Drop (Figure 4(a)), cast (Figure 4(b)), and spray (Figure 4(c)) approaches have proven, thus far, the fabrication method of choice by many. This is likely a direct consequence of their low cost and straightforward processing rather than any direct functional enhancements the fabrication techniques allow for. Drop, cast and spray techniques all require CNT inks. In each case, respectively, these are deposited either by direct dropping of the ink, spin coating or casting of the ink, or spray coating of the ink using a pressurised carrier gas. All these approaches initially require wet processing of the as-grown CNTs and it is this which, commonly to all, limits their usefulness and the consequent performance of the resulting X-ray source. The required high power sonication and vacuum unstable surfactants needed to form the stabilised, homogenous inks results in compromised temporal stability, deleterious hysteresis, and generally degraded emission.

Electrophoresis is a derivative of inking deposition. Electrophoresis involves the motion, placement, and potentially modest alignment of drop or spray deposited CNTs. Electrophoresis exploits the CNTs native anisotropic charge distribution and its interaction with an applied electric field [160,167,168]. Elsewhere it has been widely used as a means of depositing phosphor materials. It has also been applied to the deposition of bundles of SWCNTs on various substrates, such as stainless steel or doped Si [167–179]. Composite CNT films deposited by electrophoresis have been fabricated with controlled surface density with good field emission performance, current density and long-term stability under high operating voltages, which has been applied to an electron source for high-resolution X-ray imaging [160]. Wang et al. [169] deposited MWCNTs on Cr/Cu electrodes supported on an oxidised Si support with a glass interlayer adhesion promoter. The emitters were annealed at 480°C for 30 min following deposition to mobilise the glass interlayer and reduce the effective interfacial resistance. As with other solution-based fabrication techniques, in order to reduce the emitters turn-on bias they were taped to activate them; a threshold field of 3.5 V/μm was measured (for an emission current density of 10 μA/cm²). Though a simple and low-cost means of defining the pixel size, electrophorectially defined emission zones only allow for relatively large cathode diameters, of the order of 2 mm, which reduces the systems functionality. Calderón-Colón et al. [160] employed photolithography and liquid phase electrophoretic deposition to pattern disordered and unaligned MWCNTs in a homogenous solution with MgCl₂ and glass frits. Here the film thickness was broadly controlled by the deposition time and the magnitude of the applied voltage, though accurate sample-to-sample reproducibility and control over the thickness, to <10 nm surface roughness, is challenging. Such techniques are comparatively low in uniformity and reproducibility, a direct consequence of the simple macro scale processing. Moreover, the common binder matrices are far from pure, making elucidation of the underlying emission mechanisms somewhat challenging, though nonetheless functional. Indeed, as with all other inking deposition techniques, the emitters do require mechanical activation for the cathodes to be of any practical use, though electrophoretic patterning is indeed rapid and simple to implement.

Many of the FE X-ray sources demonstrated have employed screen printed or vacuum filtrated CNT thin films [156,169–171] (Figure 4(d)). Here CNT solutions are deposited at the macro-scale and form highly disordered, nominally planar, spaghetti-like networks with but a few individual CNTs standing upright, at many unregistered angles relative to one another. In the case of vacuum filtration the CNT ink is poured onto a porous mixed (nitro) cellulose ester membrane, which has a partial vacuum applied on its opposite side. Such porous membranes have controlled apertures, typically around 0.2 μm in diameter. The reduced pressure stimulates the solvent within the CNT ink to pass through the membrane, whilst the membrane stops the CNTs from passing. Once the ink reservoir is depleted the thin film is rinsed with de-ionised water to remove much of the deleterious surfactants, leaving a thin CNT film on the membrane. Note, however; that much of the surfactant still remains even after extensive rinsing using deionised water. The CNT thin film can then be transferred to an arbitrary acetone-resistant substrate by applying modest pressure (40 g/cm²) and heat (ca. 70°C) for an extended period of time (often of the order of 3 h). The membrane is then dissolved, by exposure to acetone or methanol, and the CNT thin film remains. The CNT film thickness, and hence sheet resistance and optical transparency, is controlled by adjusting the amount of CNT ink filtered through the membrane.

CNT thin film fabrication by vacuum filtration has been employed for more than a decade. It offers a rapid, low cost way to fabricate CNT thin films. Little to no infrastructure is required and the films can be processed rapidly, over large areas. Nonetheless, as in the case of drop, cast and spray, such chemi douche processing requires stabilised inks. CNTs experience high inter-tube van der Waals forces; they tend to agglomerate. Though this interaction has been exploited elsewhere to fabricate novel aligned nanostructured membranes [172], such agglomeration is problematic in producing homogenous CNT inks and solutions. As such, various, often sodium-based, surfactants such as sodium dodecylbenzene sulphate, sodium dodecyl sulphate, and Triton X are required to produce homogenous solutions with the CNTs well-dispersed throughout the solution. Further aggressive acid treatments and extended durations under high power ultrasonication significantly degrade the length, crystallinity and subsequent electronic character of the CNTs which necessarily limits the electron emission performance. These inks are then transferred to metallic disks using conventional screen printing methods or vacuum coated mixed cellulose ester membranes that are subsequently dissolved in acetone following transfer [173]. The morphology of these films is highly disordered. They typically lack high aspect ratio protrusions, resulting in poor field emission performance. FEs fabricated in this way have a number of intrinsic problems; chiefly that mechanical taping is required to activate the surfaces and enhance their field emission characteristics. Taping, using adhesive coated tape, increases the surface roughness of the CNT thin-films. It is a macro-scale process and affords very little reproducibility, Moreover; the necessary surfactants are usually vacuum unstable giving rise to emission profiles that deleteriously drift with time [170,174–176]. CNT inks can often have significant out-gassing [177] when even modestly heated, compromising the vacuum envelope. The resultant reduction in field emission performance and reduced current density has, as a result, prevented CNT-based pastes and ink from gaining commercial traction, though cathodes fabricated in this way appeal as the reduced reproducibility is off-set by the ease of fabrication. Such films are also somewhat dynamic and often have weak adhesion to the substrate. Their morphology shifts with time during the application of a high electric field due to the intrinsic torque induction within the CNTs due to the tip or root positioning of the growth catalyst particle.

