Microscopic analysis of performance variations in carbon nanotube field emission cathodes: Implications for device optimization

Authors


Abstract

Carbon nanotubes (CNTs) are considered very promising for the realization of low-cost field emission electron sources. However, despite intensive research and development efforts, the fabrication of reliable CNT cathodes for high current density (>100 mA/cm2) applications remains a formidable challenge. In this study we use scanning anode field emission microscopy (SAFEM) to investigate the microscopic origins of macroscopic emission performance variations in chemical vapor deposition (CVD) grown CNT planar field emission cathodes. The field enhancement distributions are determined and the field emission properties of individual emission sites on the cathodes are probed. Contact I(V) measurements are carried out to estimate the resistance of individual emitters. The degradation behavior of individual sites is also studied and can be related with the macroscopic cathode performances. Scanning (SEM) and transmission electron microscopy (TEM) provide additional information on the contact and structural properties of the cathodes. Our results indicate that the sample macroscopic performances depend strongly on the individual emitter field emission properties in terms of maximum current before degradation and contact resistance.

1 Introduction

Electron field emission from carbon nanotubes (CNTs) has been intensively investigated for more than 15 years because of its high potential for novel applications such as field emission displays 1, microwave amplifiers 2, or cold cathodes for X-ray tubes 3. Their intrinsic high aspect ratio, chemical and mechanical stability and electronic properties make CNTs ideal candidates for electron field emitters. However, in spite of considerable interest and research, commercial realizations are still limited1 4. CNT field emitter fabrication generally encounters issues hampering large-scale industrial applications when high emission current densities – above 100 mA/cm2 – are required. These issues include: poor spatial emission homogeneity, lack of reproducible cathode performance, and lack of long-term operation stability. Despite very impressive recent progress in the development of new carbon based field emission cathodes, e.g., on flexible substrates, usually the emission current density is of the order of 1 mA/cm2 5. Therefore, successful industrialization of high current density field emission cathodes still necessitates a better understanding of the corresponding structural deficiencies and their impact on the emission performances.

High current densities are often required for applications other than displays. For instance, replacing the conventional thermionic cathode of a high power X-ray tube by a CNT-based emitter requires the CNTs to deliver 1 A from a 1 cm2 emitting area, at an electric field below 10 V/µm for 1000 h of tube operation, and in a rather harsh environment (pressure of 10−4 mbar under operation, presence of evaporated metals, risk of sputtering, etc.). A single CNT can deliver a maximum current of 10–100 µA without breaking down 6, therefore achieving a high emission current involves the simultaneous operation of several thousand of emitters under the action of a global electric field. This puts stringent requirements on the structural homogeneity and degradation hardness of the emitter ensemble.

Different approaches have been developed for the fabrication of high current CNT-based emitters, from low-cost continuous films of randomly oriented CNTs to arrays of single CNTs of precisely controlled dimensions. However, parameters often found in the literature to quantify the field emission performance – such as threshold voltage or low current I(V) characteristics (up to some µA) – give only very limited insight into the device properties and limitations. For instance, it has been observed that emission at the macroscopic level may be dominated by a small number of individual emitters within an ensemble of CNTs 7. Therefore, the failure of one of these emitting sites may have a dramatic impact on the global emission current. It is thus necessary to understand the breakdown mechanism and the difference between the various active emission sites. A better understanding of the macroscopic properties of a CNT-based cathode requires to characterize the whole distribution of the individual microscopic emitters using a statistical approach 8.

