Available online at www.sciencedirect.

com

Physica E 21 (2004) 483 486 www.elsevier.com/locate/physe

Fabrication of double quantum dots by combining afm and e-beam lithography

Abstract In recent years several attempts have been made to fabricate coupled quantum dots as a crucial element of quantum computing devices. One important challenge is to achieve a reliable control of the interdot tunneling. For this purpose we have combined direct nanolithography by local anodic oxidation (LAO) with standard electron-beam lithography. LAO is used to produce parallel double quantum dots. Additional metallic split gates are responsible for the control of the interdot coupling. We describe our fabrication scheme and demonstrate the function in low-temperature transport measurements. ? 2003 Elsevier B.V. All rights reserved.
PACS: 73.21.La; 73.23.Hk; 73.40.Gk Keywords: Coupled quantum dots; Atomic force microscope; Local anodic oxidation; Electron-beam lithography

1. Introduction The development of several technologies for lithography on very small scales has led to great scientic activities in the eld of mesoscopic systems. During the last decade quantum dots as a solid-state-based realization of atom-like potentials have gained great interest [1]. Due to their discrete energy spectra these quasi zero-dimensional objects are often referred to as articial atoms [2]. Following this analogy two coupled quantum dots can be viewed as articial molecules [3]. Single and double quantum dots are
Corresponding author. Tel.: +49-511-762-19002; fax: +49511-762-2904. E-mail address: [email protected] (M.C. Rogge).

often mentioned as elements of quantum computing devices [4]. Several double quantum dots have been fabricated in recent years (e.g. Refs. [57]). The most common fabrication technique is the so-called split gate technique, where metallic gates dened by electron-beam (e-beam) lithography are patterned on the surface of semiconductor materials containing a two-dimensional electron system (2DES). The negative biasing of these gates locally depletes the 2DES underneath forming well-dened electronic potentials. We introduce an alternative technique for the fabrication of double quantum dots with tunable interdot coupling [8]. We combine conventional e-beam lithography with local anodic oxidation (LAO) using the tip of an atomic force microscope

1386-9477/$ - see front matter ? 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.physe.2003.11.054

484

M.C. Rogge et al. / Physica E 21 (2004) 483 486

Fig. 1. (a) Schematic view of the sample after the second lithography step. The LAO parts include in-plane gates 1 and 2, islands A and B and all ve tunneling barriers. Also the top gate stump produced by e-beam lithography is shown. (b) Schematic view after the second and nal e-beam step. The top gate is complete. (c) Inner and outer metallic markers are used for alignment. (d) AFM image of the nal sample.

(AFM) [9,10]. The negative biasing of the tip leads to a local oxidation of the sample surface depleting the 2DES underneath. Therefore LAO is a powerful tool to write dened geometries on the scale of a few 10 nm into the 2DES. Compared to common split gate structures LAO-based geometries are easier to produce and operate. They are less sensitive to electrical discharges and allow for steeper electronic potentials. We use LAO to write a basic double dot structure including in-plane gates and tunneling barriers. We enhance the geometry with an additional metallic top gate patterned on top of the LAO parts with e-beam lithography. This top gate is responsible for the tunability of the interdot tunneling nally providing a fully controllable double quantum dot. A schematic overview of our geometry can be seen in Figs. 1(a) and (b). Fig. 1(a) shows the LAO parts of the structure before adding the nal top gate. Two in-plane gates (1 and 2) leave an 800 nm wide space in between, where further oxide lines are placed to form two islands (A and B) for the two quantum dots. There are four 80 90 nm wide point contacts connecting the two dots to the leads in parallel. Another 100 nm wide point contact is placed in between the two dots. The tunneling barriers connecting island A to the leads can be tuned via in-plane gate 1, and those of island B via gate 2, respectively. Both the gates couple capac-

itively to the islands and thus allow to shift the electrochemical potentials. Since the interdot barrier is far away from the in-plane gates it is not yet tunable. Its width of 100 nm provides a connection between both islands leaving the system in a state with only one big quantum dot occupying both islands. Fig. 1(b) shows the nal device after the lithography of the metallic top gate which is placed on top of the central tunneling barrier. By biasing the top gate we are now able to tune this barrier so as to separate or merge the two dots. Since already there is an LAO-dened barrier underneath the top gate we only have to shift its height slightly to control the interdot coupling. Therefore we only need small voltages at the top gate that are not su cient to deplete the drain region and thus leave a common drain contact even with the dots separated. Since the top gate is only meant to tune the interdot tunneling barrier without depleting one dot, an exact alignment is crucial. In the following we will describe our fabrication scheme and the alignment procedure in detail. We start with a GaAs/AlGaAs heterostructure with a 2DES 34 nm below the surface, an electron mobility of = 42 m2 =Vs and a sheet density of n = 5 1015 m2 . The mesa and Ohmic contacts are dened by optical lithography. Since we need a common coordinate system for the two instruments involved in the following nano-lithography steps (electron microscope and AFM), we apply a rst e-beam step to write metallic markers consisting of 7 nm Cr and 30 nm Au. Metallic markers are necessary, because the oxide lines produced with LAO are not visible with the electron microscope when used for lithography. There are two sets of markers illustrated in Fig. 1(c). The inner markers are close enough to the center to be scanned by the AFM. Thus they allow for a su cient alignment in the later LAO step. The outer markers are intended for the nal e-beam step. Their distance to the center is large enough to be scanned safely with the electron microscope without exposing sensitive inner parts. In this rst e-beam step also the stump of the top gate (see Fig. 1(a)) is produced. It will be used as an additional marker very close to the center that allows a very high accuracy for the LAO step enormously reducing the negative in uence of a drifting AFM-piezo and bad scanning resolution at large scanning elds. In the second lithography step we align the AFM rst roughly to the inner markers and then exactly to the

