Fabrication and characterization of inkjet-printed carbon nanotube electrode patterns on paper

CARBON

5 8 ( 2 0 1 3 ) 1 1 6 –1 2 7

Available at www.sciencedirect.com

journal homepage: www.elsevier.com/locate/carbon

Fabrication and characterization of inkjet-printed carbon nanotube electrode patterns on paper Oh-Sun Kwon a, Hansu Kim a, Hyojin Ko a, Jumi Lee a, Byeongno Lee a, Chan-Hee Jung b, Jae-Hak Choi b,c, Kwanwoo Shin a,d,* a

Department of Chemistry and Institute of Biological Interfaces, Sogang University, Seoul 121-742, Republic of Korea Advanced Radiation Technology Institute, Korea Atomic Energy Research Institute, Jeongeup 580-185, Republic of Korea c Department of Polymer Science and Engineering, Chungnam National University, Daejeon 305-764, Republic of Korea d School of Engineering and Applied Sciences, Harvard University, 60 Oxford Street, Cambridge, MA 02138, USA b

A R T I C L E I N F O

A B S T R A C T

Article history:

An aqueous conductive ink of multi-walled carbon nanotubes (MWCNTs) for inkjet printing

Received 3 January 2013

was formulated. To prepare the homogeneous MWCNT ink in a size small enough not to

Accepted 18 February 2013

block a commercial inkjet printer nozzle, we used a kinetic ball-milling process to disperse

Available online 26 February 2013

the MWCNTs in an aqueous suspension with naphthyl-group-based non-ionic surfactants. When a patterned electrode was overlaid by repeated inkjet printings of the ink on various types of paper, the sheet resistance decreased rapidly following a power law, reaching approximately 760 X/sq, which is the lowest value ever for a dozen printings. Hall effect measurements at various magnetic fields and temperatures showed that the printed MWCNT paper electrode exhibited excellent semi-metallic conductivity with a p-type extrinsic semiconductive behavior such that the temperature-dependent resistivity was dominantly affected not by the lattice phonon scattering related to the mobility but by the charge impurity scattering related to the concentration of doped impurities. The Raman and Fourier transform infrared spectra revealed that the oxidation of the MWCNTs was the source of the doped impurities.  2013 Elsevier Ltd. All rights reserved.

1.

Introduction

Attempts have been made to apply carbon nanotubes (CNTs) [1], which are nano-scale structures consisting of rolled graphite sheets, to flexible conductive films [2–7] due to their superior electric and thermal conductivities, their various chemical functionalities, and their strong mechanical strengths [8–11]. Although the resistivity of an isolated CNT is low, e.g., on the order of 10 4 X cm for both a metallic single-walled carbon nanotube (SWCNT) [12] and a multi-walled carbon nanotube [13] at room temperature, obtaining a good conductive film with a group of aggregated CNTs has not been easy on a general electronic circuit scale [2–7].

Among the various deposition methods for a CNT solution, such as electrophoretic deposition [3], filtration [4,5], brush painting [6], conformal coating [6], spraying, and dip coating [14], inkjet printing [16] is an attractive option for obtaining a conductive sheet of randomly networked CNTs due to its potential direct applications in printing electronic circuit lines and patterned electrodes easily on flexible substrates, such as various types of paper and other planar substrates. Inkjet printing technology [17] has developed rapidly over the last three decades due not only to the boom in the home, office and industry printer markets but also to increased interest in coatings with thin polymer films, the fabrication of solar cells [6], and the construction of three-dimensional structures

* Corresponding author: Fax: +82 2 701 0967. E-mail addresses: [email protected], [email protected] (K. Shin). 0008-6223/$ - see front matter  2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.carbon.2013.02.039

CARBON

5 8 (2 0 1 3) 1 1 6–12 7

[18]. Because this technology relies on precise, thermal, electro-spinning or piezoelectric injection of a digital ink drop on a substrate at micrometer resolution, a conductive ink capable of replacing carbon and color dye ink solutions can be used to deposit a conductive film on a planar substrate such as paper with the advantages of inkjet printing: easy access, compact size and cheap cost. Such an inkjet printing technology would also make it possible to print a spiral conductive power line, which is required to provide a non-contacted electromagnetic-induced power source for radio-frequency identification, directly onto paper. Such a technology would thus clearly hold promise for a very wide range of applications. Recently, massive efforts have been conducted to develop conductive inks, especially for inkjet printing, based on metallic materials such as polymers [19], copper [20], and silver [21], but most efforts have not been very successful because of limitations and disadvantages such as low conductivity, oxidation of the surface, or inefficient electromigration. Thus, CNT inks have become attractive because they may provide highly conductive films without surface oxidation, electro-migration, or the need to replace polymers or use metallic pastes. Some productive works combining CNTs with inkjet technology have been performed (e.g., Fan et al. used surfactant-based inks in 2005 [15] and Korda´s et al. applied pure water dispersions in 2006 [16]), yet the actual conductivity has not improved much [22,23]. In practical applications, if a highly conductive CNT ink suitable for inkjet printers is to be realized, two obstacles must be overcome technically. The first obstacle is related to the structure of the CNT itself. A long CNT is, of course, preferable for effective physical percolation, whereas for longlasting inkjet printing without clogging, a length as short as possible is required in reality. These two contradictory aspects for CNT length might be resolved by either finding an optimized length suitable to the printer’s nozzle size or adjusting the functionality of the CNTs by incorporating suitable functional groups onto the CNT’s outer wall [23]. The second obstacle is related to the operating parameters, e.g., the method of ejecting micro-drops of CNT ink through a micronozzle (piezoelectric or thermal ejection), the evaporation of printed ink drops (solvent selection), and the aggregation of CNTs on paper substrates (surface treatment of the CNTs) for printing the ink solution with dissolved CNTs, all of which must be carefully tuned. In this research, we focused on trying to solve only the problem of the contradictory length requirement by applying a simple kinetic milling process to fabricate an inkjet ink with a smaller size distribution and a good MWCNT dispersion. We used a common home and office inkjet printer without any modifications, except for the charging process, to inject the MWCNT ink into the printer’s ink cartridge. We printed highly conductive patterned electrodes on paper substrates, even when applying overwriting printing processes only a few times. The minimum sheet resistance of the paper electrodes reached 760 X/sq for a dozen overwriting printings, which is, as far as we are aware, the lowest value yet obtained for printing MWCNTs onto paper using inkjet printers. We characterized the microscopic electric features of these paper electrodes by performing Hall effect measurements at various magnetic fields and temperatures.