Electron emission from CNTs is well described by the Fowler–Nordheim (FN) relation 9. One should be aware that the FN-law describes the experimental emitter current correctly under three nontrivial assumptions. First the resistivity of the emitter is sufficiently small, second space-charge effects can be neglected (which is generally true), and third the emitter does not degrade or change during the measurement. Using the triangular barrier approximation, the current is then given by

equation image((1))

where A is the effective emitter surface area, ϕ its work function, β its field enhancement factor, E0 the applied electric field, me the mass of an electron, and ħ is the reduced Planck constant 8. From the FN Eq. (1) we learn that the emission current of a single emitter depends on A, β, and ϕ. A macroscopic field emission cathode is then characterized by the statistical distribution of A, β, and ϕ among the individual emitters – here CNTs – present on it. Due to the homogeneous composition of the CNTs, the variation of the work function can be assumed to be negligible and a fixed value can be assigned to ϕ. Furthermore as

equation image((2))

it follows that β is the dominant statistical parameter to characterize the ensemble of the emitters. For the geometry of the CNT, the field enhancement factor can be reasonably well approximated by the aspect ratio h/r. This simple approximation overestimates the field enhancement factor to some extent (by about 30%), however, in the β range under consideration, the dependency of the field enhancement is indeed linear with the aspect ratio 10. The β distribution, f(β), can then be used to describe at the macroscopic scale a field emission cathode consisting of an array of CNT emitters: f(β)dβ gives the number of emitters per unit area with a field enhancement factor in the interval [β, β + dβ] 8.

The local field enhancement β(x,y) can be measured using scanning anode field emission microscopy (SAFEM) 8. In this technique, the sample is scanned with a micro-tip anode allowing for field emission measurements with a few µm resolution. The β distribution can then be calculated by mapping the sample and collecting β values as a function of the spatial coordinates x and y. It is also possible to carry out single measurements on local sites. This characterization method has already been used to study CNT emitters 6, 11–13.

In this study, cathodes resulting from three different fabrication protocols were compared. The preparation, growth, and treatment conditions had been previously studied and optimized for each protocol but the macroscopic field emission properties of the corresponding samples still differed significantly: one fabrication process led systematically to cathodes with maximum current densities up to 250 times higher than those of samples produced by the other methods. In order to identify the origins of these strong performance variations, SAFEM was used to characterize CNT arrays representative of these three protocols. A comparison of the microscopic behavior of the samples was carried out and related to their respective macroscopic field emission properties. Scanning electron microscopy (SEM) and transmission electron microscopy (TEM) showed substantial differences between the samples depending on the fabrication process. The aim was to obtain a better understanding of these samples in order to identify possible routes toward optimization for high current density applications.

2 Experimental

2.1 Sample preparation

In this study three CNT arrays were investigated in order to compare different fabrication processes (preparation, growth method, and post-deposition treatment) that result in cathodes with markedly different field emission performances. Two samples were fabricated using a combined DC plasma/thermal chemical vapor deposition (CVD) growth method referred to as “Black Magic” (BM)2. This process consists in a variant of thermal CVD that includes a 60 s conditioning of the iron-based catalyst stack prior to deposition 14 by a 75 W, 15 kHz rectangular DC pulse ammonia plasma discharge and subsequent CNT growth from an ammonia/acetylene gas mixture. During growth (100 s growth time), the heater temperature was set at 800 °C (resulting in a temperature of 660 °C at the sample level), the NH3 flow at 692 sccm, and the C2H2 flow at 125 sccm. The third sample (referred to as microwave oxygen treated (MWO) sample later in this study) was produced by microwave plasma enhanced CVD (MW CVD) as described in Ref. 15. The microwave hydrogen plasma power was set at 1 kW and its frequency at 2.45 GHz. After a preheating time of 25 min, CNTs were grown for 6 min using a hydrogen/methylene gas mixture (H2 flow 170 sccm, CH4 flow 20 sccm), at a temperature of 645 °C. The MWO sample was treated ex situ after deposition by an inductively coupled high density radiofrequency (RF) oxygen plasma (Copra Cube, CCR Technology, Rheinbreitbach, Germany) 16 at 5 × 10−4 mbar for 30 s with an RF input power of 500 W, and a substrate DC bias of −300 V to etch the surface of the CNTs. This step aims at removing any amorphous carbon, decreasing the CNT density, and create some roughness on the CNT bundle top surface in order to increase the β factor. The two “Black Magic” samples were produced in the same batch under identical conditions. Then one of the samples (denoted black magic oxygen treated (BMO)) was exposed in situ to a 100 W, 15 kHz oxygen plasma at a pressure of 2.8 mbar for 30 s after deposition in an attempt to mimic the RF treatment step used for the MWO sample.

