1: \documentclass[twocolumn,prb,amssymb,showpacs]{revtex4}
2: \usepackage{graphicx}
3:
4: \begin{document}
5:
6: \title{Nonohmic conductivity as a probe of crossover from diffusion to hopping in two dimensions}
7:
8: \author{G.~M.~Minkov}
9: \author{A.~A.~Sherstobitov}
10: \author{O.~E.~Rut}
11: \author{A.~V.~Germanenko}
12:
13: \affiliation{Institute of Physics and Applied Mathematics, Ural
14: State University, 620083 Ekaterinburg, Russia}
15:
16: \date{\today}
17:
18: \begin{abstract}
19: We show that the study of conductivity nonlinearity gives a
20: possibility to determine the condition when the diffusion
21: conductivity changes to the hopping one with increasing disorder.
22: It is experimentally shown that the conductivity of single
23: quantum well GaAs/InGaAs/GaAs heterostructures behaves like
24: diffusive one down to value of order $10^{-2}e^2/h$.
25: \end{abstract}
26: \pacs{73.20.Fz, 73.61.Ey}
27:
28: \maketitle
29:
30:
31: Two opposite models are used to describe the conductivity in two
32: dimensional (2D) systems at low temperature. The first one
33: considers an electron as free particle with well defined
34: quasimomentum, which sometime scatters by static scattering
35: potential. The quantum corrections to the conductivity (due to
36: interference and interaction) determine in the main the
37: temperature and magnetic field dependences of the conductivity. It
38: is clear this model is valid for $k_Fl>1$, where $k_F$ and $l$ are
39: Fermi quasimomentum and mean free path, respectively.
40:
41: The second model treats electrons as localized particles which
42: rarely hope from one localized state to another one. All the
43: results for this case are obtained in the framework of percolation
44: theory which is valid when the probability of the transitions
45: between the states is very small so that the dispersion of the
46: probabilities is exponentially large and percolation theory
47: becomes applicable. This theory gives\cite{Mott,ES}
48:
49: \begin{equation}
50: \rho(T)=\rho_0(n,T)\exp({T_0/T})^p,\; 0<p<1 \label{eq1}.
51: \end{equation}
52: where $n$ is electron density and $T_0$ is constant depending on
53: localization length. The value of $p$ depends on relationship
54: between $k_B T$ and Coulomb gap. This expression is valid when the
55: exponent $({T_0/T})^p$ is larger than $10$.\cite{ES} The value of
56: $\rho_0(n,T)$ depends on overlapping of the wave functions of the
57: localized states and hopping mechanism, but for all cases it is
58: larger than $h/e^2$. So, the expressions for hopping conductivity
59: are valid when conductivity is less than $10^{-4} e^2/h$.
60:
61: Thus the conductivity of 2D systems has to be described by theory
62: of quantum corrections when $\sigma>e^2/h$ and by theory of hoping
63: conductivity when $\sigma<10^{-4}e^2/h$. Over wide conductivity
64: range $e^2/h>\sigma > 10^{-4}e^2/h$ both approaches are not
65: applicable in the strike sense. However, it is commonly accepted
66: that the conductivity mechanism in this range is hopping and
67: practically all the experimental data are treated in framework of
68: the theory of hopping conductivity.
69:
70: One of theoretical reasoning behind this conclusion is the well
71: known Landauer formula which shows that the conductivity of one
72: open channel is $e^2/h$. For the first sight the 2D random network
73: of the open 1D channels should have the conductivity $e^2/h$ also.
74: This is true without taking into account interference. However,
75: the closed paths arise in such network and the interference of the
76: waves propagated in the different channels has to lead to
77: decreasing of the conductivity value. Therefore it seems that the
78: value of conductivity $e^2/h$ is not lower limit for the diffusive
79: conductivity when the transport is by delocalized states.
