Probing the Sensitivity of Electron Wave Interference to Disorder-Induced Scattering in Solid-State Devices B. C. Scannell,1 I. Pilgrim,1 A. M. See,2 R. D. Montgomery,1 P. K. Morse,1 M. S. Fairbanks,1 C. A. Marlow,1 H. Linke,3 I. Farrer,4 D. A. Ritchie,4 A. R. Hamilton,2 A. P. Micolich,2 L. Eaves,5 and R. P. Taylor1 1

Department of Physics, University of Oregon, Eugene, OR 97403-1274, USA School of Physics, University of New South Wales, Sydney, NSW, 2052, Australia 3 Division of Solid State Physics and Nanometer Structure Consortium (nmC @ LU), Lund University, Box 118, S-221 00, Lund, Sweden 4 Department of Physics, Cavendish Laboratory, J. J. Thompson Avenue, Cambridge, CB3 0HE, United Kingdom 5 School of Physics and Astronomy, University of Nottingham, Nottingham NG7 2RD, United Kingdom (Dated: November 13, 2016)

arXiv:1106.5823v1 [cond-mat.mes-hall] 28 Jun 2011

2

The study of electron motion in semiconductor billiards has elucidated our understanding of quantum interference and quantum chaos. The central assumption is that ionized donors generate only minor perturbations to the electron trajectories, which are determined by scattering off billiard walls. We use magnetoconductance fluctuations as a probe of the quantum interference and show that they change radically when the scattering landscape is modified by thermally-induced charge displacement between donor sites. Our results challenge the accepted understanding of quantum interference effects in nanostructures. PACS numbers: 73.63.-b, 72.20.My, 05.45.Mt, 73.63.Kv

The Aharonov-Bohm (AB) effect provides a remarkable demonstration of the fundamental interplay between quantum mechanics and electromagnetism [1]. Although its original incarnation used electrons in vacuum [1, 2], observations of the AB effect in solid-state environments demonstrated how sub-micron electronic devices are highly productive for studying interference phenomena [3–6]. The first example consisted of a nanoscale gold ring enclosing an area A, connected on opposing sides to source and drain leads [7]. An electron flowing into the ring can take the clockwise or anticlockwise arm to exit at the other side. At temperatures low enough to establish phase coherence, quantum interference between these paths leads to a sinusoidal oscillation in conductance as a function of magnetic field B applied perpendicular to the ring with a period ∆B = h/eA. Observations were extended to a range of semiconductor device geometries [8–11] in which a multitude of AB ‘loops’ arise from the many possible paths an electron can follow through the device due to scattering from both the device walls and any enclosed impurities. The result is more complex magnetoconductance fluctuations (MCF) with a spectral content determined by the landscape of the scattering potential and the distribution of loop areas [10]. MCF provide a ‘magnetofingerprint’ of electron dynamics within a device, and this was of great interest for studies of quantum chaos in the 1990s [5, 12]. By combining ultrapure semiconductor heterostructures [13] with nanofabrication techniques [11, 14, 15], it became possible to make quantum ‘billiards’—devices where electrons are confined within a geometry smaller than the mean free path for large-angle impurity scattering. By adopting an idealized ballistic model, in which electron trajectories are pictured as scattering only from the bil-

liard walls [16], the MCF were used to investigate the change from stable to chaotic electron dynamics for billiards with differently-shaped walls [17–19] and for shape transitions within a single billiard [20]. These pioneering studies attracted the attention of the broader research community due to analogous phenomena in wide quantum wells [21]; in microwave [22], optical [23, 24], and acoustic [25] cavities; and cold atoms in optical traps [26]. Here, we point out a crucial oversight in the study of semiconductor billiards, namely, the influence of smallangle scattering by remote ionized dopant atoms used to populate these devices with electrons. The presence of small-angle scattering has been acknowledged [17, 27] but considered to be a small perturbation that does not interrupt the direct connection between the MCF and the dynamics determined by the billiard walls. Recent studies using scanning gate microscopy showed that smallangle scattering produces branching of electron trajectories at length scales much shorter than the mean free path for large-angle scattering [28–30]. Here, we demonstrate that small-angle scattering is the dominant factor influencing the MCF by showing that thermal redistribution of the ionized dopants reconfigures the measured fluctuations despite the billiard geometry remaining unchanged. This result and reappraisal have crucial implications for studies of quantum chaos in semiconductors and analogous systems, and more generally highlight the limitations of realizing genuine ballistic transport in semiconductor nanostructures. To investigate the generic properties of electron interference in semiconductor billiards, we consider two widely studied material systems and two distinct methods for defining billiards. In each case, a two-dimensional elec-

2 electrostatic potential landscape of the billiard [13, 33]. Consequently, the ionized donors predominantly scatter the electrons through small angles and the associated deviations in the trajectories are not sufficient to reduce the mean free path ℓµ derived from mobility measurements (3 µm and 6 µm for the GaAs/(AlGa)As and (GaIn)As/InP systems) to below the billiard width of 1 µm. The electron densities are 2 × 1015 m−2 and 7 × 1015 m−2 , giving Fermi energies of EF = 8.4 meV and 50 meV, respectively. Accordingly, the Fermi wavelengths (50 nm and 30 nm) are significantly smaller than the billiard widths, as required for semiclassical interference phenomena [16].

