Summary…

The CNTs grown from an Fe-Ti codeposited catalyst at 700 ℃ in thermal CVD exhibited a partially immersed structure with larger contact area to the Ti film as compared with those for the conventional ones. This amazing structure provided better adhesion and lower contact resistance between the CNTs and the substrates. With better adhesion, the abrupt decreases in emission current were suppressed remarkably and the characteristics of field emission could be stabilized to obtain smooth J-E curves. The proposed samples exhibited a relative stable emission current density with 30 mA/cm² at 7.7 V/μm for 3,600 sec.

Furthermore, the reduction of contact resistance could diminish the generation of Joule heat to prevent the CNTs from heat induced destruction. By utilizing the Fe-Ti codeposited catalyst layer, the reliability of CNTs could be also improved for the application as electron emitters.

In addition, the proposed structure was utilized to growth CNTs at low temperatures of 550 ℃ in thermal CVD. The higher growth rate was exhibited in the proposal samples. The length of conventional samples was about 437 nm, but that of the proposed ones was about 2.13 μm. In the mean while, the less length variation of CNTs was also shown in the proposed samples than those in the conventional ones. It was due to smaller and more uniform catalytic particles in the proposal samples than those in the conventional ones. For field emission measurement, the current density of the proposed samples was much less than 10 μA/cm² at 6.25 V/μm. The current density of the proposed samples was as high as 3.36 mA/cm² at 6.25 V/um, and the turn-on field was 4.44 V/μm.

Chapter 3

The Field Emission Characteristics of Carbon Nanotube Pillars Synthesized from an Fe-Ti Codeposited Catalyst

3.1 Introduction

Carbon nanotubes(CNTs), a self-organized nanoscale structure, have a variety of applications such as field emission displays (FEDs) [3.1], X-raytubes [3.2], flat lamp [3.3]

and backlight units (BLU) for liquid crystal displays (LCDs) [3.4], because of their excellent field emission characteristics. Especially, BLU has been remarkably investigated for a recent few years because of development of large area LCD TV. The schematic of a typical BLU is shown in Fig. 3-1(a)[3.5] including light source, reflector, light guide, diffuser, and brightness enhancement film (BEF). The light source can be an incandescent light bulb, light emitting diode (LEDs), cold cathode fluorescent lamp (CCFL), hot cathode fluorescent lamps (HCFL).

All the backlights employ a diffuser and a BEF. The diffuser posited between the light source and the display panel is used to scatter the light for display uniformity. The BEF is used to enhance display brightness. The cost structure of materials for TFT-LCDs is described as Fig.

3-1(b). As Fig. 3-1(b) shows, the cost of BLU module for 15 inch monitor is 23 %, and the cost for 30 inch LCD TV monitor is as high as 38 %. They imply that the reduction of BLU cost is an important issue. CNTs are thus utilized to BLU not only for the reduction of cost but also for less power consumption, optical films needless, no toxic chemicals, and super color performance [3.6].

Many synthesis methods for CNTs have been investigated and reported since the discovery of CNTs. It is indispensable to lower the turn on field (which is defined as the electronic field to reach current density with 10 μA/cm²) and threshold field (which is defined

as the electronic field to reach current density with 10 mA/cm²) to achieve practically applicable field emission emitters that operate with low power consumption. For CNTs-FEA, the high density of CNTs can provide a great deal of field emission sites which can raise the emission current density (due to the increase of the emission area, α) but the density of CNTs will affect the field enhancement factor (β) which is also strongly relative to the emission properties of CNTs. For CNTs with high density, the screening effects reduce the field enhancement factor (β), therefore, suppress the field emission current density, as shown in Figure 3-2 [3.7]. Obviously, it is important to obtain an optimized density of CNTs to improve the field emission properties, such as turn-on field, threshold field, and emission current density, as shown in Figure 3-3 [3.8]. Well control of density and surface morphology of CNTs are thus required for applications in the near future. To effectively control the density of CNTs and surface morphology, the CNT pillars have been investigated. According to the prediction of Nilsson et al.[3.7], the field emission will become maximum when the interpillar distance is about two times the height of the pillar [3.9].

Uniform and small catalyst nanopartilces after pretreatment are known to be the key for growing highly aligned CNTs. It is attributed to the lasting van der Waals force since these uniform nanoparticles could lead to the grown CNTs with the equal growth rate. Therefore, the CNT pillars were easily synthesized from uniform and small catalyst nanoparticles after pretreatment.

