Molecular-based electronically switchable tunnel junction devices

Article abstract of DOI:10.1021/ja0114456

Solid-state tunnel junction devices were fabricated from Langmuir Blodgett molecular monolayers of a bistable [2]catenane, a bistable [2]pseudorotaxane, and a single-station [2]rotaxane. All devices exhibited a (noncapacitive) hysteretic current-voltage r

Full text of DOI:10.1021/ja0114456

12632
J. Am. Chem. Soc. 2001, 123, 12632-12641
Molecular-Based Electronically Switchable Tunnel Junction Devices
C. Patrick Collier, Jan O. Jeppesen, Yi Luo, Julie Perkins, Eric W. Wong,
James R. Heath,* and J. Fraser Stoddart*
Contribution from the California NanoSystems Institute and the Department of Chemistry and
Biochemistry, UniVersity of California, Los Angeles, 607 Charles E. Young DriVe East,
Los Angeles, California 90095-1569
ReceiVed June 12, 2001
Abstract: Solid-state tunnel junction devices were fabricated from Langmuir Blodgett molecular monolayers
of a bistable [2]catenane, a bistable [2]pseudorotaxane, and a single-station [2]rotaxane. All devices exhibited
a (noncapacitive) hysteretic current-voltage response that switched the device between high- and low-
conductivity states, although control devices exhibited no such response. Correlations between the structure
and solution-phase dynamics of the molecular and supramolecular systems, the crystallographic domain structure
of the monolayer film, and the room-temperature device performance characteristics are reported.
Introduction
candidates for solid-state molecular switch devices. These
molecules and supermolecules present many advantages for such
a study. First, electrochemically addressable (solution-phase)
bistability can be designed6 into these molecules (supermol-
ecules), and this bistability can be thoroughly characterized using
various optical and NMR spectroscopies.7 Second, the catenane
and rotaxane structures and pseudorotaxane superstructure can
be employed as different (supra)molecular architectures for
supporting similar molecular switching mechanisms. Third, the
chemistry of these systems is sufficiently flexible so that
amphiphilic character can be either directly, or indirectly,
incorporated into their (super)structures. This property allows
for the preparation of molecular monolayer Langmuir Blodgett
(LB) films that can subsequently be incorporated into devices.
In this contribution, we report on solid-state devices fabricated
from a bistable [2]catenane,8 a bistable [2]pseudorotaxane,4 and
a single-station [2]rotaxane9 (Figure 1). Molecular electronics
switches fabricated from the [2]catenane were previously
reported,1 and those devices are discussed more fully here. The
syntheses of the [2]pseudorotaxane and the single-station [2]-
rotaxane are described in this paper. All (super)molecules were
prepared as LB films for incorporation into solid-state devices.
Brewster angle microscopy (BAM), as well as various scanning
probe microscopies, were utilized to interrogate the structure
of the LB films. Remnant response curves, device cycling,
device volatility, and current-voltage traces were recorded for
the various solid-state devices, and those responses are correlated
with the structure of the (super)molecular switches.
Molecular electronics-based solid-state switches have been
proposed as the active components in either nonvolatile random
access memory circuits1 or as the configurable bits for a custom-
configurable logic-based computing machine.2 The basis of such
a device is a two-terminal molecular tunnel junction that can
be electrically switched between high- and low-conductivity
states. If the active device characteristics (i.e., the switching
mechanism, the device volatility, the conductance of the device
in its various states, etc.) arise from intrinsic molecular
properties, then rational design of the switching molecule can
be employed to optimize the switching characteristics. Further-
more, such molecular electronics devices should, in principle,
exhibit a self-similarity with respect to device performance
parameters, even as the devices are scaled to molecular
dimensions.3
The development of a working model for correlating molec-
ular structure/device property relationships represents a formi-
dable challenge.4 For example, there has yet to be demonstrated
a common analytical tool that can be used to correlate the
structure and dynamics of molecules in either the solution-phase
or as thin films, with the characteristics of solid-state molecular
electronic devices. Thus, only through systematic investigation
of how device performance is modified through molecular
structure variations can one hope to begin piecing such a model
together. We have been investigating molecular mechanical and
supramolecular complexes from the classes of compounds5
known as catenanes, rotaxanes, and pseudorotaxanes as potential
* Correspondence address: Dr. James R. Heath, Department of Chemistry
and Biochemistry, UCLA, 607 Charles E. Young Drive East, Los Angeles
CA 90095-1569. E-mail: heath@chem.ucla.edu.
(1) Collier, C. P.; Mattersteig, G.; Wong, E. W.; Luo, Y.; Sampaio, J.;
Raymo, F. M.; Stoddart, F. M.; Heath, J. R. Science 2000, 289, 1172-
1175.
Results and Discussion
Synthesis and Characterization. The synthesis of the [2]-
catenane 14+ has already been reported.8 Here, we will describe
(2) Heath, J. R.; Keukes, P. J.; Snider, G.; Williams, R. S. Science 1998,
280, 1716-1721.
(6) Balzani, V.; Credi, A.; Raymo, F. M.; Stoddart, J. F. Angew. Chem.,
Int. Ed. 2000, 39, 3349-3391.
(3) Reed, M. A.; Zhou, C.; Muller, C. J.; Burgin, T. P. Science 1997,
278, 252-254.
(7) Ashton, P. R.; Ballardini, R.; Balzani, V.; Credi, A.; Dress, K. R.;
Ishow, E.; Kleverlaan, C. J.; Kocian, O.; Preece, J. A.; Spencer, N.; Stoddart,
J. F.; Venturi, M.; Wenger, S. Chem Eur. J. 2000, 6, 3558-3574.
(8) Balzani, V.; Credi, A.; Mattersteig, G.; Matthews, O. A.; Raymo, F.
M.; Stoddart, J. F.; Venturi, M.; White, A. J. P.; Williams, D. J. J. Org.
Chem. 2000, 65, 1924-1936.
(4) Pease, A. R.; Jeppesen, J. O.; Stoddart, J. F.; Luo, Y.; Collier, C. P.;
Heath, J. R. Acc. Chem. Res. 2001, 36, 433-444.
(5) (a) Amabilino, D.; Stoddart, J. F. Chem. ReV. 1995, 95, 2725-2828.
(b) Raymo, F. M.; Stoddart, J. F. Chem. ReV. 1999, 99, 1643-1663. (c)
Lindoy, L. F.; Atkinson, I. M. Self-Assembly in Supramolecular Systems;
Stoddart, J. F., Ed.; Royal Society of Chemistry: London, 2000.
(9) Jeppesen, J. O.; Perkins, J.; Becher, J.; Stoddart J. F. Org. Lett. 2000,
2, 3547-3550.
