Abstract
We have performed magnetotransport measurements on gated GaAs twodimensional electron gases in which electrons are confined in a layer of the nanoscale. From the slopes of a pair of spinsplit Landau levels (LLs) in the energymagnetic field plane, we can perform direct measurements of the spin gap for different LLs. The measured gfactor g is greatly enhanced over its bulk value in GaAs (0.44) due to electron–electron (ee) interactions. Our results suggest that both the spin gap and g determined from conventional activation energy studies can be very different from those obtained by direct measurements.
Keywords:
Spin; gfactor; DisorderBackground
With the growing interest in spinbased quantum computation and spintronic applications [1], there is an increasing need to understand and accurately determine critical parameters of the electron spin degree of freedom. It is well established that when measuring an electron spin in an external magnetic field B, it can either align parallel to or antiparallel to B. The energy difference between these two discrete states, also known as the spin gap or Zeeman splitting, is given by gμ_{B}B where g is the Lande gfactor and μ_{B} is the Bohr magneton. It is worth mentioning that successful application of the wide range of possible spindependent phenomena requires effective techniques for the electrical injection of spinpolarized currents [2]. It has been demonstrated that a net spin current can be produced when
where kT and Γ are the thermal and level broadening, respectively [3].
For practical applications, it is highly desirable that the generation of the spin currents can be accomplished without requiring the use of extremely high B. Therefore, an accurate measurement of the spin gap and gfactor would allow one to ensure that only a moderate B is required so that Equation 1 holds. Moreover, the precise measurement of the gfactor [4] would shed light on the predicted divergence of spin susceptibility χ ∝ g m* and ferromagnetic ground state [5], where the system exhibits the unexpected metalinsulator transition [6]. Here m* represents the effective mass of electron (or hole). Given that the spin gap is the most important energy scale in any spin system and the gfactor is the central quantity characterizing the response of an electron or hole spin to an applied B, there have been many attempts to measure the spin gap in the literature. A standard method of obtaining the spin gap is to perform activation energy measurements at the minimum of the longitudinal resistivity , where Δ_{s} is the spin gap [7]. However, such a measurement is rather restrictive as ρ_{xx} must be very low and has to vary over at least an order of magnitude as a function of T. Moreover, Δ_{s} has to be much greater than the thermal energy kT over the whole measurement range. Most importantly, activation energy measurements yield the ‘mobility gap’, the width of the localized states in the energy spectrum. This may be quite different from the real spin gap which corresponds to the energy difference between the two maxima densities of neighboring extended states [4,8].
In this paper, we report a method to directly measure the spin gaps in twodimensional electron gases (2DEGs), in which the electrons are usually confined in layers of the nanoscale. We can change the applied gate voltage V_{g} to vary the electron density n_{2D} and hence the local Fermi energy E in our system. By studying the peak positions of ρ_{xx} at various n_{2D} and B, we can construct the Landau levels in the EB diagram. As shown later, from the difference between the slopes of a pair of spinsplit Landau levels in the EB plane, we are able to measure the gfactors for different Landau level indices n in the zero disorder limit. We find that the measured gfactors (approximately 10) are greatly enhanced over their bulk value (0.44). Most importantly, our results provide direct experimental evidence that both the spin gap and gfactor determined from the direct measurements are very different from those obtained by the conventional activation energy studies. A possible reason is that our method is conducted in the zero disorder limit, whereas activation studies are performed under the influence of the disorder within the quantum Hall system.
In the integer quantum Hall effect (IQHE), when the spin of the 2DEG is taken into consideration, in the zero disorder limit each Landau level splits into two with the corresponding energy given by
where ω_{C} is the cyclotron frequency, and n = 0, 1, 2, 3…, respectively. According to early experimental work [9], it was established that in 2D systems in a magnetic field the gfactor is greatly enhanced over its bulk value due to exchange interactions [10,11]. The precise measurement of the gfactor in 2D systems is a highly topical issue [4] since it has been predicted to be enhanced in strongly interacting 2D systems that exhibit the unexpected zerofield metalinsulator transition [6].
