}$. The inset shows the critical indices

343: $\alpha_\chi$ and $\alpha_m$ vs $n_0$. }

344: \label{fig3}

345: \end{minipage}

346: \end{figure}

347: 

348: 

349: 

350: 

351: 

352: 

353: For a given $n_0$ value, the estimated

354: critical indices  $\alpha_\chi$ and $\alpha_m$ are close to each other

355: (see the inset to Fig.~3a).

356: This might be expected:

357:  there is no experimental indication for a

358:  critical behavior of $g^*(n)$ \cite{gm,SKDK0111478},

359: therefore $\chi^* \propto g^*m^*$ and $m^*$ should exhibit the same

360: $n$- dependence (a critical behavior or otherwise).

361: 

362: 

363: The conclusion on the absence of the magnetic instability at

364: $n\approx n_c$, which we have drawn from our analysis, differs from the

365: one suggested in Refs.~\cite{shashkin_0007402,vitkalov_0009454}.

366: There might be several reasons for this disagreement. Firstly, we

367: believe that the magnetoresistance data for in-plane fields,

368: analyzed in Refs.~\cite{shashkin_0007402,vitkalov_0009454}, might be indirectly

369: related to the spin susceptibility of {\em mobile} electrons.

370: Secondly, measurements in

371: Refs.~\cite{shashkin_0007402,vitkalov_0009454} were taken in {\em

372: strong} in-plane fields. The strong fields drive a 2D system into

373: the hopping regime \cite{shashkin}; the characteristic values of

374: $B_\parallel$ go to zero as

375: $n$ approaches $n_c$ \cite{aniso}. Moreover, even moderate  fields

376: $B_\parallel < E_F/g^*\mu_B$

377:  induce non-linearity of magnetization,  i.e. the $\chi^*(B)$-dependence

378:  \cite{gm,nonlinear}. In contrast, our

379: {\em direct} measurements of $\chi^*$ have been performed

380: in the {\em low-field} linear regime.

381: 

382: 

383: 

384: To summarize, we measured  Shubnikov-de Haas oscillations in weak

385: perpendicular fields  over a wide density range $n \geq 0.77 \times

386: 10^{11}$cm$^{-2}$, which includes the sample-specific critical densities $n_c$ of the 2D

387: MIT for two different samples. It has been found that the period

388: of oscillations corresponds to the fourfold degeneracy of

389: spin/valley systems on both sides of the 2D MIT.

390: Our results demonstrate that the apparent 2D MIT

391: is not accompanied by

392: a spontaneous complete polarization of spins or valleys at zero

393: field. Moreover, the experimental data allow us to put an upper

394: limit $P_0<0.18$ on the value of a possible spontaneous

395: polarization at the transition.

396: We also

397: explored a possibility of critical behavior of

398: the renormalized spin susceptibility and the effective mass at a

399: sample-independent density $n_0$.

400: We found that the

401: divergence of both $\chi^*$ and $m^*$ is unlikely for

402: $n_0 > 0.65\times 10^{11}$cm$^{-2}$.

403: However, it may occur at lower densities

404: (significantly less than $n_c$): e.g. for $\chi^*$ -- at

405: $n_0\lesssim 0.5\times 10^{11}$cm$^{-2}$

406: and $\alpha \gtrsim 0.6$.

407: 

408: Authors are grateful to E.\ Abrahams, B.\ L.\ Altshuler, and D.\ L.\

409: Maslov for discussions. The work was supported by the NSF, ARO

410: MURI, NWO, NATO, RFBR, INTAS, and the Russian programs ``Physics of

411: Nanostructures'', ``Quantum and Non-linear Processes'',

412: ``Quantum computing and telecommunications'',

413: ``Integration of High Education and Academic Research'', and

414: ``The State Support of Leading Scientific Schools''.

415: 

416: 

417: 

418: 

419: \begin{references}

420: \vspace{-0.4in}

421: \bibitem{aks}E.\ Abrahams,  S.\ V.\ Kravchenko, and M.\ P.\ Sarachik,

422: Rev. Mod. Phys. {\bf 73}, 251 (2001).

423: 

424: \bibitem{JETPL_98a}V.\ M.\ Pudalov, G.\ Brunthaler, A.\ Prinz, G.\

425: Bauer, JETP Lett. {\bf 68}, 442 (1998). Cond-mat/9801077.

426: 

427: 

428: \bibitem{finkelstein} A.\ M.\ Finkelstein, Sov. Sci. Rev. A {\bf 14}, 3

429: (1990).