3.2 Motivation

The pillar-like CNTs are well to control the density and the morphology of CNTs. It has been reported that the screening effect of CNTs can be effectively reduced by the density control of the pillars. Therefore, the field emission characteristics can be enhanced from the compromise of the screening effect and emission sites. In Chapter 2, we utilized an Fe-Ti thin

film whose weight percentage of Fe was 64 % as a catalyst layer, the nanoparticles after pretreatment were more uniform and smaller. As a result, it could be desirable for as many CNTs per unit surface area as possible to grow with the equal growth rate. The growing CNTs can support each other due to the lasting van der Waals force to obtain the highly aligned CNTs. Therefore, the CNT pillars could be easily synthesized by the proposed method. By using CNT pillars as light source for BLUs in TFT-LCDs, high brightness and excellent uniformity could be achieved. As a result, the diffuser and BEF could not be needed for BLUs in TFT-LCDs. Therefore, the CNT pillars could be potentially applied in the BLUs for TFT-LCDs to reduce the material cost.

In this chapter, the pillar arrays of aligned CNTs were fabricated with Fe-Ti codeposited catalyst to form a device with area of 0.02 cm². By adjusting the ratio of the interpillar spacing (R) to pillar height (H) with fixed diameter, an optimized field emission was obtained.

In the mean while, the mechanism of Fe-Ti codeposited catalyst for CNTs growth was more complete via the study for the effects of growth time on the length of CNT pillars.

3.3 Experimental Procedures

Here, we also used two different catalyst thin films, conventional and proposed ones, (as the same as Chapter2). For the conventional samples, a 5-nm-thick Fe layer was deposited as the catalyst of CNTs. On the other hand, Fe co-deposited with Ti was utilized as the catalyst of CNTs for the proposed samples. The weight percentage of Fe in the co-deposited layer for the proposed samples was about 64 % with constant quantity of Fe as compared with the conventional ones. The formation of the patterns was as shown in Fig. 3-4, A photoresist was spin-coated on an N-type Si (100) substrate and the emitting sites were defined by the mask which has several 6-μm-diameter circles with interspacing of 12, 15, 20, 25, 30, and 35 μm by photolithography, as shown in Fig. 3-4(a). Afterward, a thin Ti film (50 nm) was deposited on

the Si substrate as the buffer layer which had excellent adhesion to the Si substrate, as shown in Fig. 3-4(b). A catalytic film was deposited directly on the photoresist-patterned Si substrate by dual E-gun evaporation system, as shown in Fig. 3-4(c). Then the patterns were formed after removing the photoresist by lift-off method as depicted in Fig. 3-4(d). Finally, the CNT pillars were grown selectively by thermal CVD system, as shown in Fig. 3-4(e).

The samples were pretreated at 700 ℃ in N2 and H2 (500/100 sccm) for 8 min, and then grew CNTs at 700 ℃ with C2H4, N2, and H2 (20/500/100 sccm) for different time from 8 to 120 min. The growth condition of CNTs in the thermal CVD system was shown in Fig. 3-5.

After the synthesis of CNTs, the samples were analyzed with scanning electron microscopy (SEM, Hitachi S-4700) to observe the density and the morphology of CNTs. The catalyst nanoparticles after pretreatment were analyzed by atomic force microscopy (AFM).

The field emission properties of CNTs were characterized by a high–vaccum measurement environment with a base pressure of 5×10^-6 Torr. Cathode contact was made directly on the wafer. A glass plate coated with indium-tin-oxide (ITO) was positioned 150 um above the tip of CNTs as an anode. All cables were shielded except for the ground return path to the power source. The emission current densities of CNTs were measured as a function of applied electron field, using Keithley 237 high voltage units as DC source and Keithley 238 high current units as ground source. The measurement was auto-controlled by the computer with IEEE 488 interface.

3.4 Results and Discussion

3.4.1 The Effect of Growth Time on the Length of Carbon Nanotube Pillars

The rate of growth under given reaction conditions is an important issue concerning CNTs growth: how long can CNTs be grown, and how fast do they grow? For the investigation of the issue and the mechanism of the CNTs growth with conventional and

proposed catalyst films, we measured lengths of the CNT pillars as a function of growth time.

