> -
Help for Webmasters « back to results for ""
Below is a cache of http://www.lassp.cornell.edu/lassp_data/RALPH/Pubs/Champagne_GMCBJ.pdf. It's a snapshot of the page taken as our search engine crawled the Web.
The web site itself may have changed. You can check the current page or check for previous versions at the Internet Archive. Yahoo! is not affiliated with the authors of this page or responsible for its content.
Mechanically-adjustable and electrically-gated single-molecule transistors
1
Mechanically-adjustable and electrically-gated single-molecule transistors
A. R. Champagne, A. N. Pasupathy, and D. C. Ralph
Laboratory of Atomic and Solid State Physics, Cornell University, Ithaca, NY, 14853.
We demonstrate a device geometry for single-molecule electronics experiments that
combines both the ability to adjust the spacing between the electrodes mechanically and the
ability to shift the energy levels in the molecule using a gate electrode. With the
independent in-situ variations of molecular properties provided by these two experimental
"knobs", we are able to achieve a much more detailed characterization of electron transport
through the molecule than is possible with either technique separately. We illustrate the
devices' performance using C
60
molecules.
PACS numbers: 81.07.Nb, 73.63.-b, 73.23.Hk 2
A primary challenge in the field of single-molecule electronics [1,2] is to develop
adjustable devices that can enable well-controlled, systematic experiments. If one uses
techniques that measure only a current-voltage (I-V) curve, it can be difficult to determine
even whether a molecule is present between electrodes, because nonlinear transport across
tunnel junctions or metallic shorts can easily be mistaken for molecular signals [3].
Previous efforts to overcome this difficulty have employed two separate strategies for
systematically adjusting a molecular device in situ to make changes that can be compared
with theory. Electrostatic gating permits control of electron transport through a molecule
by shifting its energy levels [4-7]. Mechanical adjustability, using scanning probes [8-11]
or mechanically-controlled break junctions [12-15], enables manipulation of the device
structure and the strength of bonding to electrodes. Here we report the implementation of
both electrostatic gating and mechanical adjustability within the same single-molecule
device. This combined capability enables a detailed characterization of electrical transport
in molecules, providing understanding that is not possible with just gating or mechanical
adjustability separately.
Our scheme for combining electrostatic gating and mechanical adjustability is to add
a gate electrode to a mechanically-controlled break-junction (MCBJ). An MCBJ [16-19]
consists of a narrow bridge of metal suspended above a flexible substrate (Fig. 1(a,b)). By
bending the substrate, one can break the bridge and then adjust the spacing between the
resulting electrodes. The main challenge in fabricating an electrically-gated MCBJ is to
minimize the molecule-gate spacing to enable useful gating. We use as our gate electrode a
degenerately-doped Si substrate, which allows us to employ standard lithographic tech-
niques to produce a molecule-gate spacing as small as 40 nm, compared to ~1 mm substrate
spacings achieved previously with MCBJs on rougher metal or glass substrates [18]. 3
To fabricate the devices, we first grow a 250-nm thick SiO
2
film on top of a 200 mm-
thick degenerately-doped silicon wafer. Using photolithography and a hydrofluoric-acid
etch, we open windows in this thick oxide film and grow a thinner 40-nm oxide for the
device regions. We then use photolithography, electron-beam lithography and liftoff to
pattern Au lines 32 nm thick and 500 nm long, with a 50-nm wide constriction in the
middle, connected to larger-area contact pads. The Au lines are positioned within the thin-
oxide windows while the contact pads lie on the thicker-oxide regions. A timed buffered-
hydrofluoric-acid etch is used to remove the SiO
2
from under the Au bridge, suspending it
above the silicon substrate. A scanning electron microscope (SEM) image of a device
before our Au bridge is broken is shown in Fig. 1(b).
We bend the substrate by placing a 15-mm long by 6-mm wide Si chip against two
supports spaced 10 mm apart, and applying a pushing screw to the middle of the chip (Fig.