The CNTs were grown onto multilayer catalyst stacks deposited using a patterned mask on HF etched Si substrates (wafer resistivity as purchased: ∼1–6 Ω cm). The catalyst stacks had a 3 × 3 µm2 square shape and were spaced by 15 µm. After exposure to growth conditions, the catalyst stacks are routinely observed to become round. They were made of 14 nm TiN, 16 nm Ti, 164 nm Cu, and an intermixed 8 nm layer of Al and Fe for the BM and BMO samples, and of 24 nm Ti, 25 nm Cu, 4 nm Al, 7 nm Fe for the MWO sample. The layer thicknesses were measured by TEM after deposition. For the BM and BMO samples, the TiN, Ti, and Cu layers were deposited by sputtering and the Al and Fe layers by evaporation, whereas all layers were evaporated for the MWO sample. The sample characteristics are summarized in Table 1. Samples were analyzed before and after SAFEM characterization by SEM using a Philips XL-30 field emission gun microscope. Macroscopic field emission measurements were carried out in DC and pulsed modes (pulse width 100 µs, frequency 5 Hz) on a number of 5 × 5 mm2 cathodes produced in these conditions. Typical macroscopic emission currents in pulsed mode reached 10–100 mA for the BM samples, less than 1 mA for the BMO samples, and ∼200–250 mA for the MWO samples.

Table 1. Characteristics of the samples analyzed in this study.
samplegrowth methodpost-deposition treatmentcatalyst stacktypical FE currents from 5 × 5 mm2 arrays
BMthermal CVDno14 nm TiN, 16 nm Ti, 164 nm Cu, 8 nm Al and Fe10–100 mA
BMOthermal CVDin situ oxygen plasma14 nm TiN, 16 nm Ti, 164 nm Cu, 8 nm Al and Fe<1 mA
MWOMW CVDex situ oxygen plasma24 nm Ti, 25 nm Cu, 4 nm Al, 7 nm Fe200–250 mA

2.2 SAFEM setup

The SAFEM setup comprises a sample stage and a micro-tip which serves as an anode, placed in a vacuum chamber at ∼10−8 mbar. The sample stage consists of stepper-motor driven translation stages, where the xy directions are position encoded with an accuracy of 100 nm and a reposition accuracy of about 1 µm. The micro-tip is made of Pt/Ir, and has a 1 µm curvature radius with an opening angle of 60°. The sample is scanned by the micro-tip anode as field emission measurements are carried out. The current and voltage are controlled and measured using a Keithley 237 unit.

The SAFEM measurements were carried out using a specially designed LabView control program. The sample can be mapped in constant current mode (CCM), leading to a voltage map, or in constant voltage mode (CVM), leading to a current map. CCM was preferentially used in this study because it gives directly an inverted map of the field enhancement. Local I(V) curves in field emission mode were measured as follows. On the local emission site of interest, the tip emitter distance was determined by constant current V(z) measurements. In the next step, the tip was positioned at a precise height – typically several µm – above the emitter and the field emission current was measured for a given voltage range. The I(V) sweeps have been measured with progressively increasing current compliance levels in order to controllably follow the emitter degradation.

