Journal of Infrared, Millimeter, and Terahertz Waves

, Volume 30, Issue 12, pp 1338–1350

Carbon Nanotube Bundle Array Cold Cathodes for THz Vacuum Tube Sources


    • Jet Propulsion LaboratoryCalifornia Institute of Technology
  • Risaku Toda
    • Jet Propulsion LaboratoryCalifornia Institute of Technology
  • Robert H. Lin
    • Jet Propulsion LaboratoryCalifornia Institute of Technology
  • Anna Liao
    • Jet Propulsion LaboratoryCalifornia Institute of Technology
  • Michael J. Bronikowski
    • Jet Propulsion LaboratoryCalifornia Institute of Technology
    • Atomate Corporation
  • Peter H. Siegel
    • Jet Propulsion LaboratoryCalifornia Institute of Technology
    • Beckman Institute, California Institute of Technology

DOI: 10.1007/s10762-009-9547-x

Cite this article as:
Manohara, H.M., Toda, R., Lin, R.H. et al. J Infrared Milli Terahz Waves (2009) 30: 1338. doi:10.1007/s10762-009-9547-x


We present high performance cold cathodes composed of arrays of carbon nanotube bundles that routinely produce > 15 A/cm2 at applied fields of 5 to 8 V/µm without any beam focusing. They have exhibited robust operation in poor vacuums of 10-6 to 10-4 Torr- a typically achievable range inside hermetically sealed microcavities. A new double-SOI process was developed to monolithically integrate a gate and additional beam tailoring electrodes. The ability to design the electrodes for specific requirements makes carbon nanotube field emission sources extremely flexible. The lifetime of these cathodes is found to be affected by two effects: a gradual decay of emission due to anode sputtering, and catastrophic failure because of dislodging of CNT bundles at high fields ( > 10 V/µm).


Field emissionCarbon nanotubeCNTsNanoklystronVacuum tubeHigh frequency sources

1 Introduction

1.1 Background

Efficient electron sources are the fundamental components of- (i) miniature vacuum tube sources for high frequency applications [1], (ii) many analytical instruments used for elemental and mineralogical analyses [2], and (iii) vacuum microelectronic devices that are radiation-insensitive and high-temperature tolerant for extreme environments electronic applications. Each application requires an electron source with specifically designed beam forming optics and emission densities that span a range of tens to hundreds of amperes per sq. cm. State-of-the art thermionic cathodes are ill-suited for miniature instruments because of their bulkiness, high temperature operation, and high power consumption. Conversely, state-of-the-art cold cathodes [3] based on atomically sharp micromachined tips [4, 5] are highly susceptible to poisoning when operated in non-UHV (10-8 to 10-9 Torr) environments. This is an important limitation because the micromachining techniques such as lithography, etching, and wafer bonding used to create vacuum cavities in miniature devices are not able to sustain vacuums better than 10-6 Torr because of small volumes containing relatively large outgassing surfaces. This vacuum tends to get worse during operation as well as due to aging because of its static nature (without continuous pumping). Moreover, presence of microcomponents inside these cavities creates a high potential for virtual leaks and surface contamination. While incorporating getters help to some extent, it does not address the fabrication process-induced inability to achieve UHV. Interest in high current density electron sources that are robust to poor vacuums and are capable of operation at low voltages is motivated by the re-emerging vacuum tube technology for high frequency sources. Intense research efforts continue with the goal to realize portable, narrow band, coherent CW sources of tunable RF power in the submillimeter wave frequency bands (300–3000 GHz) for active (and passive) sensor systems based on vacuum tube technology. One such approach based on the reflex klystron principle, called the Nanoklystron [1] merges cold cathode and silicon micromachining techniques. A simplified beam analysis performed using relations well known in previously published work [6] indicated current density requirement of ~ 1 kA/cm2 for a 500 V beam to produce ~ 3 mW output power at 1200 GHz. The device dimensions such as the beam tunnel diameter of 20 μm, the cavity diameter of 80 μm, and the overall device footprint of < 1 mm2 posed new challenges to develop a cathode technology suitable for such small devices. In addition to being small they are required to deliver tens to hundreds of amperes per sq. cm and be robust to poor vacuums.