2.

Experimental

2.1.

Preparation of MWCNT inks

117

One gram (1 g) of MWCNT powder purchased from TCI America Co. (Portland, USA) was mixed in 200-mL deionized water (electric resistivity: 18.2 MX cm) for 5 h at 80 C with 30.0 g of a dispersant. The dispersant was synthesized from a reaction of 20 wt.% 1-naphthylacetic acid (Aldrich, N0640), 20 wt.% polyethylene glycol (Kopex-PEG 600) and 10 wt.% polyethylene glycol monomethyl ether (Koremul-ME400) mixed in water, followed by mechanical stirring at 50 rpm for about 5 h at 60 C until the viscosity of the reaction mixture reached 230–260 cP. (see Scheme S1 in Supplementary Information) Because both the tangled and coiled chains of the inherent structure and the powder bunching of the long MWCNTs in the solution prevent smooth printing and may easily block the small nozzle of an ink cartridge, chemical dispersion of the MWCNTs is usually not sufficient. For this reason, we adopted mechanical ball-milling to obtain a more effective dispersion. The purpose of this process is threefold: (a) pulverizing the aggregates of MWCNT powder into smaller pieces, (b) straightening the tangled MWCNT threads and (c) forming a suspension to disperse the ink solution homogeneously. For ball-milling dispersion, an inhomogeneous mixed solution was sealed with a number of hard alumina beads that were 3 mm in diameter in a horizontal cylinder with a 12-cm inner diameter loaded on a rolling-type homogenizer. The system was rotated for 3 days at a frequency of 17 Hz, which corresponded to a maximum tangential speed of the spinning cylinder of approximately 0.1 m/s, to unravel and homogenize the aggregated and crossed-linked MWCNTs to a submicron size. Subsequently, the solution was centrifuged at 3000 rpm for 10 min 3 times to separate the non-dispersed MWCNT fractions and obtain the final MWCNT ink with a concentration of 0.15 mg/mL.

2.2.

Inkjet printing process

An EPSON Stylus T10 printer with a nozzle diameter of 20 lm and a maximum resolution of 5760 · 1440 dots per inch (dpi) was used for dispensing the MWCNT ink onto paper substrates. The printer employs the so-called drop-on-demand method [17,24–26] in which the size and ejection of a drop is controlled by the piezoelectric force. The pulsed pressure, which is driven by a piezoelectric transducer, compresses the fluid contained in a micro-capillary channel, jetting a submicron-sized (minimum size: 4 pL) drop from the orifice of the nozzle in a few microseconds. After the fabricated MWCNT ink was injected into an empty ink cartridge using a syringe, a black image of the simple patterned electrodes, which had been designed with 700 · 500 pixels per inch (ppi) using the Photoshop Elements software, was printed with a resolution of 720 dpi using an EPSON T10 printer in the grayscale text/image printing mode. The printed samples were dried in an ambient phase. In this research, two A4-size paper substrates were used: general paper for document printing (denoted as document paper) and

118

CARBON

5 8 ( 2 0 1 3 ) 1 1 6 –1 2 7

glossy-finished photo paper for inkjet photo printing (EPSON paper index: C13S042187; denoted as photo paper).

3.

Results and discussion

3.1.

Characterizations of dissolved MWCNT inks

To investigate the structural effects of the ball-milling process for homogeneous dispersion, we used transmission electron microscopy to compare the morphologies of the MWCNTs before and after the process. Fig. 1a shows a highly entangled bundle of MWCNTs. That bundle was selected, as a typical example, from those among the sporadically distributed MWCNTs. In Fig. 1b, we see that we can separate such a bundle into many smaller defragmented threads. In addition, overall, the defragmented threads can be seen to unravel

and distribute homogeneously. This indicates that, due to the milling, the networked MWCNTs are, as we expected, not only well dispersed but also cut. Similar observations can also be made based on the scanning microscopy (SEM) images of two samples, one with for ball-milled MWCNTs and the other with un-milled MWNTs, prepared in micro drops of inks and spread on a photo-paper substrate. As Fig. 1c and d and the higher magnified images in the insets show, for the samples with ball-milling times of 1 h and 3 days, respectively, two features are manifestly obvious: an increase in the uniformity of the spatial distribution and a decrease in the size of the MWCNTs making the closed packed networks under the ball-milling treatment. Quantitatively, the decrease in the size of the well-dispersed MWCNTs in the aqueous suspension was monitored using the dynamic light scattering technique. The results (see Fig. S1 in

Fig. 1 – Comparison of TEM and SEM images (a, c) before and (b, d) after the ball-milling process. Comparing images (a) and (c) to those of (b) and (d) shows an obvious increase in the degree of dispersion and defragmentation in such way that the smaller and unraveled MWCNTs formed into the closed packed homogeneous bundles as shown in the two insets.