2.3 Data analysis

Data were analyzed using the software package Igor Pro 6.1.2.13. The β maps were calculated from the voltage maps using the following equation:

equation image((3))

where d is the tip cathode distance determined by V(z) measurements at constant current. Assuming that multiwall CNTs have a work function of 4.9 eV, the local electric field, E, to reach a 20 nA emission is 3900 V/µm 17, 18. This assumption can be done by considering that the CNT emitters have all the same emitting surface and geometry. From these β maps, β distributions were extracted by an automated search and listing of local maxima in the β(x,y) maps. They were subsequently fitted using a log–normal function as follows:

equation image((4))

where A0 is the amplitude parameter, θ the location parameter, µ the scale parameter, and σ is the shape parameter 19. f(β) is thus defined for β > θ. The maximum of the log–normal distribution is reached for equation image. However, it is important to keep in mind that the measured distribution may differ from the real one, as the low β part of the measured distribution depends more strongly on the resolution of the measurement. Nevertheless, the measured distribution can be very well described using the log–normal distribution. As the low β region of the distribution is likely to result from an artifact due to the measurements conditions, an exponential fit is also used to describe the high β tail of the measured distribution;

equation image((5))

The higher C1 and C2 are, the larger is the emitter density and the more homogeneous is the β distribution. I(V) curves in field emission mode were fitted using the FN model [Eq. (1)].

2.4 TEM characterization

TEM studies were performed using an FEI TECNAI F30ST TEM operated at 300 kV. The CNTs were put on a holey carbon foil supported by a copper grid by wiping the foil over the substrate. The cross-section samples were prepared using an FIB200 (FIB stands for Focused Ion Beam) and a Nova Nanolab 200 dualbeam. Before FIB preparation, a thin Pt layer was deposited using electron beam induced deposition (EBID). After EBID a 1.5 µm Pt layer was deposited in the FIB on the region of interest to protect the sample during FIB milling.

3 Results

The microscopic structure of the samples was first characterized by SEM and Fig. 1 shows representative SEM images. All samples exhibit a relatively good uniformity as seen in the low magnification images and dense bundles of CNTs were formed where the catalyst stacks had been deposited. The CNT height ranges from 6 to 8 µm for the BM sample, from 9 to 11 µm for the BMO sample, and from 5 to 6 µm for the MWO sample. The height difference between the BM and BMO samples may be explained by a lack of uniformity of the catalyst stack deposition on the Si wafer, or of the temperature profile in the CVD growth chamber. It may also result from continued CNT growth during the in situ oxygen treatment, where amorphous carbon acted as the carbon source. This amorphous carbon could have been present at the surface of the catalyst particles or the CNTs, or inside the catalyst particles. No significant microstructural differences in terms of CNT density or CNT diameter can be discerned between the three samples by SEM.

Figure 1.

SEM images of the CNT arrays considered in this study observed with secondary electrons before SAFEM characterization. (a, d) BM sample, (b, e) BMO sample, (c, f) MWO sample.

Figure 2 shows SAFEM maps of the three samples acquired under similar conditions. 300 × 300 µm2 areas were scanned at a constant current of 20 nA. β maps were calculated from these voltage maps and are also shown. For the BM sample an initial scan at 5 µm tip–sample distance produced some contacts with the sample during the scan. For this reason the distance was increased to 13 µm and the scan has been performed in a new region of the sample. As a smaller d results in a higher resolution scan, however, we have checked that the field enhancement distributions obtained from the 5 and 13 µm distance scans are indeed in agreement. For all three samples, the resolution was good enough to reveal the CNT bundle array structure. On the β maps, the emitter positions are indicated by crosses. These crosses are indeed regularly spaced along the x and y directions as expected from the patterning of the substrates, with one emission site detected per CNT bundle. This effect is particularly striking for the BMO sample, whose emitter positions match extremely well with the fabricated pattern of CNT bundle array (Fig. 2c and d). The detection of only one dominating emitter per bundle was universally observed on all samples and results from our cathode geometry.

Figure 2.

(online color at: www.pss-a.com) Voltage maps recorded at 20 nA (left) and corresponding β maps (right). (a, b) BM sample, (c, d) BMO sample, (e, f) MWO sample.