1.2 Carbon nanotube field emitters

A new class of field emitters was identified when carbon nanotubes (CNTs) were discovered [7]. These tubular structures can be synthesized with diameters in the nanometer range, which naturally offer sharp tips for field emission. The ideal field emission follows the Fowler-Nordheim (F-N) relation [3] as shown below.
$$ \ln \left( {\frac{J}{{V^2 }}} \right) = \ln (a) - \frac{b}{V} $$
Where, J is the emission current density (A/cm2); V: biasing voltage in volts, and a, b are constants defined as,
$$ a = 1.54 \times 10^{- 6} \left( {\frac{{\gamma^2 }}{{\phi \cdot d^2 }}} \right);b = 6.8 \times 10^7 \left( {\frac{{\phi^{3/2} \cdot d}}{\gamma }} \right) $$
Where, γ = the field enhancement factor, ϕ = work function (eV), and d = gap between the anode and the cathode tip (μm). The field enhancement factor γ is an important quantity that qualifies a field emitter. As its name indicates it is a factor by which an applied field, V/d, is enhanced at the cathode tips. In a simple sense γ is proportional to the reciprocal of the tip radius. Higher values of γ are desirable in order to realize good quality field emitters that have low threshold fields and can generate high current densities. The F-N relation shown in Eq. (1) is of the form y = mx + c, when ln(J / V2) is plotted against V-1 and the quantity γ can be determined from its slope. CNTs with γ values in hundreds to a few thousands range have been reported. A single tip is known to produce tens of nA of current at very low threshold fields, less than 1 V/μm [8]. Additionally, what makes them attractive is their imperviousness to low vacuum ranges of 10-4 to 10-5 Torr. As mentioned before this is a typical vacuum one can achieve using wafer bonding [913] or solder reflow bonding [14] techniques used to fabricate micromachined vacuum cavities. However, it was found that the high current emission ability of single, isolated tip does not scale up when multiple tips are employed simultaneously due to the electrostatic screening effect [15] or “hot spots” (see Fig. 1). With multiple tips, only those CNTs with locally the highest field enhancement factor (or γ-factor) participate in electron emission. Producing high current density from a large area source is an optimization exercise of achieving the highest possible number density of emitters with lowest possible electrostatic screening. Many research groups have reported different optimum CNT arrangement [1519], to minimize these hot spots, and achieve high, uniform emission currents. But producing tens to hundreds of A/cm2 on a routine basis is still an active area of research.
Fig. 1

Photograph of a CNT sample showing isolated emission sites imaged through a transparent ITO-anode. The image shows the “hotspot” effect due to electrostatic screening.

2 Carbon nanotube bundle array cathodes

2.1 Carbon nanotube bundle arrays

In our pursuit of high-performance field emitters, we found that CNTs when arranged as arrays of 1-2 micrometer (μm) diameter bundles spaced 5 μm apart (see Fig. 2) give very high emission current densities [14, 20]. This optimum array arrangement was detected empirically while testing emission from series array combinations of different bundle diameters and inter-bundle spacing. CNT bundle arrays were fabricated using lithography to pattern Iron catalyst dots and CVD synthesis at 600°C to grow multi-walled nanotubes of 20 nm tube diameter. A flat-tip probe anode of 100-μm diameter was scanned over all different arrays recording the emission current. Figure 3 shows one such scan projected onto the sketch of the arrays. Notice that the emitted current increases by more than an order of magnitude (8 μA to 100 μA) when the probe anode is over the array of 1-2 μm diameter bundles spaced 5 μm apart. This behavior was observed in several samples. The measured current density for these types of samples was > 2 A/cm2 at 4 V/μm. Following this success, we have optimized the CNT growth process and the architecture in recent times to routinely produce 10 to 25 A/cm2 at applied fields of 5 to 10 V/µm. These tests were conducted using CNT bundle array samples that occupied a circular area of 100 μm diameter (see Fig. 4), which is considered a large area source for our applications. The repeatability of this current production was achieved over multiple samples as shown in Fig. 4. All emission tests were conducted in ~10-5 Torr vacuum.
Fig. 2

SEM micrograph of optimum CNT bundle arrays (1 μm diameter spaced 5 μm edge-to-edge). Inset shows the magnified view of one of the bundles containing hundreds of 20 nm-diameter nanotubes [14, 20].
Fig. 3

Emission current shown as a function of probe anode scan across an array of CNT bundles of different diameters and inter-bundle spacing. Two curves- one at 100 V and other at 300 V are shown. Notice the emission increases almost by an order of magnitude when the probe hits 1-2 μm diameter bundles in the 5-μm array.
Fig. 4

(a) SEM micrograph of 100-μm diameter sample with 1-μm diameter bundles. (b) An example of field emission data from multiple samples producing 10 to 15 A/cm2 at fields ranging from 4 to 9 V/μm for samples of 1, 1.5, and 2 μm diameter bundles spaced 5 μm apart (inset shows dimensions and CNT heights). Curves grouped to the left have a higher γ-factor of > 7000 while those on the right have γ-factor < 4000.

