Influence of Low Energy Barrier Contact Resistance in ChargeTransport Measurements of Gold Nanoparticle+Dithiol-Based Self-Assembled FilmsPatrick Joanis, Monique Tie, and Al-Amin Dhirani*

Department of Chemistry, University of Toronto, Toronto, Ontario, M5S 3H6, Canada

*S Supporting Information

ABSTRACT: Gold−thiol self-assembly is a widely employed strategy forengineering electronic devices using molecules and other nanostructures asbuilding blocks. However, device behavior is expected to be governed by bothbuilding block architecture and contact effects. In order to elucidate the role ofthe latter in such devices, we have studied conductance of n-butanedithiol-linkedAu nanoparticle (NP) films using different types of electrode configurations,namely, four-probe versus two-probe and break junctions before versus afterdielectric break down of contact resistance. We find that contact resistance isgoverned by transport across a small barrier which can dominate devicebehavior when temperatures and resistances of the self-assembled devices arelow. Accounting for such contact resistance reveals a more precise picture ofdevice behavior in these regimes, including in the present system film propertiesnear the onset of the percolation insulator-to-metal transition and beyond.

■ INTRODUCTIONSince electronic properties of matter at nanometer length scalesare both remarkable and controllable via synthesis, there isvigorous ongoing effort to use nanostructures as building blocksto engineer and study interesting and potentially usefulelectronic devices.1 There are, however, a number of challengesthat arise in this effort. One is that such nanostructures mustinevitably be “wired” to each other or to external electrodes.Given the small length scales involved, self-assembly, that is,exploiting a natural tendency for functional groups to interactwith each other or with surfaces, is a commonly used strategy toovercome this challenge, with thiol−gold being the most widelystudied interaction.Another key challenge is that the electronic properties of self-

assembled, nanostructured devices are influenced not just bythe nanostructures themselves but by the contacts as well;indeed, in some instances, contact resistance dominates overalldevice behavior. Conductance can vary depending on moleculebinding site and orientation2 as well as electrode geometry.3

Aromatic molecules can exhibit a transition from tunnelling tofield emission that depends on the molecular terminalfunctional group (thiol vs isocyanide) as well as the type ofmetal contact.4 Molecules can also exhibit different resistancedepending on molecular terminal group + electrode contact5,6

and cis−trans conformations.6 The terminal functional groupand the type of metal may influence the nature of a surfacedipole, the alignment of the molecular energy levels relative toFermi level of the metal, and thereby even the dominant chargecarrier (electron vs hole) flowing through the molecule.Aromatic molecules with dicarbamate terminal functionalgroups self-assembled on a gold surface exhibit 80 times higher

currents than those with thiol terminal function groups.7

Contacts can have important implications for commercialdevices. Comparing indium tin oxide, gold, and surface oxidizedgold contacts for organic light emitting diodes, surface oxidizedgold contacts yield the best performance due in part to the lowsheet resistance of the underlying gold and improved injectionbarrier provided by the surface oxide.8

Studies of nanostructured electronic devices thus far havefocused on fundamentally two-probe geometries, in which abias voltage (V) is applied across a device containingnanostructure(s), and the device’s responses (current, I, andits derivatives with voltage, dI/dV, d2I/dV2, etc.) includecontributions from contact resistances. In some cases, a three-terminal or transistor-type geometry has been employed.1

These are fundamentally two-probe devices as well, in which athird “gate” electrode can subject nanostructures beinginterrogated to an electric field. In the present study, weemploy a so-called four-probe method to interrogate electricalproperties of a self-assembled nanostructured system, probingboth the nanostructured system+contact resistance as well asjust the nanostructured system itself and thereby revealing therole of the contact resistance. Figure 1a is a diagram of our fourprobe setup, and Figure 1b is a simplified resistor networkequivalent. RF12, RF23, and RF34are the film resistances betweenprobes 1 and 2, 2 and 3, and 3 and 4, respectively. RC1, RC2, RC3,and RC4 are the contact resistances between film and probes 1,2, 3, and 4, respectively. We can use Figure 1b as an example to

Received: November 5, 2012Revised: December 24, 2012Published: January 7, 2013

Article

pubs.acs.org/Langmuir

© 2013 American Chemical Society 1264 dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−1272

