Facile fabrication of carbon nanotube network thin film transistors …€¦ · Keywords: carbon...

Int. J. Nanotechnol., Vol. 14, Nos. 1/2/3/4/5/6, 2017 505

Copyright © 2017 Inderscience Enterprises Ltd.

Facile fabrication of carbon nanotube network thin film transistors for device platforms

H.Y. Zheng and N.O.V. Plank* School of Chemical and Physical Sciences, Victoria University of Wellington, MacDiarmid Institute for Advanced Materials and Nanotechnology, School of Chemical and Physical Sciences, Victoria University of Wellington, P.O. Box 600, Wellington, New Zealand Fax: +64-4-4635237 Email: [email protected] Email: [email protected] *Corresponding author

Abstract: Carbon nanotube network thin film transistors with channel lengths of 10–40 µm are fabricated from a nanotube buckypaper with 99% semiconducting carbon nanotubes by surfactant free solution routes. The carbon nanotubes are suspended in 1,2-dichlorobenzene by ultrasonication at concentrations of 7.9 µg/ml, 3.4 µg/ml and 2.6 µg/ml, resulting in carbon nanotube field effect transistors with tube densities ranging from 5 tube/µm2 up to 32 tube/µm2. The device percolation threshold is 11 tube/µm2 with optimum device performances of on/off current ratio 7200, mobility 0.55 cm2/V·s and Vth = +1 V. Devices with tube density above the percolation threshold are typically metallic-like. By encapsulating with poly-4-vinylphenol cross-linked with poly-(melamine-co-formaldehyde), the hysteresis becomes independent of channel length and the device performance approaches the best devices at the percolation threshold.

Keywords: carbon nanotubes; thin film field effect transistor; solution processing techniques; hysteresis; encapsulation; PVP; poly-4-vinylphenol.

Reference to this paper should be made as follows: Zheng, H.Y. and Plank, N.O.V. (2017) ‘Facile fabrication of carbon nanotube network thin film transistors for device platforms’, Int. J. Nanotechnol., Vol. 14, Nos. 1/2/3/4/5/6, pp.505–518.

Biographical notes: H.Y. Zheng received her BS in Microelectronics from Sichuan University of China in 2010. She started her PhD at Victoria University of Wellington, New Zealand in September 2011 holding a Victoria University of Wellington and the Chinese Government joint Scholarship. Her interested research area is nanomaterials based electronics.

N.O.V. Plank received her BSc (Hons) in Astrophysics, MSc in Microelectronics and PhD in Microelectronics all from the University of Edinburgh. Her PhD work on Carbon nanotube Functionalisation was under the supervision of Prof Rebecca Cheung. Following this she spent three years as a postdoc in the group of Professor Sir Mark Welland at the Cambridge University Nanoscience Centre working on the use of ZnO nanowires for optoelectronic devices in collaboration with the Cavendish Optoelectronics group. She arrived in New Zealand in 2009 and worked as a postdoctoral

506 H.Y. Zheng and N.O.V. Plank

researcher on thin film fabrication of rare earth nitride materials with Dr. Ben Ruck, before embarking on an independent programme as a recipient of a Foundation for Research Science and Technology postdoctoral fellowship. She is a Lecturer in Physics at the Victoria University of Wellington and has established the Cleanroom Fabrication Facility and the Nanomaterials Device Group. Her research interests include carbon nanotube network properties, ZnO nanowires and plastic electronics for sensors. She is also a Principal Investigator of the MacDiarmid Institute for Advanced Materials and Nanotechnology.

This paper is a revised and expanded version of a paper entitled ‘Carbon nanotube field effect transistors platforms for sensors’ presented at Advanced Materials and Nanotechnology 7, Nelson, 8–12 February, 2015.

1 Introduction

There is a great deal of interest in the fabrication of robust and reliable carbon nanotubes (CNTs) networks as thin film field effect transistors (CNT FETs) for device applications [1–11]. CNTs can be thought of as a rolled up sheet of graphene. Depending on how the tube has rolled up a pristine carbon nanotube can show either metallic or semiconducting behaviour [12]. Normally for a given growth process 1/3 of the CNTs are metallic and 2/3 of the CNTs are semiconducting [12]. Efficient methods to separate out the semiconducting and metallic CNT components from the mixed samples have been developed [13] and it is now possible to buy pre-sorted CNTs as a bucky paper from Nanointegris, with 99% and above semiconducting CNTs components. As a consequence of this, CNT FETs with network thin films are now becoming a viable technology. However, there is as always a trade-off between on/off current ratio and mobility [8–11,14] due to the many junctions present in a network CNT FET. Much of the device performance is based on the ability to fabricate high quality films with a suitable density to make the device percolation threshold [7–9], while avoiding any effects of short circuits through any residual metallic path way.