1(a)). The fine-threaded screw (1/80 inch pitch) is driven by a stepper motor via a series of
reducing gears (factor of 100 reduction). Each chip contains 36 devices, of which 16 can
be wired bonded at once. The amount of bending allowed by the Si chip (~ 0.3 mm over 10
mm) is generally not sufficient to break the metal bridge by mechanical motion alone.
Therefore we first break the wires partially or fully at 4.2 K using electromigration, a
technique used previously to make nm-scale gaps for molecular transistors [20,4-7].
Electromigration is accomplished by ramping a voltage across the Au wire until the
resistance increases. If the electromigration process is stopped when the sample
conductance reaches a few times e
2
/h, mechanical motion can be used to complete the
process of breaking the wire, resulting in the stepwise reduction in conductance as observed
in other MCBJ devices [19] (inset of Fig. 1(c)). Electromigration can also be used to break
the wires completely, giving typically decice resistances ranging from 100s of kW to 100s
of M_. 4
To calibrate the motion, we performed measurements at 4.2 K on bare Au electrodes
after electromigration was used to fully break the wires. The junction resistance increases
smoothly upon bending the substrate (Fig. 1(c)). The resistance of a tunnel junction is
expected to depend on the width x of the tunneling barrier approximately as [21]




R � e
2
k
x
,

,
2
h
f
k
e
m
=
(1)
where f = 5.1 eV is the Au work function and m
e
the electron mass. From the data in Fig.
1(c), we extract a calibration of 5.57 � 0.06 pm per full rotation of the stepper motor. The
average over our devices is 5.4 � 0.3 pm/turn. The full range of motion available before the
substrate breaks is generally 5 �. A few devices (not included in the average) exhibited
much less motion per turn, together with large gate leakage currents (> 1 nA at a gate
voltage V
g
= 1 V). In SEM images, these were identified as collapsed bridges.
If the substrate bends uniformly, the expected source-drain displacement is [19]
,
3
2
y
L
ut
x
d
d
=
(2)
where u is the length of the suspended bridge (500 nm), t is the wafer thickness (200 mm), L
is the distance between the two support posts used for bending (10 mm), and dy is the
pushing screw displacement (3.1 mm per full rotation of the stepper motor). This formula
predicts dx = 9 pm per full rotation, within a factor of 2 of the calibration result. The
difference is similar to results for other MCBJs [16-19], and can be ascribed to
uncertainties in the tunnel barrier height or to non-uniform bending of the substrate [19,22].
To characterize the performance of the gate electrode, we must insert into the device
a molecule with low-lying energy levels, which can be shifted by V
g
to modulate current
flow. We chose C
60
molecules because they have been used successfully to make single- 5
molecule transistors [4,7]. To fabricate C
60
devices, we first clean unbroken Au wires in
acetone, isopropanol, and oxygen plasma. We deposit a 0.2 mM solution of C
60
in toluene,
and blow dry after 30 s. Then we cool to 4.2 K and perform electromigration until the
devices conductance falls below a quantum of conductance In approximately 30% of
samples, one or more C
60
molecules bridge between the electrodes [4]. We identify these
samples as the ones whose I-V curves display Coulomb-blockade characteristics -- with
non-negligible current only for
V
greater than threshold values that depend on the gate
voltage V
g
. Control junctions formed from bare Au electrodes did not exhibit such
Coulomb-blockade characteristics. All the measurements that we will describe were
conducted at 4.2 K.
The effect of V
g
and V on conductance, at fixed source-drain displacement, is shown
in Fig. 2 for three samples at two different source-drain distances each. Fig. 3(a) shows the
corresponding Coulomb-blockade I-V curves for sample #1 at V
g
= 2.5 V. The dark regions
on the left and/or right of each panel in Fig. 2 correspond to low-current regions of
Coulomb blockade. Bright regions denote large dI/dV, where the applied source-drain
voltage provides sufficient energy for electrons to tunnel via the molecule and initiate
current flow. The threshold
V
required for current flow depends on V
g
, which shifts the
energy of the molecular states with respect to the Fermi energy of the electrodes. The
energy to add an electron to a molecular level can be tuned to zero for a particular value of
gate voltage
deg
V
V
g
=
.
As can be seen in Fig. 2, the sensi