M.C. Rogge et al. / Physica E 21 (2004) 483 486

485

VG1 (mV)
(a) 0.6

VG1 (mV)
-60
-95 250 -65 -96

VG1 (mV)
-66 243

-120
VTG=0

G (e2/h)

VG2 (mV)

0.2

n+1 m+1 n m n m+1

0.0 (b) 0.4

G (e2/h)

A VTG=-110 mV A A B A A

B A A

220 235

(a)

(b)

213

0.2 0.0 -100

VTG=-102 mV

VG2 (mV)

CA

-20

VG1 (mV)
Fig. 2. (a) Di erential conductance G versus VG1 at VTG = 0. Equidistant Coulomb blockade peaks of a single quantum dot are observed. (b) At VTG = 110 mV two di erent peak types A and B appear corresponding to two dots A and B.

2t

(c)
205 -97 -67

(d)

VG1 (mV)
Fig. 3. Di erential conductance G versus VG1 and VG2 for di erent top gate voltages: (a) at VTG = 106 mV the splitting is very faint. n and m denote the charge on the almost separated dots; (b) at VTG = 104 mV the splitting becomes visible indicating a weak interdot coupling; (c) VTG =102 mV the splitting increases further and forms hexagonal cells and (d) the splitting is in general caused by a capacitive coupling CAB (dashed lines) and a tunnel coupling 2t forming bonding (solid lines) and anti-bonding states (dotted lines).

top gate stump and apply LAO as described above. Afterwards, we scan the geometry including the inner marker set to check our alignment. Thus we are able to compensate an eventual mismatch in the nal step. We nish our device with the second e-beam step. We align the electron microscope to the outer markers produced in the rst e-beam step and complete the top gate with the 100 nm wide metal nger. An AFM image of the nal sample is shown in Fig. 1(d). With this procedure we achieve the required alignment accuracy between LAO and e-beam parts which is better than 50 nm. We investigate the tunability of our device in transport measurements in a 3 He= 4 He-dilution refrigerator at 50 mK base temperature. In the standard lock-in technique we measure the di erential conductance using an AC-excitation voltage of 10 V at a frequency of 16 Hz. Fig. 2(a) shows the di erential conductance G versus gate voltage VG1 at in-plane gate 1 with the top gate grounded (VTG = 0). We observe the equidistant Coulomb blockade oscillations of a single quantum dot. The capacitance to gate 1, derived from the intervals of the Coulomb peaks is 28 aF. A similar capacitance of 25 aF to gate 2 (measurement not shown here) indicates the dot to be centered in the middle of the geometry. Thus at VTG = 0, the barrier between A and B is open as intended and A and B have merged to form one big quantum dot extended over both islands.

When we apply a negative voltage of VTG = 110 mV to the top gate (see Fig. 2(b)), we nd two kinds of peaks marked A and B indicating the Coulomb blockade of two quantum dots A and B. Accordingly, the capacitances to gate 1 are 21 aF for dot A and 2 aF for dot B. Sweeping gate 2 yields the capacitances to gate 2 of 8 aF for dot A and 19 aF for dot B (measurement not shown here). These values indicate dot A to be in close vicinity of gate 1 and dot B in close vicinity of gate 2. Thus in the regime of VTG = 110 mV the barrier between A and B is closed and the large quantum dot mentioned above is separated into two smaller quantum dots A and B each occupying one island. Since a voltage of VTG 190 mV is required to deplete the drain region underneath the top gate, there still is one common drain contact. A more distinct insight can be gained by measuring G as a function of both in-plane gate voltages as shown in Fig. 3. Fig. 3(a) depicts the situation for VTG = 106 mV. The lines with a higher slope correspond to dot A marking transitions from m to m + 1

VG2 (mV)