To obviate issues of weak interfacial adhesion between the substrate and the CNTs, Kim et al. [154,178], using screen printing, as illustrated in Figure 4(e), employed ball-milled pastes combined with inorganic fillers including Ag and Ni alloyed nano-particles and oxides in an ethyl cellulose powder and terpineol solvent. No surfactants were used. The pastes were deposited on indium tin oxide coated glass substrates followed by 300°C post-bake to remove residual solvents. The derived slurry showed strong adhesion to the substrate, good uniformity and reproducibility with a current density of up to 350 mA/cm². Nevertheless, ball-milling degrades the graphitic quality of the MWCNTs, which adversely impacts the long-term time stability relative to high quality crystalline material, although these devices did show excellent high-temperature operation (up to 800°C), robustness towards harsh vacuum environments, impressive spatial uniformity, a low turn on voltage of 1 V/μm, and enhanced temporal stability relative to conventional CNT pastes, although the anode current was still degraded by around 30% under DC operation over a 100 hr test period. Another fabrication problem associated with screen printing is the limitation on the pattern resolution. The sonicated CNTs within the inks are typically of the order of 5 μm in length. These can clog the screen printing mesh resulting in a low porosity and inability to print.

Formation of stabilised, homogenous inks is central to drop, cast, spray, vacuum filtration and screen printing. The formation of these inks, as highlighted above, requires deleterious ultra-sonication and aggressive acid treatments, both of which degrade the length and electronic character of the CNTs. Though the necessary wet chemistry approaches provide a facile, rapid and inexpensive route to fabricate the emitter, the constituent CNTs are coated with deleterious vacuum unstable surfactants that can only be removed following high temperature post-deposition treatments which, if not fully removed, would otherwise cause significant out-gassing during FE operation [160]. Maintaining slurry and ink consistency over time and between batches is difficult, which manifests as a reduction in device-to-device reproducibility. It is also challenging to pattern emitters fabricated in this way though screen printing (the migration of CNT inks through patterned apertures within a regular mesh), as demonstrated by Kim et al. Though screen printing is indeed large-area compatible, it is rather low resolution and thus limits the degree of control over the detailed design of the electron source. Oxygen plasma etching coupled to conventional lithographic techniques is another viable option though significant surface roughness of the CNT thin film can again compromise the maximum resolution. Moreover, plasma etching techniques are only applicable to non-organic substrates.

While there have been reports that electrophoresis can produce a degree of alignment [179,180] this is relative only to very randomly orientated screen printing and ink-based processes. Misalignment prevents any fabricated devices from realising the full field enhancement factor of the composite CNTs; in order to achieve this alternative fabrication methods capable of aligning, en masse, the CNTs must be considered, with chemical vapour deposition being the most promising method to date.

Chemical vapour deposition (CVD) (Figure 4(f)) provides a more controllable means of growing CNTs. The CNTs self-assemble from atomic units in a highly parallelised process, which when coupled with high resolution lithographic techniques to pattern the catalyst material, allows for near nano-scale engineering of the CNTs and CNFs. CVD techniques mediate the growth of chemically untreated disordered or aligned CNT thin films depending on the substrate, catalyst and growth precursors employed. In a typical implementation, Silicon is coated with a physical vapour deposited metal catalyst which is then patterned via lithographic or masking techniques by either additive or subtractive process, such as magnetron sputtering, or plasma etching, respectively. The substrate is then heated to temperatures often in excess of 500°C, and the growth of the CNTs on these sites is initiated by supplying a hydrocarbon feedstock gas, such as CH₄ or C₂H₂, combined with an a-C etching gas species, typically H₂ or NH₃ both of which readily pyrolyse to give a constant supply of carbon and atomic hydrogen. In situ plasma can also be employed to enhance the catalysis and align the CNTs during growth. Cole et al. have compiled a concise overview of the CVD of nanocarbons, available elsewhere [181].

Using thermal-CVD synthesised MWCNTs directly deposited on shadow-masked Mo discs (as well as Si, fused quartz, mica, copper, and highly ordered pyrolytic graphite), with a Al/Fe (0.5/10 nm) bilayer catalyst, Sarrazin et al. [182] have routinely measured current densities in excess of 1 A/cm². Here the MWCNTs were several microns in length and randomly orientated, with most running adjacent to the substrate; CVD techniques allow for CNT alignment, though random orientation is still possible. Very little control over the type, orientation and area packing density of the CNTs was evidenced with a largely qualitative analysis presented. Nevertheless, a turn-on bias of approximately 2 V/μm was observed. Rather surprisingly, the robustness of the cathodes toward arcing was clearly shown resulting in an increase in the voltage for a given current following repeated arcing events. They suggest that either the emitting sites are not completely destroyed or that they are efficiently replaced by other nanotubes within the film. Indeed, disordered films are structurally dynamic when under the influence of a high electric field, which can augment their emission characteristics. Though film reorganisation benefits the long term stability, in that degraded CNTs are in essence replaced, the short term temporal stability is likely to be very poor. More than 100 cathodes were tested and a poor reproducibility was indeed noted, a probable consequence of the disorder and uncontrolled microscale morphology of the emitter.

As is the case for the screen printed and vacuum filtration methods, one potential problem is the degree of adhesion between the CNT and the substrate when exposed to high electric fields. Detailed control of the underlying catalysis has shown that such emitter removal concerns can be solved. Li and Cole et al. showed very low variation in anode current of < ±0.7% for constant emission over >200 hrs using an ITO/Ni (10/1 nm) bilayer catalyst which effectively enhanced the degree of adhesion [183].