illustrate the four probe method. A known current, I, flowsfrom probe 1 to 4, through the outer contacts and the film. Thetotal voltage, V14, required to drive this current depends onboth the outer contact resistances and the film resistance. ByOhm’s law, V14 = IRt, where Rt is the total resistance of the filmplus two contact barriers, Rt = RC1 + RF12 + RF23 + RF34 + RC4.This would be a standard two probe measurement whichincludes contact contributions. Note that a large ratio ofcontact resistance to total resistance leads to a large ratio ofvoltage drop across the contacts to the total voltage applied(“voltage divider” effect). In the four-probe method, the voltagedifference between probes 2 and 3, V23, is measured using highimpedance buffers, which do not allow current to flow acrossRC2 or RC3 but rather force all of the current through the film.As a result, (1) by Ohm’s law, the voltage drops across thesecontact resistances is zero, and V23 is also the voltage dropacross RF23; and (2) the current flowing through RF23 is I.Notwithstanding the presence of contact resistances, we obtainthe film resistance in terms of measured quantities: RF23 = V23/I. This illustration assumed Ohm’s law,that is, that I varieslinearly with V and dI/dV is a constant. In general, dI/dV tendsto vary with V. Determining dI/dV in this common situationusing the four probe method is discussed in the SupportingInformation.Self-assembled nanoparticle systems can exhibit a variety of

remarkable electronic9 and optical properties10 and arecurrently of great interest.11−13 Using a “bottom up” approach,self-assembly affords control over nanoparticle size, linkerlength, and assembly architecture. This tunability enables apotential for complex architectures and behaviors, making suchsystems a favorable proving ground for a variety of applications

ranging from single-electron devices to biosensing technolo-gies.12 For the nanostructured system in the present study, weuse self-assembled films containing gold nanoparticles (NP) +n-butanedithiol linker molecules, and use thermally depositedgold films as contacts. Since this system is based on thiol-goldself-assembly, it serves as a test-bed that is of interest. Also, thissystem can exhibit a wide range of behaviors, from nonmetallicbehavior with high film resistance to metallic behavior with lowfilm resistance.9,14−18 As film thickness increases, the sizes of“clusters” of metallic, linked nanoparticles grow, until at apercolation transition there are sample spanning metallicclusters. Concurrently, film resistance drops and contactresistance becomes more significant. Previous studies of chargetransport in these systems have interrogated film propertiesusing the two-probe method. As films become thicker and filmresistance goes down (i.e., near metal−insulator transition andin the metallic regime), contact resistances become moresignificant and in the metal regime in fact can dominate. In thiswork, we study film properties as a function of temperatureusing the four probe method, enabling measurements of bothfilm + contact resistance as well as just film resistance itself. Wealso use the two-probe method to study break junctions withnanoparticle films bridging the junctions. Both electrodeconfigurations were used to study films with increasingthickness across the metal−insulator transition. Our four-probe data shows the existence of a contact resistance with asmall energy barrier, exhibiting tunnelling behavior at lowtemperatures and voltages and thermally assisted transport athigher temperature. Our break junction samples in the thickfilm regime initially exhibited the presence of the barrier, butapplication of a large voltage caused a sudden, irreversible drop

Figure 1. (a) Four-probe electrode configuration with gold electrodes and a self-assembled butanedithiol-linked gold nanoparticle film. (b) A typicalfour configuration represented by a simplified resistor network. (c) Four-probe configuration enabling a control measurement with aluminumelectrodes, aluminum oxide contact barrier, outer gold electrodes, and a thermally deposited thin gold film. Two separate measurements are possible:a two probe measurement between outer gold electrodes (minimal contact resistance) and a four probe measurement using two inner aluminumelectrodes and outer gold and aluminum electrodes. (d) Photos of (i) our sample and (ii) our control sample represented in (a) and (c),respectively. Any two inner aluminum electrodes can be used for the four-probe method; extra electrodes were deposited to increase sample yield.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721265

in resistance and resulted in clear metallic behavior. Wespeculate that this might be due to the breakdown of thecontact barrier. This work shows that the two-probe methodcan exhibit voltage and temperature dependence due to a smallbarrier tunnelling which has a potential to obscure filmproperties and that, in order to study film properties moreaccurately, especially in metallic regime and near metal−insulator transition, it is important to account for this contactbarrier using the four-probe method.

■ EXPERIMENTAL SECTIONButanedithiol used to link Au NPs was purchased from Sigma-Aldrichand used as received. Au NPs were synthesized in toluene by reducinghydrogen tetrachloroaurate with sodium borohydride.15 Tetraocty-lammonium bromide was used to stabilize the NPs. Transmissionelectron microscopy yielded an average NP diameter of 5.0 ± 0.8 nm.NP films on the glass slides with predeposited Au electrodes (preparedas described below) were self-assembled by first immersing the slidesin 0.5 mM ethanolic solutions of butanedithiol for 1 h, and thenalternately immersing the slides in Au NP solutions for 10−60 minand butanedithiol solutions for 10 min. The amount of NP depositedper immersion cycle depended on NP solution concentration andimmersion time. As a result, room temperature film resistance wasmonitored after each NP/dithiol immersion cycle, and self-assemblywas continued until a desired resistance was reached.To fashion electrodes for four-probe devices, glass substrates, 10