10.1021/ja0114456 CCC: $20.00 © 2001 American Chemical Society
Published on Web 11/22/2001

Molecular Switch Tunnel Junction DeVices
J. Am. Chem. Soc., Vol. 123, No. 50, 2001 12633
Figure 1. Molecular formulas and graphical representations of the [2]catenane 14+, the two station [2]pseudorotaxane 2‚CBPQT4+, and the single-
station [2]rotaxane 34+
.
Scheme 1. Synthesis of the Hydrophobic Stopper.
the preparation of a hydrophobic, and also a hydrophilic, stopper
before outlining the syntheses of the single-station [2]rotaxane
3
4+, incorporating both stoppers, and the two-station [2]-
pseudorotaxane 2‚CBPQT4+ where the semidumbbell compo-
nent contains only the hydrophobic stopper. Molecular formulas
and graphical representations of the [2]catenane 14+, the two-
station [2]pseudorotaxane 2‚CBPQT4+, and the single-station
[2]rotaxane 34+ are shown in Figure 1.
The hydrophobic tetraarylmethane-based stopper 6 was
obtained (Scheme 1) in 50% yield by a modification of the
procedure already reported10 in the literature. Although we
previously used the phenol 6 as its potassium salt during its
alkylation, we now prefer to isolate it directly as the phenol
after acid-catalyzed reaction between (4-tert-butylphenyl)-4-
ethylmethanol10 (4) and phenol (5). Alkylation of 6 with 2-[2-
(2-chloroethoxy)ethoxy]tetrahydropyran11 (7) in n-butanol, fol-
lowed by removal of the THP-protecting group with aqueous
HCl gave 8 in 84% yield. The alcohol 8 can either be brominated
using CBr4 and Ph3P in CH2Cl2 or tosylated using TsCl in a
mixture of THF-H2O, affording 9 or 10 in 95 or 53% yields,
respectively. Subsequent treatment of 9 or 10 with NaI gave
the corresponding iodide 11 in 95% yield or better.
For the synthesis (Scheme 2) of the hydrophilic stopper, we
employed a modified literature procedure.12 Methyl 4-hydroxy-
benzoate (12) was alkylated (K2CO3 in MeCN) with toluene-
4-sulfonic acid 2-(2-methoxyethoxy)ethyl ester13 (13). The
resulting ester 14 was reduced (LiAlH4/THF), and the benzyl
alcohol 15 was chlorinated (SOCl2/CH2Cl2), affording 16 in 60%
overall yield for the three steps. Alkylation (K2CO3/DMF) of
(10) Ashton, P. R.; Ballardini, R.; Belohradsky, M.; Gandolfi, M. T.;
Philp, D.; Prodi, L.; Raymo, F. M.; Reddington, M. V.; Spencer, N.;
Stoddart, J. F.; Venturi, M.; Williams, D. J. J. Am. Chem. Soc. 1996, 118,
4931-4951.
(11) Gibson, H. W.; Lee, S.-H.; Engen, P. T.; Lacavalier, P.; Sze, J.;
Shen, Y. X.; Bheda, M. J. Org. Chem. 1993, 58, 3748-3756.
(12) Percec, V.; Cho, W.-D.; Mosier, P. E.; Ungar, G.; Yeardley, D. J.
P. J. Am. Chem. Soc. 1998, 120, 11061-11070 and references therein.
(13) Satoru, I.; Hideo, M.; Kioshi, T. J. Chem. Soc., Perkin Trans. 1
1997, 1357-1360.
methyl 3,4,5-trihydroxybenzoate (17) with the chloride 16 gave
the ester 18 in 82% yield. Compound 20 was isolated in 68%
overall yield following (i) reduction (LiAlH4/THF) of 18 to give
the alcohol 19 and (ii) chlorination (SOCl2/DTBMP/CH2Cl2)
of this alcohol. Next, 4-hydroxybenzyl alcohol (21) was
alkylated (K2CO3/DMF) with the chloride 20 to yield the alcohol
22, which was chlorinated (SOCl2/DTBMP/CH2Cl2) in 95%

12634 J. Am. Chem. Soc., Vol. 123, No. 50, 2001
Scheme 2. Synthesis of the Hydrophilic Stopper.
Collier et al.
protected semidumbbell compound 33, which on treatment with
TsOH in THF-EtOH, gave the semidumbbell compound 2 in
98% yield.
Solution-State Characterization. The full characterization
of the [2]catenane and 14+ has already been reported.8 The [2]-
rotaxane 34+ contains only one donor stationsthe TTF units
and exhibits a charge transfer (CT) band centered around 810
nm (ꢀ ) 1400 L mol-1 cm-1) in Me2CO at 298 K. Since the
two-station [2]pseudorotaxane 2‚CBPQT4+ contains two donor
unitssa TTF unit and a NP onesit is expected to exist as a
mixture of two co-conformerssone where the CBPQT4+
encircles the TTF unit and the other where the CBPQT4+ resides
around the NP unit. In fact, a 1:1 mixture of CBPQT4+ and the
semidumbbell 2 exhibits CT bands centered around 540 nm (ꢀ
) 870 L mol-1 cm-1)sfor the NP‚CBPQT4+ CT interactions
and 785 nm (ꢀ ) 1000 L mol-1 cm-1) for the TTF-CPBQT4+
CT interaction in MeCN at 298 K which indicates that the [2]-
pseudorotaxane 2‚CBPQT4+ is indeed a mixture of two co-
conformers. UV-vis dilution experiments were carried out to
determine18 the binding constant (Ka) for 1:1 complexation of
CBPQT4+ with the semidumbbell 20. Ka Values were obtained
using both the NP‚CBPQT4+ and TTF‚CBPQT4+ CT absorption
bands as probes: They were 85000 ( 15000 and 95000 (
15000 M-1, respectively, corresponding to free energies of
complexation of just under 7 kcal mol-1
Solid-State Device Fabrication. Solid-state molecular switch
tunnel junctions were fabricated from (1) the [2]catenane 14+
.
yield to give the hydrophilic stopper as a reactive benzyl chloride
23.
,
(2) the [2]pseudorotaxane 2‚CBPQT4+, and (3) the [2]rotaxane
34+ in an effort to elucidate molecular (super)structure/device
property relationships. The starting point for all of the devices
was 100 Si wafers coated with a thick (0.1 µm) oxide (SiO2)
film. All devices were fabricated using optical- or electron-beam
lithographically defined n-type polycrystalline (poly-Si) bottom
electrodes (0.05 or 5 µm wide and resistivity of 0.02 Ω-cm).
Typically, poly-Si films formed via direct chemical vapor
deposition growth onto SiO2 are neither smooth nor defect free.
Since amorphous Si films can be very smooth,19 however,
amorphous Si was used as the starting point for the fabrication
of smooth poly-Si electrodes. This many-step process, which
has been described previously,1 was critical for achieving a high
(>90%) device yield. Single monolayer films of 14+, 2‚
CBPQT4+, and 34+ were transferred onto the poly-Si electrode-
patterned substrates as Langmuir-Blodgett (LB) films (see
below). Following LB film deposition, 0.05 or 10 µm-wide top
electrodes (50 Å Ti followed by 1000 Å Al) were deposited
onto the LB film using electron-beam evaporation. Except where
noted otherwise, devices were measured in air and at room
temperature, using a shielded probe station with coaxial probes.
Bias voltages were applied to the poly-Si electrode, and the
top electrode was connected to ground through a DL Instruments
1211 current preamplifier. For certain experiments, such as
device cycling and remnant molecular signature measurements,
a series of “write” voltage pulses were applied to perturb the
device, and the response was “read” through the current preamp
by applying some constant (small) bias. For those measurements,
a mechanical relay switch was placed between the device, and
the current preamplifier was only connected electrically to the
device when the device was actually being “read”. Otherwise,
the relay switch connected the device directly to ground.
The synthesis of the [2]rotaxane 3‚4PF6 is outlined in Scheme
3. A THF-MeOH solution of the asymmetric tetrathiafulvalene
(TTF) derivative14 was treated with 1 equiv of CsOH‚H2O to
generate a TTF-monothiolate which could subsequently be
alkylated with 1 equiv of either the bromide 9 or the iodide 11
affording 25 in 76 or 91% yields, respectively. Removal of the
tosyl-protecting group from 25 was achieved in 92% yield by
refluxing the N-tosylate in THF-MeOH containing an excess
of NaOMe. Following N-alkylation (NaH/DMF) of the pyrrole
unit in 26 with the chloride 23, the dumbbell-shaped compound
27 was isolated in 80% yield. Using this compound as a template
for the formation of the cyclobis(paraquat-p-phenylene) tetra-
cation in DMF and starting from the dicationic precursor15 28‚
2PF6 and 1,4-bis(bromomethyl)benzene (29) afforded the single-
station [2]rotaxane 3‚4PF6 in 8% yield.
The synthesis of the semi-dumbbell compound 2, prior to its
being added to cyclobis(paraquat-p-phenylene) tetrakis(hexaflu-
orophosphate)16 CBPQT‚4PF6 to form the [2]pseudorotaxane
2‚CBPQT4+ is outlined in Scheme 4. The monotosylate 30,
which was obtained (TsCl/NEt3/CH2Cl2) in 48% yield from 1,5-
bis[2-(2-hydroxyethoxy)ethoxy]naphthalene,17 was converted
(NaI/Me2CO) almost quantitatively into the iodide 31 which
was protected (DHP/TsOH/CH2Cl2) in 88% yield to give the
THP “ether” 32. Alkylation (NaH/DMF) of compound 26
(Scheme 3) with the iodide 32 afforded in 86% yield the
(14) Jeppesen, J. O.; Takimiya, K.; Jensen, F.; Brimert, K.; Nielsen, K.;
Thorup, N.; Becher, J. J. Org. Chem. 2000, 65, 5794-5805.
(15) Anelli, P.-L.; Ballardini, R.; Balzani, V.; Gandolfi, M. T.; Goodnow,
T. T.; Kaifer, A. E.; Pietraszkiewicz, M.; Prodi, L.; Reddington, M. V.;
Slawin, A. M. Z.; Spencer, N.; Vicent, C.; Williams, D. J. J. Am. Chem.
Soc. 1992, 114, 193-218.
(16) (a) Odell, B.; Slawin, A. M. Z.; Spencer, N.; Stoddart, J. F.;
Williams, D. J. Angew. Chem., Int. Ed. Engl. 1988, 27, 1547-1551. (b)
Asakawa, M.; Dehaen, W.; L’abbe´, G.; Menzer, S.; Nouwen, J.; Ramyo,
F. M.; Stoddart, J. F.; Williams, D. J. J. Org. Chem. 1996, 61, 9591-
9595.
(18) Ka values were also obtained for this 1:1 complex in Me2CO at
298 K. Using the NP‚CBPQT4+ and TTF‚CBPQT4+ CT absorption bands,
centered on 545 nm (ꢀ ) 760 L mol-1 cm-1) and 745 nm (ꢀ ) 590 L
mol-1 cm-1), respectively, a common value of 25000 ( 3000 M-1 was
(17) Ashton, P. R.; Ballardini, R.; Balzani, V.; Boyd, S. E.; Credi, A.;
Gandolfi, M. T.; Go´mez-Lo´pez, M.; Iqbal, S.; Philp, D.; Preece, J. A.; Prodi,
L.; Ricketts, H. G.; Stoddart, J. F.; Tolley, M. S.; Venturi, M.; White, A.
J. P.; Williams, D. J. Chem. Eur. J. 1997, 3, 152-170.
obtained for Ka corresponding to a free-energy complexation of 6 kcal mol-1
.
(19) Kamins, T. Polycrystalline Silicon for Integrated Circuit Applica-
tions; Kluwer Academic: Boston, 1988.

Molecular Switch Tunnel Junction DeVices
Scheme 3. Synthesis of the [2]Rotaxane 34+
J. Am. Chem. Soc., Vol. 123, No. 50, 2001 12635
Scheme 4. Synthesis of the Semidumbbell Compound 2 and
H2O or (b) 18.2 MΩ H2O containing 6.4 mM CdCl2 (aq),
adjusted to pH 8.5 with NaOH (aq). The Langmuir monolayers
were compressed at 10 cm2/min to a surface pressure of 1.0
mN/m, 5 cm2/min to a surface pressure of 14.0 mN/m, and
finally at 2 cm2/min to the deposition pressure of 30 mN/m.
The monolayers were then deposited as LB films onto the
electrode-patterned poly-Si substrates on the upstroke at 1 mm/
min, after equilibrating for 3 min. Compounds 14+ and 34+ and
complex 2‚CBPQT4+ were transferred as LB monolayers at
(supra)molecular areas of 74(5), 83(6), and 38(4) Å2/(super)-
molecule respectively, on an aqueous CdCl2 subphase, or at 125-
(8), 122(8), and 35(4) Å2/(super)molecule using a pure water
subphase.
the Formation of the [2]Pseudorotaxane 2‚CBPQT4+
.
The preparation of monolayers of the [2]catenane 14+ has
been described previously.1 The tetracationic compound was
co-deposited on the subphase with an anchoring phospholipid,
dimyristoylphosphatidic acid, as the counterion (DMPA-). The
DMPA- anion, employed initially as its monosodium salt
-
dissolved in CHCl3/MeOH (3:1), and the PF6 salt of the
tetracationic [2]catenane were dissolved initially in MeCN.
Immediately before spreading, the DMPA and catenane solu-
tions were mixed in a 6:1 molar ratio. By contrast, the
-
amphiphilic [2]rotaxane 34+ was spread directly as its 4PF6
salt from a CHCl3/MeCN (20:1) solution. Finally, the [2]-
pseudorotaxane 2‚CBPQT4+ is different from the [2]catenane
1
4+, and the [2]rotaxane 34+, insofar as its semidumbbell
component, containing the two π-electron rich recognition sites,
can dissociate easily from the tetracationic cyclophane (CB-
PQT4+) component. Monolayers of this 1:1 complex were
prepared by mixing Na‚DMPA, CBPQT4+, dissolved initially
in MeCN as its 4PF6- salt and the semidumbbell component 2
dissolved in MeCN, in a 12:2:1 molar ratio, prior to spreading.