CARBON

5 8 (2 0 1 3) 1 1 6–12 7

Fig. 2 – Spectral comparisons to check the structural deformations of the MWCNTs under the ball-milling process: (a) Raman spectra and (b) FTIR spectra.

Supplementary Information) revealed that the average length of the size distribution for the MWCNT colloidal particles was less than 1 lm, which agreed well with the TEM and the SEM images, indicating that the ball-milling homogenizer had been highly effective in dispersing the MWCNTs in the ink solution, of course, with aid of the dispersant. Conversely, based on these morphological and optical analyses, our ball-milling operation at a low frequency of 17 Hz was a highly effective way to make good cuts and to disperse random MWCNT networks. To study the damage to the MWCNTs more in detail, we used Raman spectroscopy [24] with excitation by a 514-nm red laser at a power of 100 mW (Fig. 2a). For the MWCNTs that were not treated with the ball-milling dispersion, the typical Raman features of MWCNTs were obtained (thick black line) [25]: two peaks of D (1350 cm 1) and G (1582 cm 1) bands, associated with defects and lattices in the graphene sheets, respectively. This means that the un-milled sample contained some damage as-received. The D/G ratio from the un-milled sample and that of the well-dispersed MWCNTs by ball-mill-

1

119

ing (thin red1 line) were 0.56 and 0.76, respectively. The increase in the D/G ratio after the ball-milling process indicates an increase in the number of defects, such as lattice vacancies, as well as destroyed and deformed sites, caused in the hexagonal graphene layers by broken bonds and/or by oxidation through a chemical reaction, probably with either the dispersant additive or air impurities, that occurred during the mixing and milling of the sample. To identify which source of defects, physical or chemical, was dominant in the MWCNTs, we compared the Fourier transform infrared spectra (FT-IR) in the attenuated total reflectance mode for MWCNTs coated on a gold substrate by using the spin-coating technique [16,26,27], and the results are presented in Fig. 2b. Two bands of peaks, 1650–1750 and 1460–1560 cm 1, corresponding to C@O and C@C bonds, respectively, were observed. The first band corresponds to the presence of carboxylic functional groups formed as defects on the MWCNT walls. Because the first band appeared for two samples without any significant difference in intensity or shape, we cannot, unfortunately, determine clearly whether the defects were caused by oxidation that occurred upon exposure to air during fabrication of the MWCNT ink under ambient conditions or by the reaction with the functional groups of the dispersant during the ball-milling process. Thus, we are left with a question to be studied more in the future because the answer to that question may prove to be a critical factor for not only improving the conductivity of the MWCNT ink but also for theoretically understanding the mechanism of electron behavior in our MWCNT ink printed on paper. The dynamic viscosity of the fabricated MWCNT ink was measured as 2.5 ± 0.1 cP at room temperature and 45% relative humidity, which is less than that of normal EPSON inkjet ink, i.e., 4.3 ± 0.1 cP. The surface tension measured using the Wilhelmy plate method was 45 mN/m. These optimized viscosity and surface tension for the MWCNT ink are expected to provide the improved printing quality for a quick drying time and a micro-level high resolution. The bonding of MWCNT ink on paper may be attributed to the cohesive van der Waals forces among MWCNTs and to the adhesion between the MWCNTs and the paper [28,29]. Although no additional treatments of the MWCNT ink were used, except for the viscosity and the surface tension, good ink conditions for normal printing without any clogging of the micro-sized channel was maintained for 2–3 days, after which an aggregated colloidal state due to the gradual loss of dispersion was observed.

3.2.

MWCNT pattern formation by inkjet printing process

Ejection, wetting and fixation of a submicron-sized discrete ink drop are the fundamental inkjet printing processes. Especially in drop-on-demand printing, as schematically shown in Fig. 3, the initial dynamical interaction of the ink drop with the paper substrate is a critical factor in printing our MWCNT ink properly under normal conditions [30,31]. Moreover, just after printing, the water solvent should be removed from

For interpretation of color in Fig. 2, the reader is referred to the web version of this article.

120

CARBON

5 8 ( 2 0 1 3 ) 1 1 6 –1 2 7

Fig. 3 – Schematic views of inkjet printing processes: a digital ink drop is ejected from the printer nozzle orifice by piezoelectric actuation under an applied voltage, and raster printing directions yield an array of patterned MWCNT electrodes on a paper substrate. (a) A MWCNT ink drop on paper is printed through the wetting dynamics balance among the surface energies of three phases: air, ink and paper. (b) Unidirectional MWCNT networks are formed on a paper substrate by van der Waals interactions.

the paper surface quickly at a rate similar to that of normal black ink without any special treatment, such as the use of chemical agents and additives or thermal annealing (inset (a) in Fig. 3). Therefore, the surface tension and viscosity in the capillary channel play important roles in deforming and ejecting an ink droplet of the required size, as the piezoelectric pressure acts against the surface tension and the viscoelastic force of the ink fluid. Furthermore, the unidirectional printing pattern in raster printing must be closely related to the electric properties of the corresponding patterns on the paper (inset (b) in Fig. 3). To check the hydrodynamic properties and surface interactions, we examined the variations in the wetting moments of an ejected ink drop over time using a high-speed zoom camera, and the images are shown in Fig. 4a–i. The contact angles were then measured and are summarized in Fig. 4j. To obtain these wetting data, we ejected a micro-droplet (1 lL) using a syringe and deposited it on a paper substrate to indirectly simulate the drop ejected from the printer nozzle. Although the dimensions of the drops were quite different from those of the droplets ejected from the nozzle (range: pL to nL), because the surface tension does not depend on the drop size, we may take the dynamic variations in the

adsorption, the absorption and the fixation of an ink drop on a paper substrate to be independent of the ejection and solvent evaporation times [29].