For each individual sample, it can also be seen from this data that emitters from different bundles exhibit a strong variation in field enhancement. This effect can be better analyzed using the β distributions determined from the β(x,y) maps, as shown in Fig. 3. Parameters of the log-normal and exponential fits are gathered in Table 2. The β distribution of the BMO sample is shifted toward lower values in comparison to the BM sample (equation image = 119 in the first case versus 166 in the second case), suggesting that the oxygen treatment may affect the strongest emitters. Regarding the homogeneity of the emitter strengths, the coefficient C2 from the exponential fit is the most suited to compare the samples. As already mentioned, the low β region is strongly affected by the scan resolution and therefore, the high β exponential tail is more relevant to describe the spread of the distribution. The BM and MWO samples show similar values for C2, which indicates that they are comparable in terms of homogeneity. However, C2 is higher for the BMO sample, suggesting that its distribution is somewhat more homogeneous than for the other two samples.

Figure 3.

(online color at: www.pss-a.com) β distributions calculated from the β maps shown in Fig. 2.

Table 2. Characteristic values of the β distributions shown in Fig. 3.
sampleA0 (cm−2)log–normal fitexponential fit
σΘµequation imageFWHMC1C2
BM3.419 × 106 ± 1.09 × 1050.59 ± 0.05107 ± 583 ± 5166884.77 × 105 ± 8 × 1030.01517 ± 8 × 10−5
BMO5.647 × 106 ± 1.06 × 1050.47 ± 0.0365 ± 467 ± 4119621.48 × 106 ± 3.5 × 1040.0233 ± 1.7 × 10−4
MWO3.840 × 106 ± 1.14 × 1050.70 ± 0.0459 ± 266 ± 299742.39 × 105 ± 2 × 1030.01626 ± 7 × 10−5

In Fig. 2 it can be observed that the emission sites are stochastically off-set from the regular grid position of the bundles. This off-set indicates the discrete nature of the emission site and suggests that in most of the bundles only one emitter carries the majority of the emission current. Furthermore the field enhancement (β) distribution resembles that of a random CNT film 11, the pillar structure allowing for a higher general β factor. From the β distributions, it is also clear that only a small fraction of the bundles will carry the majority of the current when the cathodes are operated, since the strongest emitters will be turned on first. As discussed above, differences are observed between the three samples but they are certainly too small to explain the strong variation in terms of macroscopic maximum current.

For each sample a number of individual emitters were locally probed by recording successive I(V) curves in field emission mode. Representative results are shown in Fig. 4 and reveal that the three samples have different micro-scale field emission properties. The BM sample is able to emit up to 1 µA per individual emission site and follows the FN model (Fig. 4a). When the voltage is further increased the emitter breaks down: the current suddenly decreases and when measurements are subsequently carried out, electron emission is still FN-type but the I(V) curve is shifted toward higher voltages with a larger threshold voltage. This suggests that emission within a bundle is dominated by one CNT and that, when this CNT degrades, the second strongest emitter then dominates the emission. The BMO sample shows poorer emission properties (Fig. 4b). The emission is FN-type only up to 0.1–0.5 µA. Above this range, significant deviation from the FN fit is observed. This phenomenon can be explained by the presence of a strong resistance along the CNTs and/or at the CNT/substrate interface. Contact I(V) curves were acquired to determine this resistance and led to values of 20–50 MΩ for the BMO sample. When the FN fits are modified according to these resistance measurements 20, a very good agreement with the experimental data is then obtained (resistor-limited FN fits).

Figure 4.

(online color at: www.pss-a.com) I(V) curves recorded in field emission mode on the different samples and corresponding FN fits and resistor-limited FN fits. (a) BM grown sample, (b) BMO sample, (c) MWO sample.