Methods
Experimental details
Magnetoresistance measurements were performed on three gated Hall bars (samples A, B and C) made from modulationdoped GaAs/Al_{0.33}Ga_{0.67}As heterostructures. For sample A, the structure consists of a semiinsulating (SI) GaAs (001) substrate, followed by an undoped 20nm GaAs quantum well, an 80nm undoped Al_{0.33}Ga_{0.67}As spacer, a 210nm Sidoped Al_{0.33}Ga_{0.67}As, and finally a 10nm GaAs cap layer. For sample B, the structure consists of an SI GaAs (001) substrate, followed by an undoped 20nm GaAs quantum well, a 77nm undoped Al_{0.33}Ga_{0.67}As spacer, a 210nm Sidoped Al_{0.33}Ga_{0.67}As, and finally a 10nm GaAs cap layer. Sample C is a modulationdoped GaAs/AlGaAs heterostructure in which selfassembled InAs quantum dots are inserted into the center of the GaAs well [12]. The following sequence was grown on an SI GaAs (001) substrate: 40nm undoped Al_{0.33}Ga_{0.67}As layer, 20nm GaAs quantum well inserted with 2.15 monolayer of InAs quantum dots in the center, a 40nm undoped Al_{0.33}Ga_{0.67}As spacer, a 20nm Sidoped Al_{0.33}Ga_{0.67}As, and finally a 10nm GaAs cap layer. Because of the lack of inversion symmetry and the presence of interface electric fields, zerofield spin splitting may be present in GaAs/AlGaAs heterostructures. However, it is expected that the energy splitting will be too small (0.01 K) to be important in our devices [13]. For sample A, at V_{g} = 0 the carrier concentration of the 2DEG was 1.14 × 10^{11} cm^{2} with a mobility of 1.5 × 10^{6} cm^{2}/Vs in the dark. For sample B, at V_{g} = 0 the carrier concentration of the 2DEG was 9.1 × 10^{10} cm^{2} with a mobility of 2.0 × 10^{6} cm^{2}/Vs in the dark. The selfassembled InAs dots act as scattering centers in the GaAs 2DEG [12,14]; thus, the 2DEG has a mobility much lower than those for samples A and B. For sample C, at V_{g} = 0 the carrier concentration of the 2DEG was 1.48 × 10^{11} cm^{2} with a mobility of 1.86 × 10^{4} cm^{2}/Vs in the dark. Experiments were performed in a He3 cryostat and the fourterminal magnetoresistance was measured with standard phasesensitive lockin techniques.
Results and discussion
Figure 1 shows the fourterminal magnetoresistance measurements R_{xx} as a function of B at V_{g} = 0.08 V for sample A. When the Fermi level is centered at a Landau level, there exists a peak in R_{xx} as shown in Figure 1. By studying the evolution of the peaks in R_{xx} at different gate voltages (and hence n_{2D}), we are able to locate the position of the Landau levels in the n_{2D}B plane. Figure 2a,b shows such results obtained from sample A and sample B, respectively. It is known that in the low disorder or high B limit, the filling factor of a resistivity (or conductivity) peak is given exactly by the average value of the filling factors of the two adjacent quantum Hall states [15]. This is equivalent to the situation when the Fermi energy coincides with a Landau level. It is worth pointing out that the peak position of magnetoresistance oscillations can be given by , where ν is the Landau level filling factor. At first glance, the peak position does not depend on either the gfactor or the effective mass of the 2D system. However, as shown later, in our case the energy of the Landau levels can be considered directly proportional to the density via the free electron expression [16], where m^{*} = 0.067 m_{e} in GaAs and m_{e} being the rest mass of a free electron. Then the effective mass should be considered when constructing the energymagnetic field diagram. Here the oscillation of the Fermi energy is not considered. It may be possible that the effective mass of the 2DEGs will increase due to strong correlation effect [17]. In order to measure the effective mass of our 2DEG, we plot the logarithm of the resistivity oscillating amplitudes divided by temperature ln (Δρ_{xx} / T) as a function of temperature at different magnetic fields in Figure 3. Following the procedure described by the work of Braña and coworkers [18], as shown in the inset to Figure 3, the measured effective mass is very close to the expected value 0.067 m_{e}. Therefore it is valid to use m^{*} = 0.067 m_{e} in our case. We can see that the Landau levels show a linear dependence in B as expected. At low B and hence low n_{2D}, the slight deviation from the straight line fits can be ascribed to experimental uncertainties in measuring the positions of the spinup and spindown resistivity peaks.