Fig. 4 – (a)–(i) Sequential images of wetting and adsorption of a MWCNT ink drop of 1 lL on paper and (j) contact angle as a function of time.

CARBON

5 8 (2 0 1 3) 1 1 6–12 7

We found that the initial wetting on the EPSON photo paper was a very quick process, on the order of milliseconds, and was comparable to that of normal printing. This result means that the surface energy of the paper substrate is sufficiently higher than that of the MWCNT ink in the air ambient phase to lead to a quick transition from partial to complete wetting states, even for a micron-sized ink drop. One noticeable feature is the exponential variation during the wetting transition, as shown in Fig. 4j. This figure shows that in the final drying, the wetting transition from the partial to the complete state happened faster because the size of the residual partially wetted drop had been reduced to pL, which was comparable to that of a real inkjet-printed drop, thus leading to high-resolution printing without any unexpected irregular spreading of the drop due to a fast drying time. At the final moment of drying, a fixation of networked MWCNTs on the paper fibers results from van der Waals interactions. The lifetime of the MWCNT ink cartridge was longer than 4 days, with normal printing operation and good printing quality maintained without any clogging of the nozzle’s orifice. Instead of the well-known high conductivity of a single CNT with respect to the chirality [8–11], the conductivity of a printed CNT electrode deposited in a single printing is not high enough for usual electronic applications, due to the inhomogeneous CNT network. The overwriting method [15,16,22,31–33] was, therefore, adopted to increase the conductivity in such a way that the electrode was printed in the same position as many times as necessary, which is clearly one of the advantages of using printers to coat films. In Fig. 5, two patterned electrodes, which were overwritten as

121

(a) a filled rectangular area and (b) a thin spiral line, are presented. For overwriting printing, however, a study of the morphological images printed in MWCNT ink is required to evaluate not only the homogeneity but also the resolution. Hence, the transmission optical microscope images of the left top corners of the rectangular electrodes in the first column in Fig. 6a were investigated to directly observe any surface variations in the sequential images of gradually deposited ink. In the first printing, as shown in Fig. 6a, compared to the y-directional printings, the x-directional printed drops (as defined in Fig. 3) were well connected serially to each other. Because the EPSON T10 printhead has 90 nozzles with a pitch of 1/120 in. for black ink (inset to Fig. 6a) and uses bi-directional printing [17] to completely print an image in 720 dpi, it needs six bi-directional raster printings with the subsequent feeding of paper in the y-direction to produce six swaths of dots for a single printing. Such a directional difference in printing quality, therefore, is due to the zig–zag raster printing (Fig. 6a) of the EPSON T10 printer. This difference occurs because the digital lateral motion of the printer head driven by an electronically controlled step motor is much more precise than the motion of the analog paper substrate in the vertical direction driven by the mechanical friction force between the paper and the roller. Conversely, a few satellite printed dots, which are observed in Fig. 6a–d [33] and are usually formed at the beginning of printing, were found in the vicinity of the edge boundary, causing a clearly lower resolution of the printed electrode. Based on the observed data, we might expect the roller to cause a lower resolution, especially in our overlaying printing

Fig. 5 – Patterned electrodes (a) in a rectangular shape in terms of the printing number n = 2, 4, 6, . . ., and 20 and (b) in a spiral shape for n = 20 for a plausible application of an antenna for radio-frequency identification.

122

CARBON

5 8 ( 2 0 1 3 ) 1 1 6 –1 2 7

Fig. 6 – Microscopic images of an electrode printed with MWCNT ink for various values of the printing number n ((a) n = 1, (b) n = 2, (c) n = 3, and (d) n = 10) for the left top corner of the rectangular patterned electrodes shown in Fig. 5. Inset to (a): configuration of the nozzles (shown are the first three only) of the printer head.

mode. Because it not only causes an irregular vertical position but also tilts when the paper is fed into the printer, overlapped printings were observed in the 1st and 2nd bi-directional swath printings, as shown in Fig. 6a. Unexpectedly, however, the overall accumulated resolution for 10 printings (Fig. 6d) was markedly reduced to roughly DY = 100 lm, which is much smaller than the DX = 200 lm that is mostly observed for satellite dots, meaning that overlay printing results in offset printing, especially in the y-direction, with improved resolution. Additionally, because the improper feeding of the paper by the roller causes paper distortion and undesired wear, the overlay printing time was limited to a maximum of approximately n = 20 times. In fact, as expected, we found that as the number of printings increased, an enhanced and noticeable gradual opaqueness was observed in the optical images, indicating that a homogenous surface had been formed due to reinforced inter-connections among individual drops. In addition, the pores among the MWCNTs disappeared, making a homogeneous electrode.

3.3.