Using rapid thermal CVD, Kim et al. [184] deposited CNTs directly on stainless steel sheets coated with TiN/Ni. The emitters showed a maximum current of 2 mA (data on the emitter area was unavailable) at an anode potential of 5 kV. The CNTs were 30 μm long, with Raman analysis suggesting a very defective material, whilst scanning electron microscopy showed vertically aligned forests with a high packing density. Though the emission showed good performance the geometry was not optimised; the dense forest results in significant shielding of the CNTs from the applied electric field – the material appears as bulk - such that the full field enhancement factor of the CNTs was not realised in this instance. Whilst individual, one dimensional nanowire and nanotube emitters have been empirically evidenced to produce the highest electric fields, the proximity of other emitters will effectively shield the field enhancement [146,185]. In a field emission device, where our interest is in the total available current density, the optimum arrangement will not be that with the highest density of emitters [186,187]. In addition, if the emission pattern of the field emitter is not uniform CVD growth can result in individual CNTs, or structures such as CNT pillar arrays and toroids [160,167,188–191]. Patterns of control electrodes may be grown, in such a way as to focus or concentrate the field from the tip.

The use of CVD to nanoengineer X-ray sources has remained in its infancy due to a number of challenges in explicating the underlying material growth. Nevertheless, following recent advances in the understanding of nanocarbon catalysis [192–194], the use of CVD-grown CNTs and CNFs in FE X-ray sources appears to be accelerating with it emerging as an emerging as an exciting candidate for viable commercialisation.

Many groups have grown CNTs, by CVD or derivatives thereof, directly on conical or other shaped tips, on common Spindt-like materials [213], Si [169,205,214] or bulk metals [200]. As such, these systems do not exploit the full geometric potential of the CNTs; they simply amplify, often only marginally, the existing field enhancement factor of the tip structures and almost always result in an induced asymmetry in the electron beam profile that is challenging to accommodate with conventional electrostatic focussing. Indeed, even conventional X-ray TE cathode assemblies will, in general, produce an anisotropic focal spot, as a result of the electron beam dimensions, imperfect focusing, and the impact of the target angle. One potential route to realising symmetrical X-ray sources with shaped beam profiles is to shape the cathode such that the emergent electron beam defines, at least in part, the spatial distribution of the emergent X-ray beam. This requires patterning of individual CNTs or CNT forests into well-defined geometries. Several groups have considered the use of cathode shaping as a mean to achieve beam profiling. Most have used elliptical cathode geometries to define a circular distribution of the emergent X-ray beam [160]. Two examples are as shown in Figure 5(a,b). Shaping has also been shown to enhance the FE performance by exploiting the natively high field enhancement factor at the shaped emitter’s edges. In particular, the group of Milne have shown field emission properties of individual CNFs and patterned forests [185,215,216]. No systematic studies on the effect of shaped cathode eccentricity, or other controlled geometrical parameter, on the symmetry or profiling of the emergent X-ray beam appear to have been considered to date. Indeed, the electrophoretically deposited and photolithographically patterned elliptical cathodes of Calderón-Colón et al. [160] showed impressive emission characteristics though no detailed assessment of the emergent beam shape was reported, making is difficult to assess the success of such beam shaping techniques (Figure 5a).

Figure 5

Shaped CNT cathodes. Control over the electron beam distribution, and subsequent symmetry of the X-ray beam, can be achieved by shaping the electron emitting areas on the cathode by conventional lithographic techniques. (a) A screen printed elliptical shaped electron emitter and subsequent X-ray image of a leaf [160] (Scale bars: 2 μm, 1 mm). Copyright Institute of Physics. (b) Scanning electron micrographs of an elliptical electron source, formed from tepee-like CNTs, demonstrated by Ryu et al. (Scale bar: 500 μm, 300 μm) Adapted from [205], Copyright 2012, Multidisciplinary Digital Publishing Institute.

Using Si-supported CNT arrays, Ryu et al. demonstrated the potential of CVD in X-ray source applications [205]. A resist-assisted patterning process was used to produce a cathode assembly consisting of a gate and focusing electrode. This assembly was approximately 0.5 mm × 2.0 mm and gave an isotropic focal spot [205]. Ryu et al. also described a further shaping technique for post processing the as-grown vertically aligned CNTs. Here they used hydrofluoric acid to produce an array of conical pillars (Figure 5b), rather reminiscent of earlier Spindt emitters. Hydrofluoric acid, though cheap and readily available, is rather hazardous to handle and poses a significant health risk. This somewhat limits the commercial viability of the compaction technique presented. Nevertheless, they demonstrated anode currents of up to 90 mA at 7.8 V/μm, with a turn-on electric field of approximately 3 V/μm. To increase the geometric uniformity of the emitters the team used an electrical aging treatment, which degrades the taller tubes and bundles that would dominate the emission. For a high performing emitter this aging, also termed seasoning, is not desirable and CNTs of very uniform height and diameter are preferred, as evidenced by Teo et al. [166], to obviate the need for such post-growth treatments. Nevertheless, part of the Ryu et al. ageing process [205] was to remove the Ni catalyst particles at the CNTs apex, which they argued reduced the emission current. Detailed analysis and theoretical studies surmising the emission implications of the metallic nanoparticle in the emitter tip are lacking in the literature.

High beam current, and therefore current density, is desirable. This gives rise to higher photon flux, and hence shorter detector integration times, in addition to improved signal-to-noise ratios. Combined with the use of gate control to pulse the X-ray source, high beam currents can be very advantageous to system designers. Recently, micro-fabricated Spindt-like emitters have been used for applications such as static tomography [217]. Here each cathode was formed from approximately 50,000 tips covering an area of about 1 mm². These emitters had a measured current capacity greater than that reported for CNTs, though CNTs will almost certainly exceed this once the technology matures. For current CNT-based sources the emission current density is several orders of magnitude less than that of equivalent area Spindt emitters. Nevertheless, for CNT-based emitters, current densities of up to 1 A/cm² have been reported [18], with studies elsewhere suggesting that individual CNTs are capable of conducting current densities corresponding to 5×10⁶ A/cm² [218], which is significantly larger than any attained by Spindt-like emitters fabricated to date.