Figure 1. Magnetoresistance measurements R_{xx }( B ) at V_{g }= 0.08 V for sample A at T = 0.3 K. The maxima in R_{xx }occur when the Fermi energy lies in the nth spinsplit Landau levels as indicated by n = 3↓ and n = 3↑, n = 2↓ and n = 2↑, and n = 1↓ and n = 1↑, respectively.
Figure 2. The Local Fermi energy E and the corresponding 2D carrier density n_{2D }for different Landau levels. (a) Sample A and (b) sample B at T = 0.3 K. Circle, 3↓ and 1↓; square, 3↑ and 1↑; star, 2↓; triangle, 2↑.
Figure 3. Logarithm of the amplitudes of the oscillations. The logarithm of the amplitudes of the oscillations divided by T ln(Δρ_{xx }/ T) as a function of temperature at different magnetic field for sample C at V_{g }= 0. The curves correspond to fits described by [18]. The inset shows the measured effective mass at different magnetic fields.
In our system as the spinsplit resistivity peaks are not observed at the same magnetic field, we need to describe the method of measuring the gfactors as follows. Equation 2 can be rewritten as
where we consider the effective Lande gfactor g^{*}. We can see that Equation 3 corresponds to two straight line fits through the origin for a pair of spinsplit Landau levels in the EB plane as shown in Figure 2a,b. Such an approach was applied to a GaNbased 2DEG in our previous work [19]. We note that our method does depend on the exact functional form of the Landau band since the peak positions of the Landau level is only related to the carrier density in our system.
Let us now consider the region ν = 3 between the two linear fits corresponding to two spinsplit Landau levels n = 1↓ and n = 1↑. According to Equation 3, the difference between the slopes of the spinsplit Landau levels is given by g^{*}Φ06Δ_{B}B. Thus we are able to measure g^{*} for different Landau level indices (n = 1, 2, 3,…). In our system, the spin gap value is proportional to the magnetic field with good accuracy and corresponds to a constant g^{*} for a pair of given spinsplit Landau levels. Figure 4 shows the measured g^{*} as a function of Landau level index n for samples A and B. In all cases, the measured g^{*} is greatly enhanced over its bulk value in GaAs (0.44). We ascribe this enhancement to exchange interactions. We suggest that the determined g^{*} is in the zero disorder limit since the positions of the spinsplit Landau levels are located using Equation 2.
Figure 4. The measured g^{* }as a function of Landau level index n. The measured g^{* }as a function of Landau level index n for samples A and B at T = 0.3 K.
It is worth mentioning that conventional activation energy studies are not applicable to our data obtained on sample A, sample B as well as the GaNbased 2DEG in our previous work [19]. The reason for this is that the values of the R_{xx} (and σ_{xx}) minima are high; therefore, it is not appropriate to speak of electrons being thermally activated from the localized states to the extended states. In order to provide further understanding on the measurements of the spin gap, we have studied the slopes of the spinsplit Landau levels in the EB plane and have also performed conventional activation energy measurements on sample C over the same magnetic field range. Sample C is a more disordered device compared with samples A and B thus we can only perform measurements in the regime where the ρ_{xx} corresponding to a spinsplit ν = 3 state is resolved. Figure 5 shows the evolution of the n = 1↓ and n = 1↑ resistivity peaks at different magnetic fields for sample C. From the difference between the two slopes of n = 1↓ and n = 1↑ spinsplit Landau levels, the exchangeenhanced gfactor for the n = 1 Landau level is measured to be 11.65 ± 0.14, which is in close agreement with those obtained on a much higher mobility in samples A and B. We note that in a dilute GaAs 2DEG, the enhancement factor of g can decrease from about 6 to 3 as the density is reduced [20]. It may be possible that as our 2DEG density is considerably higher than those reported in the seminal work of Tutuc, Melinte, and Shayegan. Therefore we do not see such a trend in our system.
Figure 5. Local Fermi energy E and the corresponding 2D carrier density n_{2D}. The local Fermi energy E and the corresponding 2D carrier density n_{2D }for n = 1↓ and n = 1↑, Landau levels as a function of B for Sample C at T = 0.3 K.