It is with this in mind that we look towards simple fabrication routes of CNT network FET devices, using a simple suspension of CNT buckypaper and no surfactants. In our work here we suspend pre-purchased sorted CNTs buckypaper from Nanointegris with 99% semiconducting components in 1,2-dicholorobenzene by ultrasonication to investigate the device thresholds. The suspension concentrations investigated are 7.9 µg/ml, 3.4 µg/ml and 2.6 µg/ml and the time of substrate submersion ranges from 5 min up to 16.5 h. For both the 3.4 µg/ml and 2.6 µg/ml the device percolation threshold is 11 tubes/µm2 and on/off current ratios on the order of 103 can be achieved. However, tube density above the percolation threshold shows metallic-like behaviour. We report that by encapsulating the metallic-like CNT FETs with cross linked poly-4-vinylphenol (PVP) polymer, the on/off current ratio is improved and approaches the performance of the devices fabricated at the percolation threshold. These results open the way to make use of CNT FETs fabricated from unknown concentrations of CNTs in the original suspensions. We put forward that CNT bucky paper suspensions in 1,2-dichlorobenzene can be a fast and efficient means of fabricating CNT thin film FETs with device performances comparable to the current literature [9,11,14].

Facile fabrication of carbon nanotube network thin film transistors 507

2 Experiment

2.1 Carbon nanotube suspension

To produce uniform films of CNTs, we first of all create a suspension of CNTs in anhydrous 1,2-dichlorobenzene (DCB) from Sigma-aldrich. Small amounts of 99% semiconducting CNTs from a NanoIntegris (IsoNanotube-S 99%) bucky paper were weighed out in sequence using a Metter Toledo UMX-5 precision scale with a resolution of 0.1 µg, a repeatability of 0.5 µg and a maximum linearity error of 3 µg. A precise volume of DCB was added to the CNTs using a precision micropipette to create suspensions with concentrations of 7.9 µg/ml, 3.4 µg/ml and 2.6 µg/ml CNTs in DCB. The CNTs were then dispersed into the DCB using the Sonorex ultrasonic water bath for 0.5 h up to 1 h to create the suspension to the point where there are no obvious particles visible by the naked eye. The bottles all show suspensions with a light pink colour containing 99% CNTs in a DCB suspension with CNT concentrations of 7.9 µg/ml, 3.4 µg/ml and 2.6 µg/ml respectively, Figure 1.

Figure 1 Segments of CNT buckypaper (IsoNanotube-S 99%) are suspended in DCB at concentrations: (a) 7.9 µg/ml, (b) 3.4 µg/ml, (c) 2.6 µg/ml. After 0.5 h to 1 h sonication, no particles are visible to the naked eye (see online version for colours)

2.2 CNT thin films

Devices were all prepared with a global back gate geometry, using SiO2/Si substrates from Silicon Quest International Inc. with highly doped p-type Si and an oxide thickness of 100 nm. SiO2/Si substrates are cleaned by sonicating in acetone for 3 min and followed by thorough rinsing in IPA and dried in clean N2. In order to functionalise the SiO2 surface, a polydimethylsiloxane (PDMS) (Sylgard 184) stamping method is used [15] to transfer a thin layer of 2-thiolpyridine onto SiO2 surface. To do so, 2-thiolpyridine (Sigma Aldrich) is dissolved in ethanol at 10 mg/ml concentration while PDMS is cleaned in 50W O2 plasma (Plasma Etcher-50) for 1 min. The thiolpyridine ethanol solution is spin coated onto clean PDMS at 2000 rpm for 40 s. SiO2/Si substrates are then directly contacted with the thiolpyridine covered PDMS for 5 min. The thiolpyridine functionalised SiO2 substrates are then submerged into the respective premade CNT DCB suspensions for 5 min up to 16.5 h. The SiO2/Si substrates are then taken from CNT suspension and cleaned in ethanol for 5 min to remove excess solvents before being dried in clean N2.