Using electrophoretically deposited SWCNTs [167], high electron beam currents have been achieved using a triode configuration by Yue et al. [149]. Here, the gate assembly was mounted approximately 100 μm adjacent to the CNT electron emitter resulting in a device geometry capable of sustaining a beam current of several milliamps, although this would likely prove challenging to manufacture in a practical X-ray source. In addition, they demonstrated peak pulse currents of up to 28 mA (emitted area not stipulated), without vacuum breakdown, at an anode potential of 14 kV. Using CVD grown CNT arrays, Ryu et al. [205] reported emission currents of up to 90 mA (emitted area not stipulated) in a diode configuration, although in a gate-controlled triode arrangement, the beam currents were substantially reduced.

Some improvement towards increasing the current density has also been achieved by controlling the emitter morphology. Toroidal CNT arrays, which have a central void [191], are one such example. By subsequent surface treatments, it is also possible to enhance the native FE characteristics of such arrays by means of the formation of nano tips, tepees and micro cones, as outlined previously [190,205]. The use of emitter forming post-treatments has also been widely investigated. One leading example is the use of conical CNTs which enhanced the beam current and stabilised the emitter geometry. Another alternative towards higher beam currents is to augment the electronic character of the CNTs, via the dry or wet deposition of various adlayers. Such adlayers adjust the interfacial characteristics at the critical emitter-vacuum interface. This often leads to a decrease in the turn-on voltage and increase the emission current density at a given anode bias. As was highlighted above, FE is highly surface sensitive; at 10⁷ A/cm² a 1% change in the surface work function, due to intentional adlayer sputtering or indirect sputtering from reactive gas species in the vacuum cavity, can increase the current density by more than 10% [152]. Little work has been reported on the use of adlayers to enhance the emission performance of CNT-based FE X-ray sources. Nevertheless, significant advantages may be achieved by carefully designing the emitter-vacuum interface to provide high emission currents, limit adverse effects of vacuum leakage, prevent unintentional work function shifts, as well as robustness towards local ionisation and plasma etching, all of which are critical in ensuring long term stability.

The failure mechanisms of Spindt and CNT based emitters, which are similar in many respects, have been widely studied [219], most notably by Bonard et al. [220]. Thermal migration, field sharpening and subsequent avalanche breakdown are perhaps the most common failure mechanism in such whisker-like emission geometries [221]. Although recent work on CNT-based emitters has shown impressive lifetimes of 7000 hours, with only a 10% emission degradation [222], it remains an on-going challenge to design and fabricate high temporally stable emitters. As noted previously, the electron emission is sensitive to the bulk emitting material and the emitters’ surface chemistry. CNTs have a high sublimation temperature and high maximum current densities, making them resilient towards arcing events. Pristine CNTs are also largely inert, although when defects are added to the graphitic lattice, often through plasma processing, they are readily damaged and the formed dangling bonds bolster the CNTs emission performance, though often at the expense of the long term stability. The enhanced emission is due to the augmented surface characteristics. This effect is often noted as a shift to higher turn-on bias during hysteresis studies. Indeed, many plasma and some dopants, such as oxygen when the emitter outgases, can be particularly damaging to the CNTs. Local Ohmic (I²R) heating can also increase the rate of emitter degradation, as can vacuum breakdown. It has been shown that the axial resistance of CNFs increases with increasing temperature. Interestingly this intrinsic property helps protect the CNFs from degradation during local heating and helps prevent thermal run-away deleteriously increasing the emission current [221]. Vacuum breakdown results in the emission of extremely high current densities which may cause a plausible local oxygen micro-plasma formation – this etches the CNTs, particularly at their apex. This shaping stimulates the formation of defects in the graphitic lattice, which preferentially emit electrons, thereby enhancing the emission, however; this results in a temporally unstable emission profile. The CNTs can be sharpened with time, which tends to increase the emission current, or can alternatively be entirely ablated, which reduces the emission current due to the reduced number of electron emitters available.

There have been various methodologies proposed which attempt to reduce temporal instabilities. Thermal annealing or electrostatic seasoning, to remove residual surfactants and non-uniform emitter profiles, are perhaps the most common and certainly the most simple and readily implemented [160]. During thermal annealing the cathode assemblies are heated to temperatures in excess of 300°C under high vacuum conditions. This out-gases the emitter, removing weakly surface bound chemisobed species. This increases the work function uniformity across the surface of the emitter. Such out-gassing techniques are also useful in emitter recovery following an arcing event. Arcing events stimulate high current flow which heats and subsequently out-gases the emitter. This out-gassing can lead to further transient arcs which, if allowed to continue, will degrade the emitter. If the emitter is initially well out-gassed any local arcs will only marginally increase the cavity pressure and the emitter will stabilise more rapidly. Annealing is also employed to enhance the pressure of the vacuum cavity, making local plasma formation increasingly unlikely. In the case of electrostatic seasoning, the extraction electric field is slowly ramped up to around 80% of the nominal emission voltage. The emitters are then left emitting for tens to hundreds of hours to increase the surface smoothness of the emitter and hence, stabilise the emission current. Such approaches are critical in achieving intrinsic emitter stability, and though feedback techniques have been employed to artificially control the stability, engineering intrinsic stability remains central to the formation of a long-term stable emitter. In feedback based systems, in the same way as conventional TE generator designs, the anode current is monitored and the extraction voltage adjusted accordingly to maintain a known, safe, emission current. Though a viable and widely adopted approach to ensuring emitter stability, the slow response times of the feedback loop cannot entirely remove transient effects, such as arcing events, and only careful design of the electron source can facilitate this.