The individual emitters of the MWO sample exhibit much better field emission properties (Fig. 4c). An FN behavior is observed up to ∼10 µA with no deviation of the experimental I(V) curves from the fits. For the 6th and 7th curves (IV6 and IV7), some slight resistor limitation appears at 10 µA, and around 100 µA the emitter breaks down. The 8th curve (IV8) shows also some resistor limitation above ∼5 µA. A 210 ± 3 kΩ resistance was measured on this emitting site by contact I(V) curves. Other contact measurements were carried out on various emitters of this sample and led to values ranging from 50 to 300 kΩ. The FN fit of curve IV7 was modified accordingly (resistor limited FN fit) but did not result in a better agreement with the experimental data in the high current region. However, this may be due to the fact that, in this range, the emitter is also about to break down – as indicated by the abrupt decrease of the emission current – and emission might not be anymore FN-type. It is anyway worth noting that, in comparison to the BM and BMO samples, the individual emitters of the MWO sample exhibit outstanding emission properties and very low contact resistances.

The MWO sample was further investigated by recording voltage maps on the same region, first at 20 nA and then at 10 µA in order to emulate the situation of high emission current density. The corresponding β maps were calculated and results are depicted in Fig. 5. As the current is increased, more emitters are detected (59 instead of 39). The voltage and β maps at high current show no signs of degradation and the emitters can be individually resolved. This is remarkable as each emitter was forced to produce 10 µA of current! 59 emitters were detected on a surface of 2 × 10−4 cm2, which translates to an emission site density of 2.95 × 105 cm−2. Taking into account an emission current of 10 µA per emitter, this amounts to a potential emission current density of 2.95 A/cm2.

Figure 5.

(online color at: www.pss-a.com) Voltage maps recorded on the MWO sample at 20 nA and 10 µA and corresponding β maps.

SEM characterization was carried out after the SAFEM experiments, especially on the parts which were scanned at very high single emitter currents (up to 30 µA: results are not shown here). The SEM images shown in Fig. 6 illustrate the different degradation behaviors. For the BM and BMO samples, a large number of CNT bundles are missing or damaged, as can be seen from the low magnification images (Fig. 6a and b). At closer inspection, the damaged CNT bundles seem to have exploded: they are deformed or even destroyed and CNT debris is scattered across the substrate surface. The substrate has even melted in some regions, suggesting a strong release of energy (Fig. 6d and e). On the other hand, the substrate of the MWO sample has not been damaged. However, a large area of this sample has been affected by the measurements (Fig. 6c). In the damaged region, all the bundles seem to have been pulled in the same direction (Fig. 6c) and some have even been torn off their metallic stack. It is particularly notable that, in contrast to the BM and BMO samples, the CNT bundles have not exploded but have kept their integrity (Fig. 6f).

Figure 6.

SEM images of the CNT arrays after SAFEM characterization observed with secondary electrons. (a, d) BM sample, (b, e) BMO sample, (c, f) MWO sample.

The microscopic structure of BM, BMO, and MWO samples was further investigated by TEM. It revealed that the CNT microstructure is different depending on the sample type, as shown in Fig. 7. The CNTs of the BMO sample are straight, well-graphitized but present a bamboo-type wall structure: the inner graphitic planes are joined in a cup shape (Fig. 7a). Similar results were obtained for BM samples. This type of defect is present every ∼20 nm along the same CNT, indicating a large defect density. On the other hand, the CNTs of the MWO sample are curly but exhibit well-aligned, defect-free multiwalls (Fig. 7b). All sample types display nanotubes with a large variety of diameter and number of walls.

Figure 7.

Bright field TEM images of the CNT samples. Inset: high resolution (HR) TEM images. (a) BMO sample. The arrows show the positions of defects referred to as bamboo structure. (b) MWO sample. The HRTEM image reveals the CNT multiwall structure.

