High-Performance Transparent Conductors from Networks of Gold

One well-known problem with networks of silver or copper NWs is that the ...... of Ag Nanowire 2D Mass-Spring Networks Prepared by Floating Compressio...
0 downloads 0 Views 3MB Size
LETTER pubs.acs.org/JPCL

High-Performance Transparent Conductors from Networks of Gold Nanowires Philip E Lyons,† Sukanta De,† Jamil Elias,‡ Matthias Schamel,‡ Laetitia Philippe,‡ Allen T Bellew,§ John J Boland,§ and Jonathan N Coleman*,† †

Centre for Research on Adaptive Nanostructures and Nanodevices and School of Physics and §Centre for Research on Adaptive Nanostructures and Nanodevices and School of Chemistry, Trinity College Dublin, Dublin 2, Ireland ‡ Laboratory for Mechanics of Materials and Nanostructures, Empa — Materials Science & Technology, Feuerwerkstrasse 39, CH-3602 Thun, Switzerland ABSTRACT: We report the deposition and characterization of thin networks of gold nanowires on plastic substrates. The average nanowire diameter was 47 nm, while the networks had mean thicknesses in the range of 35750 nm. The conductivity of networks with mean thicknesses below 100 nm was controlled by percolation, as characterized by the percolation exponent, n = 0.8, and the percolative figure of merit, Π = 28. However, networks with thicknesses above 100 nm had thickness-independent direct current conductivity of σDC,B = 5.4  105 S/m. The conductivity was limited by junction resistances, which we estimate at ∼3 kΩ. Networks of all thicknesses were described by an optical conductivity of σOp = 1.0  104 S/m. These electrical and optical properties result in networks with sheet resistance and transmittance very close to industry requirements, that is, 49 Ω/0 @ 83%. These values are superior to almost all reported carbon nanotube networks and are competitive with silver nanowire networks. We measured the work function of these networks to be 4.6 eV, suggesting them to be suitable for hole injection or collection in electronic devices. SECTION: Nanoparticles and Nanostructures

T

he last 7 years have seen intensive research with the aim of replacing brittle doped metal oxides (e.g., ITO) with flexible nanostructured thin films for use as transparent conductors (TCs) in large-area or flexible displays.1 Although much of the early work focused on networks of carbon nanotubes24 or solution-exfoliated graphene flakes,5,6 it is becoming clear that the most promising candidates are vapor-grown graphene films7 or solution-cast networks of metallic nanowires (NWs).816 While both of these candidates surpass the minimum industry standards for transmittance, T, and sheet resistance, Rs, (T g 90% and Rs e 100 Ω/0) NW networks may have advantages in terms of cost and ease of deposition at low temperatures over large areas. At present, the literature has focused on silver NW (AgNW) networks,2,8,1012,15 with an additional three papers describing copper NW (CuNW) networks.13,14,16 Much progress has been made with a recent report showing that AgNWs can be spraydeposited over large areas to give networks with T = 90% and Rs = 50 Ω/0.15 However, a number of problems surround these materials. In particular, both CuNWs and AgNWs are prone to oxidization or tarnishing, leading to stability issues. In addition, their work functions are not high enough to act as efficient hole injectors/collectors in current-driven displays or solar cells. Finally, they are completely unsuited for operation in harsh environments such as those found in dye-sensitized solar cells.17 These problems are serious in that they limit the application areas suitable for NW electrodes. However, each of these problems could be addressed by using NWs made from an inert metal such r 2011 American Chemical Society

as gold or platinum. In fact, there is one report that mentions a single gold NW network with T ≈ 87% and Rs = 200 Ω/0.18 However, this single data point is insufficient to accurately assess performance or benchmark this material against other candidates. In this work, we have prepared gold NWs (AuNWs) by template-assisted electrodeposition. The resultant wires could be dispersed in a solvent and deposited to form thin networks. We show that these networks have optical and electrical properties superior to the vast majority of nanotube networks and are competitive with networks of AgNWs. Electrodeposition of AuNWs was performed in a threeelectrode electrochemical cell. Platinum spiral wire and a saturated calomel electrode (SCE) were used as the counter and reference electrodes, respectively. The working electrode (cathode) consisted of a commercial ion track etched polycarbonate membrane (Maike Pieper, Germany) coated from one side with about a 100 nm thin layer of silver by sputtering (Balzers Union SCD 040 sputtering machine). The thickness, nominal pore size, and density of these membranes were 6 μm, 10 nm, and 6  108 pores/cm2, respectively. This cathode was placed at an equal distance (about 1 cm) between the other two electrodes. The electrolyte consisted of a commercial gold bath of 10 g 3 L1 of potassium dicyanoaurate (I) (Puramet 402, Received: October 19, 2011 Accepted: November 18, 2011 Published: November 18, 2011 3058