Though CVD is certainly coming to the fore as the most reproducible fabrication technique with the finest degree of control over the emitter design, it, like other techniques, is faced with issues of tip-to-tip uniformity. Poor uniformity in effective surface roughness is known to de-stabilise the temporal stability. Indeed, wet chemistry ink approaches to emitter fabrication result in much greater surface roughness, and this exacerbates and further compromises their temporal stability. Small height variations between tips can instigate preferential emission from a small proportion of the longer CNTs, which consequently burn-out. It has been shown elsewhere that individual CNTs can in practice pass a current of several microamps [223]. Currents in excess of this threshold cause the CNT to sublime, which manifests as a temporal instability in the anode current, and subsequent X-ray emission. In the case of CVD-synthesised CNT emitters one solution – originally described for application with Spindt emitters [224] and initially proposed for general electron emission applications by [183] – consisted of integrating a ballast resistor micro-fabricated in series with the electron emitter. Here resistive deposited layers are fabricated in series with the electron emitters. The series resistance ballasts the emission, functioning as a current limiting resistance, preventing emitter sublimation, subsequently enhancing temporal stability. Yet to be applied immediately to the design of an X-ray source, the group of Milne have developed a novel field effect transistor (FET) ballasted field emission source, where each individual CNF electron source is equipped with its own dedicated FET ballast layer which is capable of limiting, and electronically controlling, the emission current to prevent emitter sublimation and significant temporal instabilities [183]. Other thin film deposition techniques have also been considered [204,225] though these too have yet to gain any significant interest.

Individual CNTs are fragile. Their propensity toward tip or root growth results in an axial asymmetry in their magnetic susceptibility which manifests as rotational torque induction when exposed to high electric fields - an effect which is typically exploited during electrophoretic alignment. However, during FE this can result in the removal and transfer of the CNTs from the cathode to the anode. This severely impacts the lifetime and stability of any field emission device into which they are incorporated. Several approaches have been taken to improve the robustness of the emitting elements by increasing the degree of adhesion between the CNTs and the supporting substrate. Bundles of vertically aligned CNTs - typically with a height of around 10 μm or less, and a diameter of 30 μm, have been shown to be one such robust assembly, that is not readily damaged or easily removed. However, FE preferentially occurs at the periphery of the patterned CNT forest or array [188,189] suggesting that if some CNTs are in fact removed the emission stability may simply be recovered by new CNTs contributing the emission current of those removed; emitter areal design can in part accommodate and engineer out temporal instabilities. Nevertheless, there is a continuing effort to engineer the CNT-support interface to enhance the degree of adhesion and hence enhance the temporal stability. In the case of ink approaches various solution additives have been exploited, such as glass frits [160], though the exact implications of such approaches with regards to the turn-on field and maximum current density remain unclear.

Elsewhere there has been significant work on the use of surface treatments to stabilise the temporal stability of the electron emission [205,226]. As discussed previously, such low work function adlayers do indeed increase the emission current density, though they also, depending on the adlayer material and means of deposition, hermetically seal the CNTs – which act as a high aspect support structure – thereby increasing the emitters stability [225].

FE sources have the potential for extremely high spatial modulation. By controlling, the electron emitter location at the nanoscale, coarse control over the position of the electron beam, and subsequent X-ray beam are possible. A range of focussing techniques has also been developed to reduce the electron beam focal spot size. Standard X-ray tubes have a focal spot size of typically around 1 mm diameter. For greater spatial resolution, “mini-focus” X-ray tubes have been developed, with a spot size down to 50 μm, as well as “micro focus” X-ray tubes with a spot size of ~5 μm [227]. Almost exclusively, X-ray sources with effective focal spot sizes of <10 μm, which are commercially available, employ electrostatic or magnetic focussing and require high beam current pulsing techniques, to prevent limitation in their imaging resolution by motion-induced blurring [206]. The size of the electron beam at its source and latterly as it impacts the target material, contributes, in part, to the size of the X-ray focal spot, which itself impacts on the resolution of the resultant X-ray image. A large electron beam focal spot will create a penumbra effect where the X-ray spot subtends a significant angle at the subject; this blurs the resultant image. Conventional X-ray tubes will use a lensing cup around the heated filament assembly. This, to a first order, electrostatically focuses the electrons onto the centre of the anode. Careful design of the cathode and supporting electrostatic lenses will minimise the size of the focal spot [228,229]. Additionally, further electrostatic rings or focussing coils may also be employed to redirect the flow of electrons emanating from the cathode, to further reduce the spot size, where the focal spot size is determined by measuring the Point Spread Function using standard methods [230].

In just such a way, a transmission CNT X-ray tube with a solenoid focussing unit was constructed by Heo et al. to produce a 5 μm focal spot [163]. Liu et al. developed a single electrostatic focusing lens to produce a source with a nominal focal spot size of 65 μm, which was further reduced to 35 μm by limiting the dimension of the cathode assembly [231], a concept yet to be fully exploited. A similar mini-focus tube, used for small animal CT work, was also described. This used two focussing electrodes to produce a focal spot size of about 100 μm, using a mesh gate electrode to maximise the electron emission from the cathode [232]. FE tubes with cathodes constructed from CNFs have also incorporated conventional three stage electrostatic Einzel lenses. Such systems have achieved focal spot sizes down to 40 μm [200].