Through real time SEM observation, the reason for high emission from bundle arrays is understood to be the free-ends and outliers in each bundle that rearrange themselves under an applied field [21] causing high field enhancements (γ -factor), which in turn cause efficient field emission. Similar results have been presented by other studies [22]. Typically we have measured γ-factor values of 2,000 to 8,000 from CNT bundle array samples, which are high compared to other field emitters. In large area sources, the initial emission occurs from the highest γ -factor tubes (which accounts for the low threshold fields: 1-3 V/μm) followed by additional emission from those CNTs with lower γ-factors as the applied field is increased. Because of the latter effect, the average γ-factor of the entire sample decreases at higher fields, somewhat flattening the higher field part of the Fowler-Nordheim curve [23] (see Fig. 5).
Fig. 5

Experimental demonstration of the averaging effect of field enhancement factor (γ-factor) for a large area sample (multiple emitters), which has a flattening effect on the Fowler-Nordheim curve at higher fields. In this example, the average γ-decreases from 4000 to 2000 at higher fields.

Using a theoretical relation, Bonard et al. [24] have shown that for an isolated nanotube, the γ-factor increases as a function of its aspect ratio before approaching a plateau. Similar behavior was observed for large area CNT bundle array samples as well. Measured γ-factors for different bundle arrays (1 and 2-μm diameters, 5 and 10 μm spacing) as a function of bundle height are shown in Fig. 6. For a given diameter bundle array, the γ-factor (and consequently the emitted current) increases towards saturation as the height of the bundles is increased (data is shown up to a bundle height beyond which, the weight of the bundles overcomes their stiffness to conduct accurate measurements). The low turn-on and higher current emission as a function of bundle height is shown in Fig. 7.
Fig. 6

Experimental verification of the variation of the γ-factor as a function of CNT-bundle height. As the aspect ratio increases the γ-factor approaches saturation.
Fig. 7

Field emission curves for a 1-μm diameter, 5-μm spaced array of CNT bundles with different bundle heights from 25 μm to 100 μm plotted as a function of normalized voltage. Notice that the samples with taller CNTs turn on at lower fields and reach higher currents at faster rates.

2.2 Monolithic electrode integration

While gate integration with CNT field emitters has been reported before [25], the process presented here [26], and its extension (as in many-layered-SOI substrates) allow monolithic integration of multiple electrodes for electron beam shaping to produce a highly compact field emission electron gun. In addition, this process offers improved dimensional precision, and can be used to integrate gates with single CNT bundles or with an array as needed for an application. Figure 8 shows a cross-sectional view of the CNT field emitter design. At the center is an array of cylindrical CNT bundles 2 μm in diameter and 5 μm in edge-to-edge spacing. An integrated gate electrode is overhanging near the tips of the CNT bundle array with only a few microns in lateral separation. The gate electrode is 2 μm in thickness and the cavity is 12 μm in vertical depth. The sidewall of the hole is laterally recessed by a 10–15 μm undercut. This recess helps prevent stray electric fields due to charges on the sidewall surface from influencing the emission field. This recessed sidewall could also be biased for shaping field emission beam profile.
Fig. 8

Schematic of the monolithically gate-integrated CNT bundle array emitter showing different parts of the double-SOI design.

Figure 9 illustrates the fabrication process of the field emitter device. The starting material is a dual-layer SOI wafer comprising, from top layer to the bottom, a single crystal silicon device layer (1), a silicon dioxide (buried oxide, or BOX) layer (2), another single crystal silicon device layer (3), another BOX layer (4) and a silicon handle wafer (5). The double-SOI wafers were prepared by a commercial vendor using a multiple bond-and-etch back SOI (BESOI) process. In the future triple or more layered SOI wafers may be used to increase the number of gate electrodes. In Step 9 of Fig. 9, Xenon difluoride (XeF2) gas is used to etch a recess undercut by 10 to 15 µm. In Step 15, photoresist is spin-coated to form a bridging membrane over the recessed gate. The bridging photoresist is patterned with a contact mask aligner (Step 16). Catalyst for the CNT CVD growth is deposited and lift-off patterned (Step 17). Finally, the CNTs are grown on the catalyst.
Fig. 9

Fabrication process schematic (not to scale).