The cross-section views of the CNT pillars for the conventional and proposed samples with different growth time were taken by scanning electron microscope and displayed in Fig. 3-6 and Fig. 3-7, correspondingly. The lengths of the CNT pillars synthesized from the conventional method for different time were about 5.6 µm for 8 min, 15.7 µm for 15 min, 22.4 µm for 30 min, 47 µm for 60 min, and 69.8 µm for 120 min. On the other hand, the lengths of the CNT pillars grown from the proposed method for different time were about 4.48 µm for 8 min, 12 µm for 15 min, 19.4 µm for 30 min, 52.2 µm for 60 min, and 118 µm for 120 min. The relationship between the length and the growth time for both samples were plotted in Fig. 3-8. As shown in Fig. 3-8, the growth rate of CNT pillars in conventional samples seemed to saturate when the growth time was over 60 min and the growth rate of CNT pillars in proposed samples sustained almost a constant in 120 min. For the conventional sample, the growth rate of the CNT pillars was about 0.78 μm/min in the first 60 min, and appeared to decrease to 0.39 μm/min in the last 60 min. For the proposed samples, the length of CNT pillar increased linearly with growth rate of 0.98 μm/min.

The observation of CNT pillars for the conventional samples revealed growth and

saturation steps. It could be possible that the inactivation of the metal catalyst nanoparticles included overcoating with carbon or conversion of the metal into a metal carbide or other non-catalytic form [3.10]. Slowing or complete stoppage of nanotube growth with increasing growth time has been attributed to catalyst deactivation, overcoating, or coalescence.

From Fig. 3-8, the growth rate of CNT pillars for the conventional samples was slightly faster than that for the proposed ones in the initial few minutes, about first 15 min. It could be explained by the possible mechanism as shown in Fig. 3-9. The conventional samples had more carbonaceous diffusion path due to more nanoparticle surfaces exposed to the reaction gases in the first few minutes. In the contrast, the proposed samples had a lot of nanoparticles

which were merged in the Ti thin. There were less carbonaceous diffusion path for the proposed samples. As a result, the conventional samples had higher growth rate than the proposed ones in the initial few minutes. After that, the catalyst nanoparticles were covered by the CNTs no matter the nanoparticle was in the bottom or top of CNTs. The CNTs growth rate was dominated by the carbonaceous volume diffusion but the surface diffusion. The smaller nanoparticles had shorter volume diffusion path, and resulted in higher growth rate of CNTs pillars. The proposed samples had smaller nanoparticles, as shown in Fig. 3-10.

Therefore, the growth rate of CNT pillars for the proposed samples increased, and was higher than that for the conventional ones.

For the proposed samples, the grown CNTs were denser and aligned, as shown in Fig.

3-11. It could be due to that the proposed samples had smaller nanoparticles. It confirmed our previous opinion which was the grown CNTs per unit surface area with the equal growth rate by using an Fe-Ti catalyst layer for CNT pillars synthesis.

3.4.2 The Optimization of Spacing to Height Ratio for Carbon Nanotube Pillars

The fabricated CNT pillars in the proposed samples for growth time of 15 min were shown by SEM images in Fig. 3-12. The pillars were aligned perpendicular to the substrate, and their height (H) was about 12 μm. There were six different spacing (R) between pillars of 12, 15, 20, 25, 30, and 35 μm. The R-to-H ratio (R/H) was 1, 1.25, 1.67, 2.08, 2.5, and 2.92, respectively. Each pillar can be regarded as an individual emitter since the electron filed emission can be neglected inside the pillar due to the screening effects. The emitter device was diode structure with the area of 0.02 cm². The emission current versus voltage (I-V plot) of the pillar arrays was shown in Fig. 3-13(a) with the spacing 150 um between the anode and the cathode. It showed that the emission current was as high as several million Ampere at the operating voltage. The corresponding Fowler-Nordheim (FN) plots were shown in Fig. 3-13(b)

and the linearity of the F-N plot confirmed the field emission phenomenon.

The emission current of device was divided by the device area of 0.02 cm² to get turn-on field and threshold field of one device and the results were shown in Table 3-1. The turn-on field (Eon) and the threshold field (Eth) were defined as the field for current density of 10 μA/cm² and for 10 mA/cm², respectively. The curve of turn-on field versus R was plotted in Fig. 3-14 and showed that the turn-on field remain almost a constant for R/H ratio larger than 2.5. As the simulation results shown in the work of Nilsson, et al., the screening effect is greatly reduced when the interspacing of emitters is two times of its height. The local enhancement factor will not be improved obviously even the interspacing between emitters was increase greatly. It also hints that the turn-on field will not be decreased remarkably with increasing the interpillar spacing more.

The reliability of the pillar array was determined by a stress test at voltage with 800 V (5.33 V/μm) for 1 hour. The current versus time plot was shown in Fig. 3-15(a). The coefficient of standard variation (CV) and the mean current density (Jmean) were shown in Table 3-2. The CV was defined as the standard variation to the mean value of current. The lower CV means higher reliability of field emission for pillar-like CNTs. Table 3-2 revealed that the lowest CV (20.59 %) and the highest Jmean (18.94 mA/cm²) occurred in R of 30 μm.

The Jmean versus R was plotted in Fig. 3-15(b). As shown in this figure, the emission current increased rapidly with enlarging R from 12 μm to 25 μm and then increased slowly even a slightly decreased with larger R. The increasing of emission current resulted from the suppression of screening effect by enlarging the R, however, the improvement in screening effect was getting unobvious and emission areas was also getting small that caused a trade-off for the total emission current of device. The trade-off between the suppression of screening effect and the reduction of emission sites leaded to an optimal R to obtain a maximum emission current density and here was 30 μm, about 2.5 times of the H, in our experiments.

The fluorescent images of field emission with R of 30 μm under the stress test were shown in Fig. 3-16. It obviously observed that the CNT pillars had high brightness even through a period time of 1 hr, as shown in Fig. 3-16(b). In addition, the field emission fluorescent images of the CNT pillars with H of 12 μm and R of 30 μm and the full plane CNTs with emission area of 0.01 cm² for different voltages were shown in Fig. 3-17 and Fig. 3-18. The brightness was enhanced with the increasing voltage form 400 V to 800 V. The brightness of the CNT pillars was much higher than that of the full plane CNTs under the equal voltage.

The field emission current density versus time of CNT pillar and that of full plane CNTs for different voltages were plotted in Fig. 3-19(a) and 3-19(b), respectively. They obviously showed the current density of the CNT pillar was much higher than that of the full plane CNTs under the equal voltage. The plot of the mean current density versus field under field emission stress test for 120 sec for the CNT pillars and the full plane CNTs was shown in Fig.

3-20. The mean current density increased by the increasing field.

The fabricated CNT pillars in the proposed samples for growth time of 8 min were shown by SEM images in Fig. 3-21. The height (H) of the CNT pillars was about 6.5 μm.

There were also six different spacing (R) between pillars of 12, 15, 20, 25, 30, and 35 μm.

The R-to-H ratio (R/H) was 1.84, 2.31, 3.07, 3.85, 4.62, and 5.38, respectively. The emitter device was also a diode structure with the area of 0.02 cm². The emission current versus voltage (I-V plot) of the pillar arrays was shown in Fig. 3-22(a) with the spacing 150 um between the anode and the cathode. It showed that the emission current was several micron Ampere at the operating voltage. The corresponding Fowler-Nordheim (FN) plots were shown in Fig. 3-22(b). The results of the field emission were shown in Table 3-3. An optimal R to obtain a maximum emission current density and here was 15 μm, about 2.31 times of the H (6.5 μm) for the growth time of 8 min due to the trade-off between the suppression of screening effect and the reduction of emission sites. The curve of maximum field emission

current density versus interpillar spacing for the CNT pillars was plotted in Fig. 3-23.

3.5 Summary

The CNT pillars synthesized by using the proposed method exhibited a linear growth

rate of 0.98 μm/min in 2 hours but the growth rate of the CNT pillars in the conventional samples tended to saturate at 0.39 μm/min after about 40 min. This phenomenon might be due to the smaller dimension of nanoparticles in the proposed samples could hold its activity at the same temperature better than those in the conventional samples. Although, the length of CNT pillars for the conventional sample was longer than those for the proposed ones due to more carbonaceous diffusion paths in the initial few minutes. Moreover, the CNTs in the proposed samples also showed a straighter morphology as compared with those in the conventional ones which might result from that the high density of nanoparticles in the proposed samples restricted the growth direction of the carbon nanotubes. It is helpful to form pillars with better uniformity in direction and then gain a more uniform emission current.

Additionally, we showed that the R/H played a crucial role for the field emission properties. The optimal interpillar spacing of 30 μm and 15μm were also found for the pillars with H of 12 μm and 8μm, respectively, to obtain a largest emission current density. They were also found that a trade-off between the suppression of screening effect and the reduction of emission area caused by decreasing the R was the main factor for the optimization of the R.

The optimal R/H was about from 2.3 to 2.5. For the CNT pillars with growth time of 15 min, a low turn-on field of 1.01 V/μm was obtained for the CNT pillars with R of 30 μm (H=12μm). An excellent reliability for the CNT pillars was also shown in R/H of 2.5 at 800 V

The optimal R/H was about from 2.3 to 2.5. For the CNT pillars with growth time of 15 min, a low turn-on field of 1.01 V/μm was obtained for the CNT pillars with R of 30 μm (H=12μm). An excellent reliability for the CNT pillars was also shown in R/H of 2.5 at 800 V