FE sources have the potential for unprecedented temporal modulation. Compared to other one-dimensional nanomaterials, CNTs allow for near ballistic conduction making them ideally suited for such high-speed applications. Pulse X-ray sources are practically beneficial when imaging moving objects. The principle of operation is analogous to stroboscopic lighting in optical imaging. Conventional TE X-ray sources are generally not capable of rapid control; they cannot be pulsed much more rapidly than a few tens of milliseconds, without sophisticated gating mechanisms. The means of controlling the beam current in a TE source is by adjusting the filament current and hence temperature, which introduces time delays due to the finite cooling and heating time of the filament. Pulse sources are therefore generally implemented by switching the anode voltage, or by means of mechanically controlled shutters, the latter only allowing for relatively course control over the pulse shape, on period, and mark-to-space ratio, and will inevitably create a trapezoidal pulse. In the case of existing FE and TE sources, the short duration of the pulse must be compensated by increased power levels, resulting in significant engineering design constraints with regards to the resilience of the emitting material; conventional electron emitting materials degrade rapidly when used in such high power applications.

In FE X-ray sources often the means of controlling the beam current is via the gate voltage. Depending on the exact emitter geometry, this voltage can be considerably lower than the anode voltage and so at a much lower power level; it can be switched virtually instantaneously. Gate electrodes can certainly be incorporated into more traditional TE sources also, though much more control over the on/off current ratio and pulsing performance can be achieved when using FE sources as the gate and anode can be controlled simultaneously, with both responding rapidly. The result is a FE X-ray source that may readily be pulsed with a high on/off current ratio. This lowers the total thermal dissipation, reduces the total amount of emitted radiation – allowing for safer medical diagnostics, and eliminates the need for bulky mechanical components.

Most devices fabricated to date operate in either DC or low frequency (tens of Hz) mode [148,150,151,154,160,163,169,171,182,184,159,200,202,204,205,207–209,211,212] with only a very limited set of electronically controlled devices operating at frequencies in excess of a few hundred Hz to a few kHz [149,198,203,210]. The highest electronic pulse rate achieved to date is of the order of 10⁷ Hz, as reported by Cheng et al. [201]. Nevertheless, the pulse rate achieved by Thales [233] in their optically gated devices exceeds all other devices by a significant margin. Figure 6(a) shows the anode current and X-ray intensity temporal response by Reyes-Mena et al. [198]. Here the X-ray beam 0-90% turn-on time is <50 μs and the system is used to image a rotating cooling fan. Images were acquired at a constant fan angular velocity with decreasing X-ray emission pulse width. As the X-ray pulse width decreases the fan blade sharpness increases evidencing the usefulness of beam pulsing in acquiring high resolution images in moving systems. The pulsing of the gate voltage has also been shown to produce turn on times of the order of 100 μs [149,198]. At an early stage, Yue et al. [149], as shown in Figure 6(b), demonstrated peak beam currents of up to 28 mA, at anode potentials of 14 kV, without degradation of the field emitters. An interesting artefact of their rapid pulsing was the chaotic and rather noisy X-ray emission, which showed an approximate periodicity of 100 μS. The noise here must certainly be reduced if a practical device is to be realised.

Figure 6

Pulsed Sources. (a) Anode current and X-ray intensity temporal response plots showing an X-ray beam turn-on time of <50 μs, and X-ray images of a rotating computer cooling fan acquired at a constant fan angular velocity with decreasing X-ray emission pulse width. Note the increased sharpness of the fan blades with increasingly rapid beam pulsing [198]. Copyright 2005, Joint Committee on Powder Diffraction Standards. (b) Bias voltage and X-ray emission temporal response plots with an acquired X-ray image of a users hand [149,156]. Copyright 2003, Elsevier and American Institute of Physics. (c) The time dependence of the cathode voltage with integrated pull-up resistance showing a rise time of <3 μs [154]. Copyright 2013, SPIE. (d) Schematic depiction of a red-source photo-gated CNF-based electron source. Each vertically aligned CNF bundle sits on top of a pin⁺ photodiode coated with a TiN diffusion barrier coated and a Ni catalyst, as illustrated in the scanning electron micrograph taken after the CNT growth (Scale bar: 1 μm). Adapted from [233]. Copyright 2014, Thales Electron Devices.

The limiting factor in many pulsed systems is often the capacitance of the gate assembly. Kim et al. [154] showed that pulsing performance could be improved and demonstrated <3 μs pulse periods, as indicated by the cathode voltage, via the use of a simple high voltage MOSFET pull-up circuit to switch the cathode, proposing that their pulsed source may find application in angiography. However it is worth noting that such a measure does not correlate with the X-ray photon flux [198]. Liu et al. [206], by pulsing the gate bias using a pulse generator with a constant anode bias of 40 kV, showed a stable 0.3 mA temporal response over 15 h with an approximate pulse rise time of <5 ms. Certainly electronic control over the pulsing performance has some use. Though the bandwidth is dramatically reduced, by at least an order of magnitude, the on/off ratio can be increased and the emitter fabrication simplified. Nevertheless, this is necessarily at the expense of other technological challenges, chiefly the associated RC constant of the vacuum cavity. This RC constant induces intrinsic time delays that are not experienced in the optically stimulated case. As illustrated in Figure 6(c), Kim et al. [154] employed 10 kΩ pull-up resistances to negate RC lag issues. The pulsing performance of arrayed sources has also been considered. Wang et al. [169] showed millisecond temporal response in a matrix source designed for micro radiotherapy. This had an array of 5 × 10 pixels achieving an emission current of 3 mA per pixel.

In order to obviate issues associated with the devices RC time constants, as depicted in Figure 6(d), Thales Electron Devices exploited integrated photodiodes and direct optical excitation. Vertically aligned CNFs were grown on p-i-n⁺ photodiodes on a back etched Si membrane. The n+ doped areas were defined by ion implantation in a 5 μm thick intrinsic layer, whilst the p+ doped silicon wafer was subsequently thinned to obtain the 7 μm thick membrane [233]. These arrayed emitters where then irradiated, over the entire emitter area, with pulsed red laser light allowing them to attain very high bandwidth and on-off ratio of around 10, in addition to a maximum on current density of around 100 μA/cm².