In Step 15 of Fig. 9, the spin-coated photoresist forms a membrane over the recessed gates. This membrane is found to be appropriate for patterning catalyst dot patterns with a 2 μm circle feature size. Bridge-coating and patterning was successful for circular gates with diameters up to 150 μm. Figure 10 shows a SEM image of successfully fabricated field emitter devices. 10(a) is a varying transconductance design (array of bundles) and Fig. 10(b) is a more traditional gate-integrated single bundle design. The gate aperture diameter is 25 μm and the CNT bundle length is approximately 10 µm. Preliminary field emission current test data are presented in Fig. 11. The measurement was done in a triode configuration. The anode is positioned at a separation of approximately 460 μm. As the gate voltage Vg is turned on, the field emission onset shifts significantly. The emission efficiency was calculated to be in the range of 70% - 80%.
Fig. 10

Monolithically gate-integrated CNT field emitters. (a) SEM image of multi-bundle design (25 μm diameter) gate, (b) SEM image of a single-bundle design (20 μm diameter gate).

2.3 Lifetime issues

We have found that, when operated in a continuous mode, the performance of CNT bundle cathodes is affected by two effects- (i) a gradual decay of emission due to anode sputtering and CNT tip loss at lower fields (< 8–10 V/μm), and (ii) dislodging of CNT bundles from the substrates at higher fields ( > 8–10 V/μm) because of poor adhesion of CNTs to substrates. The former effect has been observed in traditional cold cathodes. But, the latter is unique to CNT field emitters. . Over several samples we have observed that the emission current starts above 1 mA, then eventually settles to around 200 μA within 20 minutes (starting from t = 0.0), as shown in Fig. 12(a). These experiments were conducted using copper and glassy carbon anodes. EDX analysis of a CNT sample tested using Cu showed sputtered copper particles settled on the CNTs. This analysis is shown in Fig. 12 (b). For glassy carbon a similar sputtering phenomenon is shown in the SEM micrograph as a circular etched region (Fig. 12(c)) corresponding in dimension and location to that of cathode (accounting for the electron beam divergence).
Fig. 11

Field emission current measured in a triode configuration.
Fig. 12

(a) Graph showing a gradual decay of field emission current in continuous bias mode. Top curve indicates bias voltage (held constant after t=0); (b) EDX spectrum of the cathode sample tested using a Cu-anode. Other than C and Si, the main peaks occur for Cu showing sputtered anode on CNTs; (c) SEM micrograph of glassy carbon anode showing etched circular region corresponding to CNT cathode area.

When the field is increased to > 10 V/μm, we start seeing the second failure mechanism. The force exerted at these fields appears large enough to dislodge individual CNT bundles. The adhesion strength of CNTs to substrates as measured in our laboratory is in the rage of ~ 35 to 65 kPa. Assuming bundles and anode plate as parallel plate capacitors we can estimate the force exerted by the applied field on a CNT bundle. For a 2-μm diameter bundle with γ of 1000, a conservative estimate shows that the exerted force is greater than 1 mN at 10 V/μm applied field. This translates to pressure in the MPa range, which is clearly capable of dislodging CNT bundles whose adhesion strength to the substrate were measured to be in the kPa range. This failure mode is shown in Fig. 13. The graph shows a continuously increasing emission current even though the bias voltage is kept constant, indicating slowly rising CNT bundle(s) that have been dislodged but are bound to the surrounding bundles. A sharp drop in the emission current indicates the bundle(s) completely losing any contact with the cathode. Figure 13 also shows SEM micrograph of an anode tip that has dislodged CNT bundles attached to it. We are currently developing techniques to mitigate both failure modes.
Fig. 13

Graph showing the high field failure of field emission due to CNTs dislodging from the surface. The SEM micrograph on the right side shows dislodged CNTs stuck to the probe anode.

3 Conclusions

We have developed CNT bundle array cathodes that show promise as high-current density cold cathodes, robust to poor vacuums and capable of operating at low voltages. A monolithic electrode integration technique using a double SOI process was presented that enables multiple level electrode integration to realize a miniature electron gun. Additionally, two failure modes of these CNT bundles were presented. The first is a gradual decay of emission due to CNT tip failure as well as the anode sputtering effect. The second corresponds to dislodgement of CNT bundles due to force exerted by the anode at high fields. Enhanced adhesion of CNT to substrates can overcome the latter failure.


This research was carried out at the Jet Propulsion Laboratory, California Institute of Technology, under a contract with National Aeronautics and Space Administration (NASA). This work was funded by JPL’s Research and Technology Development Fund and partly by Defense Advanced Research Project Agency. We thank Dr. Paula Grunthaner, Dr. Barbara Wilson, Dr. James Cutts, and Dr. Timothy Krabach for their advice on the cathode development as part of the internal research and development effort. We thank Dr. Mark Rosker of Defense Advanced Research Projects Agency for his support of this work. We thank Dr. Ken Dean of Motorola, Dr. John Hong of JPL (at the time of this work), and Dr. Dev Palmer of US Army Research Office for useful discussions.

Copyright information

© US Government 2009