mm × 10 mm × 1 mm in size, were cut from microscope slides andcleaned by immersion in a hot piranha solution (3:1 H2SO4/H2O2) for30 min. The cleaned slides were then functionalized by immersion intoa boiling 40 mM toluene solution of 3-mercaptopropyltrimethox-ysilane for 20 min. Next, four ∼6 mm long, ∼200 nm thick Auelectrodes separated by 2 mm were thermally deposited on to the glassslides using shadow masks. Copper magnet wires were attached to theAu electrodes using indium solder before NP film self-assembly. Figure1d (i) shows our sample after NP film self-assembly. Control samplesshown in Figure 1d (ii) were made similarly. Glass substrates, 20 mm× 10 mm × 1 mm in size, were cleaned and functionalized as above.Then five 8 mm long, 250 μm wide and 150 nm thick aluminumelectrodes were thermally deposited 3 mm apart on the glass slidesusing shadow masks and left in air to grow a surface oxide layer. A thingold film, 200 μm wide and 21 nm thick, was then deposited across thealuminum electrodes, and subsequently, two gold pads 150 nm thickwere deposited on the outer edges of the gold film. Copper magnetwires were attached using indium.To interrogate NP film properties over smaller length scales, we also

studied films employing a two-probe, break junction electrodeconfiguration. To fashion these devices, glass substrates, 4 mm × 8mm × 1 mm in size, were cut from microscope slides and cleaned by

immersion in a hot piranha solution (3:1 H2SO4/H2O2) for 30 min.The slides were then functionalized with mercaptosilane by immersioninto a boiling solution of 3-mercaptopropyltrimethoxysilane for 20min. The substrates were next placed in a vacuum chamber where a100 μm wide and 8 mm long Au wire was deposited by metalevaporation using a shadow mask. A 100 μm stretch in the middle ofthe wire was 16 nm thick, and the rest of the wire was 150 nm thick.Electrodes were connected to the electronics using copper magnetwires and indium solder before NP self-assembly. The sample wasimmersed in liquid nitrogen and a gradually increasing voltage wasapplied until the wire broke due to electromigration.

Figure 2a and b, respectively, shows scanning electron microscopeimages of break junctions before and after self-assembly of a NP film.Break junctions exhibited gaps with widths that ranged from a fewmicrometers to a few nanometers (the latter is estimated frommeasurements of tunnel currents). A thick NP film is clearly visible inFigure 2b. It obscures the break junction gap, and its edges can be seenas excess material to the sides of the electrodes has been wiped away.

Two- and four-probe methods were to used measure film+contactand film properties, respectively. To improve signal-to-noise, (differ-ential) conductance (g = dI/dV) was measured directly by using lockintechniques, rather than by measuring I versus V across the film andnumerically differentiating (see the Supporting Information). Asinusoidal voltage modulation (V, 10 mV, 10−200 Hz depending onsample resistance) from a lockin was summed with a DC bias, V, usinga summing amplifier. The sum was applied the first electrode of thesample. The resulting current was converted to a voltage using anotheramplifier. This voltage was input to a data acquisition card (DAC,National Instruments PCIMIO 16X) to measure its DC componentand to the lockin to measure its sinusoidal component. The lockin’soutput was recorded using a second channel of the DAC. In the four-probe method, the DC voltage drop across middle electrodes 2 and 3was measured using a third channel of the DAC via an operationalamplifier to ensure no current was drawn. The sinusoidal componentof the voltage drop across the middle electrodes was measured using asecond lockin, whose output was measured using a fourth channel ofthe DAC. Low pass filters were sometimes used to reduce noise. In thefour-probe method, these various inputs were used to determine theconductance of the film as a function of voltage drop across the film(not the total voltage drop) as described in the SupportingInformation. Conductances versus voltages for our samples weremeasured at different temperatures (T) down to ∼2 K using a physicalproperty measurement system (PPMS). Measurements for controlsamples were performed down to ∼77 K by slowly lowering/raisingthe samples in a L-N2 dewar. Two four-probe and five two-probesamples were studied in detail up to ∼30 and ∼60 dithiol/NPexposures, respectively. No hysteresis was observed when voltagesweeps were repeated.

To further test the electronics and software, an additional controlmeasurement was performed using commercial resistors arranged as

Figure 2. Scanning electron microscope images of break junctions (a) before and (b) after self-assembling butanedithiol-linked Au nanoparticles.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721266

Figure 3. Normalized conductance (g) versus temperature (T) at zero bias from 77 to 300 K and normalized conductance versus voltage (V) at 77 Kfor a thermally deposited gold film obtained using various electronic configurations. (a, b) Two-probe measurements with minimal contact resistanceexhibiting film properties. (c, d) Two-probe measurements with aluminum oxide tunnel junction in series exhibiting combined film and contactresistance properties. (e, f) Four-probe measurements with aluminum oxide tunnel junction in series exhibiting film properties. Normalization factorsare (a) 1.2 × 10−2 S at 290 K, (b) 1.9 × 10−2 S at 2 V, (c) 3.8 × 10−5 S at 290 K, (d) 3.7 × 10−5 S at 3 V, (e) 1.5 × 10−2 S at 290 K, and (f) 1.7 ×10−2 S at 0 V.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721267

per Figure 1b. To simulate metallic samples with high contact barrier,“contact resistances” (RC’s) of 10 kΩ ± 10% and a “film resistance”(RF) of 1 kΩ ± 10% were used. These values were confirmed using amultimeter. Using the four-probe method, RF23 was found to be 0.98kΩ.