The availability of rapidly pulsed sources opens up the potential for high-speed, real-time inspection technologies, capable of coupling high throughput, on-line manufacturing with real-time inspection. Indeed, with controlled high pulse rates the potential to perform medical diagnostics without strict patient restraints is accessed, allowing inspection of dynamic organs without image blurring. Intrinsic motion and the associated quasi-periodic respiration and heart beats, for example, allow one means of gating such high speed pulsed systems. One possible approach considered is to lock the gate electrode pulsing to the patients’ cardiac or respiration signal. Such motion-induced artefacts can thus be reduced as the X-ray exposure is synchronised with a patients physiological rhythm, or indeed an objects motion [206].

X-ray based Computed Tomography (CT) imaging is fundamental to medical diagnostics. However, at present such imaging devices are affected by long scan times. Enabled by advances in computer and detection technologies, integration time issues are exacerbated when considered in relation to the concept of real-time 3D X-ray imaging. Multi-pixel sources offer one viable solution. Multi-pixel sources present a unique opportunity in next-generation inspection techniques, especially when one considers the physically small CNT FE cathodes, and the potential for nanoengineering and rapid switching. Pulsed multi-pixel sources give rise to a series of previously challenging applications typically relating to real-time medical diagnostics [234]. Tomography relies on the acquisition of large numbers of images, whilst viewing a subject at a range of angles. These systems, at present, incorporate multiple sources which are sequentially switched. This allows the system to capture an array of independent images without the necessity to physically rotate components, patients, or samples on an expensive and bulky gantry. The emerging alternative is to have multiple, distributed X-ray sources. Here, the subject remains stationary with the added benefit of more rapid data acquisition and potentially higher image quality. Indeed, such systems are largely immune to image blurring, which persists in conventional moving-source systems.

A range of different approaches to the application of tomography or tomosynthesis have been presented. The University of North Carolina, as illustrated in Figure 7(a), have demonstrated high-resolution stationary tomosynthesis using 31 individually addressed CNT X-ray sources distributed within a single large vacuum enclosure [234]. Each source had a 0.6 mm focal spot size, was operated at 28 kV at an anode current of 38 mA. Though an excellent demonstrator of the potential of CNT-based multi-pixel sources, the system is extremely large and fails to fully exploit the potential of emerging CVD techniques to pattern multiple sources on a single chip as a means to facilitate miniaturisation. Nevertheless, the system is at present one of the most advanced CNT-based FE X-ray sources available. It is based on physically separate CNT ink deposited thin films with inter-emitter pitches of the order of a few tens of centimetres. The system has also, rather impressively, demonstrated real-time 3D image reconstruction. Figure 7(b) shows another stationary computed tomography system by the same team. This system consists of 4 cathode banks, with a total of 52 separate cathode assemblies. Though rather large and limited to stationary or slow moving objects, this prototype system is the first in a new wave of CNT-based real-time 3D tomosynthesis scanners.

Figure 7

CNT-based tomosynthesis systems. (a) A distributed CNT tomosynthesis unit consisting of 31 individually addressable X-ray sources enclosed in a single vacuum chamber, and below a schematic the corresponding drive circuitry. Adapted from [234]. Copyright 2012, American Association of Physicists in Medicine. (b) An actively pumped multi-pixel X-ray source consisting of 52 individually controlled sources. Courtesy of the University of North Carolina at Chapel Hill [235].

High spatial resolution requires a small X-ray spot size. However, practical systems also require a wide field of view in order to be able to inspect suitably large areas. Thus, to realise high resolution large area scans the X-ray source must be constructed from multiple cathodes, each with dedicated focusing electrodes. One way to maximise the field of view is to carefully engineer the vacuum chamber to maximise the total angular field of view. A 5 pixel source system has been constructed by Zhang et al. [236]. This system provided sequentially acquired images over a finite angular range, each fitted with an electrostatic focusing electrode to control the size and scanning of the focal spot, which had a width of less than 300 μm. Operating at 40 kV, they showed that the electronic switching time was largely negligible with high emission reproducibility, both in FE curves and beam diameter, between the individual electron sources. Yang et al. demonstrated a multiple source digital breast tomosynthesis unit using an array of 25 cathodes, each individually programmed using analogue controls. Their sources employed CNT-based inks, each with a dedicated gate and an electrostatic focusing ring to control the focal spot size to approximately 200 μm [237]. In this system an impressive total angular field of view of up to 48° was realised, and the performance compared favourably with commercially available moving-source systems. A similar system, termed a multi-beam FE X-ray source, described a similar geometry [203]. A simple linear multiple source using an array of 31 CNT FE cathodes has also been constructed, for the purposes to digital breast tomosynthesis, which was shown to be capable of a total scan angle of 30° [171]. In a further extension, a digital chest tomosynthesis scanner was demonstrated [238]. This system used 75 CNT FE sources, operating at up to 80 kV with a 20° beam angle.

In the majority of the multi-pixel systems discussed the electron sources have largely been constructed using individual cathodes fabricated in linear arrays. A two dimensional geometry based on a series of linear arrays in a square or hexagonal format of physically separate detectors and sources, has been evaluated using simulation. Such a system is highly manufacturable [239], however; the inspection technique relies on close alignment of the generators and detectors which is likely challenging. Guzilov et al. [240] proposed circular arrays of FE emitters to emulate the classical rotary gantries used for CT, although their practical implementation has been limited to staggered linear arrays. Two dimensional arrays have a number of advantages over their linear counterparts, chief amongst which is the potential for ad hoc beam shaping of the emergent X-rays. Micro-integration of two dimensional source arrays allows for ever higher spatial resolution and subsequent beam shaping. Such micro-integrated two dimensional FE source arrays have been realised, though only few such devices have been investigated. Figure 8(a) depicts a 5 × 10 electron beam array fabricated by Wang et al. [169], as mentioned above. Deposited using inking techniques (Figure 8(b)), the 50 individually addressable electron beams allow for arbitrarily defined, and controllable electron, and hence X-ray beam shapes, for application in radiotherapy. An example of one such ad hoc beam shaped, electronically defined, X-ray intensity distribution is illustrated in the bottom image of Figure 8(c). A dose rate of 1.24 Gy/min at the centre of the irradiated object was obtained at an emission current of 3.0 mA [169]. High energy sources for tomosynthesis based on such multi-pixel sources, with high anode potentials of up to 160 kV, have also been demonstrated by Sprenger et al. [241]. This was achieved using a square array of 4 × 13 electron sources for applications including image guided radiotherapy.