dx.doi.org/10.1021/jz201401e | J. Phys. Chem. Lett. 2011, 2, 3058–3062

The Journal of Physical Chemistry Letters

LETTER

Figure 1. (A) Photograph of a thin AuNW network with T = 87% (t ≈ 35 nm). (B) Optical transmittance spectra for a subset of the networks studied. (C) SEM image of a AuNW network (t ≈ 300 nm). The scale bar is 1 μm. (D) NW diameter as measured by SEM. The mean NW diameter was 47 nm.

Doduco) with a small amount of gelatin (2 wt %), pH 7.3.19 The solution was heated up to 55 C and then diluted with an equal amount of water. The electrodeposition of gold inside the template was performed potentiostatically at 1.145 V versus SCE under slow agitation with a magnetic stirrer. The complete filling of the membrane pores without overgrowth under these conditions was optimized to occur after about 5 min. This will result in a total weight of approximately 5.5 μg of Au NWs per 1 cm2 of template. The separation and the dispersion of AuNWs in solution was achieved by preferentially dissolving the sputtered silver layer (from the filled membrane) using H2O2/ NH4OH (1:1 in volume). Afterward, the NW-containing polycarbonate membrane was dissolved in a dichloromethane solution. It was found that the NWs remained dispersed in this solution as long as the polymer was not removed. The final NW concentration was CAuNW = 0.001 mg/mL. Gold NW films were prepared by vacuum filtration of the above dispersions using porous filter membranes (MF-Millipore Membrane, mixed cellulose esters, hydrophilic, 0.2 μm, 47 mm). The deposited films were washed with 200 mL of Millipore water followed by a wet transfer to a polyethylene terephthalate (PET) substrate using heat and pressure.3 The cellulose filter membrane was then removed by treatment with acetone vapor and subsequent acetone liquid baths followed by a methanol bath.3 The film diameter was 16 mm. We note that because the AuNW dispersion contained polycarbonate, it is likely that some polycarbonate remains trapped in the final films. We attempted to remove this by washing the final films in dichloromethane. However, this washing step was not found to improve the electrical properties of the films. Shown in Figure 1A is a photograph of a thin (t ≈ 37 nm, see below) AuNW network on PET. This film was transparent and extremely uniform. However, we note that thicker networks have a distinct redish hue. The transmittance spectra (Varian Cary 6000i with PET as reference) for networks with a range of thicknesses (∼35750 nm) measured shortly after preparation

Figure 2. (A) Transmittance (550 nm) as a function of sheet resistance measured for networks at a range of thicknesses. The main data set describes networks measured immediately after preparation (closed red squares). However, we also show data for networks stored in ambient conditions for 130 days before measurement (open red squares). The dashed line describes bulk-like behavior and is a fit to eq 1, while the solid line describes percolative behavior and is a fit to eq 2. Also shown are data representing the state of the art networks of both AgNWs15 and SWNTs4 and the single previously reported AuNW network.18 (B) The transmittance and sheet resistance data from (A), replotted to illustrate the bulk and percolation regimes as described by eqs 1 and 2, respectively. (C) SEM image of the edge of a AuNW network on Si/SiO2 after cleavage. (D) Transmittance (550 nm) plotted as a function of approximate network thickness, t.