Solid-State Device Performance. There are certain figures
of merit that characterize the quality of a two-terminal switching
or storage device, and the rank-order of those characteristics is
dictated by the targeted application. For these devices, we are
concerned with fabricating cross-point memory and logic
Langmuir molecular monolayers of (1) the [2]catenane 14+
,
(2) the [2]pseudorotaxane 2‚CBPQT4+, and (3) the [2]rotaxane
34+ were prepared on the water surface of a temperature-
controlled NIMA Type 611 trough. For all Langmuir films, the
compounds were deposited onto the water surface at 20 °C and
allowed to equilibrate for 30 min prior to transferring them to
the patterned substrates. The subphase was either (a) 18.2 MΩ

12636 J. Am. Chem. Soc., Vol. 123, No. 50, 2001
Collier et al.
structures, meaning that the devices would be arranged in a two-
dimensional lattice of crossing wires, similar to an expanded
tic-tac-toe board. A disadvantage of the crosspoint architecture
is that all devices are electrically interconnected, a feature which
implies that the voltages that are used to address (open or close)
the devices must be sharp and well-defined. Otherwise, it is
very difficult to change the state of one device in the circuit
without perturbing neighboring devicessa problem known as
“half-select”. Second, it is important to have a large signal
differencesresistance or current, for examplesbetween the
“open” and “closed” states. Given a sharp address voltage, it is
then the magnitude of this signal difference that ultimately
dictates the size of the crosspoint structure that can be reliably
used. Third, the volatility of the device should be low, since a
highly volatile device requires constant refreshing and will
correspondingly increase the power consumption and the
overhead requirements for operating such a circuit. Fourth,
diodesor current rectifying charactersof the molecular tunnel
junction is an important consideration since diode behavior can
lend noise immunity to the various devices within a crosspoint
circuit. Such rectification may be either a molecular or an
electrode characteristic; this issue, as it relates to molecules, is
discussed in some depth below. The final consideration is the
switching speed of the device. The fact that speed is the least
important consideration is perhaps surprising; indeed, for certain
other applications, device-switching speed may be much more
vital. However, the cross-point architecture is one that lends
itself naturally to parallel addressing, and this factor eases
substantially the demands on device-switching speeds.
Figure 2. BAM images of films of the [2]pseudorotaxane 2‚CBPQT4+
,
taken on different subphases on the Langmuir trough, and as transferred
LB films onto glass substrates. (a) Pure water subphase (18.2 MΩ, pH
) 5.5). (b) CdCl2 (aq) subphase (pH ) 8.5). (c) Pure water subphase,
transferred to a cleaned glass substrate. (d) CdCl2 (aq) subphase,
transferred to a cleaned glass substrate. The scale bar is 100 µm.
on the length scales accessible with BAM. The divalent Cd2+
cations have been demonstrated to lessen the electrostatic
repulsion between the phospholipid headgroups.21 This sub-
phase-dependent domain structure was retained in the LB films
that were transferred to cleaned glass substrates (Figure 2, c
and d), thereby establishing BAM as a structural feedback tool
for correlating device performance with monolayer structure.
Solid-state devices prepared with the [2]pseudorotaxane trans-
ferred from CdCl2 (aq) subphases, as compared with a pure
aqueous subphase, did behave more uniformly from one device
to the next. However, even when a CdCl2 (aq) subphase was
utilized, domain formation was still a factor, although the
domains were much smaller in size and thus were not resolved
by the BAM. The micrometer and submicrometer structures
observed for these films are discussed below.
For all of the devices discussed in this contribution, we carried
out measurements that were designed to characterize the
sharpness of the address voltages, the magnitude of the switching
response, the device volatility, and molecular-dependent current
rectification characteristics. The switching speeds of all of the
devices arise from a complex interplay of molecular properties
and device architecture. Optimization of the switching speed
would require an atomic scale engineering of the tunnel barriers
that separate the molecules from the electrodes. No such effort
was made for this study.
Structures of the Monolayers. In this work, it was not just
of importance to obtain a uniform coverage of the LB film on
the electrode, but it was also necessary to correlate the 2D
crystallographic structure of the monolayer with device perfor-
mance. Therefore, both Langmuir monolayers (on aqueous
subphases) and LB (transferred) films were interrogated using
Brewster angle microscopy20 (BAM, Nanofilm Technology).
The BAM technique is a polarized light-imaging probe that is
sensitive to small changes in the refractive index and can
therefore detect structural variations in a molecular monolayer,
although, in our case, only to a resolution of approximately 4
µm2. For certain LB films, smaller features were interrogated
using scanning probe microscopy techniques (Digital Instru-
ments SPM).
In Figure 2, BAM images reveal the effects of varying the
nature of the subphase on the structure of a Langmuir monolayer
and LB monolayer films of the [2]pseudorotaxane. Figure 2a
shows characteristic domains indicative of the phase separation
of the pure DMPA- anions from the mixture of these anions
with CBPQT4+ and the half-dumbbell component 2 of the [2]-
pseudorotaxane. Pure DMPA- spread on the water subphase
under the same conditions (pH and temperature) showed almost
identical patterns. Switching to a CdCl2 (aq) subphase (Figure
2b) resulted in considerably more homogeneous films, at least
BAM images were also collected from Langmuir monolayers
of the [2]catenane 14+ and the [2]rotaxane 34+. Images of the
[2]catenane on pure water versus CdCl2 (aq) showed, in a similar
fashion, the disappearance of structural domains. On the other
hand, for the [2]rotaxane, images from Langmuir monolayers
formed on pure water and CdCl2 (aq) appeared similar and
featureless. The transport characteristics of the [2]rotaxane-based
solid-state devices fabricated with this molecule also did not
exhibit a dependence upon subphase. This correspondence
appears to highlight the importance of using CdCl2 (aq) as the
subphase when DMPA- is used as an anchoring counterion.
Three modes of scanning probe microscopy (SPM), noncon-
tact topographic, lateral contact (friction) force, and surface
potential measurements were employed to image the submi-
crometer structure of LB films transferred onto poly-Si elec-
trodes. Lateral force images revealed structure such as domain
boundaries, and this technique exhibited a higher contrast than
the noncontact AFM mode. However, only with surface potential
imaging were we able to assign chemical identification to the
components in the domains. Figure 3 shows simultaneously
captured topographic (Figure 3a) and surface potential (Figure
3b) images of a film of 2‚CBPQT‚4DMPA transferred onto
poly-Si from a CdCl2 (aq) subphase. Whereas the BAM image
(21) Losche, M.; Mohwald, H. J. Colloid Interface Sci. 1989, 131, 56-
67.
(20) Mobius, D. Curr. Opin. Colloid Interface. Sci. 1998, 3, 137-142.

Molecular Switch Tunnel Junction DeVices
J. Am. Chem. Soc., Vol. 123, No. 50, 2001 12637
from 14+, 2‚CBPQT4+, and 34+, we expected to observe some
switching in the solid-state devices. The general solution-phase
mechanism for all three molecules is very similar, and we
discuss this picture first before moving to the solid-state devices.
In the ground state of the molecule (supermolecule), the
tetracationic cyclophane encircles the TTF binding site, although
for the [2]pseudorotaxane, there is competition between the TTF
and NP sites, as discussed above. For all three systems, the first
oxidation of the molecule (or complex) corresponds to the
removal of an electron from the TTF site, leading to Coulombic
repulsion between the TTF•+ site and the positive charged
CBPQT4+ ring. The [2]catenane 12+ and the [2]pseudorotaxane
2‚CBPQT4+ are both bistable systems since both the TTF and
the 1,5-dioxynaphthalene (DN) unit can serve as an electron
donors. Thus, in those systems, molecular mechanical motion
(circumrotation of the crown ether in 14+ and translation of the
cyclophane in 2‚CBPQT4+) is the observed solution-phase
switching mechanism. However, in the case of the [2]rotaxane
Figure 3. “Tapping mode” topographic (a) and surface potential (b)
SPM images of an LB film prepared as a Langmuir monolayer on a
CdCl2 (aq) subphase with a composition of [2]pseudorotaxane 2‚
CBPQT4+/6DMPA-/2Na+ and transferred to a poly-Si substrate. BAM
images of this film were featureless, as the domain structure is finer-
grained than the imaging resolution of the BAM. The surface potential
measurements indicate that the circular domains are rich in the [2]-
pseudorotaxane component, while the topographic measurements
indicate some aggregation within the (bright) DMPA--rich regions.
3
4+, the dumbbell component contains only one binding site.
Therefore, while oxidation of the TTF unit should induce a small
amount of motion on the part of the CBPQT4+ ring, the
switching characteristics of this molecule are expected to be
poor relative to either 14+ or 2‚CBPQT4+. When the TTF unit
is reduced back to the charge-neutral state, the molecule
(supermolecule) will revert back to its ground-state co-conformer
on a time-scale that varies from milliseconds to a few seconds,
depending on the system.
appeared to be homogeneous, the SPM images revealed that
domain structure was still present in this film, although at much
smaller length scales than was observed for the pure water
subphase.
To measure the surface potential, an oscillating voltage, Vac
sin ωt, was applied directly to the cantilever tip, and this created
an oscillating electric force at frequency ω. This oscillating force
has the amplitude F ) dC/dz ∆VdcVac, where dC/dz is the vertical
derivative of the tip-sample capacitance, ∆Vdc ) Vtip - Vsample
is the dc voltage difference between the tip and the sample,
and Vac is the oscillating voltage applied to the tip. When the
tip and sample surface are at the same dc potential, the cantilever
feels no oscillating force; in surface potential mode, the
microscope adjusts the voltage on the tip until the oscillating
force on the cantilever vanishes. The tip voltage is then equal
to the effective surface potential.
The surface potential measured this way is directly propor-
tional to the density of dipole moments directed perpendicular
to the surface.22 Thus, bright areas in the surface potential image
correspond to a higher density of vertically directed dipole
moments. A component of the film that is richer in the
[2]pseudorotaxane will possess a higher dipole density than a
component richer in phospholipid (DMPA-) counterion. There-
fore, the bright areas in the surface potential image correspond
to a domain that is at least dominated by the [2]pseudorotaxane.
The corresponding topographic image shows these regions to
be dark, with a height difference of about 5 nm between the
bright and dark regions. This observation implies that the
component rich in DMPA- is aggregating or polymerizing, and
this result can be corroborated with surface pressure/area
isotherms from the Langmuir trough. A film of the [2]-
pseudorotaxane using a freshly prepared solution of Na‚DMPA
yielded a lower molecular area (Å/molecule) at a particular target
pressure during transfer than an older Na‚DMPA solution.
Nevertheless, it was not possible to remove this heterogeneity
completely from these films, and some amount of microscopic
phase segregation was observed for every [2]pseudorotaxane/
DMPA LB film. Topographic images of a film of the [2]-
pseudorotaxane using “fresh” Na‚DMPA did reveal a greater
uniformity in the structural features of the film, and the height
difference between the phase-segregated domains was reduced
to about 2 nm.
We previously reported on the characteristics [2]catenane 14+
-
based solid-state devices, and we briefly review and expand
upon those results here. The [2]catenane devices were stable
switches, exhibiting roughly a factor of 100% change in junction
resistance between the 0 and 1 states, and could be cycled at
least a few hundred times. Moreover, the voltages required to
open or close the switches were stable from one cycle to the
next and from device-to-device. Monostable [2]catenane control
devices (two DN groups in the crown ether ring, rather than
one DN and one TTF group), as well as other controls, did
exhibit a switching response. The mechanism for the bistable
[2]catenane had at least one thermally activated component, and
no switching response was observed at temperatures below 230
K. In addition, as the device was cooled stepwise from 330 K
to 230 K, the voltages required to open or close the switch
systematically increased in amplitude. This was understood as
follows: In a single-molecule thick-tunnel junction, when the
molecule is oxidized, that charge is balanced by image charges
in the electrodes. At lower temperatures, the polysilicon
electrode becomes less polarizeable and therefore less able to
support an image charge. These results taken together implied
that at least some critical aspect of the solution-phase redox-
activated switching mechanism was retained in the solid-state
device.
There were also certain obvious differences between the
solution-phase and solid-state switching mechanism, and our
previous report1 on the [2]catenane 14+ highlighted some of
those differences. In particular, for both the switch-open and
switch-closed states of the device, the TTF group is charge-
neutral. Oxidation of the TTF group of the ground (switch-open)
state co-conformer does activate circumrotation of the ring and
thereby conversion of the device to the switch-closed state.
However, in a conducting-tunnel junction, the junction does not
remain charged after the bias is removed. For the solid-state
device, circumrotation of the crown ether ring back to the
ground-state co-conformation is greatly accelerated (by a factor
of 103) by applying a reducing bias to the device. One possibility
The Switching Characteristics of Molecular Electronic
Tunnel Junction Devices. In the case of the devices fabricated
(22) Mohwald, H. Annu. ReV. Phys. Chem. 1990, 41, 441-476.

12638 J. Am. Chem. Soc., Vol. 123, No. 50, 2001
Collier et al.
Figure 4. This graphical representation depicts a molecular mechanism
that is consistent with the observed voltage cycling of the [2]-
pseudorotaxane-based solid-state devices. It is not clear which co-
conformer represents the ground (switch-open) state of the device and
which represents the metastable (switch-closed) state. However, note
that it is necessary to charge the dumbbell component (oxidation to go
forward, reduction for the reverse pathway) to cycle the device between
the two (dumbbell charge neutral) co-conformers.
is that the activation barrier for circumrotation back to the
lowest-energy co-conformer is lowered by a partial reduction
of the anchoring DMPA- counterions. As discussed below, a
similar phenomena is observed for the 2‚CBPQT4+ device, and
here again, translation of the cyclophane ring from the DN site
to the TTF site may be aided by partial reduction of the
cyclophane ring. Previously, we proposed a possible mechanism
for the [2]catenane 14+ device. In Figure 4, we present a
switching mechanism for the 2‚CBPQT4+ device that is
analogous to what we hypothesized for the [2]catenane 14+
device.1 Note that both the open and closed states of the
dumbbell component of the molecular tunnel junction switch
are charge neutral, but that the pathways between the two states
involve charging the dumbbell component. For the 2‚CBPQT4+
device, one of the two co-conformers will obviously be the
ground state, while the other will be metastable. The reduction
of the CBPQT4+ ring is a component of the switch-cycling
mechanism that is not observed in the solution phase for any
of these systems. Other substantial variations between the solid-
state and solution-phase switching mechanisms are also likely.
For example, interactions between the molecular monolayers
and the electrodes may modify the relative stabilities of the
various co-conformers. One potentially significant issue is the
interaction between the molecular monolayer and the evaporated
Ti top electrode material. However, we have previously reported
measurements on a number of control molecules, including
simple linear-alkyl amphiphiles,23 as well as the monostable [2]-
catenane mentioned above, and others. If the Ti/molecular
interface impacts device performance, then that effect is subtle
and is not readily observed in the control experiments. In
addition, the crystallographic packing of the LB monolayers,
as discussed above, will also impact the solid-state device
characteristics.
Figure 5. Remnant molecular signatures of (a) the [2]catenane 14+
(b) the [2]pseudorotaxane 2‚CBPQT4+, and (c) the [2]rotaxane 34+
,
.
The [2]catenane and the [2]rotaxane devices yielded reproducible curves
from experiment-to-experiment, and from device-to-device. The [2]-
pseudorotaxane devices, however, exhibited large-amplitude switching
signatures that fluctuated from measurement-to-measurement, and two
representative measurements are plotted. The remnant molecular
signatures were monitored at -0.1 V, +0.1 V, and +0.2 V for
molecules 14+, 2‚CBPQT4+, and 34+ respectively.
current-voltage trace by some amount. In fact, such a voltage
offset is the signature of charge storage in a junction. However,
we are not looking to construct capacitors, but rather molecular
switch tunnel junctions. Therefore, to make certain that any
observed hysteretic response did not originate from device
capacitance, careful measurements, at a resolution of 10 mV
and 10-100 pA, were made to ascertain that the current passed
from a negative to a positive value at zero-bias. In any event,
the actual hysteresis characteristics (switching voltage/switching
magnitude) of a device are quantified using a measurement that
we have described previously1 as a remnant molecular signature
(RMolS). This experiment, which is similar to the remnant
polarization measurement used for ferroelectrics,24 consists of
applying a series of probe and read voltage pulses so that the
probe voltages cycle through a loop (from -2.5 to +2.5 V,
and back again, for example), and the read voltage is some
voltage near zero bias (+0.2 V for example). In this way, the
response of the device to some series of applied voltage is
interrogated in a relatively nonperturbing fashion, and device
capacitance is largely canceled out.
The signature of molecular switching in a (two-terminal)
molecular electronic tunnel junction is a hysteretic response in
the current-voltage characteristics. Device capacitance actually
provides a more general mechanism for generating a hysteretic
current/voltage response, since separating charge across a tunnel
junction will lead to a dc voltage offset, thereby shifting the
The RMolS curves for the [2]catenane 14+, the [2]pseudoro-
taxane 2‚CBPQT4+, and the [2]rotaxane 34+ are presented in
Figure 5, a, b, and c, respectively. For the [2]pseudorotaxane
device we measured the remnant molecular signature curves as
a function of the time spent writing to the device (not shown).
(23) Collier, C. P.; Wong, E. W.; Belohradsky, M.; Raymo, F. J.;
Stoddart, J. F.; Kuekes, P. J.; Williams, R. S.; Heath, J. R. Science 1999,
285, 391-394.
(24) Lines, M. E.; Glass, A. M. Properties and Applications of Ferro-
electrics and Related Materials; Clarendon: Oxford, 1977.

Molecular Switch Tunnel Junction DeVices
J. Am. Chem. Soc., Vol. 123, No. 50, 2001 12639
The experiment roughly corresponds to a measurement of the
device switching speed. We found that the device could operate
from 0.1 to 100 Hz. It is clearly a slow device, exhibiting a
collapsing RMolS trace for the faster measurements. Note that
if capacitance were responsible for this hysteresis loop, then
the RMolS traces would be expected to broaden at faster times.
In any case, a simple calculation, based on the amount of current
that flows through the device, reveals that the nature of this
limitation is likely a combination of molecular properties and
device construction. Our 100 Hz (and slower) pseudorotaxane
device was about 5 µm × 10 µm in size, and based on the
Langmuir isotherm measurements, contained about 5 × 106
molecules. At the voltage required to close the device (around
+1.7 V), the current level through the “switch-open” device is
approximately 10 nA. At 100 Hz device cycling, only about
1% of the electrons that actually pass through the device cause
device switching. Thus, if every electron event resulted in
oxidation of the TTF unit and translation of the cyclophane ring,
the device could only be operated at about 10 kHz. In a
molecular switch tunnel junction device, the current flow
through the device, as well as the residence time of a charge
on a molecule, is largely limited by the widths and energy
heights of the tunnel barriers that separate the redox center of
the molecule from the electrodes, although molecular relaxation
may also contribute. The probability that a charging event will
lead to translation of the cyclophane ring, or, in the case of the
[2]catenane, to circumrotation of the crown ether ring, is limited
by a combination of electron tunneling rates and molecular
properties, such as the activation barrier to ring translation or
circumrotation. For the [2]pseudorotaxane device, that product
probability is about 1%. Separating the molecular switching
properties from the device architecture remains an existing
experimental challenge.
Figure 6. Device cycling for (a) the [2]catenane 14+, (b) the
[2]pseudorotaxane 2‚CBPQT4+, and (c) the [2]rotaxane 34+. Figure 5a
also shows control experiments for the bistable [2]catenane‚DMPA layer
performed by substituting: (i) DMPA as its monosodium salt, (ii) the
tetracationic cyclophane CBPQT4+ anchored with DMPA, and (iii) a
singly stable degenerate [2]catenane (CBPQT4+ interlocking bispara-
phenylene[34]crown-10) anchored with DMPA.
of a bulky dendritic hydrophilic stoppersalong with a hydro-
phobic stoppersonto the dumbbell component, thereby increas-
ing the area/molecule in the device. However, the single-station
character of the backbone also makes this device effectively a
control experiment.
The two (gray and black) RMolS traces (Figure 5b) for a
device made from 2‚CBPQT4+, the two-station [2]pseudoro-
taxane, reveal a much more dramatic response than that obtained
for the [2]catenane 34+ (Figure 5a). The switching voltages are
very sharp and are accompanied by a large magnitude change
in current through the device. Repeated RMolS measurements
of this and other [2]pseudorotaxane devices consistently repro-
duced the large current change upon switching. However, while
this molecular switch tunnel junction always closed at a positive
bias and opened at a negative bias, the actual voltages required
for switching the devices fluctuated from experiment-to-
experiment. We attribute this phenomenon to the above-
discussed domain structure of the supramolecular monolayer.
Relative molecular motion in a tightly packed monolayer, such
as the translation of the cyclophane ring, is likely to be
accompanied by a reorganization of the monolayer. Thus, the
switching characteristics of this device are dominated by both
the relative motions of individual components of the [2]-
pseudorotaxane, and also by the reorganization of the supramo-
lecular domains. Domain switching will be accompanied by the
statistical fluctuations that accompany any nucleation phenom-
enon, and is clearly not a desirable characteristic in a molecular
switch tunnel junction device.
Device Cycling and Volatility. The RMolS traces yielded
information that could then be used to cycle the devices between
the switch-open and switch-closed states, such as the voltages
that can open or close a device as well as nonperturbing voltage
levels that can be utilized to interrogate the device. In device
cycling experiments, voltages of sufficient magnitude to either
close or open the molecular switch are applied and, after each
application of such a voltage, the state of the device is read at
some small, nonperturbing voltage. In Figure 6a-c we present
device cycling results for 14+, 2‚CBPQT4+, and 34+, respec-
tively. Device cycling of the [2]catenane 14+ has been reported
previously.1 Its cycling behavior was robust, exhibiting a
consistent on/off ratio of approximately 2-3 for a few hundred
cycles, and it could be repeatedly sampled over a period of about
2 months. Included with the results for 14+ (Figure 6a) are
control experiments performed by substituting the components
described in the Figure caption for the bistable [2]catenane/
DMPA film in the devices. Although a very small amount of
switching (<1% of the 14+ bistable [2]catenane device) was
observed for the DMPA film, such switching was attributable
to a small amount of device capacitance, as measured by a
nonzero current at zero applied bias. Bistability was not observed
in a RMolS trace for this control.
The RMolS trace (Figure 5c) for a device made from 34+
,
The current/voltage traces for devices made from the [2]-
pseudorotaxane 2‚CBPQT4+ were consistent with the RMolS
traces in which a large amplitude molecular switching signal
was apparently strongly coupled with the domain fluctuations
in the 2D film. For the device cycling experiments, the change
in current signal through this device was higher than 102 for
some cycles, but cycle-to-cycle reproducibility was poor. The
single-station [2] rotaxane 34+ exhibited reasonably stable, but
the single-station [2]rotaxane, reveals only a very weak switch-
ing signature, but one that was highly reproducible from device-
to-device. However, even this minimal switching response was
only observed for a few device cycles, after which the RMolS
trace collapsed. This compound represents our first attempt
synthetically at removing domain switching from the device by
rendering the [2]rotaxane amphiphilic through the incorporation

12640 J. Am. Chem. Soc., Vol. 123, No. 50, 2001
Collier et al.
Figure 7. The temporal stability of the “closed” and “open” states of
molecular switch tunnel junctions fabricated from 14+ (gray) and 2‚
CBPQT4+ (black).
Figure 8. The different molecular architectures [(a) the [2]catenane,
(b) the [2]rotaxane, and (c) the [2]pseudorotaxane] within the context
of a donor/acceptor picture for the molecular switch tunnel junctions.
The gray shaded area represents the “net” acceptor site on the (super)-
molecule, where the tetracationic cyclophane encircles one of the
π-electron rich stations.
low-amplitude, switching characteristics for very few cycles,
after which the performance of this device degraded rapidlys
consistent with the RMolS measurements on this device. The
origin of this switching is not clear. Two possibilities are the
redox-driven translation of the cyclophane off of the TTF site
or field-driven reorganization of the molecular components.
However, the small amount of switching that did occur did not
arise from device capacitance.
prediction has been verified26,27 experimentally. Current recti-
fication does, in fact, appear to be a device characteristic that
is under synthetic control at the molecular level. The molecules
and supermolecule discussed here are all based on donor-
acceptor complexes and units. Hence, it is instructive to analyze
the I/V characteristics of these devices in the context of donor
and acceptor units.
For the purpose of this discussion, we will consider the TTF
unit and 1,5-dioxynaphthalene ring system in 14+, 2‚CBPQT4+
and 34+ as electron donors of varying strengths and the
bipyridinium units of CBPQT4+ as electron acceptors. Within
this context, it is possible that when the tetracationic cyclophane
binds with one of the donor sites, and then the entire recognition
center becomes a “net” acceptor site. Thus, we can view the
[2]catenane, the [2]pseudorotaxane, and the [2]rotaxane as the
rather idealized entities shown in Figure 8.
In Figure 9, we plot the current-voltage (I/V) traces of all
three devices for both the switch-closed and the switch-open
states. To highlight the rectifying character of these devices,
we have expanded these traces near the region of zero-bias. The
first thing to note is thatsas discussed in the previous sections
all traces cross zero-current at zero-bias, indicating that capaci-
tance effects are not being measured here. Second, while the
various data discussed earlierssuch as the RMolS traces,
exhibited some variation from sample to sample, it is the I/V
traces that appear to be the most sensitive to the molecular
variations from device-to-device. The third common feature is
that all devices, in their switch-closed states, exhibit higher
current flow at both negative and positive bias. This observation
is consistent with the proposed model in which the excited-
state co-conformer of these (super)molecules has a higher
density of states and is therefore a generally better conductor.
The I/V traces (Figure 9a) for the [2]catenane reveal a slight
diode effect at positive bias for both the switch-closed and
switch-open states. Although this molecule is made up of
interlocked donor and acceptor portions, it does not meet the
criterion of having the donor and acceptor portions well-
separated in the path between the electrodes. In fact, the actual
geometry of this device is such that the entire [2]catenane is
located at the poly-Si end of the molecular junction, while the
phospholipid (DMPA-) anions sit on top of the [2]catenane and
The two best molecular switch tunnel junctions reported
heresthose made from 14+ and 2‚CBPQT4+sexhibit reasonably
good volatility characteristics. Devices made from 34+ relaxed
from the switch-closed state within a couple of minutes,
consistent with the picture that true bistability does not exist in
this device. In Figure 7, we present volatility measurements for
both the [2]catenane- and [2]pseudorotaxane-based devices. In
these experiments, the devices were initially set to the “switch
closed” state by applying a voltage of +2 V for a couple of
seconds, and then the current through the devices was monitored
for 1000 or 2000 s, respectively. Applying a bias of -2 V for
a few seconds then opened the device, and the current was
monitored again. The switch-closed state 1/e decay times of
the [2]catenane and the [2]pseudorotaxane-based devices were
20 and 60 min, respectively, while the switch-open states for
both devices did not exhibit a time dependence. For the [2]-
pseudorotaxane device, the switch-closed state is substantially
noisier than the switch-open state, and sudden jumps in the
current through the device were recorded as a function of time.