2.3 Device fabrication

After submersion, the entire SiO2 surface is coated with a uniform CNT thin film. To remove CNTs at the edges of the SiO2 and control the area of CNTs in the channel the films are O2 plasma etched through lithographically defined areas using an Oxford Plasmalab 80 Plus etcher. The pressure is set to 600 mTorr, power is set to 200 watts, O2 flow is set to 40 sccm and an etch time of 3 min is used. The residual photoresist is then removed. In our experiment, 300 µm by 300 µm areas of CNT thin films are left at controlled locations on the substrate surface. These CNT regions then form the active channel of the FET. We use the Karl Suss MJB3 UV lamp mask aligner for photolithography. Metallisation is achieved using the Angstrom Engineering Nexdep Evaporator. Contacts of 3 nm Cr and 35 nm Au are deposited directly onto the CNT thin films as the source and drain electrodes, while highly doped p-type Si coated with Cr/Au forms the global gate. After a lift-off process, the samples consists of CNT FETs with channel lengths varying between 10 µm, 20 µm, 30 µm to 40 µm and fixed channel width of 100 µm.

For the devices encapsulated with PVP, we spin coat poly-4-vinylphenol (PVP) cross linked with poly-(melamine-co-formaldehyde) (PMF) in propylene glycol monomethyl ether acetate (PGMEA) onto a CNT FET at 5000 rpm for 40 s. PVP (average Mn ~ 25,000), PMF (average Mn ~ 432) in 1-butanol and PGMEA solvent are all purchased from Sigma Aldrich. The ratio of PVP in PGMEA is 20% measured by weight. The CNT FET is baked at 200°C on a hot plate in ambient air for 30 min to cross link the polymer layer [16,17].

2.4 Characterisation techniques

After the CNT FETs are fabricated, scanning electron microscopy (SEM) is performed using a Jeol JSM 6500F to determine the tube density of CNT thin films. To characterise the electrical performances of CNT FETs, we use a 4156C Agilent parameter analyser with Rucker and Kolls probes station. Images are analysed using Image J. Veeco Dektak 50 is used to measure material thickness.

3 Results and discussions

In our study CNT FETs are fabricated using three different concentrations of CNT suspension in DCB: 7.9 µg/ml, 3.4 µg/ml and 2.6 µg/ml. Usually 1 h of sonication time is required to produce uniform CNT suspensions. The CNTs in DCB remain in a suspension and do not visibly aggregate over a period of a few days. In order to assemble films of CNTs from the DCB suspension functionalisation of the SiO2 surface is required. We use a thiolpyridine stamping method, where the thiol bonds with the SiO2 leaving the surface coated with a uniform distribution of pyridine. The CNTs in the DCB suspension are then able to attach to the pyridine via pi-pi interactions, resulting in a uniform film of CNTs on the substrate. The concentration of the thiolpyridine is crucial, as if not high enough little CNT attachment takes place at the surface. However if it is too high, the thiolpyridine can stack causing the pi-pi interactions between neighbouring pyridines to be in the direction perpendicular to the plane of the substrate [15]. When the functionalised substrates are submerged into the CNT suspensions, CNTs immediately


deposit onto the substrate surface, and uniform films can be achieved in as little as 5 min of submersion time.

In order to control the CNT FET performance we have investigated the role of tube density in the CNT thin films, based on the initial suspension concentration and the submersion time. In the CNT FET devices being investigated we have a large channel length of 20 µm which requires many CNTs to bridge the electrode gap. A typical SEM image of a CNT FET channel is shown in Figure 2. To determine the tube density we use a counting method based on analysing the SEM images in Image J and applying a simple stick model [7,18], giving an estimate of tube densities and length distribution.

Figure 2 SEM image of a CNT thin film deposited from the 3.4 µg/ml DCB suspension for 10 min

The tube density for submersion times ranging from 5 min up to 16.5 h for the 7.9 µg/ml, 3.4 µg/ml and 2.6 µg/ml CNT suspensions have been calculated. The tube density for the CNT thin films as a function of submersion time for the three different concentrations of CNT submersions are shown in Figure 3, and the percolation threshold for conduction through the FET channel is indicated with an arrow. Figure 3 also indicates whether percolation was achieved and if the resulting CNT FET showed metallic or semiconducting characteristics. CNT FETs made by 7.9 µg/ml and 3.4 µg/ml CNT DCB suspensions have faster ‘deposition’ rate as compared to the 2.6 µg/ml CNT DCB suspension. The apparent CNT film deposition rates are similar at any given submersion time in the range of 5–60 min for the 7.9 µg/ml and 3.4 µg/ml, resulting in tube densities in the range of 5 tubes/µm2 to 21 tubes/µm2. It is difficult to determine the tube density from the 7.9 µg/ml concentration after 60 min of submersion as the CNTs form a dense mat coverage on the sample surface. For the 3.4 µg/ml suspension the tube density continues to increase up to a maximum of 32 tubes/µm2 after 16.5 h of submersion. In contrast the deposition rate of CNTs from the 2.6 µg/ml suspension is roughly flat with submersion times from 5 min up to 60 min, all resulting in tube densities of around 7 tubes/µm2. At 90 min submersion time the tube density reaches 11 tubes/µm2 and a maximum of only 17 tubes/µm2 is reached after a full 16.5 h submersion. We find the