are shown in Figure 1B. The transmittance tends to fall off toward lower wavelengths probably due to light scattering or plasmonic effects. In general, the transmittance falls rather slowly with film thickness, indicating that the extinction of these networks is not extremely large (see below). We also analyzed the networks by SEM (Zeiss Ultra SEM, Figure 1C). The networks appear reasonably uniform and closely resemble networks of silver NWs.15 One obvious difference is that these AuNWs often display bends and kinks that might be expected to induce additional disorder within the network. From the SEM images, we could measure the NW diameter. As shown in Figure 1D, we observed a broad distribution with a mean of ÆDæ = 47 nm and a fwhm of ∼20 nm. These diameters are very slightly smaller than those commonly observed for AgNWs in the TC application area. In addition to the optical transmittance described above, we measured the sheet resistance for networks of a range of thicknesses (four-probe technique with silver electrodes and a Keithley 2400 source meter). The measured transmittance (550 nm) is plotted as a function of sheet resistance in Figure 2A. Both Rs and T increase steadily as the network thickness is reduced. The two thinnest films shown in Figure 2A displayed (Rs,T) values of 3059

dx.doi.org/10.1021/jz201401e |J. Phys. Chem. Lett. 2011, 2, 3058–3062

The Journal of Physical Chemistry Letters

LETTER

(49 Ω/0, 83%) and (119 Ω/0, 87%). These are reasonably close to the minimum industry standards for transparent electrodes of (100 Ω/0, 90%).20 We have also included on this graph the only reported data point for an AuNW network.18 This falls reasonably close to our thin-film data. Also shown in Figure 2A are the state of the art data for AgNW15 and SWNT4 (acidtreated) networks. It is clear from this graph that our AuNW data are competitive with these relatively well studied materials. One well-known problem with networks of silver or copper NWs is that the wires tarnish or oxidize over time, resulting in degradation of the electrical properties.13,21 Due to their inert nature, this problem should not apply to networks of AuNWs. To partially test this, we stored three networks in ambient conditions for 130 days after fabrication. After this period, we applied electrodes, noting that no visible change to the network had occured. We then measured T and Rs (Figure 2A), detecting virtually no change. This suggests that AuNW networks are much more stable than those of NWs of other metals commonly used as transparent conductors. For thin conducting films, the optical transmittance (at a given wavelength), T, is related to the sheet resistance, Rs, by2,22,23 !2 Z0 σ Op T ¼ 1 þ ð1Þ 2Rs σ DC;B where Z0 is the impedance of free space (377 Ω). Here, the ratio of direct current to optical conductivity, σDC,B/σOp, can be considered a figure of merit, with high values giving the desired properties (high T coupled with low Rs). We can test for this behavior by plotting the data as T1/2  1 versus Rs/Z0 (Figure 2B). Here, a straight line on a loglog plot with slope of 1 is characteristic of bulk behavior. This is indeed the case, allowing us to obtain σDC,B/σOp = 54 ( 2 (using this value, we have plotted eq 1 in Figure 2A for comparison). This value is significantly larger than those reported for all published nanotube networks,24 with the exception of one paper on acid-treated networks,4 which found σDC,B/σOp = 60. However, it is smaller than values reported for other metallic NW networks (83 < σDC,B/ σOp < 453; see a recent review24 for tabulated data). However, eq 1 only fits the data for networks with Rs/Z0 < 0.05, with the data diverging for thinner networks. This is a relatively common phenomenon8,2528 and has been attributed to percolation effects.23 For such thin networks, a new relationship between T and Rs has been proposed23 " #2   1 Z0 1=ðn þ 1Þ T ¼ 1 þ ð2Þ Π Rs where Π is known as the percolative figure of merit " #1=ðn þ 1Þ σDC;B =σOp Π¼2 ðZ0 tmin σOp Þn

ð3Þ

Here, tmin is the transition thickness, below which the DC conductivity becomes thickness-dependent (i.e., eq 2 applies below tmin). Analysis of these equations shows that large values of Π coupled with low values of n are desirable to achieve low Rs and high T.23 Furthermore, we showed empirically that networks of NWs have values of tmin that scale closely with the wire diameter, D, tmin ≈ 2.33D.23 In fact, eq 2 fits the high Rs/Z0 data in Figure 2B very well, giving fit values of Π = 28 ( 6 and