430: 

431: \bibitem{note1} The parameter

432: $r_s = 2.63\times (10^{12}{\rm cm}^{-2}/n)^{1/2}$

433: which characterizes the electron-electron interactions,

434: is the  ratio of the Coulomb interaction energy $E_{ee}$ to the Fermi

435: energy $E_F$. The bare values of the Fermi energy $E_F$, the effective mass $m_b=0.19$,

436: and the dielectric constant $\kappa =7.8$ at the

437: Si-SiO$_2$ interface are used in its definition.

438: 

439: 

440: \bibitem{isihara_82}For reviews, see:  G.\ Senatore, S.\ Moroni, D.\ Varzano,

441: Sol. St. Commun. {\bf 119}, 333 (2001). A.\ Isihara and L.\ C.\

442: Ioriatti, Jr.,  Phys. Rev. B {\bf 25}, 5534 (1982).

443: 

444: \bibitem{shashkin_0007402}A.\ A.\ Shashkin, S.\ V.\ Kravchenko, V.\

445: T.\ Dolgopolov, and T.\ M.\ Klapwijk, Phys. Rev. Lett. {\bf 87},

446: 086801 (2001).

447: 

448: \bibitem{vitkalov_0009454}S.\ A.\ Vitkalov, H.\ Zheng, K.\ M.\

449: Mertez, M.\ P.\ Sarachik, and T.\ M.\ Klapwijk,  Phys. Rev. Lett.

450: {\bf 87}, 086401 (2001).

451: 

452: \bibitem{okamoto}T.\ Okamoto, K.\ Hosoya, S.\ Kawaji, and A.\ Yagi,

453: Phys. Rev. Lett. {\bf 82}, 3875 (1999).

454: 

455: \bibitem{gm}V.\ M.\ Pudalov, M.\ E.\ Gershenson, H.\ Kojima, N.\ Butch, E.\ M.\

456: Dizhur, G.\ Brunthaler, A.\ Prinz, and G.\ Bauer,

457: Phys. Rev. Lett. {\bf 88}, 196404 (2002).

458: 

459: \bibitem{crossed}M.\ E.\ Gershenson, V.\ M.\ Pudalov, H.\ Kojima,

460: N.\ Butch, E.\ M.\ Dizhur, G.\ Brunthaler, A.\ Prinz, and G.\ Bauer,

461: Physica E {\bf 12}, 585 (2002).

462: 

463: \bibitem{termination94}V.\ M.\ Pudalov, M.\ D'Iorio, J.\ W.\

464: Campbell, JETP Lett., {\bf 57}, 608 (1993). Surf. Sci. {\bf 305}, 107 (1994).

465: 

466: \bibitem{kravSSC2000}S.\ V.\ Kravchenko, A.\ A.\ Shashkin, D.\ A.\

467: Bloore, T.\ M.\ Klapwijk, Sol. State Commun. {\bf 116}, 495 (2000).

468: 

469: 

470: \bibitem{prl93} V.\ M.\ Pudalov, M.\ D'Iorio, S.\ V.\ Kravchenko, J.\ W.\

471: Campbell, Phys. Rev. Lett. {\bf 70}, 1866 (1993).

472: 

473: \bibitem{pud_92}M.\ D'Iorio, V.\ M.\ Pudalov, and

474: S.\ G.\ Semenchinsky, Phys. Lett A, {\bf 150}, 422 (1990).

475: Phys. Rev. B {\bf 46}, 15992 (1992).

476: 

477: \bibitem{SdH}I.\ M.\ Lifshitz and A.\ M.\ Kosevich,

478: Zh. Eks. Teor. Fiz. {\bf 29}, 730 (1955).

479: A.\ Isihara, L.\ Smr$\check{c}$ka,

480:  J. Phys. C: Solid State Phys. {\bf 19}, 6777 (1986).

481: 

482: \bibitem{twomasses}The two values for the effective mass (open and closed symbols)

483: are its upper and lower border \protect\cite{gm} and

484: originate from lacking of adequate theory for the temperature

485: dependence of $T_D$ and $m^*$ under conditions of $k_Fl \sim 1$.

486: 

487: \bibitem{shashkin}A.\ A.\ Shashkin, S.\ V.\ Kravchenko,

488: and T.\ M.\ Klapwijk, Phys. Rev. Lett. {\bf 87}, 266402 (2001).

489: 

490: \bibitem{SKDK0111478}A.\ A.\ Shashkin, S.\ V.\ Kravchenko,

491: V.\ T.\ Dolgopolov, and T.\ M.\ Klapwijk, cond-mat/0111478.

492: 

493: \bibitem{aniso} V.\ M.\ Pudalov, G.\ Brunthaler, A.\ Prinz,

494: and G.\ Bauer,  cond-mat/0103087.

495: 

496: \bibitem{nonlinear}V.\ M.\ Pudalov, M.\ E.\ Gershenson, H.\ Kojima,

497: to be published elsewhere.

498: 

499:  \end{references}

500: \end{multicols}

501: \end{document}

502: