Confidential: not for distribution.

Submitted to IOP Publishing for peer review 11 March 2011

Hetero-nanojunctions with atomic size control using a lab-onchip electrochemical approach with integrated microfluidics
P Lunca Popa1,3, G Dalmas1, V Faramarzi1, J F Dayen1, H Majjad1, N T Kemp2 and B Doudin1
1

Institut de Physique et Chimie des Matriaux de Strasbourg, UMR 7504 CNRS-UdS, 23 rue du

Loess, BP 43, 67034 Strasbourg, France

Department of Physics, The University of Hull, Cottingham Road, Kingston-upon-Hull, HU6 7RX,

United Kingdom

E-mail: petpo@ifm.liu.se and bernard.doudin@ipcms.u-strasbg.fr

Abstract. A versatile tool for electrochemical fabrication of hetero-nanojunctions with nanocontacts made of a few atoms and nanogaps of molecular spacing is presented. By integrating microfluidic circuitry in a lab-on-chip approach, we keep control of the electrochemical environment in the vicinity of the nanojunction and add new versatility for exchanging and controlling the junctions medium. Nanocontacts made of various materials by successive local controlled depositions are demonstrated, with electrical properties revealing sizes reaching a few atoms only. Investigations on benchmark molecular electronics material, trapped between electrodes, reveal the possibility to create nanogaps of size matching those of molecules. We illustrate the interest of a microfluidic approach by showing that exposure of a fabricated molecular junction to controlled high solvent flows can be used as reliability criterion for the presence of molecular entities in a gap.

1. Introduction

Understanding and mastering electronic properties of devices reaching molecular and atomic sizes are among the top challenges for studying fundamentals of ultimate miniaturized electronics components [1]. When the size of metallic contacts becomes of the order of magnitude of the Fermi wavelength, (typically few angstroms in metals, corresponding to the interatomic distance), new interesting and promising quantum phenomena appear [2]. For example, new properties are envisioned by taking advantage of the spin degree of freedom of charge carrier [3]. Studying these effects is however Present address: Thin Film Physics Division Department of Physics Chemistry and Biology Linkping University SE-581 83 Sweden
3

hindered by challenge of fabricating of such small contacts, especially when a significant lifetime is needed for experimental characterization. As a result, devices lack reproducibility and stability, and investigations are therefore mostly limited to statistical averaged properties or require cryogenic environments. To date, mechanical breaking, electromigration and electrochemistry are the main techniques used for fabricating planar devices of ultimate small sizes [4], each method having its own advantages and disadvantages. Electrochemistry methods for nanocontacts fabrication are ideally suited when wet chemistry compatible strategies are needed. They also provide the freedom to selectively deposit different materials on either electrode (heteronanojunction) or the ability to synthesize desired molecules directly within the nanocontact region. Another important advantage of electrochemistry is the reliable control of the surface states of the electrodes. Here we refer to oxidation or other contamination processes, which are very important when magnetic properties of transition metals nanocontacts are studied [5]. A number of reports involving nanocontacts obtained via electrochemistry can be found in the literature [6] with some were very spectacular results [7,8], remaining however controversial [9]. The lack of reproducibility form one team to the other requires careful handling of the sample environment control. In this paper we propose an integrated approach for the fabrication of nanogaps and heteronanojunctions using an elegant multi-valve microfluidic system. This approach, implementing microfluidic techniques and a lab-on-chip strategy, allows improvements in device fabrication regarding the control of the chemical environment of the nanojunction and its mechanical stability, and brings convenience in terms of miniaturization. By achieving very smooth and controlled transition between different electrolytes or fluids coming into contact with the junction area, we can optimize the strains affecting the fragile small device, while keeping control of its electrochemical environment. We demonstrate the interest and reliability of this approach by presenting some examples of applications for metallic-type junctions and heterojunctions, showing the versatility and the nearly atomic control of the deposited material achieved by this fabrication method. The related realization of atomic-size gaps between electrodes is also detailed, and its applicability for molecular electronics is illustrated by studies on benchmark alkane molecules.

2. Experimental section

The devices fabrication process can be divided into three main steps. First, optical lithography is used to pattern few micrometers-spaced electrodes. Then, the inter-electrode distance is reduced down to a nanogap with spacing below 50 nm using e-beam patterning. Finally, the on-chip electrochemical cell is integrated with PDMS (Polydimethylsiloxane), encapsulating the electrodes and allowing controlled electrolyte flow and reversible heteronanojunctions deposition.

2.1. Patterning initial electrodes

An initial pair of electrodes separated by a 5 microns gap is patterned over Si/SiO2 wafer by standard optical UV lithography using reversible AZ5214 resist. A double layer of Ti(5nm)/Au(50nm) is then e-beam evaporated, followed by lift-off in acetone, to obtain the metallic electrodes. The interelectrode distance is then decreased down to a few tens of nm only, by patterning two tips face to face separated by a 50 nm gap, using e-beam lithography. We used a Zeiss Scanning Electron Microscope (SEM) and a Raith EBL (Electron Beam Lithography) tool, with a double resist layers process (MMA/PMMA - MethylMethAcrylate/PolyMethylMethAcrylate). Reproducible nanogaps around 30 nm (figure1(c)) are routinely obtained.

Figure 1. (a) Schematic of the silicon chip-PDMS assembly. The microfluidic channels are aligned on top of the initial nanogap. (b) Optical image of the micro-electrochemical cell; the darker area corresponds to electrolyte flowing through the channel. The working electrodes (left and right), the counter electrode (large bottom square) and the reference electrode (middle line) can be observed. The horizontal bar scale corresponds to 5 microns. (c) Scanning electron microscopy of initial gold nanoelectrodes after e-beam lithography. The horizontal bar scale corresponds to 200 nm

2.2. Building on-chip electrochemistry micro-cell

Fabrication of the electrochemical cell starts by creating an SU-8 master template defined by optical lithography, where a reversed image of the microfluidic channels is patterned in SU-8 spin coated (with a height of 40 microns) onto a Silicon wafer. A commercially available PDMS elastomer, Sylgard 184 from Dow Corning-USA, is then poured over the SU-8 master. The elastomer is mixed

with a baking agent in 10 to 1 ra and a vacuum pump desiccator is used to rem atio move air bubbles from the mixture. The viscous PDMS solution is poured over the master and cured at 650 C for one hour, t after which the replica is peeled away from the master. Access to the microfluidic circuitry is made by c using a biopsy punch to make ho in the PDMS. The most critical step in the process is the bonding oles p of the PDMS microfluidic to the silicon oxide wafer. To ensure a strong sticking between the PDMS e g and the sample, oxygen plasma under RF power is used to activate surface che emical bonds in both PDMS and Silicon [10]. Plasma activation also changes the PDMS from hydrop phobic to hydrophilic, helping in wetting the small mi icro-channels. Alignment of the microfluidic in channel with the nlet patterned gold structure on Si/ /SiO2 is then performed under a microscope, to ensure a precise positioning of the channel over the electrodes (figure 1(b)). The electrochemical cell built this way is t 100 micrometers wide and 40 mi icrometers high (figure 1(a)). Finally, plastic tube are introduced into es the access holes, with UV polym merizing glue is used to prevent any leaks if n necessary. The micro electrochemical cell is attached onto a copper cold finger and contact pads are wire bonded from a chip to a PCB (printed circu board). The PCB is connected to the electrical circuit for uit e electrochemistry via coaxial conn nectors as depicted on figure 2.

Figure 2. The lab on chip micro electrochemical cell. Microfluidic tubes (containi a pink solution) ing entering into the PDMS cell and electrical connections (white wires) are visible. The PCB (green) is T mounted onto the end of a cryost tat.

The flow of electrolyte in the cell is driven by a syringe pump system correlated with external microd valves that permit exchange of up to 6 different fluids. Thus a key advantage of ou experimental setup p ur is the ability to easily interchang between electrolytic solutions for fabricating heteronanojunctions, ge solutions for cleaning and rinsing (D.I. water), gases for drying (N2 gas) and solu g utions of molecules to be trapped in the heterojunctions One of the most important advantages of the microfluidic technique s. m

presented in this work is the ability to achieve all these critical steps without any mechanical perturbation, thus providing a stable environment to study electronic transport through nanoscale entities.

2.3 Reversible controlled electrodeposition in nanojunctions.

Our experimental setup (figure 3) is based on a technique developed by Morpurgo et al. [11]. The nanojunctions are obtained by depositing metal (eg. Au, Ni, Co, Pt) that fills the patterned gap between electrodes. One side of the junctions is used as working electrode, at ground potential, separated by a small AC voltage from the other side of the junction. A potentiostat is used to control the deposition so that constant working potential conditions are kept versus a reference electrode. To monitor changes of the impedance of nanojunction a small ac excitation (4mV) is connected in series with the nanojunction and two 1k resistors. The differential voltage (VX), measured across one of the 1k resistors by a lock-In amplifier provides an accurate measure of the impedance amplitude and phase of the nanojunction. Both sides of the junctions are at the same DC potential, and referred below as working electrodes.

Figure 3. Electrical equivalent circuit of the electrochemistry setup. Shaded area represents the electrochemical cell with four electrodes. Z1, Z2 and ZX are the corresponding impedances for baths between counter electrode and working electrode 1, counter electrode and working electrode 2 and the impedance between the two working electrodes respectively. The lock-in reads the voltage drop, VX, across the1k resistance.

At the start of the experiment, i.e. when no metal has been deposited between the two nanoelectrodes, the impedance between the two working electrodes is the same as the electrochemical bath impedance, of the order of hundreds of k as measured by electrical impedance measurements. With application of the adequate reduction potential, specific to the metal used, deposition occurs on the working

electrodes. The distance between the working electrodes decreases until contact between the two electrodes occurs, readily detected by the significant AC voltage drop through the 1k additional resistor. The process is fully reversible (except for Pt) i.e. by applying the oxidation potential one can dissolve the contact. The process can also be stopped at any time, allowing the exchange of electrolytic solutions so that different metals can be deposited. For molecular electronic studies, the electrolyte can be replaced with a solution containing the molecules of interest. The electrochemical cell also facilitates the ability to synthesize organic/inorganic entities directly in the nanogap using standard electrochemical synthesis techniques. Having the ability to manage the potential of the electrodes and through careful redox control, one can avoid oxidative processes that can occur at the surface of some electrode materials (eg. Ni).

3. Results and discussions

3.1. Controlled fabrication of nanogaps and a few atoms contacts into microfluidic circuit.

When the active area of the contact region between two nanoelectrodes is reaching sizes comparable with the Fermi wavelength, the ordinary Drude model of conduction no longer holds and a ballistic regime is achieved (i.e. electrons are not scattered while passing through the contacts) [12]. The concept of transmission channels is introduced [12, 13] and, the resistance exhibits a step-like behavior corresponding to opening or closure of these channels [14] revealing the signature of a few atoms contacts. The plateaux and steps are of the order of conductance values multiples of G0 = 2e2/h the quanta of conductance. A resistance value of the order of h/2e2 13k corresponds to a

perfectly transmitted single channel of conduction, where the related wave function exhibits no reflection at the contact position, and the resistance corresponds to contact resistance related to scattering towards equilibrium in the source and drain macroscopic electrodes. Because the Fermi wavelength for metals is only a few Angstroms (corresponding to interatomic distances) the fabrication of such contacts is very challenging. Atomic size contacts where obtained for several metals, as revealed by occurrence of conductance plateaux, thanks to fine tuning of the electrochemical deposition and dissolution reactions. For each of these experiments, the working electrodes were gold, the counter electrode the same as the deposited material and the reference electrode was from platinum. The electrolytes used are aqueous solutions containing the metallic ions. Exact compositions for all electrochemical baths used are given in appendix. The plating potential was conveniently tuned in order to obtain a slow plating regime. In this way we were able to obtain a small atomic deposition rate, in order to observe and hence to control slow conductance changes. The flow rate for the electrolyte, controlled by the microfluidic system, was chosen to be below 1ml per hour corresponding to a velocity of electrolyte in the vicinity

of the nanocontact around mm/s. Instabilities were observed for flow rates exceeding 10 ml per hour, possibly because of mechanical disturbances of the nanocontact.

Figure 4. Conductance plateaux and jumping steps for nickel (a) and cobalt (b) The conductance is measured in units of quantum conductance G0 = 2e2/h ( 1/13k)

Steps and plateaus of conductance were observed for nickel and cobalt as shown in figure 4. What should be emphasized here is the long lifetime of some plateaus of conductance, as illustrated by the 1 and 2 G0 plateaus for nickel from figure 4. The ability of this process to achieve such long life time for low conductivity plateaux is rarely reported. The results in the figure 4 indicates that a regime of conduction involving a few atoms only can be achieved and maintained for relatively long periods of time in our system. We found critical the use of electrodes initially separated by small-enough distances, i.e. below 100 nm. Thicker deposits needed for connecting electrodes with wider separations did not exhibit stable low-conductance values. We speculate that the width of the approaching apex of the electrodes (related to their initial distance), is an important geometrical parameter for optimizing the stability of the contact. The process of closing a nanocontact can be reversed by adjusting the potential to the metal oxidation potential, promoting a dissolution process at the electrodes. In this case, atoms are removed from the contact and after a period of time the contacts breaks and a nanogap is formed. The voltage can then be tuned at an intermediate value where neither deposition nor dissolution take place. Using the microvalves system the electrolyte can be replaced by pure water or nitrogen for rinsing and respectively drying the nanogap. The electrodes can also be kept at a given potential to avoid the formation of different unwanted chemical species as long as the contact is in solution. Since the whole electrochemical cell is incorporated within a PDMS enclosure the electrodes are also protected from ambient conditions. Note however that PDMS remains porous for oxygen diffusion, making an inert atmosphere enclosure necessary if surface oxidation is critical (the miniaturization of our setup makes

inert atmosphere very accessible experimentally). We consistently obtained nanogaps of sizes limited to a few nanometers as shown in figure 5.

Figure 5. SEM picture of a nanogap with size below 2 nm (the resolution of our SEM). The horizontal bar corresponds to 20 nm. The arrow is indicating where the contact closure (confirmed by electrical measurements) occurs.

The whole process of closing and opening a nanocontact can be repeated by switching between two values of potential corresponding to deposition and dissolution, as illustrated by Figure 6. The upper figure is the time evolution of the phase of the AC current in the shunt measuring circuit. It shows clear and reproducible jumps between values corresponding to an open circuit (capacitive behavior) and a closed one (resistive behavior)

Figure 6. Successive opening and closure of a nanojunction under potentiostatic conditions. Bottom graph: time evolution of conductance. Upper graph: time evolution of the AC current in the shunt measuring circuit. Switching between a resistive behavior (close contact) and a capacitive one (open contact with electrochemical bath between nanoelectrodes) is observed and reproduced with optimum long term reproducibility and stability.

This controlled reversibility is one of the great assets of this method, in contrast with the nanojuctions built by electromigration where reversibility can only be obtained under specific conditions and for specific materials [15].

3.2. Fabricating heteronanojunctions

3.2.1. Fabrication of metallic heteronanojunctions

Our fabrication method based on microfluidic technology can be used to design Au/Ni/Ag/Ni/Au nanoheterojunction, demonstrating the versatility of the method, and its potential for studying novel 1-D atomic structures. The use of a ferromagnetic buried layer can be used for magnetic trapping purposes [16], and Ag has been reported as an ideal electrodeposited material exhibiting conductance values at simple integer multiples of G0 [17], mostly owing to the mechanical stability of atomic configurations related to a few G0 conductance [18]. We started by depositing nickel using a Ni sulphamate solution by applying a plating potential of -1V vs Ag/AgCl. This potential corresponds to a very slow deposition therefore the number of Ni ions deposited in unit time is relatively small. In this way a sudden and very strong closure of a gap is avoided. After the contact is made, a short dissolution voltage of -0.45V is applied until the conductance reaches a value corresponding to a few nm sized gap. Then the freshly formed Ni nanogap is rinsed and dried by replacing the electrolyte with a buffer solution and high purity nitrogen gas respectively. This is a necessary process because the electrolytes solutions can chemically react resulting in precipitates that can affect the quality of the junction or sometime block the micrometric channels.

Figure 7. Controlled switching between two very low conducting states of a silver nanocontact. The conductance switches (upper graph) by switching the applied potential (bottom graph)

A silver contact is then deposited using a nitrate solution. The controlled electrodeposition of silver is quite sensitive because the difference between plating and dissolution potentials is only few tens of millivolts. After the junction closure, a pulsed potential between the two values corresponding to

dissolution and deposition of the silver is applied. Figure 7 shows the time evolution of conductance while switching between those potentials. The nanocontact is controllably switched between two states corresponding to the opening and closing of additional transmission channels responding to controlled modification of the electrodepositing potential. Reproducing the results of Xie et al. [17] allows us to use Ag as benchmark material for showing that the tunability of electrodeposited nanocontacts kept in microfluidic environment. We can straightforwardly extend the structuration of electrodes with

successive layers of materials, making for example possible the construction of multilayers contacts.

3.2.2. Fabrication of molecular nanojunctions

Opening nanocontacts provides the possibility to create nanogaps with widths that match molecular lengths, making molecular nanojunctions possible. Fabrication starts by narrowing the gap below one nanometer using the desired metal followed thereafter by the insertion of molecules of interest. In the example presented here we start by electrodepositing Au nanocontacts reaching a conductance of typically a few G0. We found that long-term relaxation of the as-deposited gap allows opening of the junction for approximately 40% of the samples. If spontaneous opening does not occur, we found that applying a dissolution potential for a few seconds, with subsequent relaxation waiting, can promote the opening. The electrolyte is then evacuated and replaced by the solvent used for the molecule to study (ethanol in our particular case). The leakage current is then checked, typically below 1 nA under 3 V bias. The sample is left overnight in a glove box under N2 atmosphere with constant impedance monitoring of the nanogap. After drying, the sample impedance reaches around 1012 overnight, with a current below 10 pA under 3 V bias, revealing further Au reconstruction at the electrodes apex and possibly further opening of the nanogap.

Figure 8. I (V) measurements of decanedithiol molecules in ethanol, brought into contact with two Au electrodes. Empty circles correspond to the curve taken with only ethanol while the filled circles

correspond to the curve taken after the insertion of molecules. Inset representation of a model for a molecule attached between two electrodes.

The alkanes molecules in solution are then introduced in the microfluidic channel using an ethanol solvent (typical concentration 2 mMol), under impedance monitoring. At 3 V bias, the current reaches tens of nA values after typically 1 hour. The sample is left overnight in the glove box, with slow solvent evaporation. The resulting I (V) curve shows (fig 8) tens of nA current values at 3 V bias, of the same order of magnitude that the data found after one hour immersion (usually 10-50 % smaller after overnight exposure). Removal of excess molecules is then performed by ethanol perfusion. We monitored the sample current at 2 V bias, and observe an immediate drop of the junction current by one order of magnitude, down to a few nA values. Using extensive washing, specifically exposing the junction to significant ethanol flow (reaching 10-1 m/s), we observe opening of the junction, with the current dropping below 100 pA (figure 9). The nA-range intrinsic conductance is re-established after a few seconds following the release of the ethanol flow. One can therefore interpret the observed 1-2 nA at 2 V bias as corresponding to molecules strongly bound to the electrodes. We propose that occurrence of the lower current value relates to breaking on one side of the junction. The roughly 103 104 M related impedance of the closed molecular junction can be favorably compared to what is

obtained in other experiments, for example by AFM probing of a self-assembled alkane monolayer [19]. Comparison with a large body of experimental data also indicates that a typical change of two orders of magnitude of the conductance is expected when comparing mono-thiols and di-thiols species. This also compares favorably with our observed loss of a factor of typically 20 when using our rather aggressive cleaning process. We also performed check experiments on decanethiols, which exhibited conductivities typically more than one order of magnitude lower, without clear signature of reversible changes under flow of solvent. This allows us to draw the rather nave picture inset of fig. 8, where we can suppose that dithiols alkanes can effectively bridge the two electrodes. More systematic and statistical studies would ne needed to fully confirm this model, keeping in mind that the presented data mostly aims at validating the experimental methodology. We consider that this type of experiment clearly reveals the potential interest of the integrated microfluidic setup for molecular electronics purposes. A platform where the sample can be immersed, dried, exposed to large solvent flow, or kept for long periods of time in a solvent or an adequate chemical solution, opens many possibilities for monitoring the impedance of molecular junctions under variable and tunable environments. Our proof-of-principle experiments provide an interesting basis for performing systematic studies, and identifying stability and robustness criteria for molecular electronics studies.

Figure 9. Evolution of the junction during cleaning with ethanol under high solvent flow and showing initial detachment of the molecular contact, followed by re-connection after the flow is stopped.

3. Conclusions

We have presented a new integrated and small-sized tool for fabricating nanoheterojunctions via electrochemistry, where microfluidic integration opens new possibilities. The key advantages are the flexibility to modify the (electro) chemical environment, under constant monitoring of the electrical properties and the very good mechanical stability while fabricating very fragile entities formed by a few atoms only. Examples of junctions containing just a few atoms with relatively long lifetimes have been routinely obtained. Such high stability is for example particularly useful for magneto-electronic transport studies. The strategy for fabricating nanogaps in the nanometer range has been detailed, and possible molecular electronics studies shown on benchmark alkane molecules confirm that the realized gaps are of size compatible with small molecules lengths. Molecular electronics applications can benefit from the integrated PDMS microfluidics which offers protection of the sample from ambient condition and from testing under severe solvent cleaning conditions. Further developments where chemical reactions can be performed inside the channels, in order to construct more elaborate chemical synthesis can be envisioned. For example, step-by-step construction of molecules, or click chemistry to chemically close a molecular device, can be readily implemented [20]. The presented methodology is also compatible with mechanical break junctions (MBJ), where the inter-electrode distance is tuned by substrate bending, making therefore possible investigations on heterogeneous materials by MBJ tuning. Combining electrochemical junctions with electromigration opening of nanocontacts opens the possibility of new reversible electromigration studies.

Acknowledgements This research was partly supported by French Ministry ANR grants (ANR-08-NANO-009-04 SUD and ANR-07-BLAN-0274-01 MOSE). Technical help of D. Spor for electronics, B. Leconte and A. Boulard for the buildup of experimental setups, as well as assistance of M. Acosta and F. Chevrier are gratefully acknowledged.

Appendix. Compositions of electrochemical baths

a) Nickel bath: Nickel sulphamate tetra hydrate [Ni(SO3NH2)2 4H2O]-1.8 M, Nickel chloride hexa hydrate [NiCl2 6H2O]- 50mM and boric acid [H3BO3]-0.65 M; b) Cobalt bath: Cobalt(II) sulfate hepta hydrate [CoSO4 7H2O]-0.45M and boric acid-0.65 M; c) Silver bath: Silver nitrate [AgNO3]1mM and nitric acid [HNO3]-0.1M; d) Gold Commercial bath: ECF61 from Metalor containing Au 0.05g/l and Potassium sulfite [K2SO3]-0.22M

References

[1] Tao N J 2006 Nature Nano. 1, 173 [2] Ohnishi H, Kondo Y and Takayanagi K 1998 Nature 395 780 [3] Wolf S A, Chtchelkanova A Y and Treger D M 2006 IBM J. Res. & Dev. 50 101-110 [4]van Ruitenbeek J M 2001 Naturwissenschaften 88 59-66. [5] Wei H X, Wang T X, Wang H, Han X F, Bari M A and Coey J M D 2007 International Journal of Nanotechnology 4 21-31 [6] Doudin B and Viret M 2008 J Phys.: Condens. Matter 20 083201 [7] Garcia N, Munoz M and Zhao Y W 1999 Phys. Rev. Lett. 82 2923 [8] Hua Z S and Chopra H D 2003 Phys. Rev. B 67 060401 [9] Egelhoff Jr W F et al 2004 J. Appl. Phys. 95 7554 [10] Tang K C, Liao E, Ong W L, Wong J D S, Agarwa A, Nagarajan R and Yobas L 2006 Journal of Physics: Conference Series 34 155161 [11] Morpurgo A F, Marcus C M and Robinson D B 1999 Appl. Phys.Lett. 74 2084 [12] Landauer R 1970 Phil.Mag. 21 863 [13] Agrait N, Yeyati A L and van Ruitenbeek J M 2003 Physics Reports 377 279

[14] van Wees B J, Kouwenhoven L P, Willems E M M, Harmans C J P M, Mooij J E, Houten H van, Beenakker C W J, Williamson J G and Foxon C T 1991 Phys. Rev. B 43 12431 [15] Ittah N, Yutsis I, and Selzer Y 2008 Nano Lett. 8 3922 [16] Long D P et al. 2005, Appl. Phys. Lett. 86, 153105 [17] Xie F Q, Nittler L, Obermair C and Schimmel T 2004 Phys. Rev. Let. 93 128303 [18] Xie F Q et al. 2008 Nano Lett. 8 4493 [19] Akkerman H B and de Boer B 2008 J. Phys.: Condens. Matter 20 013001 [20] Chen X, Braunschweig A B, Wiester M J, Yeganeh S, Ratner M A and Mirkin C A, 2009 Angew. Chem. Int. Ed. 48 5178