80:
81: Experimentally, the conclusion on the conductivity mechanism comes
82: from an analysis of the temperature dependence of the
83: conductivity. When $\sigma<e^2/h$, the experimental dependences
84: $\rho(T)$ can be satisfactorily described by Eq.~(\ref{eq1}) with
85: $0.3\lesssim p\lesssim 0.8$.\cite{wrh1,wrh2,wrh3,wrh4,wrh5,wrh6}
86: Relaying on such value of power $p$, practically all the authors
87: conclude that the variable range hopping regime takes place.
88:
89: We emphasize that a number of unusual features is observed in such
90: a 2D hopping regime as compared with the well-studied 3D variable
91: range hopping conductivity. The are the following.
92: \begin{enumerate}
93: \item The less than unity value of $p$ is observed up to
94: relatively high temperature, $T\simeq 5-10 $~K. The transition to
95: the nearest-neighbor hopping regime ($\epsilon_2$-regime) or to
96: the conductivity by free carriers ($\epsilon_1$-regime) is not
97: observed usually.
98:
99: \item The value of $\rho_0$ is close to $h/e^2$ practically in all
100: cases independent of the structure type and carrier density.
101:
102: \item The Hall coefficient is temperature independent, its value
103: is determined by the electron density, $n$, down to very low
104: conductivity values of order $10^{-2} e^2/h$: $R_H=(en)^{-1}$.
105:
106: \item The negative magnetoresistance is observed down to $\sigma
107: \simeq (0.1-0.01) e^2/h$ and its shape is very close to that
108: observed for high conductivity $\sigma\gg e^2/h$, when it is
109: caused by suppression of interference quantum correction.
110: \end{enumerate}
111:
112: Another approach to interpret the conductivity in this
113: intermediate range $e^2/h>\sigma > 10^{-4}e^2/h$ was used in
114: Ref.~\onlinecite{ourWLSL}. The authors attempted to trace the
115: changes in electron transport with decreasing conductivity caused
116: by increasing disorder starting from $\sigma\gg e^2/h$. They
117: analyzed not only the temperature dependences of the conductivity
118: but the magnetic field dependences of the conductivity tensor
119: components at low and high magnetic field. A surprising conclusion
120: has been arrived: all the transport effects are well described
121: within the framework of the quantum correction theory down to
122: conductivity significantly less than $e^2/h$.
123:
124: In the present paper we demonstrate that the study of nonohmic
125: conductivity can elucidate the low temperature conductivity
126: mechanism when $\sigma<e^2/h$. An idea of experiment will be clear
127: after consideration of the nature of conductivity nonlinearity in
128: diffusive and hopping regimes.
129:
130:
131: {\it Diffusive regime}. At low temperature, changing of the
132: conductivity $\sigma=j/E$ with electric field and, thus, with
133: injected power $Q=jE$ originates in this case from electron
134: heating. The heating leads to growth of the conductivity via
135: lowering of the quantum corrections magnitude which is determined
136: by electron temperature.\cite{foot}
137:
138: For steady state the injected power $Q$ is equal to the energy
139: relaxation rate $P$ which depends on the electron and lattice
140: temperature, $T_e$ and $T_l$, respectively. This equilibrium
141: determines the value of $T_e$ for given value of $T_l$. When the
142: energy relaxation rate $P(T_e,T_l)$ is governed by the interaction
143: with phonons it is equal to the difference of two identical
144: functions, one of them depends on the electron temperature and
145: another one depends on the lattice temperature:\cite{Price}
146: \begin{equation}
147: P(T_e,T_l)=F(T_e)-F(T_l).
148: \end{equation}
149: From this equation it follows that $\partial P(T_e,T_l)/\partial
150: T_e$ does not depend on $T_l$
151: \begin{equation}
152: { \partial P(T_e,T_l) \over \partial T_e}={ \partial F(T_e) \over \partial
153: T_e}.