percolation threshold of the 7.9 µg/ml suspension to be 5 tubes/µm2 whereas for the 3.4 µg/ml and 2.6 µg/ml the percolation threshold is 11 tubes/µm2, which can easily be explained by an increase in average length due to bundling for the 7.9 µg/ml. The simple stick model used cannot always clearly determine whether there is a single CNT or a bundle of CNTs that is forming the network. This is an obvious weakness of the stick model but it allows us to make a meaningful comparison.

Figure 3 The dependency of tube density on submersion time for CNT thin films fabricated by submersion in CNT DCB suspensions at 7.9 µg/ml (squares), 3.4 µg/ml (circles) and 2.6 µg/ml (triangles) concentration. The arrows point out the percolation threshold at 11 tubes/µm2 for both the 3.4 µg/ml and 2.6 µg/ml concentrations. The colours indicate whether the resulting CNT FETs show metallic, or semiconducting behaviour. As well as the case when the network is not percolative (see online version for colours)

To understand how the bundles and/or rope-like structures influence the CNT FET performances, we compare their electrical properties when the tube density is 11 tubes/µm2 (as determined by SEM analysis) for the thin films fabricated from all three initial concentrations in Figure 4. All three devices are fabricated using the same device processing steps and metallisation. The channel length is 20 µm and width is 100 µm, with a total area of CNT conductive film of 300 µm × 300 µm as patterned by the etch step described in the experimental section. The transistor fabricated from the 3.4 µg/ml suspension was submerged for 10 min and the transistor fabricated from the 2.6 µg/ml suspension was submerged for 90 min to reach their percolation threshold 11 tubes/µm2 and give on/off current ratios of 103 and above for both devices. The transistor fabricated from the 7.9 µg/ml suspension which is also submerged for 10 min results in tube density of 11 tubes/µm2 across the thin film. It is immediately apparent that the device fabricated from the 7.9 µg/ml shows very little gate dependence and the high current, in the region of µA, is significantly larger than the tens of nA reached by the other two devices. Therefore we state that at a 7.9 µg/ml concentration, the CNTs are not well suspended in DCB by the sonication procedure. Instead we have made films of bundles of carbon nanotubes that although well dispersed, will maintain significant metallic CNT contamination from the 1% metallic component in the bucky paper (IsoNanotube-S 99%).


In contrast, suspensions made from 3.4 µg/ml and 2.6 µg/ml suspensions have been well dispersed, and although some bundling may occur, the metallic components are insignificant and good semiconducting properties are achieved (Figure 3). Although initially this can be seen as a detrimental set of outcomes when making a CNT FET, it is clear that the thin films of CNTs must be highly uniform and well dispersed for a 1% component of metallic tubes to short circuit the FET over a 20 µm gap. As is evident in the discussion above a small variation in the CNT concentration, from 2.6 µg/ml up to 3.4 µg/ml in the initial CNT DCB suspension can result in widely varied electrical performance.

Figure 4 The electrical performance of CNT FETs when tube density is 11 tubes/µm2 for all three thin films. The CNT FETs are fabricated from CNT DCB suspensions with 7.9 µg/ml after 10 min submersion (squares), 3.4 µg/ml after 10 min submersion (circles) and 2.6 µg/ml after 90 min submersion (triangles). L = 20 µm and W = 100 µm with Vd = –0.1 V (see online version for colours)

Key performance indicators for CNT FETs are the on/off current ratio of the devices and the mobility. In order to calculate the on/off current ratio in devices where there is no clear off, we have taken the ‘off current’ at Vg = +3 V which is above Vth = +1 V; and the ‘on current’ at Vg = –15 V gives the maximum current in this measurement range, where Vg is the gate voltage and Vth is the threshold voltage. To calculate hole mobility we refer to the standard CNT FET modelling procedure given in equation (1) assuming the device to be in the linear regime [7–11,19,20].

d 1d

ds

g ds

I LV W CV

µ =

(1)

where L is channel length, W is channel width, Vd is source drain voltage, Vg is gate voltage, Ids is output current and C is SiO2 capacitance per unit area. Here the channel length and width are L = 20 µm, W = 100 µm, and the thickness of the SiO2 is 100 nm. We are making the assumption that the dominant conduction paths in these CNT FETs depend on CNTs directly bridging the channel, with no stray conduction paths through the CNTs 300 × 300 µm region of CNT films that were fabricated. The standard MOSFET mobility model stated in equation (1) assumes 100% coverage of well aligned CNTs over the channel area and takes the Capacitance per unit area as [20]


0 oxCd

ε ε= (2)

where εox of SiO2 = 3.9, εo is the vacuum permittivity = 8.85 × 10–12 F/m and dielectric thickness d = 100 nm. However, the actual CNT coverage is not 100% in our percolative networks. The current model would therefore underestimate the actual mobility of the CNT FET devices here. We can recalculate the mobility using a modified capacitance based on the CNT coverage [8–11,19].