In order to gain more insight about the differences in contact resistance and/or adhesion, the composition and microstructure of the multilayer catalyst stacks were also studied by high angle annular dark field (HAADF) scanning TEM and energy dispersive X-ray spectroscopy (EDX). Results are presented in Fig. 8 and reveal strong differences between the BM and the MWO samples. The low magnification HAADF images show that, on the contrary to the MWO sample, the stack of the BM sample is discontinuous and poorly adhering to the Si substrate. For this latter sample, the different layers can be distinguished and the stack composition was probed by EDX at a location where the stack was still attached to the substrate (Fig. 8b). A thin native silicon oxide layer (1.7 ± 0.2 nm) was detected at the bottom of the stack, then TiN (6.7 ± 0.2 nm), Ti (9.7 ± 0.2 nm), Cu (169.0 ± 0.2 nm) and, on top, Al and O (11.9 ± 0.2 nm). Some Cu has diffused into the Si substrate, indicating that the TiN diffusion barrier was not efficient. From the composition profile, the Ti and Cu layers also appear to have intermixed because of Cu diffusion into the Ti layer. The Al layer has oxidized and no Fe was detected, suggesting that it was all used to catalyze the CNT growth (tip-growth mode). The multilayer catalyst stack of the MWO sample shows a two-layer structure separated by a thin dark layer (Fig. 8d). EDX analysis identifies it as a 3.3 ± 0.2 nm interfacial oxide layer. A 52 ± 6 nm TiSix layer has been formed at the bottom, whereas the top layer consists of a mixture of Si, Ti, O, Al, Cu, and Fe and is 24 ± 6 nm thick.

Figure 8.

HAADF scanning TEM cross-sectional images of the CNT samples. (a, b) BM sample, (c, d) MWO sample.

4 Discussion

This study emphasizes how the fabrication method (catalyst preparation, CNT growth, post-deposition treatment) can significantly affect the field emission properties of CNT-based emitters. The results presented here indicate that the macroscopic field emission performances of CNT cathodes are correlated to the microscopic field emission properties of their individual emitters. If we compare the samples used in this study, we indeed see that the emission current per individual emission site before degradation reaches up to 10–100 µA for the MWO sample, 1 µA for the BM sample, and only 0.1–0.5 µA for the BMO sample. This is in agreement with the respective macroscopic emission currents of the samples: the MWO 5 × 5 mm2 cathodes can produce emission currents 10–250× higher than the BM and BMO ones. Moreover, although the in situ oxygen treatment results in a narrower β distribution, it does not improve the field emission properties but, on the contrary, deteriorates them. The individual emitter stability plays also an important role in the device behavior. When comparing the BM, BMO and the MWO samples, we see that the latter one is able to emit repeatedly high currents without breaking down, whereas the BM and BMO samples degrade earlier. In short, the microscopic field emission properties of the individual emitters are a good indication of the sample macroscopic performances.

The microscopic field emission properties can in their turn be related to the respective sample contact resistance (20–50 MΩ for the BMO sample, in comparison to 50–300 kΩ for the MWO sample). Thus, the contact quality appears to be a critical parameter for good field emission performances since the presence of a high resistance will induce a deviation from the FN model hindering efficient electron emission. Besides, because of these high contact resistances, the power dissipated during electron emission is considerably higher for the BM and BMO samples than for the MWO sample: the power density in a CNT bundle emitting 1 µA can reach several hundreds of W/cm2. This may explain the explosive degradation mechanism and the melting of the Si substrate for the BM and BMO samples, as observed by SEM. This is in contrast with the low contact resistance MWO sample, for which breakdown was much less catastrophic.

The CNT and multilayer stack microstructure characterization provides additional information. A high density of bamboo-like defects in the CNT multiwalls, as observed for the BM and BMO samples, could lead to poorer electrical conductivity. However, though the MWO sample CNTs exhibit well-aligned graphitic walls, they are less straight. This may suggest that the CNT structure is not the limiting factor for field emission, provided that the CNTs show sufficiently graphitized walls. The lack of substrate adhesion and continuity of the BM sample stack may result in high resistances. Moreover, contrary to what was expected in this case, the TiN layer did not prevent the diffusion of Cu into the Si substrate. On the other hand, the MWO sample stack adheres well to the substrate, remains continuous after growth and the formation of TiSix layer could improve its electrical conductivity 21. This suggests that, although the stack composition leads to satisfying results in the MWO case, further investigation is required to obtain good stack adhesion and conductivity, when using the thermal CVD growth method.