154: \end{equation}
155:
156: Experimentally, the derivative $\partial P(T_e,T_l)/\partial T_e$
157: can be straightforwardly obtained from the dependences $\sigma(T)$
158: measured at low injected power and $\sigma(Q)$ measured at fixed
159: lattice temperature. Differentiating these dependences we can find
160: the experimental value of relative nonlinearity introduced as
161: \begin{equation}
162: \eta=\left.{\partial\sigma(Q)/\partial Q \over
163: \partial\sigma(T)/\partial T}\right|_{\sigma(Q)=\sigma(T)},
164: \label{eqNL}
165: \end{equation}
166: which is exactly equal to $(\partial F(T_e)/\partial T_e)^{-1}$.
167: Indeed,
168: \begin{eqnarray}
169: \left. {\partial\sigma(Q)/\partial Q \over
170: \partial\sigma(T)/\partial T}\right|_{\sigma(Q)=\sigma(T)}&=&
171: {({\partial\sigma/\partial T_e})( {\partial T_e/\partial Q}) \over
172: \partial\sigma/\partial T}\\ \nonumber &=&{\partial T_e \over \partial Q} =
173: \left[{\partial F(T_e) \over \partial
174: T_e}\right]^{-1}.\label{eq4}
175: \end{eqnarray}
176:
177:
178: \begin{figure}
179: \includegraphics[width=\linewidth,clip=true]{f1.eps}
180: \caption{Temperature and power dependences of the conductivity for
181: $\sigma(1.5\, K)=2.2~e^2/h$ (a) and
182: $\sigma(1.5\,K)=6\times10^{-4}~e^2/h$ (b). }
183: \label{f1}
184: \end{figure}
185:
186:
187: {\it Hopping regime}. First of all, the conductivity in this
188: regime depends both on the lattice and electron temperature.
189: Second, the changing of the conductivity with electric field
190: results not only from the electron heating but from the increasing
191: of the probabilities of hops and impact ionization of localized
192: states as well. Finally, the energy distribution function of the
193: electrons in electric field can deviate from the Fermi-Dirak
194: function and the approximation of electron temperature can fail.
195: All these effects have to lead to changing in relative
196: nonlinearity and, thus, the value of $\eta$ will depend on both
197: electron and lattice temperature.
198:
199: Thus, we suggest a simple experimental way to distinguish the
200: diffusive conductivity from the hopping one. As long as the
201: conductivity remains diffusive the $T_e$-dependences of the
202: relative nonlinearity $\eta$ measured at different lattice
203: temperatures $T_l$ have to fall on common curve and this property
204: has to disappear when the conductivity becomes hopping.
205:
206:
207: Experimentally, we investigated the relative nonlinearity of the
208: conductivity for two types of the single quantum well
209: heterostructures GaAs/InGaAs/GaAs. The first one consists of 0.5
210: mkm-thick undoped GaAs epilayer, a Sn $\delta$-layer, a 9 nm
211: spacer of undoped GaAs, a 8 nm In$_{0.2}$Ga$_{0.8}$As well, a 9 nm
212: spacer of undoped GaAs, a Sn $\delta$-layer, and a 200 nm cap
213: layer of undoped GaAs. The electron density and mobility were
214: $n=4\times 10^{15}$~m$^{-2}$ and $\mu=0.65$~m$^2$/(V~s),
215: respectively. The samples were mesa etched into Hall bars on which
216: basis field-effect transistors with an Ag or Al gate electrode
217: were fabricated. It was very important to use thick insulator
218: between gate electrode and 2D channel to decrease the influence of
219: voltage drop over the channel. We used the $10$~mkm organic
220: insulator.
221:
222: \begin{figure}
223: \includegraphics[width=\linewidth,clip=true]{f2.eps}
224: \caption{The relative nonlinearity $\eta$ as a function of
225: electron temperature $T_e$ measured at different lattice
226: temperature $T_l$ for various conductivities. Dashed line in (a)
227: is calculated according to Ref.~\onlinecite{Price}. Solid lines
228: are provided as a guide for the eye.} \label{f2}
229: \end{figure}
230:
231:
232: Heterostructures of another type were analogous except that they
233: had no doping layers. The conductivity and electron density of
234: such samples varied by illumination from $ 10^{-4}$ to $10~e^2/h$
235: and from $1\times 10^{15}$ to $4\times 10^{15}$~m$^{-2}$,
236: respectively. Results obtained for both types of the structures
237: were close.