We believe that this erratic behavior is, once again, a signature
of domain switching. In all of the devices that we have
measured, the switch-open state represents the initial state of
the device and likely represents the ground state of the molecule
or supermolecule. The excited-state co-conformation of both
14+ and 2‚CBPQT4+ represents a molecular structure or
superstructure with a narrower HOMO-LUMO gap. Such a
gap translates into a higher density of states in the molecular
tunnel junction of the solid-state device and, in principle, higher
current flow through the device. The finite volatility of the
switch-closed states of these devices is consistent with this
picture.
Diode Behavior and its Molecular Origins. One of the
earliest suggestions for a molecular electronic device was
Aviram and Ratner’s proposal25 for a molecular rectifier, based
on a molecular electron donor/acceptor complex that bridges
across two electrodes. If the donor and the acceptor portions of
the complex are well separated between the electrodes, then
current flow in the acceptor to donor direction encounters less
resistance than the current flow in the opposite direction. This
(26) Metzger, R. M. Acc. Chem. Res. 1999, 32, 950-957.
(27) Metzger, R. M. J. Mater. Chem. 2000, 10, 55-62.
(25) Aviram, A.; Ratner, M. R. Chem. Phys. Lett. 1974, 29, 277-283.

Molecular Switch Tunnel Junction DeVices
J. Am. Chem. Soc., Vol. 123, No. 50, 2001 12641
similar. One explanation is that only a fraction of the [2]-
pseudorotaxane domains are actually switching. Given that
domain fluctuations are certainly key to understanding the
performance of this device, this possibility is not unlikely.
Nevertheless, the current-voltage traces presented here were
highly reproducible from device-to-device. Although the donor-
acceptor picture does provide some insight, it is probably not
the complete story. Other experiments, however, such as
exploring different donor sites and extending the physical
separation between the donor sites in a rotaxane or pseudoro-
taxane (supra)molecular architecture, are suggested by these
results.
Conclusions
Solid-state molecular tunnel junction switching devices were
fabricated by sandwiching LB monolayers of a bistable [2]-
catenane, a bistable [2]pseudorotaxane, and a single-station [2]-
rotaxane between an n-type polySilicon bottom electrode
(adjacent to the hydrophilic end of the LB film) and a titanium/
aluminum top electrode. The electrical properties were found
to be highly dependent on supramolecular structure, the presence
of bistability within the (super)molecule, and the crystallographic
organization of the LB film. The [2]catenane-based devices
exhibited stable operation over many device cycles, but with
only a relatively small conductance change (factor of 2-3)
between the switch-open and switch-closed states. Large
conductance changes (102) between the switch-closed and the
switch-open states could be observed for the [2]pseudorotaxane-
based device, but the overall characteristics of that device were
complicated by the presence of molecular domains within the
LB film. The single-station [2]rotaxane was designed with a
bulky, hydrophilic stopper, so as to generate a large area/
molecule, and therefore to avoid the complexities of domain
switching. While domain switching was not observed for those
devices, the general performance characteristics of those devices
were poor. This result was not unexpected since bistability was
not incorporated into the design of the [2]rotaxane molecule.
The current/voltage characteristics of these devices were
considered within the framework of a donor acceptor model.
This model provides a reasonably consistent framework for
understanding these switching devices. It also suggested a
number of new target molecular switch architectures.
Figure 9. Current-voltage traces of the three molecular switch tunnel
junction devices: (a) the [2]catenane 14+, (b) the [2]rotaxane 34+, and
(c) the [2]pseudorotaxane 2‚CBPQT4+. The gray curves are the switch
“closed” configuration, while the black curves are the switch “open”
configuration, and represent the initial state of the device.
contact the top Ti/Al electrode. Thus, we conclude that the slight
asymmetry of these I/V curves arises largely from the asym-
metric architecture of the molecular junction, rather than from
a donor-acceptor interaction.
The single-station [2]rotaxane 34+ is simply an acceptor
separated by the electrodes at the hydrophilic and hydrophobic
ends of the molecule. This molecular junction should exhibit
the most symmetric I/V responses of all the devices discussed
here, and indeed, in both the switch-closed and in the switch-
open state, the response (Figure 9b) is quite symmetric about
zero-bias. Control experiments using amphiphilic long chain
alkanes were carried out in order to quantify any rectification
behavior that arose from just the asymmetric nature of the metal/
insulator/semiconductor junction. A small amount of asymmetry
in the current/voltage trace was observed, and it nearly exactly
maps onto the asymmetry of the [2]rotaxane 34+ device.
Finally, the most revealing of these I/V traces (Figure 9c) is
that of the two station [2]pseudorotaxane 2‚CBPQT4+. This
device exhibits particularly strong current rectification properties
in the switch-open, positive-bias mode. In this bias, electrons
are flowing from the Ti/Al electrode, through the supermolecule,
and into the poly-Si electrode. According to the model of
Aviram and Ratner,25 this measurement implies that the ground
(switch-open) state of the pseudorotaxane is the co-conformer
that has the CBPQT4+ ring associated with the 1,5-dioxynaph-
thalene site. When the switch is closed, the junction becomes
conducting at either bias. On the basis of the donor-acceptor
argument, one might expect some current rectification at a
negative bias in the switch-closed state, but no such rectification
is observed. In fact, at negative bias voltages below a few tenths
of a volt, the conductance (dI/dV) (but not the current magnitude)
of both the switch-closed and switch-open states is actually quite
This paper represents a significant first step toward elucidating
molecular structure/device property relationships for active
molecular electronics devices.
Experimental Section
Full experimental details and compound characterizations along with
electrode fabrication and monolayer formation protocols are provided
in the Supporting Information.
Acknowledgment. This work was funded by the Defense
Advanced Research Projects Agency, the Semiconductor Re-
search Corporation, and the Office of Naval Research. C.P.C.
is an employee of the Hewlett-Packard Corporation, and the
NMR spectrometer used in this work was supported by the
National Science Foundation. We thank the University of
Odense for a Ph.D. Scholarship to J.O.J..
Supporting Information Available: Full experimental
details and compound characterizations; also, electrode fabrica-
tion and monolayer formation protocols (PDF). This material
is available free of charge via the Internet at http://pubs.acs.org.
JA0114456