Let us now turn our attention to the activation energy measurements. Figure 6 shows ln (ρ_{xx}) as a function of 1/T for eight different carrier densities while maintaining the filling factor at ν = 3 for sample C. The resistivity shows activated behavior . Figure 6 shows the activation energy Δ_{s} determined from a leastsquare fit to the experimental data shown in Figure 5. We can see that the spin gaps Δ_{s} drops approximately linearly to zero at a critical magnetic field B_{c} ~ 3.47 T. The spin gap is expected to have the form Δ_{s} = g_{0}μ_{B}B + E_{ex} = g^{*}μ_{B}B[12], where E_{ex} is the manybody exchange energy which lifts the gfactor from its bare value (0.44 in GaAs) to its enhanced value g^{*}. Figure 7 shows that the measured Δ_{s} is greatly enhanced over the single particle Zeeman energy (shown in the dotted line), yielding g^{*} = 4.64 ± 0.30. Moreover, the exchange energy shows a roughly linear B dependence. The disorder broadening Γ_{s} can be estimated from the critical magnetic B_{c}[12]. From this we obtain a quantum lifetime of Γ_{s} = 0.71 ps, in qualitative agreement with the value 0.40 ps obtained from the Dingle plot. For the lowfield regime where Δ_{s} < Γ_{s}, the manybody interactions are destroyed by the disorder, and there is no spinsplitting for the magnetic field less than B_{c}. As shown in Figure 7, the ‘spin gap’ measured by the conventional activation energy studies is very different from that measured by the direct measurements (shown in the dashed line). This is consistent with the fact that activation energy studies yield a mobility gap which is smaller than the real spin gap in the spectrum. Moreover, the measured by studying the slopes of the n = 1 spinsplit Landau levels is approximately 2.4 times larger than that determined from the activation energy studies. Our data shows that both the spin gaps and g^{*} measured by the activation energy studies are very different from those determined from direct measurements. A possible reason for this is that there exists disorder within 2D system which is indispensable to the observation of the IQHE. The direct measurements are performed in the zero disorder limit. On the other hand, in the activation energy studies, the disorder within the quantum Hall system must be considered. As shown in the inset of Figure 7, the spin gap in the zero disorder limit is the energy difference between neighboring peaks in the density of states N(E) which is larger than the energy spacing between the edges of the localized states given the finite extended states. We suggest that further theoretical studies are required in order to obtain a full understanding of our results on the spin gaps and g^{*}.
Figure 6. Logarithm of ρ_{xx}( B )(ν = 3) versus the inverse of temperature 1/ T . The logarithm of ρ_{xx}(B)(ν = 3) versus the inverse of temperature 1/T at different gate voltages (and hence B) for sample C. From left to right: B = 5.72 (pentagon), 5.46 (star), 5.21 (hexagon), 4.97 (diamond), 4.70 (inverted triangle), 4.55 (triangle), 4.39 (heptagon) and 4.25 (square) T, respectively. The slopes of the straight line fits Δ_{s }are shown in Figure 7.
Figure 7. The experimentally determined Δ_{s}/k_{B }at various B. The straight line fit is discussed in the text. The dotted line is the bare Zeeman energy assuming g_{0 }= 0.44. The dashed line corresponds to the spin gap using the measured g^{* }= 11.65 by the direct measurements. The inset corresponds to a schematic diagram (density of states N(E) versus E) showing the spin gap Δ_{s} as a result of the activated behavior from the localized states (hatched areas) to the extended states (in blue). The spin gap in the zero disorder limit Δ_{s} is the energy difference between the neighboring peaks in N(E).
Conclusions
In conclusion, we have performed direct measurements of the spin gaps in gated GaAs 2DEGs by studying the slopes of spinsplit Landau levels in the energymagnetic field plane. The measured gfactor is greatly enhanced over its bulk value (0.44). Since disorder exists in any experimentally realized system, conventional activation energy studies always measure the mobility gap due to disorder which is different from the real spin gap as shown in our results. As the spin gap is one of the most important energy scales and governs the electron spin degree of freedom, our experimental results provide useful information in the field of spintronics, spinrelated phenomena, and quantum computation applications.
Competing interests
The authors declare that they have no competing interests.
Authors’ contributions
TYH and CTL performed the measurements. CTL, YFC, and GHK coordinated the projects. MYS and DAR grew the samples. TYH, YFC, and CTL drafted the paper. All authors read and approved the final manuscript.
Acknowledgments
TYH, CTL and YFC were supported by the NSC, Taiwan and National Taiwan University (grant no. 102R890932 and grant no. 102R75522). The work at Cambridge was supported by the EPSRC, UK. This research was supported by the World Class University program funded by the Ministry of Education, Science and Technology through the National Research Foundation of Korea (R3210204).
References