238:
239: In what follows the results obtained for different electron
240: density will be referenced by the value of $\sigma$ measured at
241: $T=1.5$~K.
242:
243: Figure~\ref{f1} shows the temperature dependence of the
244: conductivity measured in Ohmic regime and power dependences of
245: $\sigma$ for two significantly different conductivity values. The
246: dependences $\sigma(Q)$ were measured at sweeping of the current
247: through the sample from minus $j$ to plus $j$ during several
248: minutes. Within experimental error the $\sigma$-versus-$Q$ plots
249: were identical for both current directions.
250:
251:
252: At $\sigma> e^2/h$ the temperature dependence of the conductivity
253: is close to the logarithmic one [Fig.~\ref{f1}(a)]. Such a
254: temperature dependence is determined by the quantum corrections
255: and was discussed in detail in our previous
256: papers.\cite{ourWLSL,ourEE} The increase of the conductivity with
257: increasing injected power is result of electron heating and rise
258: of electron temperature from $T_l$ to $T_e$. For this case, the
259: electron temperature $T_e$ at given power $Q$ can be determined by
260: a standard manner\cite{standard} using the experimental
261: dependences $\sigma(T)$ measured at $Q\to 0$ and $\sigma(Q)$
262: measured at fixed temperature $T$ as shown in Fig.~\ref{f1}(a) by
263: arrows: for each value of $Q$, $\sigma$ is used as a thermometer
264: giving $T_e$ as the temperature $T$ at which the same value of
265: $\sigma$ is measured for $Q\to 0$.
266:
267:
268: To obtain the relative nonlinearity we have numerically
269: differentiated the experimental dependences $\sigma(T)$ and
270: $\sigma(Q)$ and found $\eta$ in accordance with Eq.~(\ref{eqNL}).
271: This quantity as a function of electron temperature taken at
272: different lattice temperature is presented in Fig.~\ref{f2}.
273:
274:
275: First we consider the results for highest conductivity $\sigma =
276: 10~e^2/h$ which unambiguously corresponds to the diffusion regime
277: [Fig.~\ref{f2}(a)]. It is clearly seen that $\eta$-versus-$T_e$
278: data obtained for different lattice temperatures fall on common
279: curve as it has to be in the diffusive regime. This curve is close
280: to theoretical dependence $\partial F(T_e)/\partial T_e$ when the
281: interaction with deformation and piezoelectric potential of the
282: acoustic phonons is taken into account.\cite{Price}
283:
284: Such a data processing was carried out over the whole conductivity
285: range and the results are presented in Figs.~\ref{f2}(b)-(e). It
286: is evident that $\eta$-versus-$T_e$ data obtained for different
287: lattice temperatures fall on common curve down to $\sigma=
288: 0.22~e^2/h$ and only at $\sigma = 1\times 10^{-2}e^2/h$ some
289: deviation arises. This deviation becomes drastic at $\sigma
290: =4.7\times 10^{-4}e^2/h$. It points to the fact that one or more
291: specific features of the hopping regime discussed above occur: the
292: conductivity becomes dependent both on the lattice and electron
293: temperature; the electric field leads to the conductivity change
294: not only via electron heating but through the increase of the
295: probability of hops and impact ionization; the approximation of
296: electron temperature fails. In means also that the way of finding
297: of the electron temperature can be incorrect at $\sigma< 10^{-2}
298: e^/h$, and the quantity $T_e$ in Fig.~\ref{f2} is some effective
299: temperature.
300:
301:
302: In summary, we have proposed the way how studying the conductivity
303: nonlinearity one can determine the condition when the diffusion
304: regime changes to the hopping one. We have shown that for single
305: quantum well GaAs/InGaAs/GaAs heterostructures the conductivity
306: behaves like diffusive one down to $\sigma\approx (2-3)\times
307: 10^{-2}e^2/h$. The conclusion that the low-temperature
308: conductivity of the disordered 2D systems remains diffusive down
309: to $\sigma\ll e^2/h$ agrees with that obtained from the studies of
310: the quantum corrections to the conductivity at decreasing of
311: $k_Fl$ carried out in Ref.~\onlinecite{ourWLSL}.