This study represents an important step toward the identification of the sample characteristics that will lead to good device properties. It was previously commonly accepted that a narrow β distribution was one of the most important factors to obtain good field emission cathodes. A sample with a homogeneous distribution would indeed have its emitters all turned on at the same threshold and emitting the same amount of current. This would lead to a better control of the emission current and less risk of damaging the sample. Therefore, much of previous research focused on the geometry of the emitters 22, 23. This study shows however that, though a homogeneous f(β) is a desirable feature, it is not necessarily the most significant parameter. The maximum current that can be drawn from the individual emitters before breakdown and/or resistor limitation, and the contact resistance are more determining factors for good macroscopic performances. Therefore these features must be optimized for technological applications. The geometry of the samples investigated here represents also an interesting solution as the bundle array structure (i) can be easily implemented, (ii) allows for high β factors in comparison to those achieved using randomly distributed CNT films, and (iii) provides a protection against a total bundle breakdown, as for each bundle other CNTs can start emitting if the strongest emitter is degraded.

5 Conclusions

SAFEM characterization was carried out on CNT-based electron emitter arrays grown using different methods – thermal CVD and microwave plasma CVD – and fabrication processes. Valuable information on the microscopic properties of the samples was obtained and strong differences were observed depending on the preparation, growth method and post-deposition treatment. With an emission current of 10–100 µA from individual emitters, the MW CVD sample shows better features than the arrays produced by thermal CVD. These results were found to correlate with the respective macroscopic field emission properties. Possible explanations were identified such as the presence of a strong resistance (20–50 MΩ) along the CNTs and/or at the substrate interface for the thermal CVD samples. Different degradation characteristics were also observed by SEM after SAFEM characterization with a catastrophic explosive breakdown for the thermal CVD samples. TEM investigation revealed significant differences between the two types of samples in terms of CNT and stack microstructure. This study gives a detailed insight into the functioning and properties of CNT-based electron emitters. It allows for the identification of crucial parameters – the maximum current that can be drawn from individual emitters and the contact resistance – controlling the sample macroscopic performances. This will definitely help the future development and optimization of CNT arrays for high current applications.

Acknowledgements

The research leading to these results has received funding from the European Community's Seventh Framework Programme FP7/2007-2013 under grant agreements no 215399 (FINELUMEN) and No. 228579 (TECHNOTUBES). Z. X. Chen's research is jointly supported by the NSFC (61071027), by MEC (Y02020023901057), and by STDSC (2009GZ0005).

Biographical Information

Sarah Berhanu graduated from the Ecole Nationale Supérieure des Mines de Nancy (France) in 2006 with a major in materials science. The same year she obtained a Master of Research in Nanomaterials and in 2010 she received her PhD, both degrees from Imperial College London (UK). Her thesis was entitled “Nanostructured templates for donor/acceptor interface engineering in organic solar cells” and focused on the use of colloidal crystals for the growth of organic semiconductor nanocomposites. From 2010 to 2011 she worked as a Marie Curie Fellow in Philips Research Laboratories in Aachen (Germany) on carbon nanotube synthesis for innovative field emission cathodes. Her research interests lie in nanostructured materials for optoelectronic applications.

original image

Biographical Information

Oliver Gröning studied physics and mathematics at the University of Fribourg (Switzerland) where he graduated in experimental physics in 1994. He stayed at the University of Fribourg to pursue his PhD thesis where he investigated the field emission mechanisms and properties of carbon thin films and carbon nanostructures. In the following he developed the scanning anode field emission microscope (SAFEM) to relate the macroscopic emission properties of planar cathodes to the microscopic properties of individual emitters. Since 2001 Dr. Gröning is researcher at the Swiss Federal Laboratories for Materials Science and Technology, where he is leading a research group in the field of surface nanostructures. His main research focus resides in molecular self-assembly, structural and electronic properties of carbon nanostructures and surfaces of complex metallic alloys.

original image

Ancillary