Figure 8

Multi Pixel Sources. (a) Electrophorectically deposited multiple X-ray source chip and the corresponding (b) scanning electron micrograph of deposited CNTs, and (c) he resulting electronically reconfigurable X-ray distribution. Adapted from [169]. Copyright 2011, American Institute of Physics.

X-ray generators are often bulky devices, as a result of the thermal disspiation and subsequent cooling requirements. Certain user specifications necessitate significant miniaturisation. Miniaturised sources find use in hand-held non-destructive testing, X-ray spectroscopy, electrically controlled brachytherapy and in situ radiation therapy. Miniature TE sources have been available for many years, but have proven challenging to miniaturise [242]. Amptek has launched a series of miniature X-ray sources [243]. Sealed FE sources, on the other hand, lend themselves to these applications, and in particular are well suited for portable X-ray applications. Though difficult to manufacture, miniaturised sources are emerging on the market place. As in the case of pulsed and multi-pixel sources, much progress has been made towards a miniaturised CNT-based FE X-ray source though, to the best of the authors’ knowledge, no such devices have, as yet, made it to market.

One of the smallest tubes fabricated, at the time of publication, as far as the authors’ are aware, is that of the Electronics and Telecommunications Research Institute, Korea (Figure 9(a)). This cylindrical diode source, which is smaller than the average human finger, has an impressive outer diameter of only 6 mm and a length of 31 mm, and is based on CNTs deposited by a paste printing technique [208,244]. Developed for miniaturised X-ray diffractometry, the X-ray source of Sarrazin et al. [182], Figure 9(b), can support current densities of up to 1 A/cm², whilst Heo et al. [207], who demonstrated a 10 mm diameter and 50 mm long miniature source, as illustrated in Figure 9(c), derived around 0.12 A/cm². The sealed device can be operated up to 70 kV. The cathode was formed from sintered SWCNTs mixed with Ag nanoparticles to form a paste which was deposited on a flattened, 0.8 mm diameter, W wire. Here a conically machined Be window was magnetron sputtered with a 1.5 μm W thin film forming a transmission type X-ray target. The vacuum brazed alumina tube with Kovar electrodes and non-evaporable getters weighs just 14.5 g. The bias required to extract 10 mA/cm² was 29 kV, with a maximum current density of 0.12 A/cm², giving a nominal stability of ±2% at 50 kV, with a similar variation in the X-ray dose. Radially, the X-ray dose rate was 108 Gy/min and was broadly isotropic (20% variation) with a typical focal spot size, as defined by EU standard EN 12543–5, of the order of 3–4 mm. Serious heating prevented the source from being operated for significant lengths of time at the maximum anode bias. Kim et al. [154], as shown in Figure 9(d), have developed a similar brazed compact sealed source, measuring just 3.5 cm in diameter and 7.5 cm in length. This pentode device - consisting of a cathode, gate, two focusing electrodes, and an anode – had a focal spot of the order of 300 μm in diameter when operated with an anode current of 50 mA. The CNT cathode was just 0.15 cm² and was formed form a screen-printed and ball-milled complex CNT paste combined with inorganic fillers, oxides, metallic nano particles and a photo-initiator monomer with organic components including acrylates. The CNT films were relatively macroscopically uniform between samples, though rather perturbed and irreproducible microscopically. No active pumping system was integrated, though a non-evaporable getter was fabricated in an attempt to maintain the vacuum environment. The tube pressure was only of the order of 10⁻⁶ mbar upon sealing. The cavity pressure during emission was not noted. Pulse mode operation (10% duty cycle) was opted for – at 30 kV and 10 mA, to reduce thermally stimulated outgassing. Nevertheless, the tube still operated at approximately 200°C which almost certainly compromised the vacuum, even for a previously well-outgassed cathode assembly. Figure 9(e) shows the Oxford Instruments X-ray Technology Inc. miniature MWCNT FE X-ray tube with a 2 mm diameter cathode and integrated gate. The device develops a somewhat asymmetrical and diffuse focal spot with a major axis length of the order of 700 μm [182].

Figure 9

Miniature CNT-based FE X-ray sources. (a) A disordered miniature CNT, brazed triode X-ray source fabricated by ball milling, and firing of a CNT paste on the apex of a 0.6 mm Kovar rod, and operated at 12 kV [244]. Copyright 2013, Electronics and Telecommunications Research Institute [208]. Copyright 2013, American Institute of Physics (b) A miniature X-ray source fabricated by thermal CVD capable of supporting current densities of up to 1 A/cm² [182]. Copyright 2004, Joint Committee on Powder Diffraction Standards. (c) A sintered SWCNT paste transmission-type miniature CNT cathode X-ray source, weighting 14.5 g, and operating up to 70 kV with uniform spatial dose distribution [207]. Copyright 2012, Springer. (d) A 7.5 cm long selaed CNT-based X-ray source operating at 30 kV. [154]. Copyright 2013, Society of Photo-Optical Instrumentation Engineers. (e) An Oxford Instruments X-ray Technology Inc. miniature field emission X-ray tube with gated MWCNT 2 mm diameter cathode. Adapted from [182]. Copyright 2004, Joint Committee on Powder Diffraction Standards.