■ RESULTS AND DISCUSSION

Figure 3 shows g versus T and g versus V data for our controlsample. Control measurements to determine the thermallydeposited Au film’s properties were performed using a two-probe configuration using outer gold electrodes which haveminimal contact resistance (see Figure 3a and b). Two-probemeasurements including film and tunnel oxide contact effectswere performed using the outer electrodes (i.e., gold andaluminum outer electrode, see Figure 3c and d). Four-probe

measurements to determine films resistance despite contactresistances were performed using an outer gold electrode todrive current, an outer aluminum electrode with an oxidebarrier to measure current, and two central aluminumelectrodes with oxide barriers to measure voltage drop acrossthe film (see illustration in Figure 1c and data in Figure 3e,f).Figure 3a and b, obtained with minimal contact resistance,shows that the gold film exhibits typical metallic behavior.Namely, conductance decreases as temperature increases, andconductance is independent of bias. Figure 3c and d, obtainedwith an aluminum oxide tunnel junction in series with the goldfilm, indicates the presence of an energy barrier sinceconductance increases with temperature and with bias. Figure3c inset shows an Arrhenius plot that yields an activationtemperature of 1.64 K. Figure 3e and f shows that g versus T

Figure 4. Normalized conductance g versus temperature T data obtained using a nonmetallic, 12 exposure cycle Au NP/butanedithiol film. (a) Four-probe data showing film conductance (main panel) and ln g versus 1/T with a linear fit (inset). The conductance at 250 K is 1 × 10−5 S, and theslope of the inset fit is 24.0 K. (b) Two-probe data showing conductance of the film and contacts combined in series (main panel), and ln g versus 1/T with a linear fit (inset). The conductance at 250 K is 3 × 10−6 S, and the slope of the inset fit is 39.5 K. Temperature values: 2 to 10 K in 1 K steps,10 to 30 K in 2 K steps 30 to 50 K in 5 K steps, 50 to 100 K in 10 K steps, 100 to 200 K in 20 K steps, 200 to 250 K in 5 K steps.

Figure 5. Normalized conductance g versus temperature data T obtained using a metallic, 18 exposure cycle Au NP/butaneditiol film and a four-probe electrode configuration. (a) Four-probe data showing metallic film conductance (main panel) and ln g versus 1/T with a linear fit (inset). Theconductance at 250 K is 2 × 10−4 S, and the slope of the inset fit is 11.4 K. (b) Two-probe data showing conductance of the metallic film andcontacts combined in series (main panel) and ln g versus 1/T with a linear fit (inset). The conductance at 250 K is 4 × 10−5 S, and the slope of thelinear fit is 14.8 K. Temperatures values are the same as in Figure 4.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721268

and g versus V data of the film itself obtained via the four-probemethod behave similarly as data in Figure 3a and b obtainedwith minimal contact resistance. Film resistance at 290 K wasfound to be ∼81 Ω between the gold pads with minimalcontact resistance using the two-probe configuration and ∼66Ω between aluminum electrodes with aluminum oxide tunneljunction contact resistance using the four-probe configuration.Since the length of the film between aluminum electrodes isabout 0.8 that between the gold pads, the film resistancebetween the aluminum is expected to be ∼81 Ω × 0.8 or ∼65Ω, in reasonable agreement with the value determined using thefour-probe method. Using the four-probe method, we are ableto obtain film behavior even with a large (∼30 kΩ) contactresistance in series. Note that the resistivity of gold is 2.214 ×10−8 Ω·m at 293 K.19 Taking our film dimensions as ∼16 mmlong, ∼ 0.2 mm wide, and ∼21 nm thick, the expectedresistance measured in the two-probe configuration is expectedto be ∼84 Ω, in reasonable agreement with our results.Figure 4a and b shows conductance versus temperature data

obtained for a 12-NP/dithiol exposure cycle film without andwith contact resistance, respectively, using the same four-probedevice. That is, data in Figure 4a show 4-probe data using theouter and inner pairs of electrodes of a four-probe device; datain Figure 4b show two-probe data obtained using just the outerpair electrodes. Film conductance (Figure 4a) tends to zero astemperature drops, implying that the film is insulating: thedensity of states vanishes at the Fermi level and electrons haveto overcome an energy barrier in order to conduct. Thepresence of the energy barrier in the film is further evidencedby conductance increasing as temperature increases. The figureinset shows an energy barrier of 24 K. The ln g versus 1/T plotslevel off at low temperatures due to the conductance beingapproximately zero. A number of studies have found that NPfilms can exhibit energy barriers arising from single electroncharging of isolated NPs or clusters of linked NPs (“Coulombblockade”).16 The data in Figure 4b, which includes contacteffects, is similar to that in Figure 4a. The Arrhenius plot shownin the inset yields a slope of 40 K for film + contact. Since film

and contact resistances add in series, when film resistance ishigh, film resistance can dominate over contact resistance. As aresult, conductances of the film with and without contactsbehave similarly.Figure 5a shows film conductance versus temperature for the

same sample after 18 NP/dithiol exposures. This data wasobtained using the four-probe electrode configuration toremove contact effects. Note that, this time, film conductancetends to a finite, nonzero value at low temperatures, implyingthat the film satisfies the definition of a metal: the density ofstates at the Fermi level does not vanish and the electrons arenot required to overcome an energy barrier in order to conduct.Nevertheless, as temperature increases, conductance stillincreases. The figure inset shows an energy barrier correspond-ing to 11 K. Typically, for metals, conductance decreases withincreasing temperature due to increasing phonon density andincreasing electron−phonon scattering. Here, the film has justcrossed the percolation threshold (there is at least one samplespanning metallic pathway for current to flow), and the film isonly “just” metallic.16 That is, the film is a mixture of mostlynonmetallic pathways and some metallic pathways. At highertemperatures, conductance is dominated by the larger numberof nonmetallic pathways. As temperature drops, these pathwaysshut down rapidly, and conductance drops. At the lowesttemperatures, just the metallic pathways conduct, conductancetends to a nonzero value, and the film satisfies the definition ofa metal.Figure 5b shows two-probe conductance versus temperature

data of the same film measured using just the outer twoelectrodes of the four-probe device. In contrast with the four-probe data shown in Figure 5a that exclude contact effects andreveal metallic film behavior, the data in Figure 5b contain bothfilm and contact effects in series, and the conductance exhibitsnonmetallic behavior: it tends to zero as temperature drops.This indicates that the contact resistance comprises an energybarrier, and Figure 5b inset yields an energy barriercorresponding to 15 K. It should be noted that it is difficultto extract a reliable value for the energy barrier of the contact

Figure 6. Normalized conductance g versus temperature T data obtained using a metallic, 25 exposure cycle Au NP/butaneditiol film and a four-probe electrode configuration. (a) Four-probe data showing metallic film conductance (main panel), and ln g versus 1/T with linear fits at high andlow temperatures (inset). The conductance at 230 K is 3 × 10−3 S, and the slopes of the linear fits are 0.29 K above 10 and 0.032 K from 2 to 5 K.(b) Two-probe data showing conductance of the metallic film and contacts combined in series (main panel), and ln g versus 1/T with linear fits athigh and low temperatures (inset). The conductance at 230 K is 6 × 10−4 S, and the slopes of the linear fits are 2.61 K above 10 and 0.35 K from 2 to5 K. Temperature values are the same as in Figures 4 and 5 up to 230 K.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721269

from this data since film and contact conductances arecombined in series and both include significant thermallyassisted components. However, as the film becomes moremetallic and contact barrier effects become more prominent, wecan use the results from film + contact to get a clearer picture ofcontact behavior, as discussed further below.Figure 6 shows conductance versus temperature obtained

using the same sample with 25 dithiol-Au NP exposure cycles.Figure 6a and b, respectively, shows data from four-probe andouter two-probe measurements. In both cases, conductancetends to a finite, nonzero value as temperature decreases. In thecase of the four-probe measurement (Figure 6a), as temper-ature decreases, film conductance increases for most of the

temperature range by only ∼2%. This behavior can beattributed to temperature dependent electron−phonon scatter-ing as seen in typical metals. Others20 have also attributed thischange to a combination of tunnelling and an increasingtunnelling distance with increasing temperature due to thermalexpansion. For a portion of the temperature range at the lowerend, conductance drops by 4% as temperature drops, as thisfilm likely still contains some nonmetallic pathways thatcontribute to the conductance and that shut down at lowertemperatures. In the case of the two-probe measurements(Figure 6b), as temperature decreases, film+contact con-ductance increases slightly for a portion of the temperaturerange and then drops over a majority of the range by ∼30%.

Figure 7. Conductance behavior of a NP/butandithiol film obtained using two-probe break junction electrodes. (a) Conductance behavior for anonmetallic 10 dithiol/Au NP cycle film before break down of contact resistance. Conductance versus voltage at various temperatures (main panel),conductance vs temperature at zero bias (right inset) and Arrhenius plot (left inset). The conductance at 300 K is 1 × 10−5 S. Temperature values: 1K steps from 2 to 10 K, 2 K steps from 10 to 20 K, 16 steps spaced linearly in 1/T from 20 to 300 K. A linear fit to the Arrhenius plot above 10 K hasa slope of 98 K. Below 10 K, the conductance is almost zero, and a linear fit yields a flat line. (b) Conductance behavior for a metallic 50 dithiol/AuNP film before break down of contact resistance. Conductance versus voltage at various temperatures (main panel), at 2 K (right inset) and from 80to 180 K (left inset). The conductance at 300 K is 7 × 10−4 S. Temperature values: 1 K steps from 2 to 10 K, 2 K steps from 10 to 20 K, 14 stepsspaced linearly in 1/T from 20 to 110 K, 10 K steps from 110 to 300 K. (c) Conductance versus temperature (inset) and Arrhenius plot ofconductance at zero bias (main panel) for the same sample as (b). A linear fit to the Arrhenius plot above 20 K has a slope of 12 K and from 2 to 10K a slope of 0.03 K. (d) Conductance behavior for the same metallic 50 dithiol/Au NP film as in (b) and (c) after breakdown of contact resistance.Conductance versus voltage at various temperatures (main panel) and versus temperature at zero bias (inset). The conductance at 300 K is 10−2 S.Temperature values: 1 K steps from 2 to 10 K, 2 K steps from 10 to 30 K, 5 K steps from 30 to 50 K, 10 K steps from 50 to 100 K, 20 K steps from100 to 300 K.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721270

The significantly larger drop in the two-probe conductance datacompared to the four-probe data evidently reflects the influenceof the thermally assisted contact barrier. We note that the two-and four-probe conductance data are not completely Arrheniusanywhere. Estimates for high and low temperature barriers are0.35 K (film) and 2.6 K (film and contact) above 10 K; 0.032 K(film) and 0.35 K (film and contact) from 2 to 5 K. This resultis consistent with studies by others7 which found, using two-probe measurements, that ethanedithiol-linked gold nano-particle film can exhibit metallic behavior and barriers of ∼2K above 80 K.The above data obtained using macroscopic electrodes

comparing two- and four-probe measurements show that two-probe measurements can be strongly influenced by barriereffects at lower film resistances. We show a similar conclusioncan be reached using break junctions and just two-probemeasurements (Figure 7a−d). In these samples, electrodes areclosely spaced and thus contact resistance is emphasized,especially for very thick films for which film resistances arerelatively low and a significant fraction of the total appliedvoltage is dropped across the contact resistances. Figure 7a wasobtained using a film prepared using 10 dithiol/Au NPimmersion cycles. The right inset shows that at temperaturesdrops, conductance tends to zero rapidly, and this film istherefore likely nonmetallic. An Arrhenius plot (left inset)yields an energy barrier corresponding to 98 K at hightemperatures. Conductance versus voltage data in the mainpanel show that conductance increases with increasing voltage,consistent with the presence of energy barrier(s) to currentflow. Figure 7b and c shows film behavior of the same sampleafter 50 dithiol/Au NP immersion cycles. Conductance versustemperature tends to a finite, nonzero value as temperaturedrops, implying that the film has become metallic, butconductance decreases with decreasing temperature, implyingthe presence of an energy barrier. Although current may flowvia thermally assisted pathways in the film, it is likely that asmall barrier across which tunnelling can occur at lowtemperatures and voltages is present, because as temperaturedrops, conductance plateaus at the lowest temperatures andbecomes approximately constant, a hallmark of tunnelling. Also,conductance versus voltage exhibits supralinear behavior at lowtemperatures (Figure 7b, right inset). However, as temperatureincreases, eventually by 160 K, the barrier is overcomethermally, and conductance transitions to metallic filmbehavior: conductance versus voltage becomes approximatelyohmic up to ∼1 V (Figure 7b, left inset). An Arrhenius plotyields barriers corresponding to 12 K above a temperature of 20and 0.03 K below a temperature of 10 K.For samples with sufficiently thick films, as voltage is

increased, we eventually observe that the conductances of breakjunction devices can suddenly and irreversibly increase 10−100fold. We attribute this to dielectric break down of the contactthat is induced in such samples since thick films possessrelatively low film resistance and most of the voltage is thendropped across the contacts. Breakdown is much morecommon in narrow break junctions rather than macroscopicdevices due to the higher current densities that arise. Metallicfilm behavior is easily identified after this breakdown (seeFigure 7d). As temperature drops, conductance increases overmost of the temperature range, plateaus, and then tends to afinite nonzero value at low temperature (Figure 7d, inset). Also,conductance versus voltage is ohmic for low voltages (Figure

7d, main panel). This behavior was observed with all five breakjunction samples studied.The above-discussed four-probe measurements show that a

contact resistance arises at the interface between the NP filmand the butanedithiol-functionalized gold electrodes. Four-probe conductance of the 25 NP/dithiol exposure metallic film,as shown in Figure 6a, exhibits small barriers in the film itselfthat correspond to temperatures ranging from 0.3 to 0.03 K.Given that the film contains clusters of linked particles withdifferent sizes and likely different background charges, theexistence of a range of energies and the above-mention valuespoint to single electron charging barriers (Coulomb blockade)in the film. The corresponding two probe measurements, whichinclude both film and contact effects, exhibit significantly largerbarriers that correspond to temperatures ranging from ∼3 to0.3 K. Again, the existence of a range of barriers, the gradualchange of activation energies from high temperature to lowtemperature (see Arrhenius plot in Figure 6b) and their valuesagain suggest Coulomb blockade effects. Such barriers can arisefrom linked nanoparticle clusters near the electrode interfacethat have poor ohmic contact with nanoparticle film.It is also possible that contact barriers can arise due to

molecular contacts. A number of experimental and theoreticalstudies of alkanedithiols ranging from butanedithiol todecanedithiol have shown that self-assembled monolayers,formed from either vapor or liquid phase, can exist in mixedlying-down and standing-up phases on thermally deposited,polycrystalline Au(111) surfaces and on single crystals.21−23

These studies have also shown that as alkane chain lengthdecreases, the lying-down phase becomes more favored, and forbutanedithiol only the lying-down phase is formed. As NPs self-assemble on a butaneditiol-functionalized Au electrode surface,at least one of the thiols has to weaken its interaction with thesurface and interact with the NP. In principle, a weakinteraction with the NP, caused by strong butanedithiol−surface interaction or steric hindrance due to other molecules,could lead to nonohmic contact with the gold NP film andresult in a small contact resistance. A small barrier can arisefrom a surface dipole and contact potential generated at thecontact with the Au NP film.21 Due to the large curvature of theNPs arising from their small radius, the standing-up phaseshould be more favored by butanedithiols on NPs. This isconsistent with the observation that conductance versustemperature behavior of thick NP films is metallic, ratherthan through- or over-barrier that would be expected if inter-NP contacts were poor. This is also consistent with faciledielectric breakdown occurring in the break junction samples.Had there been inter-NP barriers throughout the film, voltagedrops would occur throughout the film as well, while if thebarriers were located primarily at the interface between the NPfilm and deposited Au electrodes (as the combined two- andfour-probe data indicate), voltage drops would be much moreconcentrated at these regions and, therefore, much moreeffective in generating breakdown. These results suggest that, inorder to reduce such contact resistance, short molecules thatfavor a standing-up phase are preferable when self-assemblingarchitectures on electrodes. At the same time, they alsodemonstrate that four-probe methods are effective in over-coming such contact resistances and thereby enabling studies ofthe electronic properties of such architectures themselves.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721271

■ CONCLUSIONIn summary, we have found that if two-probe conductanceversus temperature measurements of a material (1) containcontributions from a small contact barrier as well as metallicand nonmetallic components of the material; and (2)conductance tends to a finite, nonzero value as temperaturedrops (satisfying the definition of a metal), then one can safelyconclude that the material is a metal. If, however, conductancetends to zero, the role of contact resistance must be carefullyanalyzed to determine the nature of the material. Using a two-probe break junction with sufficiently thick butanedithiol-linkedAu NP films, we showed that we could observe a breakdown ofcontact resistance and thereby study conductance of metallicfilms themselves down to low temperatures where barriereffects would have otherwise dominated. Also, by using a four-probe electrode geometry, we could distinguish film versus film+contact behavior, enabling identification of metallic filmscloser to the insulator-to-metal transition. This study showsthat being able to study conductance of nanoengineeredmaterials unobscured by contact effects is desirable as it canbetter reveal behaviors exhibited by nanoengineered materialsthemselves.

■ ASSOCIATED CONTENT*S Supporting InformationAdditional experimental details as discussed in the text. Thismaterial is available free of charge via the Internet at http://pubs.acs.org.

■ AUTHOR INFORMATIONCorresponding Author*E-mail: adhirani@chem.utoronto.ca.

NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThe authors thank the Natural Sciences and EngineeringResearch Council for financial support.

■ REFERENCES(1) Tao, N. J. Electron transport in molecular junctions. Nat.Nanotechnol. 2006, 1, 173−181.(2) Hu, Y.; Zhu, Y.; Gao, H.; Guo, H. Conductance of an ensembleof molecular wires: a statistical analysis. Phys. Rev. Lett. 2005, 95(156803), 1−4.(3) Luzhbin, D. A.; Kaun, C.-C. Origin of high- and low-conductancetraces in alkanediisothiocyanate single-molecule contacts. Phys. Rev. B2010, 81, 035424.(4) Beebe, J. M.; Kim, B. S.; Frisbie, C. D.; Kushmerick, J. G.Measuring relative barrier heights in molecular electronic junctionswith transition voltage spectroscopy. ACS Nano 2008, 2, 827−832.(5) Fu, M.-D.; Chen, I.-W. P.; Lu, H.-C.; Kuo, C.-T.; Tseng, W.-H.;Chen, C.-H. Conductance of Alkanediisothiocyanates: Effect ofHeadgroup-Electrode Contacts. J. Phys. Chem. C 2007, 111, 11450−11455.(6) Li, C.; Pobelov, I.; Wandlowski, T.; Bagrets, A.; Arnold, A.; Evers,F. Charge Transport in Single Au | Alkanedithiol | Au Junctions:Coordination Geometries and Conformational Degrees of Freedom. J.Am. Chem. Soc. 2008, 130, 318−326.(7) von Wrochem, F.; Gao, D.; Scholz, F.; Nothofer, H. G.; Nelles,G.; Wessels, J. M. Efficient electronic coupling and improved stabilitywith dithiocarbamate-based molecular junctions. Nat. Nanotechnol.2010, 5, 618−624.