( ) 1

0 100

0

sinh 2 /1 1 ln2Q ox

dC

C Rπ

πε ε π

−

− Λ Λ = + Λ

(3)

where R is radius of CNT and CQ is the quantum capacitance per unit area [10,11,19]. For aligned CNTs Λ0 is the space between each tube [8,9,19] which for a random network CNT film can be estimated using the tube density [10,11]. Taking the average CNT diameter to be 1.5 nm, the resulting mobility using the standard capacitance equation (2) and the modified capacitance, equation (3) are shown in Table 1 for the 20 µm channel CNT FETs fabricated from the 3.4 µg/ml suspension.

Table 1 The mobility of the 20 µm channel CNT FET fabrication from the 3.4 µg/ml suspension

Tube density (tubes/µm2) Mobility (cm2/V·s) C from equation (2)

Mobility (cm2/V·s) C from equation (3)

11 0.39 0.55 14 1.66 2.14 16 2.54 3.15 21 6.99 8.13 27 8.78 9.77 32 5.62 6.10

There is an increase in the value of the mobility calculated for the CNT FETs using the modified capacitance from equation (3), as we are able to more accurately determine the actual contribution from the CNTs present. The approximate quantities of CNTs are 22,000 and 42,000 at density of 11 tubes/µm2 and density of 21 tubes/µm2 respectively. Giving CNT coverage of 0.99% for the 11 tubes/µm2 CNT FET and 1.89% for the 21 tubes/µm2 CNT FET. However, owing to the random network of the CNT films we may underestimate the actual mobility as the model assumes alignment. We also do not consider any resistive junctions from the CNT-CNT interfaces in the network.

The on/off current ratios, and hole mobility based as a function of the carbon nanotube densities are shown in Figure 5, for the devices fabricated from the 3.4 µg/ml suspension. These values were calculated using the modified capacitance values determined by equation (3) which we assume to be the best fit for our devices. This assumption has also been made by other teams working on randomly orientated CNT FETs [9,11,14].


Figure 5 The on/off current ratio and mobility as a function of tube density for CNT FETs fabricated from CNT thin films from the 3.4 µg/ml concentration CNT DCB suspension when tube density changes from 11 tubes/µm2 to 32 tubes/µm2 are shown on the left and right scale respectively. In on/off current ratio calculation, Ion is from Vg = –15 V and Ioff is from Vg = +3 V (see online version for colours)

Although single CNTs have measured mobility of up to 4000 cm2/V·s [21], our 20 µm CNT FETs have mobility ranging from 0.55 cm2/V·s to 9.77 cm2/V·s. These mobility values are very close to other random network CNT FETs with 99% semiconducting components. For example in the work by LeMieux et al. mobilities of 0.9–17 cm2/V·s were achieved when the tube density increases from 5 tubes/µm2 to 18 tubes/µm2 for 50 µm length and 1000 µm width devices [14]. The other major performance indicator for CNT FETs is the on/off current ratio. As observed in our plots, Figure 4, we have reliable on currents in the 10–8 Amps range and off currents around 10–12 Amps giving us on/off current ratios of 7200 for the 11 tubes/µm2 devices. Similar results are also found in works by LeMieux which has shown on/off current ratio decreases from 104 to 102 as mobility increased [14]. In work by Rouhi et al. [9] random network CNT FETs show mobility increases from 1 cm2/V·s to 90 cm2/V·s, where the channel length is 200 µm and the widths range from 10 µm to 100 µm. Their highest mobility devices, 90 cm2/V·s is achieved at 100 µm length and 200 µm width and high tube densities of 100 /µm2. However, the mobility comes at the expense of the on/off current ratio which decreases from 105 to less than 10 as the mobility increased [9]. Our devices with lengths on the order of 20 µm and low densities between 11 tubes/µm2 up to 32 tubes/µm2 compare favourably to literature data within the same device geometry and tube density regions [9,11,14].

CNT FETs fabricated with tube density around the percolation threshold have by far the highest on/off current ratios and realising tube densities just around the percolation threshold is the preferred option for the optimum device performance. Figure 6 shows the channel dependent FET characteristics for the device at the percolation threshold tube density of 11 tubes/µm2 fabricated from the 3.4 µg/ml CNT suspension. The on/off current ratio does not show any change with channel length. These characteristics show that the CNT FETs fabricated from the DCB suspensions are suitable for device platform applications. There is a commonly observed hysteresis for CNT FETs as indicated by the observed shift in threshold voltage between the forward and reverse transfer characteristics [22–25]. The mechanism for hysteresis is thought to be due to hydroxyl bonds from water molecules in ambient environment which absorbed onto the SiO2


surface. These hydroxyl bonds remain on the SiO2 surface and after application of gate voltage, charge injection can occur between the SiO2 and the CNTs [22–25].

Figure 6 Transfer characteristics for a CNT FET at the device percolation threshold of 11 tubes/µm2. L = 10 µm, 20 µm, 30 µm and 40 µm on the same chip (see online version for colours)

Although these devices show suitable on/off current ratios and working currents it may not always be possible to use a precision balance to ensure being at the correct CNT concentration for device fabrication. Furthermore, the error in the measurement of the balance with resolution of 0.1 µg, a repeatability of 0.5 µg and a maximum linearity error of 3 µg can be a large contribution to variations during measurement on different days. All of the CNT FETs fabricated from the 7.9 µg/ml DCB suspensions result in large tube density and metallic conduction properties that block the field dependence of the devices. Figure 7(a) shows pristine CNT FET transfer characteristics with different channel length 10–40 µm. The device with tube density 21 tubes/µm2 was fabricated by 60 min submersion in 3.4 µg/ml CNT in DCB suspension. As a means to improve the device performance we have encapsulated the CNT FET by spin coating a layer of PVP polymer (approximately 800 nm) and cross linking via a 200°C bake as described in the experimental section. The device characteristics for the encapsulated CNT FET are shown in Figure 7(b). Across all four devices with channel lengths of 10–40 µm on the same chip show a transition from previous metallic conduction towards semiconductor characteristics. The electrical properties of PVP encapsulated CNT FET are similar to the devices fabricated at the percolation threshold with on/off current ratio of 1360 and mobility of 3.7 cm2/V·s for the 20 µm channel FET. The pre-encapsulated CNT FET had on/off current ratio of 6.42 and mobility of 8.13 cm2/V·s showing an increases of 211 times for the on/off current ratio and a decrease in mobility by a factor of two post PVP encapsulation. Similar encapsulation methods for CNT FETs have been investigated to either reduce hysteresis [23] or enhance device performance [26].

The PVP coating appears to block some conductive (metallic-like) pathways in the percolative CNT network. The PVP must also similarly block some semiconducting pathways yet as short circuits from the 1% metallic CNT pathways dominated the FET transfer characteristics the effect is far more pronounced for the metallic pathways. The effective tube density of percolative network CNTs has been reduced bringing the


devices closer in performance to those at the percolation threshold. Owing to the limited effect on the semiconducting components we conclude that a simple blocking of the inter-tube metallic contacts is the dominant effect. It is therefore shown that PVP coating of non-optimised CNT FETs is a useful fabrication route to create a CNT FET device platform, provided the PVP layer does not interfere with the final functionality of the device.

Figure 7 Transfer characteristics for the CNT FET with tube density of 21 tubes/µm2 fabricated from the 3.4 µg/ml CNT suspension. L = 10 µm, 20 µm, 30 µm and 40 µm on the same chip. (a) Pristine CNT FET with the CNT channel directly exposed to air. (b) The same CNT FET as in figure (a) now encapsulated by a thin layer of crossed link PVP (see online version for colours)

4 Conclusion

In this paper, we have successfully fabricated reproducible uniform CNT thin film transistors by using simple surfactant free solution fabrication routes. The concentration of CNTs suspended in DCB investigated were 7.9 µg/ml, 3.4 µg/ml and 2.6 µg/ml. We were able to control the density of the CNT thin films by altering the submersion of functionalised SiO2 substrates in the CNT DCB suspensions. In doing so tube densities between 5 tubes/µm2 to 32 tubes/µm2 were achieved. The device percolation threshold was found to be 11 tubes/µm2 when L = 20 µm for CNT FETs fabricated from both the 3.4 µg/ml and 2.6 µg/ml suspensions. At around the percolation threshold, all devices