CNT network structures and electric conductivity

In a microscopic dynamic viewpoint of this gradual printing process, we can consider a model in which the colloidal clusters of MWCNTs in an ink drop should percolate well into the empty space among the cellulose fibers in the document

paper substrate and then subsequently be deposited in the form of a randomly crossed-linked network as the printing is repeated. Conversely, the MWCNT particles and clusters only penetrated the resin layer that was top-coated on the photo paper and filled with submicron, porous particles. The homogeneity of the MWCNT ink on the top surface in the overwriting printing process was confirmed with surface images obtained using atomic force microscopy (AFM), as shown in Fig. 7. The figures show that the nano-sized pores on the top surface (for n = 0, i.e., clean photo paper) were screened significantly by the MWCNTs and that the MWCNT network became homogenized as the number of printings increased (for n = 1–12), making a good conductive film in a densely packed form. The average roughnesses were less than 15 nm and 100 nm for n = 0 and n > 0, respectively. The homogeneity of the printed MWCNTs images at n = 12 is very similar to that of the close-packed MWCNT network of macro drops on the photo paper substrate shown in the SEM images in Fig. 1d. We believe that it is this close-packed MWCNT network that is formed on the photo paper surface and leads to the noticeably high conductivity, which is the highest ever. To obtain the conductivity of the electrodes printed on the two types of paper in rectangular shapes, we measured the sheet resistance using a linear four-point probe ohmmeter composed of four thin tungsten tips and a Keithely 2400 sourcemeter under ambient conditions of 45% relative humidity at

CARBON

5 8 (2 0 1 3) 1 1 6–12 7

123

Fig. 7 – AFM images for MWCNT ink printed on paper for various printing numbers n. The dimensions of all images are 5 lm · 5 lm.

Fig. 8 – Sheet resistance of a MWCNT film electrode printed on photo paper: (a) data fitted well to a power law in the printing number corresponding to the film’s thickness (inset) and (b) demonstration of a varying sheet resistor for n P 12 for lighting a LED.

room temperature. Fig. 8a presents the sheet resistance of the MWCNT electrode printed on photo paper as a function of the number of printings (n). The sheet resistance decreased rapidly as n was increased until n = 12, implying uniform deposition of the printed films. Such a uniform film deposition was confirmed using an Alpha-step IQ surface profiler to measure

the thickness, as shown in the inset. However, for n > 12, the sheet resistance began to slowly vary and asymptotically reached a minimum of 760 X/sq for n = 20. For document paper, the sheet resistance exhibited a similar behavior but was higher by a factor of 10 (not shown). The difference may be attributed to the difference in surface roughness

124

CARBON

5 8 ( 2 0 1 3 ) 1 1 6 –1 2 7

between the two types of paper. Fig. 8b shows light from a light-emitting diode (LED) when it was connected to a 3 V battery at two arbitrary contact points on the surface of a half-folded sheet electrode of 2 cm · 3 cm for n P 12, corresponding to a maximum bulk resistance of 100 kX. According to Gracia-Espino et al., the dependence of the conductivity of the inkjet-printed functionalized SWCNTs the film’s thickness followed well the universal power law for percolating networks with conducting paths [34]. Applying the same law to our data, as shown in Fig. 8, with fitting parameters R0, nc and j, we found that the resistance of the MWCNTs film also showed good agreement for R0 = 12.07 ± 0.13 kX/sq, j = 1.02 ± 0.06 and nc = 0.86 ± 0.13 (Fig. 8a). Because the last parameter is the critical printing number nc (see Fig. 7) corresponding to the critical film thickness tc  1 lm (inset in Fig. 8a), whenever n  1 or t  tc, the power law can be considered to be applicable, and the MWCNT networks can be considered to be continuous connections of bundle-to-bundle, tube-to-bundle and tube-totube Ohmic contacts, like those in the functionalized SWCNTs films in Refs. [35,36]. Therefore, when the printing number (i.e., the film thickness) is larger than 12, a saturated value with minimal resistance, Rs,min, is obtained because as the film’s thickness becomes larger than the critical thickness, the percolated MWCNT networks become less and less important for vertical percolated deposition, so the resistance reaches the bulk value. It is interesting to note that the noticeably high conductivity is due to the nature of such a quick (since nc = 1) development of such critical percolating networks of printed MWCNTs and that this is possible only on the novel photo paper, unlike on general document paper, by using inkjet printing as a method of MWCNT film deposition.

3.4.

Fig. 9 – Resistivity data for selected printing numbers n = 4, 8, 12, 16 and 20 as functions of (a) the magnetic field and (b) the temperature. Inset: sheet resistance for the overall average resistivity of the magnetic field of (a).

Hall effect measurements of CNT film on paper

To investigate the electric properties of the rectangular-patterned MWCNT film electrodes on paper, we measured the Hall effect using the van der Pauw method with a Lakeshore 7600 system, which can measure the resistivity of the electrodes as a function of the concentration, the mobility and the sign of the charge carriers [37]. Because the Hall effect for common conductor and/or semiconductor films varies with the applied external magnetic field and temperature, these two factors were studied separately. First, the resistivity as a function of the magnetic field measured at room temperature for five different printing numbers is shown in Fig. 9a. The magnetic-field dependence of the resistivity is unobservable, at least on the order of lX cm, regardless of the field strength for fields less than 0.6 T. When the printing number was n = 5, the measured resistivity was between 0.6 and 0.8 X cm. These values are similar to those of a MWCNT film deposited using a chemical vapor method [38], indicating that inkjet printing does not modify the intrinsic CNT properties much. However, the absolute value was very different from those of films fabricated using other methods [3,5,6,13,38– 40], indicating that the resistivity measurements are highly process-dependent. When the film thickness t is considered, the bulk sheet resistance Rs can be obtained from the geometrically inde-

pendent resistivity q, which was obtained from the Hall effect measurement using the simple relationship Rs = qt under only one assumption of film homogeneity. As shown in the inset in Fig. 9a, one interesting feature is that the sheet resistance deduced from the general van der Pauw method, which uses four rectangular-configured probes, was consistent with that measured using the previous four-probe method, in which the probes were configured collinearly in the same space (Fig. 8a) [41]. The thickness of a printed MWCNT electrode is thus fairly uniform such that any disconnections, which include macroscopic holes, pores and cavities, in the film should be considered negligible, supporting our assumption of homogeneity for our films. Conversely, the resistivity decreased as n was increased. Consequently, we may conclude that the deposition of MWCNTs using repeated inkjet printings at the same place should provide a uniform coating and may rapidly produce a homogeneous, planar, conducting surface with a depth that increases in proportion to the number of printings. Next, Hall effect measurements of the temperature dependence of the resistivity were carried out in the temperature range of 100–360 K for various n at a fixed magnetic field of 0.3 T, which was selected as the median of the tested field strengths. As shown in Fig. 9b, the trend of decreasing resistivity with increasing n at each temperature is similar to that

CARBON

5 8 (2 0 1 3) 1 1 6–12 7

at each magnetic field and corresponds to the increasing film thickness. At room temperature, if we exclude the data for n = 4, the bulk sheet resistances are consistent with those from the magnetic field data, as shown in the inset of Fig. 9a. However, the resistivity is very sensitive to the temperature, unlike the lack of sensitivity to the magnetic field. To see this effect more in detail, we investigated the average Hall parameters, the Hall coefficient, the mobility, the concentration of microscopic charge carriers in the MWCNT film, and the resistivity for the specimen with n = 5. All of the results are collectively shown in Fig. 10a–d. As shown in Fig. 10a, the MWCNT film’s electrode has a nonmetallic negative temperature coefficient of resistivity, dq/dT < 0 [42]. Such a nonmetallic temperature dependence points to a semimetallic behavior of the MWCNT film. In fact, a semi-metallic resistivity of a MWCNT film has been reported by de Heer and co-workers, who used a variable-range hopping model [43], and by Ska´kalova´ et al., who conjectured that such an unexpected nonmetallic conductance behavior with changing temperature for CNTs was related to the effect of fluctuation-assisted tunneling [44]. Although a single intrinsic MWCNT with a large tube diameter is well known to be a metallic material and when semiconducting is referred to as semi-metallic due to its narrow band gap energy [41], the semiconducting nature of films with MWCNT networks, not single tubes, is still not well understood, and whether the

Fig. 10 – Hall parameters at a fixed magnetic field of 0.3 T as a function of temperature for the average printing number: (a) resistivity, (b) Hall coefficient, (c) mobility and (d) charge carrier concentration.

125

conductivity is p-type or n-type due to contamination or unwanted impurities, such as oxidation, is still a subject of controversy [9,45–48]. The Hall coefficients over the tested range of temperatures seemed to be somewhat irregular, but all were clearly positive, as shown in Fig. 10b. Although fully describing the electron and hole densities separately is not possible due to the lack of microscopic structures for the MWCNT film electrodes, the dominant charge carriers are not electrons, but holes, indicating that the semiconducting film electrode is a p-type electrode [45–48]. For the above reason, some of the fluctuations shown in the three subplots in Fig. 10b–d can be attributed to an instability of the Ohmic contact due to the high Schottky barrier junction between the four metallic Hall probes and the semiconducting sample’s surface at low temperatures, not to the quantum effects in the MWCNT films, such as confinement and magnetic resistance along the chiral direction of the MWCNTs [45,47]. Moreover, our applied magnetic strengths and temperatures are neither strong nor low enough to cause any quantum effects. The Hall coefficient at a magnetic field of 0.3 T and at room temperature was 0.03 cm3/C, which is quite similar to reports in the literature. Although it is 103 times larger than the value for common metals, we can still consider the MWCNT film electrode to be a quasi-metal having positively charged carriers and holes because the corresponding carrier concentration is still much higher than those of metals. Actually, the corresponding mobility and carrier concentration at room temperature were 0.75 cm2/Vs and 3.73 · 1019 cm 3, respectively, as shown in Figs. 10c and d. The absolute carrier concentration is much higher than the typical value of 1.4 · 1010 cm 3 for intrinsic semiconducting silicon, but it is comparable to those of the common, heavily doped, degenerate, extrinsic semiconductors. Consequently, we may expect our MWCNT film electrode to have some doped impurities. The high number of defects for our MWCNTs observed in the Raman data shown in Fig. 2a might be responsible for these impurities. Bandaru noted that exposure to oxygen resulted in a reversion to p-type from n-type conductivity for SWCNTs due to their narrow band gap energy [9]. Derycke et al. demonstrated that CNT field-effect transistors exhibited unipolar p-type behavior when the doping and the oxygen adsorption on the CNT surfaces were controlled [49]. Conversely, this behavior suggests that if the conductivity is to be improved, oxidation should be eliminated either during the ink fabrication process or after printing. Due to the dominant one-type hole carrier in the MWCNT film, the constant mobility can be simply attributed to the inverse proportionality of the conductivity r, which is simply the inverse of the resistivity q, and to the Hall coefficient RH as a function of the temperature because the mobility is given simply by l = rRH = RH/q. Hence, based on the measured mobility, the Drude scattering model for charge carriers in the MWCNT film is introduced, and the collision time s and the mean free path l can be roughly estimated: s = m*l/q and l = vTs, where m*, q and vT are the effective mass, the electric charge and the thermal velocity of a hole carrier, respectively [41]. With the simple approximation of using the equivalent effective electron mass for a hole, we obtained s  1.3 · 10 12 s and l  130 nm. Interestingly, the mean free path is about

126

CARBON

5 8 ( 2 0 1 3 ) 1 1 6 –1 2 7

one-tenth the average length of a MWCNT (1 lm). This result implies that the doped impurities scattered by the hole carriers were embedded in the CNT tubes roughly at this distance, which might be confirmed by a theoretical calculation with the charge carrier concentration given in Fig. 10d, provided the scattering cross sections of the impurity scatters are known [41]. Thus, from a microscopic viewpoint, we may establish a plausible kinetic model for the hole current: the p-type hole carriers flow along the tube surfaces of MWCNTs, collide with either the impurities or the boundary of the junction between two tubes, and scatter into other tubes. The latter collision suggests a type of hopping-like jumping model for the transport of hole carriers between the MWCNT tubes in the printed film.