Bader SD, Parkin SSP: Spintronics. Annual review of condensed matter.

Shen C, Trypiniotis T, Lee KY, Holmes SN, Mansell R, Husain M, Shah V, Li XV, Kurebayashi H, Farrer I, de Groot CH, Leadley DR, Bell G, Parker EHC, Whall T, Ritchie DA, Barnes CHW: Spin transport in germanium at room temperature.
Appl Phys Lett 2010, 97:162104. Publisher Full Text

Watson SK, Potok RM, Marcus CM, Umansky V: Experimental realization of a quantum spin pump.
Phys Rev Lett 2003, 91:258301. PubMed Abstract  Publisher Full Text

Khrapai S, Shashkin AA, Dolgopolov VT: Direct measurements of the spin and the cyclotron gaps in a 2D electron system in silicon.
Phys Rev Lett 2003, 91:126404. PubMed Abstract  Publisher Full Text

Attaccalite C, Moroni S, GoriGiorgi P, Bachelet GB: Correlation energy and spin polarization in the 2D electron gas.
Phys Rev Lett 2002, 88:256601. PubMed Abstract  Publisher Full Text

Abrahams E, Kravchenko SV, Sarachik MP: Metallic behavior and related phenomena in two dimensions.
Rev Mod Phys 2001, 73:251. Publisher Full Text

Nicholas RJ, Haug RJ, von Klitzing K, Weimann G: Exchange enhancement of the spin splitting in a GaAsGa_{x}Al_{1  x}As heterojunction.
Phys Rev B 1988, 37:1294. Publisher Full Text

Dolgopolov VT, Shashkin AA, Aristov AV, Schmerek D, Hansen W, Kotthaus JP, Holland M: Direct measurements of the spin gap in the twodimensional electron gas of AlGaAsGaAs heterojunctions.
Phys Rev Lett 1997, 79:729. Publisher Full Text

Fang FF, Stiles PJ: Effects of a tilted magnetic field on a twodimensional electron gas.
Phys Rev 1968, 174:823. Publisher Full Text

Janak JF: g Factors for an interacting electron gas.
Phys Rev 1969, 178:1416. Publisher Full Text

Nicholas RJ, Haug RJ, von Klitzing K, Weimann G: Exchange enhancement of the spin splitting in a GaAsGa_{x}Al_{1  x}As heterojunction.

Kim GH, Nicholls JT, Khondaker SI, Farrer I, Ritchie DA: Tuning the insulatorquantum Hall liquid transitions in a twodimensional electron gas using selfassembled InAs.
Phys Rev B 2000, 61:10910. Publisher Full Text

Thomas KJ, Nicholls JT, Simmons MY, Pepper M, Mace DR, Ritchie DA: Possible spin polarization in a onedimensional electron gas.
Phys Rev Lett 1996, 77:135. PubMed Abstract  Publisher Full Text

Liang CT, Lin LH, Chen KY, Lo ST, Wang YT, Lou DS, Kim GH, Chang YH, Ochiai Y, Aoki N, Chen JC, Lin Y, Huang CF, Lin SD, Ritchie DA: On the direct insulatorquantum Hall transition in twodimensional electron systems in the vicinity of nanoscaled scatterers.
Nanoscale Res Lett 2011, 6:131. PubMed Abstract  BioMed Central Full Text  PubMed Central Full Text

Goldman VJ, Jain JK, Shayegan M: Nature of the extended states in the fractional quantum Hall effect.
Phys Rev Lett 1990, 65:907. PubMed Abstract  Publisher Full Text

Glozman I, Johnson CE, Jiang HW: Fate of the delocalized states in a vanishing magnetic field.
Phys Rev Lett 1995, 74:594. PubMed Abstract  Publisher Full Text

Nomura S, Yamaguchi M, Akazaki T, Tamura H, Maruyama T, Miyashita S, Hirayama Y: Enhancement of electron and hole effective masses in backgated GaAs/Al_{x}Ga_{1  x}As quantum wells.

Braña AF, DiazPaniagua C, Batallan F, Garrido JA, Muñoz E, Omnes F: Scattering times in AlGaN/GaN twodimensional electron gas from magnetoresistance measurements.
J Appl Phys 2000, 88:932. Publisher Full Text

Cho KS, Huang TY, Huang CP, Chiu YH, Liang CT, Chen YF, Lo I: Exchangeenhanced gfactors in an Al0.25Ga0.75N/GaN twodimensional electron system.
J Appl Phys 2004, 96:7370. Publisher Full Text

Tutuc E, Melinte S, Shayegan M: Spin polarization and g factor of a dilute GaAs twodimensional electron system.
Phys Rev Lett 2002, 88:036805. PubMed Abstract  Publisher Full Text