have 103 on/off current ratio and electrical properties from length scale 10-40 µm are independent with channel length. We have demonstrated that high yield CNT FETs can be fabricated from a simple solution process from the 3.4 µg/ml and 2.6 µg/ml CNT DCB suspensions over a wide range of submersion times and processing parameters. The best devices achieve good FET performance with on/off current ratios >103, Vth = +1 V and mobility of 0.55 cm2/V·s at 20 µm length scale. However, it is also clear that successful fabrication sits on a knife edge, with all CNT FET devices fabricated from the 7.9 µg/ml CNT DCB suspension and CNT FET fabricated from the 3.4 µg/ml CNT DCB suspension after 60 min submersion being short circuited by metallic conduction paths from the 1% metallic CNT component in the original buckypaper. Encapsulating those CNT FETs with a spin coated layer of PVP results in 211 times higher on/off current ratio compared to the pristine CNT FETs with devices performance approaching that of devices fabricated at the percolation threshold.

Acknowledgements

H.Y. Zheng thanks for the Victoria University of Wellington and the Chinese Government joint Scholarship. N.O.V. Plank thanks the Foundation for Research Science and Technology project number VICX0911. Both N.O.V. Plank and H.Y. Zheng thank The MacDiarmid Institute for Advanced Materials and Nanotechnology. Thank you to Vladimir Bubanja and Greg Reid for providing access to use the precision scale in Callaghan innovation.

References 1 Opatkiewicz, J.P., LeMieux, M.C., Liu, D., Vosgueritchian, M., Barman, S.N., Elkins, C.M.,

Hedrick, J. and Bao, Z. (2012) ‘Using nitrile functional groups to replace amines for solution-deposited single-walled carbon nanotube network films’, ACS Nano, Vol. 6, No. 6, pp.4845–4853.

2 Pimparkar, N., Cao, Q., Kumar, S., Murthy, J.Y., Rogers, J. and Alam, M.A. (2007) ‘Current–voltage characteristics of long-channel nanobundle thin-film transistors: a ‘bottom-up’ perspective’, IEEE Electron Device Lett., Vol. 28, No. 2, pp.157–160.

3 Snow, E.S., Novak, J.P., Campbell, P.M. and Park, D. (2003) ‘Random networks of carbon nanotubes as an electronic material’, Appl. Phys. Lett., Vol. 82, No. 13, pp.2145–2147.

4 Cao, Q. and Rogers, J.A. (2009) ‘Ultrathin films of single-walled carbon nanotubes for electronics and sensors: a review of fundamental and applied aspects’, Adv. Mater., Vol. 21, No. 1, pp.29–53.

5 Sun, D., Timmermans, M.Y., Tian, Y., Nasibulin, A.G., Kauppinen, E.I., Kishimoto, S., Mizutani, T. and Ohno, Y. (2011) ‘Flexible high-performance carbon nanotube integrated circuits’, Nat. Nanotechnol., Vol. 6, No. 3, pp.156–161.

6 Duan, Y., Holmes, N.E., Ellard, A.L., Gao, J. and Wei, X. (2013) ‘Solution based fabrication and characterization of a logic gate inverter using random carbon nanotube networks’, IEEE Trans. Nanotechnol., Vol. 12, No. 6, pp.1111–1117.

7 Choi, W.J., Park, D.W., Park, S., Jeong, D.W., Yang, C.S., Kim, B.S., Kim, J.J. and Lee, J.O. (2013) ‘High-performance carbon nanotube network transistors fabricated using a hole punching technique’, J. Mater. Chem. C, Vol. 1, No. 26, p.4087.


8 Rouhi, N., Jain, D. and Burke, P.J. (2011) ‘High-performance semiconducting nanotube inks: progress and prospects’, ACS Nano, Vol. 5, No. 11, pp.8471–8487.

9 Rouhi, N., Jain, D., Zand, K. and Burke, P.J. (2011) ‘Fundamental limits on the mobility of nanotube-based semiconducting inks’, Adv. Mater., Vol. 23, No. 1, pp.94–99.

10 Wang, C., Zhang, J., Ryu, K., Badmaev, A., De Arco, L.G. and Zhou, C. (2009) ‘Wafer-scale fabrication of separated carbon nanotube thin-film transistors for display applications’, Nano Lett., Vol. 9, No. 12, pp.4285–4291.

11 Wang, C., Zhang, J. and Zhou, C. (2010) ‘Macroelectronic integrated circuits using high-performance separated carbon nanotube thin-film transistors’, ACS Nano, Vol. 4, No. 12, pp.7123–7132.

12 Dresselhaus, M.S., Dresselhaus, G. and Saito, R. (1995) ‘Physics of carbon nanotubes’, Carbon, Vol. 33, No. 7, pp.883–891.

13 Arnold, M.S., Green, A.S., Hulvat, J.F., Stupp, S.I. and Hersam, M.C. (2006) ‘Sorting carbon nanotubes by electronic structure using density differentiation’, Nat. Nanotechnol., Vol. 1, No. 1, pp.60–65.

14 LeMieux, M.C., Sok, S., Roberts, M.E., Opatkiewicz, J.P., Liu, D., Barman, S.N., Patil, N., Mitra, S. and Bao, Z. (2009) ‘Solution assembly of organized carbon nanotube networks for thin-film transistors’, ACS Nano, Vol. 3, No. 12, pp.4089–4097.

15 Plank, N.O.V., Ishida, M. and Cheung, R. (2005) ‘Positioning of carbon nanotubes using soft-lithography for electronics applications’, J. Vac. Sci. Technol., B, Vol. 23, No. 6, pp.3178–3181.

16 Chiu, C.J., Pei, Z.W., Chang, S.P. and Chang, S.J. (2012) ‘Influence of weight ratio of poly(4-vinylphenol) insulator on electronic properties of InGaZnO thin-film transistor’, Journal of Nanomaterials, Vol. 2012, pp.1–7.

17 Kim, S.H., Nam, S., Jang, J., Hong, K., Yang, C., Chung, D.S., Park, C.E. and Choi, W-S. (2009) ‘Effect of the hydrophobicity and thickness of polymer gate dielectrics on the hysteresis behavior of pentacene-based field-effect transistors’, J. Appl. Phys., Vol. 105, No. 10, p.104509.

18 Kocabas, C., Pimparkar, N., Yesilyurt, O., Kang, S.J., Alam, M.A. and Rogers, J.A. (2007) ‘Experimental and theoretical studies of transport through large scale, partially aligned arrays of single-walled carbon nanotubes in thin film type transistors’, Nano Lett., Vol. 7, No. 5, pp.1195–1202.

19 Sarker, B., Shekhar, S. and Khondaker, S.I. (2011) ‘Semiconducting enriched carbon nanotube aligned arrays of tunable density and their electrical transport properties’, ACS Nano, Vol. 5, No. 8, pp.6297–6305.

20 Liyanage, L., Lee, H., Patil, N., Park, S. and Mitra, S. (2012) ‘Wafer-scale fabrication and characterization of thin-film transistors with polythiophene-sorted semiconducting carbon nanotube networks’, ACS Nano, Vol. 6, No. 1, pp.451–458.

21 Javey, A., Guo, J., Wang, Q., Lundstrom, M. and Dai, H. (2003) ‘Ballistic carbon nanotube field-effect transistors’, Nature, Vol. 424, No. 6949, pp.654–657.

22 Kar, S., Vijayaraghavan, A., Soldano, C., Talapatra, S., Vajtai, R., Nalamasu, O. and Ajayan, P.M. (2006) ‘Quantitative analysis of hysteresis in carbon nanotube field-effect devices’, Appl. Phys. Lett., Vol. 89, No. 13, p.132118.

23 Kim, W., Javey, A., Vermesh, O., Wang, Q., Li, Y. and Dai, H. (2003) ‘Hysteresis caused by water molecules in carbon nanotube field-effect transistors’, Nano Lett., Vol. 3, No. 2, pp.193–198.

24 Ong, H.G., Cheah, J.W., Zou, X., Li, B., Cao, X.H., Tantang, H., Li, L-J., Zhang, H., Han, G.C. and Wang, J. (2011) ‘Origin of hysteresis in the transfer characteristic of carbon nanotube field effect transistor’, J. Phys. D: Appl. Phys., Vol. 44, No. 28, p.285301.


25 Qu, M., Qui, Z., Zhang, Z., Li, H., Li, J. and Li, S. (2010) ‘Charge-injection-induced time decay in carbon nanotube network-based FETs’, IEEE Electron Device Lett., Vol. 31, No. 10, pp.1098–1100.

26 Zhao, J., Lin, C., Zhang, W., Xu, Y., Lee, C.W., Chan-Park, M.B., Chen, P. and Li, L-J. (2011) ‘Mobility enhancement in carbon nanotube transistors by screening charge impurity with silica nanoparticles’, J. Phys. Chem. C, Vol. 115, No. 14, pp.6975–6979.

Facile fabrication of carbon nanotube network thin film transistors …€¦ · Keywords: carbon...

Documents

Transcript of Facile fabrication of carbon nanotube network thin film transistors …€¦ · Keywords: carbon...