4.

Conclusion

A homogenous, aqueous and conductive MWCNT ink was formulated and applied on paper by using a commercial inkjet printer. After a mechanical ball-milling dispersion process, TEM and SEM images showed that well-dispersed printable ink was prepared, and the Raman and the FT-IR spectra revealed that doped impurities were present, presumably due to the surface oxidation of the MWCNTs. The fluidic properties of the ink, such as its viscosity, wetting, adsorption and fixation on the paper substrate were satisfactory for quick drying, and overlaying printing with fine resolution was possible up to 20 times at the same position on the paper substrate. The minimum sheet resistance achieved after 12 overlaid printings was 760 X/sq, especially for photo paper. Because Hall effect measurements revealed that the printed MWCNT film exhibited excellent semi-metallic conductivity and retained a p-type extrinsic semi-conductive behavior such that the temperature-dependent resistivity was dominantly affected by charged impurity scattering related to the concentrations of doped impurities, the process can be used for conductive films in numerous possible applications due to the excellent semi-metallic conductivity.

Acknowledgments This work was supported by the Nuclear Research R&D Program, the Mid-career Researcher Program (2011-0017539), the Global Frontier Research Program (2011-0032155), the Advanced Research Center for Nuclear Excellence, and a GISTNCRC grant (R15-2008-006) funded by Ministry of Education and Science & Technology, Korea.

Appendix A. Supplementary data Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/ j.carbon.2013.02.039.

R E F E R E N C E S

[1] Iijima S. Helical microtubules of graphitic carbon. Nature 1991;354(6348):56–8.

[2] Wu Z, Chen Z, Du X, Logan JM, Sippel J, Nikolou M, et al. Transparent, conductive carbon nanotube films. Science 2004;305(5688):1273–6. [3] Pei S, Du J, Zeng Y, Liu C, Cheng H-M. The fabrication of a carbon nanotube transparent conductive film by electrophoretic deposition and hot-pressing transfer. Nanotechnology 2009;20(20):235707–13. [4] Zhang M, Fang S, Zakhidov AA, Lee SB, Aliev AE, Williams CD, et al. Strong, transparent, multifunctional, carbon nanotube sheets. Science 2005;309(5738):1215–9. [5] Wang D, Song P, Liu C, Wu W, Fan S. Highly oriented carbon nanotube papers made of aligned carbon nanotubes. Nanotechnology 2009;19(7):075609–75614. [6] Hu L, Choi JW, Yang Y, Jeong S, Mantia FL, Cui LF, et al. Highly conductive paper for energy-storage devices. Proc Natl Acad Sci USA 2009;106(51):21490–4. [7] Sethi S, Dhinojwala A. Superhydrophobic conductive carbon nanotube coatings for steel. Langmuir 2009;25(8):4311–3. [8] Rao CNR, Satishkumar BC, Govindaraj A, Nath M. Nanotubes. ChemPhysChem 2001;78(2):78–105. [9] Bandaru PR. Electrical properties and applications of carbon nanotube structures. J Nanosci Nanotechnol 2007;7(3):1–29. [10] Bernholc J, Brenner D, Nardelli MB, Meunier V, Roland C. Mechanical and electrical properties of nanotubes. Annu Rev Mater Sci 2002;32(10):347–75. [11] Komarov FF, Mironov AM. Carbon nanotubes: present and future. Phys Chem Solid State 2004;5(3):411–29. [12] Thess A, Lee R, Nikolaev P, Dia H, Petit P, Robert J, et al. Crystalline ropes of metallic carbon nanotubes. Science 1996;273(5274):483–7. [13] Dai H, Wong EW, Lieber CM. Probing electrical transport in nanomaterials: conductivity of individual carbon nanotubes. Science 1996;272(5261):523–6. [14] de Andrade MJ, Lima MD, Ska´kalova´ V, Bergmann CP, Roth S. Electrical properties of transparent carbon nanotube networks prepared through different techniques. Phys Status Solidi 2007;1(5):178–80. [15] Fan Z, Wei T, Luo G, Wei F. Fabrication and characterization of multi-walled carbon nanotubes-based ink. J Mater Sci 2005;40(18):5075–7. [16] Korda´s K, Mustonen T, To´th G, Jantunen H, Lajunen M, Soldano C, et al. Inkjet printing of electrically conductive patterns of carbon nanotubes. Small 2006;2(8–9): 1021–5. [17] Le HP. Progress and trends in ink-jet printing technology. J Imaging Sci Technol 1998;42(17):49–62. [18] Sachs E, Cima M, Cornie J, Brancazio D, Bredt J, Curodeau A, et al. Three-dimensional printing techniques: the physics and implications of additive manufacturing. CIRP Ann Manuf Technol 1993;42(1):257–60. [19] Sirringhaus H, Kawase T, Friend RH, Shimoda T, Inbasekaran M, Wu W, et al. High-resolution inkjet printing of all-polymer transistor circuits. Science 2000;290(5499):2123–6. [20] Park BK, Kim D, Jeong S, Moon J, Kim JS. Direct writing of copper conductive patterns by ink-jet printing. Thin Solid Films 2007;515(19):7706–11. [21] Lee KJ, Jun BH, Kim TH, Joung J. Direct synthesis and inkjetting of silver nanocrystals toward printed electronics. Nanotechnology 2006;17(21):2424–8. [22] Wei T, Ruan J, Fan Z, Luo G, Wei F. Preparation of a carbon nanotube film by ink-jet printing. Carbon 2007;45(13):2712–6. [23] Denneulin A, Bras J, Blayo A, Khelifi B, Roussel-Dherbey F, Neuman C. The influence of carbon nanotubes in inkjet printing of conductive polymer suspensions. Nanotechnology 2009;20(38):385701–10. [24] Dresselhaus MS, Dresselhaus G, Saito R, Jorio A. Raman spectroscopy of carbon nanotubes. Phys Rep 2005;409(2):47–99.

CARBON

5 8 (2 0 1 3) 1 1 6–12 7

[25] Jung C-H, Kim D-K, Choi J-H, Nho Y-C, Shin K, Suh D-H. Shortening of multi-walled carbon nanotubes by r-irradiation in the presence of hydrogen peroxide. Nucl Instrum Methods Phys Res Sect B 2008;266:3491–4. [26] Cheng X, Zhong J, Meng J, Yang M, Jia F, Xu Z, et al. Characterization of multiwalled carbon nanotubes dispersing in water and association with biological effects. J Nanomater 2011;2011(14):1–12. [27] Najafi E, Kim J-Y, Han S-H, Shin K. UV-ozone treatment of multi-walled carbon nanotubes for enhanced organic solvent dispersion. Colloids Surf A 2006;284–285:373–8. [28] Denneulin A, Bras J, Carcone F, Neuman C, Blayo A. Impact of ink formulation on carbon nanotube network organization within inkjet printed conductive films. Carbon 2011;49(8):2603–14. [29] Israelachvili JN. Intermolecular and surface forces. 3rd ed. New York: Elsevier; 2011. [30] Wijshoff H. The dynamics of the piezo inkjet printhead operation. Phys Rep 2010;491(45):77–177. [31] Shin K-Y, Hong J-H, Jang J. Micropatterning of graphene sheets by inkjet printing and its wideband dipole-antenna application. Adv Mater 2011;23(18):2113–8. [32] Azoubel S, Shemesh S, Magdassi S. Flexible electroluminescent device with inkjet-printed carbon nanotube electrodes. Nanotechnology 2012;23(34):344003–8. [33] Huang L, Huang Y, Liang J, Wan X, Chen Y. Graphene-based conducting inks for direct inkjet printing of flexible conductive patterns and their applications in electric circuits and chemical sensors. Nano Res 2011;4(7):675–84. [34] Gracia-Espino E, Sala G, Pino F, Halonen N, Luomahaara J, Maklin J, et al. Electrical transport and field-effect transistors using inkjet-printed SWCNT films having different functional side groups. ACS Nano 2010;4(6):3318–24. [35] Bekyarova E, Itkis ME, Cabrera N, Zhao B, Yu A, Gao J, et al. Electronic properties of single-walled carbon nanotube networks. J Am Chem Soc 2005;127(16):5990–5. [36] Gruner G. Two-dimensional carbon nanotube networks: a transparent electronic material. Mater Res Soc Symp Proc 2006;905E. 0905-DD06-05.1–11. [37] Bube RH. Electrons in solids: an introductory survey. 3rd ed. New York: Academic Press; 1992.

127

[38] Singjai P, Changsarn S, Thongtem S. Electrical resistivity of bulk multi-walled carbon nanotubes synthesized by an infusion chemical vapor deposition method. Mater Sci Eng A 2007;443(1–2):42–6. [39] Marinho B, Ghislandi M, Tkalya E, Koning CE, de With G. Electrical conductivity of compacts of graphene, multi-wall carbon nanotubes, carbon black, and graphite powder. Powder Technol 2012;221:351–8. [40] Baker-Jarvis J, Janezic MD, Lehman JH. Dielectric resonator method for measuring the electrical conductivity of carbon nanotubes from microwave to millimeter frequencies. J Nanomater 2007;2007(1):1–4. [41] van der Pauw LJ. A method of measuring specific resistivity and Hall effect of discs of arbitrary shape. Philips Res Rep 1958;13(1):1–9. [42] Charlier J-C, Issi J-P. Electronic structure and quantum transport in carbon nanotubes. Appl Phys A 1998;67(1): 79–87. [43] de Heer WA, Bacsa WS, Chatelain A, Gerfin T, HumphreyBaker R, Forro L, et al. Aligned carbon nanotube films: production and optical and electronic properties. Science 1995;268(5212):845–7. [44] Ska´kalova´ V, Kaiser AB, Woo Y-S, Roth S. Electronic transport in carbon nanotubes: from individual nanotubes to thin and thick networks. Phys Rev B 2006;74(8):085403–10. [45] Li Z, Kandel HR, Dervishi E, Saini V, Xu Y, Biris AR, et al. Comparative study on different carbon nanotube materials in terms of transparent conductive coatings. Langmuir 2008;24(6):2655–62. [46] Petrik L, Ndungu P, Iwuoha E. Hall measurements on carbon nanotube paper modified with electroless deposited platinum. Nanoscale Res Lett 2010;5(1):38–47. [47] Heinze S, Tersoff J, Martel R, Derycke V, Appenzeller J, Avouris P. Carbon nanotubes as Schottky barrier transistors. Phys Rev Lett 2002;89(10):106801–4. [48] McEuen PL, Fuhrer M, Park H. Single-walled carbon nanotube electronics. IEEE Trans Nanotechnol 2002;1(1):78–85. [49] Derycke V, Martel R, Appenzeller J, Avouris P. Controlling doping and carrier injection in carbon nanotube transistors. Appl Phys Lett 2002;80(15):2773–5.