(8) Helander, M. G.; Wang, Z.-B.; Greiner, M. T.; Liu, Z.-W.; Qiu, J.;Lu, Z.-H. Oxidized Gold Thin Films: An Effective Material for High-Performance Flexible Organic Optoelectronics. Adv. Mater. 2010, 22,2037−2040.(9) Zabet-Khosousi, A.; Dhirani, A.-A. Charge transport innanoparticle assemblies. Chem. Rev. 2008, 108, 4072−4124.(10) Kelly, K. L.; Coronado, E.; Zhao, L. L.; Schatz, G. C. The opticalproperties of metal nanoparticles: The influence of size, shape, anddielectric environment. J. Phys. Chem. B 2003, 107, 668.(11) Kinge, S.; Mercedes, C.-C.; Reinhoudt, D. N. Self-assemblingnanoparticles at surfaces and interfaces. Chem. Phys. Chem. 2008, 9,20−42.(12) Boker, A.; He, J.; Emrick, T.; Russell, T. P. Self-assembly ofnanoparticles at interfaces. Soft Matter 2007, 3, 1231−1248.(13) Prasad, B. L. V.; Sorensen, C. M.; Klabunde, K. J. Goldnanoparticle superlattices. Chem. Soc. Rev. 2008, 37, 1871−1883.(14) Brust, M.; Bethell, D.; Schiffrin, D. J.; Kiely, C. J. Novel Gold-Dithiol Nano-Networks with Nonmetallic Electronic-Properties. Adv.Mater. 1995, 7, 795−797.(15) Fishelson, N.; Shkrob, I.; Lev, O.; Gun, J.; Modestov, A. D.Studies on charge transport in self-assembled gold-dithiol films:Conductivity, photoconductivity, and photoelectrochemical measure-ments. Langmuir 2001, 17, 403−412.(16) Trudeau, P.-E.; Escorcia, A.; Dhirani, A.-A. Variable singleelectron charging energies and percolation effects in molecularly linkednanoparticle films. J. Chem. Phys. 2003, 119, 5267−5273.(17) Musick, M. D.; Keating, C. D.; Keefe, M. H.; Natan, M. J.Stepwise construction of conductive Au colloid multilayers fromsolution. Chem. Mater. 1997, 9, 1499−1501.(18) Wessels, J. M.; Nothofer, H. G.; Ford, W. E.; von Wrochem, F.;Scholz, F.; Vossmeyer, T.; Schroedter, A.; Weller, H.; Yasuda, A.Optical and electrical properties of three-dimensional interlinked goldnanoparticle assemblies. J. Am. Chem. Soc. 2004, 126, 3349−3356.(19) CRC Handbook of Chemistry and Physics, 93rd ed.; Haynes, W.M., Ed.; CRC Press: Boca Raton, FL, 2013.(20) Muller, K. H.; Herrmann, J.; Wei, G.; Raguse, B.; Wieczorek, L.Temperature Dependence of Electrical Resistance in Films of GoldNanoparticles Linked by Organic Molecules. J. Phys. Chem. C 2009,113, 18027−18031.(21) Kobayashi, K.; Horiuchi, T.; Yamada, H.; Matsushige, K. STMstudies on nanoscopic structures and electric characteristics ofalkanethiol and alkanedithiol self-assembled monolayers. Thin SolidFilms 1998, 331, 210−215.(22) Millone, M. A. D.; Hamoudi, H.; Rodriguez, L.; Rubert, A.;Benitez, G. A.; Vela, M. E.; Salvarezza, R. C.; Gayone, J. E.; Sanchez, E.A.; Grizzi, O.; Dablemont, C.; Esaulov, V. A. Self-Assembly ofAlkanedithiols on Au(111) from Solution: Effect of Chain Length andSelf-Assembly Conditions. Langmuir 2009, 25, 12945−12953.(23) Franke, A.; Pehlke, E. First-principles study of 1,4-butanedithiolmolecules and radicals adsorbed on unreconstructed Au(111) andAu(100). Phys. Rev. B 2010, 81 (075409), 1−16.

Langmuir Article

dx.doi.org/10.1021/la304386j | Langmuir 2013, 29, 1264−12721272