n = 0.82 ( 0.27 (again, these values have been used to plot eq 2 in Figure 2A for comparison). This value of Π is higher than that found for all nanotube networks (with the exception of one paper29 that we reanalyzed24 finding Π = 29.2). It is also close to the median value of 31.7 for metallic NW networks.24 Importantly, the percolation exponent is extremely low and comparable to the lowest values found for metallic NW networks.24 This implies that the distribution of internanowire junction resistances is reasonably narrow (although, it does not give information about the mean junction resistance).23 Both σDC,B/σOp and Π are ultimately controlled by σDC,B, σOp, tmin, and n.23 The first three parameters reflect the properties of the wires themselves but require information about the network thickness if they are to be estimated. We estimated the thickness of the networks by SEM analysis of fractured films (see Figure 2C for an example). We note that measurement of thickness using this method can be difficult because of the subjective nature of the identification of the film surface. For each network, the thickness was measured at ∼10 different positions, and the mean and standard deviations were measured. As can be seen from Figure 2D, there was a wide variation in the film thickness for these AuNW networks, possibly due to suboptimal dispersion quality. Shown in Figure 2D is the transmittance (550 nm) as a function of the estimated network thickness, t. For thin metallic films, T is related to t by22 T ¼ ½1 þ Z0 σOp t=22

ð4Þ

Fitting this equation to the data in Figure 2D gives a value of σOp = 10021 ( 500 S/m. This is lower than values typically found for nanotube networks (1.52  104 S/m)27,3032 but higher than that typically found for AgNW networks (∼5000 S/m).8,15,33 Combining this value with the value for σDC,B/σOp described above gives a value of σDC,B = 5.4 ( 0.5  105 S/m. While this value is similar to the best reported ones for acid-treated SWNT networks (∼510  105 S/m),4,34 it is below those values reported for AgNW networks (>106 S/m).8,33 This is the reason why the measured value of σDC,B/σOp is relatively low compared to that for most networks of metallic NWs. It has been previously reported that the conductivity of networks of conducting rods is limited by junction resistances 3 such that σDC,B = kV2f R1 J ÆDæ , where k is a constant, Vf is the network volume fraction, and RJ is the junction resistance.35,36 Using previously reported data for nanotube networks (σDC,B  RJ = 6.3  1011 m1, Vf ≈ 0.33, ÆDæ ≈ 20 nm),27,36 we calculate k = 4.6  1011 m2 (no equivalent data exist for metallic NW networks). We use this information to estimate the mean junction resistance for these AuNW networks to be ∼3 kΩ. This is about 3 orders of magnitude smaller than the value of 2.7 MΩ reported for pristine SWNT networks36 but considerably higher than the one of ∼50 Ω estimated for AgNW networks.10 We suggest that this junction resistance may be artificially high due to the presence of residual polycarbonate from the membrane dissolution after electrodeposition. It is likely that this could be more fully removed, resulting in lower RJ and higher σDC,B. It can be seen from Figure 2A that the crossover from bulk (dashed line) to percolative (solid line) behavior occurs at a transmittance of Tx = 71%. We note that, as in most cases,24 this means that the technological relevant films with T ≈ 90% lie in the percolative regime. As a result, the high performance described here (e.g. Rs = 49 Ω/0 @ T = 83%) is set by the values of the percolation parameters, that is, the combination of a 3060

dx.doi.org/10.1021/jz201401e |J. Phys. Chem. Lett. 2011, 2, 3058–3062

The Journal of Physical Chemistry Letters relatively high value of Π combined with relatively low n (compared to other reported networks). Of these parameters, we suggest that the low value of n is particularly important because the sheet resistance at a given value of T scales exponentially with n.15 We note that n is controlled by network properties. While it is not clear why n is so low here, we speculate that it may be related to the nature of inter-nanowire interactions or the possible presence of a residual polycarbonate coating. We can use the measured values of Tx and σOp coupled with eq 4 to estimate the network thickness associated with this transition to be tmin = 99 ( 25 nm. This gives a value of tmin/ÆDæ = 2.1 ( 0.9, in very good agreement with the previously reported value of 2.33.23 Finally, a significant advantage of using networks of AuNWs would stem from the chemical inertness and high work function of gold. However, it is necessary to confirm that these properties are retained for low-diameter NWs. In fact, it has recently been shown that for metallic NWs, electronic properties such as work function and surface electron density retain bulk values as long as the wire diameter exceeds the Fermi wavelength.37 As the Fermi wavelength of typical metals is