0.4

VTG=-106 mV

VTG=-104 mV

486
235

M.C. Rogge et al. / Physica E 21 (2004) 483 486

VTG=-102 mV

0.6
(b)
60

(a) 205 -97

-67

0.0 -107

-95

VG1 (mV)

VTG (mV)

Fig. 4. (a) The fractional peak splitting is dened as F = 2 VS = VP were VS is the distance between the split triple points and VP the distance between two equivalent triple points. (b) F as a function of VTG (left legend) grows from 0:1 at VTG = 106 mV to 0.55 at VTG = 96 mV. With the assumption that CAB is very small we can derive the energy splitting 2t (right legend).

electrons on dot A. The lines with a lower slope correspond to transitions on dot B from n to n + 1 electrons. At the intersections both dots are in resonance. Fig. 3(a) shows almost no splitting at these intersections, i.e. both dots are still very well separated (at VTG =110 mV we actually see no splitting within the accuracy of measurement). When we apply slightly more positive voltages of VTG =104 mV (Fig. 3(b)) we observe a faint splitting of the triple points as a result of a beginning coupling. With growing VTG (e.g. 102 mV in Fig. 3(c)) this splitting becomes clearly visible forming hexagonal cells. The splitting can be understood as a combination of two e ects [5] (see Fig. 3(d)). The electrostatic coupling CAB between the islands A and B introduces an energy shift of the ground state of one dot each time an electron enters the other dot. This leads to a splitting as indicated by the dashed lines. The tunnel coupling between both dots in analogy to real molecules leads to the formation of binding and anti-binding states that di er in energy. This e ect leads to a rounding of the triple points and thus to a further splitting of 2t (solid lines for binding and dotted lines for anti-binding states). The splitting is quantitatively characterized by the fractional peak splitting F [11] (for a denition see Fig. 4(a)). F is 0 for two separated dots and 1 for two totally merged dots. Fig. 4(b) shows the fractional peak splitting F as a function of VTG (left legend). The linear dependence over a wide range of F from 0.1 at

VTG = 106 mV (0 at VTG = 110 mV) to 0.55 at VTG = 96 mV and the fact that we observe one single large quantum dot at VTG =0 impressively demonstrates the tunability of the interdot tunnel coupling with top gate voltage. Due to the fact that we do not observe any capacitive coupling at VTG = 110 mV we assume CAB to be very small and thus the splitting to be governed by tunnel coupling. Therefore we can calculate the molecular energy splitting 2t (right legend). With increasing tunnel coupling the energy splitting grows from 17 to 61 eV. To summarize we have introduced an alternative fabrication scheme for complex mesoscopic systems. We have combined conventional e-beam lithography with LAO using an AFM. With our alignment procedure, we are able to minimize the mismatch between e-beam and LAO parts to less than 50 nm. We have demonstrated our technique by fabricating a parallel double quantum dot with tunable interdot coupling. Transport measurements under low-temperature conditions conrm the tunability of our device. We thank D. Pfannkuche, F. Hohls and U. Zeitler for helpful discussions and BMBF for nancial support. References
[1] L.P. Kouwenhoven, C.M. Marcus, P.L. McEuen, S. Tarucha, R.M. Westerfeld, N.S. Wingreen, in: L.L. Sohn, L.P. Kouwenhoven, G. Schon (Eds.), Mesoscopic Electron Transport, (Eds.) Vol. 345 Series E Kluwer, Dordrecht, 1997, pp. 105 214. [2] M.A. Kastner, Rev. Mod. Phys. 64 (1992) 849. [3] W.G. van der Wiel, S.D. Franceschi, J.M. Elzermann, T. Fujisawa, S. Tarucha, L.P. Kouwenhoven, Rev. Mod. Phys. 75 (2003) 1. [4] D. Loss, D.P. DiVincenzo, Phys. Rev. A 57 (1998) 120. [5] R.H. Blick, D. Pfannkuche, R.J. Haug, K.V. Klitzing, K. Eberl, Phys. Rev. Lett. 80 (1998) 4032. [6] A.S. Adourian, C. Livermore, R.M. Westervelt, Appl. Phys. Lett. 75 (1999) 424. [7] A.W. Holleitner, R.H. Blick, A.K. Huttel, K. Eberl, J.P. Kotthaus, Science 297 (2002) 70. [8] M.C. Rogge, C. Fuhner, U.F. Keyser, R.J. Haug, M. Bichler, G. Abstreiter, W. Wegscheider, Appl. Phys. Lett. 83 (2003) 1163. [9] M. Ishii, K. Matsumoto, Jpn. J. Appl. Phys. 34 (1995) 1329. [10] U.F. Keyser, H.W. Schumacher, U. Zeitler, R.J. Haug, K. Eberl, Appl. Phys. Lett. 76 (2000) 457. [11] J.M. Golden, B.I. Halperin, Phys. Rev. B 54 (1996) 16757.

VG2 (mV)

2t (eV)