FIG. 1. (Color) Schematics of (a) GaAs/(AlGa)As and (b) (GaIn)As/InP billiards. The 2DEG (white) is located near the lower GaAs/(AlGa)As interface for (a) and in the (GaIn)As layer for (b). The modulation-doped layer (red) consists of Si donors grown in (AlGa)As for (a) and InP for (b). The billiard walls (light blue) are defined by surface gates (yellow) in (a) and by etched trenches in (b). (c) Simulation of the energy band structure for the GaAs/(AlGa)As heterostructure, plotting the conduction band edge EC v. distance from the 2DEG. The blue dots symbolize the position and energy levels of Si donor DX centers located close to the band edge (see text). (d) Magnetoconductance traces for a 1 µm (GaIn)As/InP billiard measured at temperatures T of (top to bottom) 12 K, 5 K, 1 K, and 237 mK. The traces are offset for clarity. Inset: A plot of the electron phase coherence length ℓφ as a function of T , and associated fit line, for the (GaIn)As/InP billiard.

tron gas (2DEG) forms within a heterostructure (Fig. 1(a, b)). For the GaAs/Al0.33 Ga0.67 As system, electrostatic depletion generated by patterned surface gates defines ‘soft’ billiard walls in the 2DEG (Fig. 1(a)) [31]. For the Ga0.25 In0.75 As/InP system, wet etching generates ‘harder’ walls with electrostatic gradients 20 times steeper than the GaAs/(AlGa)As billiards (Fig. 1(b)) [32]. The fabrication of high-quality billiards involves the modulation doping technique shown in Fig. 1(c). Silicon donors are separated from the 2DEG plane by a spacer layer in order to minimize perturbations in the

Billiards support a distribution of AB loop areas, leading to range of frequencies in the magnetoconductance [16, 17, 27, 32]. The bottom trace of Fig. 1(d) shows the MCF generated by wave interference, superimposed on a classical conductance background for a square (GaIn)As/InP billiard measured at T = 240 mK (base temperature). Raising T reduces the electron phase coherence length ℓφ [32]: by T = 12 K, ℓφ is smaller than the billiard width, suppressing the MCF and leaving only the classical background [34]. To investigate the impact of the ionized donors on the AB loops, we warmed the billiards to temperatures high enough to supply the thermal energy necessary to relocate electron charge between dopant sites. Within the 1 µm2 area of the (GaIn)As/InP and GaAs/(AlGa)As billiards, there are approximately 30,000 and 50,000 of these donor sites, respectively. If the trajectories forming the AB loops are determined purely by the geometry of the billiard walls, as traditionally believed, then the MCF will be immune to this shift in charge between dopants. We annealed the billiard devices by warming to an intermediate temperature Ti for a fixed time t = 30 minutes and then cooled the billiards back to base temperature. We then compared the pair of MCF traces taken at base temperature before and after this thermal cycle. Figure 2 shows three pairs of MCF traces measured on the (GaIn)As/InP billiard. The bottom pair of traces were recorded back-to-back (i.e., Ti = 240 mK) and demonstrate the well-known reproducibility of MCF traces at low temperatures. The middle pair of traces correspond to Ti = 115 K and reveal the same high degree of correlation as the bottom pair. However, the top pair have visibly decorrelated after thermally cycling to Ti = 300 K. Figure 3 shows the equivalent MCF measured for a GaAs/(AlGa)As billiard, demonstrating that this decorrelation is a generic effect for semiconductor billiards. We note that the MCF of the GaAs/(AlGa)As billiard have a smaller high-frequency component than the (GaIn)As/InP billiard due to the former billiard’s smaller ℓφ of 3.6 µm at T = 240 mK [32]. To quantify the change in MCF as a function of Ti , we applied the following correlation function [35, 36] to each pair of traces G1 (B) and G2 (B):

3

FIG. 2. (Color) Comparison of MCF for the (GaIn)As/InP billiard after being warmed to three intermediate temperatures Ti . Top to bottom, the Ti values for the pairs of traces are 300 K, 115 K, and 240 mK. The traces are offset for clarity. Top: Magnified comparison of traces taken with Ti = 115 K (black and green) and Ti = 300 K (black and red).

F =

v u u t D

1−

E D 2 [G1 (B) − G2 (B)]

B

N

N = [Gx (B) − Gy (B)]

2

E

B

.

,

where

(1) (2)

The symbol h iB represents an average over the 640 data points spanning −Bc < B < Bc , where Bc is the field at which the cyclotron diameter matches the billiard width: 0.16 T and 0.28 T for the GaAs/(AlGa)As and (GaIn)As/InP billiards. The normalization constant N is calculated by averaging the correlations of 15 pairs of traces (Gx (B) and Gy (B)) that have been thermally cycled to Ti = 300 K. Adopting this normalization, the correlation scale varies between 1 for mathematically identical traces to 0 for decorrelation induced by a Ti = 300 K cycle. In Fig. 4, we show the results for the (GaIn)As/InP (blue) and GaAs/(AlGa)As (green) billiards. For comparison, we also show previously reported decorrelations of the MCF for a GaAs wire (red) [35]. The wire was 10 µm long, 50 nm high, and 90 nm wide, and was heavily doped with silicon donors (n = 5 × 1024 m−3 , EF = 128 meV) distributed uniformly throughout

FIG. 3. (Color) Comparison of MCF for the GaAs/(AlGa)As billiard after being warmed to three intermediate temperatures Ti . The details are identical to those of Fig. 2.

the wire’s cross-section. Due to the absence of modulation doping, the large-angle scattering off these dopants results in ℓµ = 0.3 µm. The thermally-induced decorrelation of the MCF for this diffusive scattering system is clearly similar to that of the two modulation doped billiards. For all three systems, we attribute the MCF decorrelation to the relocation of electron charge among silicon dopants. Consider the GaAs/(AlGa)As system, and the conduction band diagram of Fig. 1(c), in more detail. If the electrons in the 2DEG are thermally excited over the energy barrier at the GaAs/(AlGa)As interface, DX centers associated with the silicon dopant ions can capture the electrons and change the DX charge state. The energy level for electron capture at the DX center is ∼10 meV above the Al0.33 Ga0.67 As conduction band edge (see Fig. 1(c)) [37, 38]. This charge relocation changes the electrostatic landscape in the nearby 2DEG [33, 39], which in turn reshapes the distribution of AB loops within the billiard. The probability P for electrons to traverse the barrier depends upon the available thermal energy according to P ∼ exp(−β/kB Ti ), where β represents the activation energy for charge relocation. The charge redistribution will also be governed by the time t the billiard is held at Ti . We model this behavior using an annealing formula F = exp[−η exp(−β/kB Ti )], where η = t/τ0 and τ0 is a characteristic time for charge

4 transfer [35]. The decorrelation data is fitted with this formula in Fig. 4. The insets in Fig. 4 are simulations demonstrating how the form of the fall-off depends upon β and η: β (upper) and η (lower) are varied while keeping the other parameter fixed. The GaAs/(AlGa)As billiard’s decorrelation commences at a higher Ti than for the other systems, indicating a higher β value. As expected, the GaAs/(AlGa)As fit value of β = 230 ± 50 meV matches the height of the interface energy barrier above EF . The β value of 80 ± 20 meV observed for the GaAs wire is also close to the expected activation energy of 100 meV, which is set by the height of the DX center above EF for this system [40]. The value of β for the (GaIn)As/InP billiard is 60 ± 20 meV. The dopants in this heterostructure are less well understood than for the other systems, but the data suggest that the mechanism for decorrelation of the MCF is similar. We plan future MCF experiments and analysis to study charge trapping and its detailed dependence on the entire thermal annealing cycle, which would allow more accurate η values to be measured.

Our results emphasize the central role that disorder plays for wave interference in solid-state environments: the associated lack of reproducibility between devices and thermal cycles has implications not only for fundamental research but also for future nanoscale electronic devices. We expect this to apply to novel systems such as graphene [44, 45] and carbon nanotubes [46]. In addition to highlighting the role of disorder, our results address fundamental aspects of quantum interference. Semiclassical theory predicts that changing just one scattering site can have a significant effect on the MCF [47]. The gradual decorrelation observed here further indicates that the total number of altered scattering sites is important and has a cumulative effect on the MCF.

[1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14] [15] [16] [17] [18] [19]

FIG. 4. (Color) A plot of F (see text) v. Ti for the GaAs wire (red squares), the (GaIn)As/InP billiard (blue circles), and the GaAs/(AlGa)As billiard (green diamonds). The insets show the annealing function’s form as β and η are varied (see text).

In conclusion, we have shown that quantum interference in semiconductor billiards is determined by electron scattering off ionized donors. The role of the billiard walls is to reflect the electrons back towards these scatterers. This contradicts previous models in which the quantum dynamics was pictured as being determined purely by the walls [5, 12, 16–18, 20, 41] and which predict that the MCF should be robust to thermal cycling. Our results highlight the need for quantitative investigations aimed at determining the relative contributions of walls and ionized donors to both the MCF and the classical background magnetoconductance [34, 42, 43].

[20] [21] [22] [23] [24] [25] [26] [27] [28] [29] [30] [31] [32] [33] [34]

Y. Aharonov and D. Bohm, Phys. Rev., 115, 485 (1959). R. G. Chambers, Phys. Rev. Lett., 5, 3 (1960). B. Schwarzschild, Phys. Today, 39, 17 (1986). C. J. B. Ford et al., Superlattices and Microstructures, 4, 541 (1988). R. V. Jensen, Nature, 373, 16 (1995). A. Yacoby et al., Phys. Rev. Lett., 74, 4047 (1995). R. A. Webb et al., Phys. Rev. Lett., 54, 2696 (1985). S. B. Kaplan and A. Hartstein, Phys. Rev. Lett., 56, 2403 (1986). W. J. Skocpol et al., Phys. Rev. Lett., 56, 2865 (1986). R. P. Taylor et al., Surf. Sci., 196, 52 (1988). T. J. Thornton et al., Phys. Rev. B, 36, 4514 (1987). E. J. Heller and S. Tomsovic, Physics Today, 46, 38 (1993). R. Dingle et al., Appl. Phys. Lett., 33, 665 (1978). B. J. van Wees et al., Phys. Rev. Lett., 60, 848 (1988). R. P. Taylor, Nanotechnology, 5, 183 (1994). R. A. Jalabert, H. U. Baranger, and A. D. Stone, Phys. Rev. Lett., 65, 2442 (1990). C. M. Marcus et al., Phys. Rev. Lett., 69, 506 (1992). J. P. Bird et al., Phys. Rev. B, 51, 18037 (1995). I. V. Zozoulenko and K.-F. Berggren, Phys. Rev. B, 56, 6931 (1997). R. P. Taylor et al., Phys. Rev. Lett., 78, 1952 (1997). P. B. Wilkinson et al., Nature, 380, 608 (1996). H. J. St¨ ockmann and J. Stein, Phys. Rev. Lett., 64, 2215 (1990). C. Gmachl et al., Science, 280, 1556 (1998). P. B. Wilkinson et al., Phys. Rev. Lett., 86, 5466 (2001). K. Schaadt, A. P. B. Tufaile, and C. Ellegaard, Phys. Rev. E, 67, 026213 (2003). C. Zhang et al., Phys. Rev. Lett., 93, 074101 (2004). A. S. Sachrajda et al., Phys. Rev. Lett., 80, 1948 (1998). M. A. Topinka et al., Nature (London), 410, 183 (2001). R. Crook et al., Phys. Rev. Lett., 91, 246803 (2003). M. P. Jura et al., Nat. Phys., 3, 841 (2007). A. P. Micolich et al., Appl. Phys. Lett., 80, 4381 (2002). C. A. Marlow et al., Phys. Rev. B, 73, 195318 (2006). J. A. Nixon and J. H. Davies, Phys. Rev. B, 41, 7929 (1990). C. W. J. Beenakker and H. van Houten, in Solid State Physics, Vol. 44, edited by H. Ehrenreich and D. Turnbull

5 (Academic Press, Boston, 1991). [35] R. P. Taylor et al., Can. J. Phys., 70, 979 (1992). [36] R. P. Taylor et al., Phys. Rev. B, 56, R12733 (1997). [37] P. M. Mooney, N. S. Caswell, and S. L. Wright, J. Appl. Phys., 62, 4786 (1987); P. M. Mooney, ibid., 67, R1 (1990). [38] D. K. Maude et al., Phys. Rev. Lett., 59, 815 (1987). [39] C. Berven et al., Phys. Rev. B, 50, 14639 (1994). [40] T. N. Theis, in Inst. Phys. Conf. Ser., Vol. 91 (1987) Chap. 1.

[41] [42] [43] [44]

A. L¨ ofgren et al., Phys. Rev. Lett., 92, 046803 (2004). C. J. B. Ford et al., Phys. Rev. Lett., 62, 2724 (1989). J. Spector et al., Appl. Phys. Lett., 56, 1290 (1990). A. K. Geim and K. S. Novoselov, Nat. Mater., 6, 183 (2007). [45] J. H. Chen et al., Nat. Phys., 4, 377 (2008). [46] J.-C. Charlier, X. Blase, and S. Roche, Rev. Mod. Phys., 79, 677 (2007). [47] S. Feng, P. A. Lee, and A. D. Stone, Phys. Rev. Lett., 56, 1960 (1986).