312:
313:
314:
315: This work was supported in part by the RFBR through Grants
316: No.~01-02-17003 and No.~03-02-16150, the INTAS through Grant No.
317: 1B290, the Program {\it University of Russia}, the CRDF through
318: Grant No.~REC-005, and the Russian Ministry of Education through
319: Grant No.~A03-2.9-521.
320:
321: \begin{thebibliography}{}
322:
323:
324: \bibitem{Mott} N.~F.~Mott and W.~D.~Twose., Adv. Phys. {\bf 10} 107
325: (1961).
326:
327: \bibitem{ES} B.~I.~Shklovski and A.~L.~Efros, Electronic Properties of
328: DopedSemiconductors, Springer, Berlin, 1984.
329:
330: \bibitem{wrh1}
331: H.W. Jiang, C.E. Jonson, K.L. Wang. Phys. Rev. B {\bf 46}, 12830
332: (1992).
333:
334: \bibitem{wrh2}
335: H.W. Jiang, C.E. Jonson, K.L. Wang, S.T Hannahs. Phys. Rev. Lett.
336: {\bf 71}, 1439 (1993).
337:
338: \bibitem{wrh3}
339: T. Wang, K.P. Clark, G.F. Spenser, A.M. Mack, W.P. Kirk. Phys.
340: Rev. Lett. {\bf 72}, 709 (1994).
341:
342:
343: \bibitem{wrh4}
344: F.V. Van Keuls, X.L. Hu, H.W. Jiang, A.J. Dahm Phys. Rev. B {\bf
345: 56}, 1161 (1997).
346:
347:
348: \bibitem{wrh5}
349: C.H. Lee, Y.H. Chang, Y.W. Suen, H.H. Lin. Phys. Rev. B {\bf 58},
350: 10629 (1998).
351:
352: \bibitem{wrh6}
353: M. E. Gershenson, Yu. B. Khavin, D. Reuter, P. Schafmeister, and
354: A. D. Wieck. Phys. Rev. Lett. {\bf 85}, 1718 (2000).
355:
356:
357:
358: \bibitem{ourWLSL}
359: G.~M.~Minkov, O.~E.~Rut, A.~V.~Germanenko, A.~A.~Sherstobitov,
360: B.~N.~Zvonkov, E.~A.~Uskova, and A.~A.~Birukov, Phys. Rev. B {\bf
361: 65}, 235322 (2002).
362:
363:
364: \bibitem{foot} Estimations show
365: that approximation of electron temperature is valid at $n>
366: 10^{14}$~m$^{-2}$.\cite{Dif}
367:
368: \bibitem{Dif}
369: I.~L.~Drichko, A.~M.~Dyakonov, V.~D.~Kagan, A.~M.~Kreshchuk,
370: T.~A.~Polyanskaya, I.~G.~Savel'ev, I.~Yu.~Smirnov, and
371: A.~V.~Suslov. Fiz. Tekn. Poluprov. {\bf 31}, 1357 (1997) [
372: Semicond. {\bf 31} 1170 (1997)].
373:
374: \bibitem{Price}
375: P.~J.~Price, J.Appl. Phys. {\bf 53}, 6863 (1982).
376:
377: \bibitem{ourEE}
378: G.~M.~Minkov, O.~E.~Rut, A.~V.~Germanenko, A.~A.~Sherstobitov,
379: V.~I.~Shashkin, O.~I.~Khrykin, and B.~N.~Zvonkov, Phys. Rev. B
380: {\bf 67}, 205306 (2003).
381:
382: \bibitem{standard}
383: R.~Fletcher, Y.~Feng, C.~T.~Foxon, and J.~J.~Harris. Phys. Rev. B
384: {\bf 61}, 2028 (2000).
385:
386:
387:
388: \end{thebibliography}
389:
390:
391: \end{document}
392: