Ultraviolet light stimulated water desorption effect on emission performance of gated field emitter array

Nano-vacuum devices^1–3 are being considered for harsh environment electronics^4,5 as they are capable of a stable operation at high temperature and in high radiation environments.^6,7 Nanoscale vacuum devices are the modern, compact versions of early vacuum tubes,⁸ without most of the issues such as high voltage requirement and large dimensions. Thus, nanoscale structures with field emitters^9–11 could be an alternative for modern, small scale, extreme environment electronics, which could be used in nuclear reactors, satellites, etc. Because they operate in vacuum, field emission based nanoscale devices have many advantages¹² over semiconductors. Electrons can travel in-between nanogaps ballistically, which can be combined with miniaturized fabrication and integration. Also, like its predecessors (large scale vacuum tubes), nanoscale vacuum devices are resistant to radiation and can operate at high temperatures, as the field emission mechanism, which follows the Fowler–Nordheim¹³ (F–N) model, is developed upon tunneling of electron into vacuum. With state-of-the-art technology, these vacuum transistors should have the precedence of vacuum tubes with the compact dimension of modern age semiconductor devices. The concept of a nano-vacuum devices⁹ is not new and has been exhibited with an wide range of materials such as molybdenum Spindt arrays,¹⁴ silicon,^9,15 diamond,^4,5,16,17 tungsten,¹⁸ carbon nanotubes,¹⁹ and graphene.^20,21 However, modern fabrication methods²² have reinvigorated interest in the nanoscale vacuum devices that can have a comparable dimension and electrical operation parameters to semiconductor devices.

Several methods are being studied to fabricate devices with three terminals. In a recent work, the upright structure²³ was used in a vacuum nanochannel transistor.²⁴ Several other efforts have proposed different types of Nano-Vacuum Channel Transistors (NVCTs) with a vertical design, where a direct out of the plane emission of electrons can be obtained, e.g., Spindt-type¹⁴ or the slit-type²⁵ NVCT. However, studies are needed to find the influence of residual gas adsorption and desorption on emission and dielectric surfaces. This effort will improve our understanding of the overall device performance, especially when atmospheric pressure operation^26,27 of some NVCTs is being proposed.

In our previous work,²⁸ we have measured the field emission performance of silicon gated field emission arrays (GFEAs) with 1000 × 1000 tips. The gate and collector were studied as a function of temperature ranging from 25 up to 400 °C and then back to 25 °C. From that work, between 300 and 400 °C, observed current at the gate reduced by greater than ten times, whereas measured current at the collector enhanced by greater than ten times. These results were repeated for multiple devices. To understand the cause, during the high temperature I-V characterization, in situ measurements of the outgassing from the test structure were carried out using a residual gas analyzer (RGA). From the experiment, it was clearly observed that water desorption took place. A high-resolution temperature test was also performed between 300 and 400 °C with an increment of 10 °C. From the high-resolution test, it was observed that the maximum water desorption took place at 350 °C and that the transitions of gate and collector currents took place over that range. The correlation of the transition with respect to water vapor desorption very clearly demonstrated that water vapor desorption reduces²⁹ the gate leakage current and improves field emission from the tips. A subsequent test, where a wafer section was kept in room air for a few days, showed that the gate and collector currents went back to their previous values when tested again; however, the arrays again returned to the enhanced collector current and reduced gate current upon heat treatment. Those results demonstrate that removal of residual gases from surfaces of the emitter and the dielectric surface between the gate and emitters has a strong effect on emission performance. Hence, operation of NVCTs without a clean surface could affect emission performance negatively.

To mitigate issues associated with direct heating of the sample, other viable options such as a UV light-based water desorption system³⁰ could be adopted. A UV light lamp, mounted inside the chamber system, can be used to desorb water molecules and clean up surfaces, thus allowing testing inside the vacuum chamber^31–33 without baking. The absorbed water molecules will absorb the UV light until they become sufficiently excited to overcome the molecular bonds and desorb from the surface.^31,33 The UV light is reflected internally from the vacuum chamber and other surfaces so that a direct line-of-sight exposure is not an essential requirement. UV excitation is a very efficient method of energy transfer as the energy is absorbed directly by the water molecule, and only negligible amounts of heat are generated.

In this new work, the same Si based gated field emission arrays^9,22,34,35 used in our prior work²⁸ have been characterized with respect to the UV exposure time to interpret the pertinence of the dielectric and emitter surface conditions³⁶ on the performance and the efficacy of UV exposure to replicate the results of high temperature (350 °C) bakeout. The setup and configuration for the experiment will be discussed first followed by test results from device testing under UV light exposure. The device characterization measurements will be analyzed in terms of adsorption/desorption of residual gases.

The field emitter array fabrication process and scanning electron microscope images are described in detail elsewhere.^22,34 Briefly, an n-type, single crystal, silicon wafer was used to fabricate the arrays. A 200 nm wide, 10 μm long silicon nanowire was used to form the emitter structures. An annular polysilicon gate with apertures of ≈350 nm diameter surrounds the tip. The tip radius varies from 5 to 7 nm. With an active emission area of 1 × 1 mm², the 1000 × 1000 array can generate a current density of ≈100 A/cm² with an exhibited lifetime of greater than 100 h.³⁴ A stainless-steel vacuum chamber system was used to carry out the I-V characterization test. The system is fitted with electrical feedthroughs and a 3D manipulator rod. As shown in Fig. 1(a), the system includes an Extorr Inc. XT100 RGA device and a UV light-based water desorption system (RBD Instrument, MiniZ). The UV wavelength is 185 nm, and the output power is 350 μW/cm². The operating UV lamp in the vacuum chamber is shown in Fig. 1(b). A turbomolecular pump with a dry scroll pump connected in series was used to achieve a high vacuum of ≈5 × 10⁻⁸ Torr inside the chamber. The pressure inside the chamber is measured by both an ion gauge and the RGA.

The test setup (Fig. 2) includes a multiaxis manipulator arm consisting of a gate pad probe pin and a stainless-steel rod as collector, a substrate jig, and the die under test. The substrate holder test jig [Fig. 2(a)] includes a low temperature co-fired ceramic (LTCC)³⁷ bottom and top plate to electrically isolate the wafer section. The top LTCC plate has a rectangular grooved cutout for the wafer section to sit-in as well as a backside ground electrical connection for the silicon emitter tips. Electrical connections from the electrical feedthroughs to the test structure terminals were made using Kapton coated metal wire. Figure 2(b) shows the wafer section under test. The wafer has multiple types of field emitter arrays. Only the 1000 × 1000 arrays were characterized for this work, as they can produce a large current at low voltage.

For these experiments, the pressure was kept below 5 × 10⁻⁸ Torr without UV excitation. For the gate pad and current collector connection, a molybdenum micromanipulator and stainless-steel rod mounted on the same manipulator probe arm, were used, respectively. However, this configuration also restricts the independent control of the collector to emitter gap which is ≈2–4 mm. An optical microscope was used to view the manipulator probe arm and gate pin to make the gate pad connection. A source-measure unit (Keysight B2902A) was used to bias the gate and collector voltages and to measure the gate and collector currents. The source-measurement unit has a voltage range of ±210 V and is capable of measuring current with a resolution of 10 fA.

I-V sweeps, without the UV and then under the UV exposure, were carried out to study whether the UV could affect the gas desorption on the GFEA and to measure the effects of the UV exposure on the leakage current to the gate and on the collector current in comparison with our prior high temperature tests.²⁸ A fixed DC voltage of 100 V was applied to the collector, and a sweep voltage of 40 V DC was applied to the gate with an increment of 0.05 V. The total sweep time was 5 min, and the sweep voltage was kept below 40 V to avoid over heating the collector.

The I-V measurement graph for the 1000 × 1000 array device is shown in Fig. 3(a). First, without the UV light exposure, I-V sweeps were carried out, and the case after the current saturation was plotted and denoted as 0 min. The UV light shown in Fig. 1(b) was then turned on, and I-V sweeps were performed at 20, 40, 60, 80, and 100 min of UV exposure. After each exposure time and shortly before each I-V sweep, the UV light was turned off to avoid unwanted photoemission. The collector I-V curves at each time point is shown in Fig. 3(a). As seen in the plot, before UV exposure, the collector current was ≈94 μA at 40 V. After 70 min of UV exposure, the collector current increased (more than ten times) to ≈1.33 mA at 40 V. Corresponding emission current density, which was 0.0094 and 0.133 A/cm² for before and after exposure, respectively, was calculated using the active emission area of 1 × 1 mm². After 100 min, the increased collector current remains constant and is shown in Fig. 3(a). The leakage current between the emitter and gate also largely reduced from 6 to 0.46 mA (more than ten times) after 70 min of UV exposure as can be seen in Fig. 4(c). The transition of gate and collector currents occurs at 20–40 min into UV exposure. Also shown in the Fig. 3(b) is the Fowler–Nordheim¹³ plot of the collector current, which will be discussed later.

These improvements in collector current can be ascribed to desorption of water vapor and possibly other adsorbates under UV exposure.³³ The adsorbates increase the work function of the emitter, thereby decreasing emission current from the tips, which is a well-known phenomenon.^29,38,39 The enhanced electron transport through surfaces along the walls between the emitters and the gate can also be attributed to adsorbates,^20,40,41 which consequently increases gate leakage. This effect was observed in our previous work.²⁸ 

Effects of the water desorption on field emission performance was observed more clearly in Fig. 3(b) which shows the F–N collector current plot before and after UV cleaning. The field emission characteristics are far less linear pre-exposure but show a much linear F–N plot after UV cleaning indicating a cleansed (fewer adsorbates) surface.

To verify that this process is indeed due to adsorbates, a separate set of tests were carried out. After an array was measured under UV light, it was taken out from the vacuum chamber and kept in the atmosphere for three days. This process was replicated several times for the same array. Results of one of the test sets is summarized in a bar chart in Fig. 4, which shows that the increased emission current and decreased gate leakage current were not permanent. After sitting in the atmosphere, the higher gate and lower emission current return, but the changes are again recovered with subsequent UV exposure. The process can be replicated each time the device is exposed to air followed by UV. These results confirm that the adsorption and desorption of residual gases (likely water) from the emitter and gate/emitter dielectric surface are the likely cause of the emission and leakage current changes.

Outgassing measurements were also performed using the RGA to study the desorption versus UV exposure time. The I-V measurements and pressure measurements were carried out at every 20 min of UV exposure. Figure 5(a) shows a bar chart of the partial pressure of water vapor inside the chamber over the entire UV exposure time range. Figure 5(b) shows the gate and collector currents versus UV light exposure at a 40 V gate bias. The water partial pressure starts increasing after 20 min of exposure with the peak at 60 min. The transition in the gate and collector currents vary almost linearly with exposure time until leveling off around 80 min. Note that both currents show a similar transition curve indicating the same phenomenon is affecting both. The correlation of the transition of gate and collector current with water vapor outgassing is very clear. This result exhibits that water vapor desorption increases field emission from the tips and reduces²⁹ the gate surface leakage.

To analyze the outgassing and desorption from the vacuum chamber and components, additional RGA measurements were performed. Figure 6 shows the entire mass spectra for 0 (before UV exposure), 60, and 100 min of UV exposure captured using the RGA. The dominant species in the desorbing gases is the water vapor as expected. However, additional mass spectra measurement with respect to exposure time points were also carried out and analyzed to compare the desorption of gas species with and without the field emission array wafer section. Figure 7 shows a bar chart of the partial pressure for additional gases in the chamber for the time points of before UV, 20, 40, 60, 80, and 100 min with and without the wafer section.

From Fig. 7(a), it can be observed that the water vapor partial pressure at 60 min was ≈1 × 10⁻⁶ Torr, when the wafer section was present in the chamber and ≈3 × 10⁻⁷ Torr without the wafer as observed in Fig. 7(b). To confirm that the difference in partial pressure is due primarily to the desorption from the test die, the outgassing rate was calculated for with die and no die cases. The outgassing rate⁴² was calculated using the volume (3.737 l) and inner surface area (1297.17 cm²) of the test chamber and the partial pressure data shown in Fig. 7. The calculated outgassing rate from 40 to 60 min of UV exposure was 9.81 × 10⁻¹³ Torr l s⁻¹ cm⁻² when the die was in the chamber and 1.96 × 10⁻¹³ Torr l s⁻¹ cm⁻² without the section, which is almost five times less.

To improve our understanding of the effects of the treatment on emitter performance, the F–N data from Fig. 3(b) was fit to the F–N¹³ equation

where J is the emission current density in A/cm², V_GE is the applied gate potential in V, and a_FN and b_FN are the F–N coefficients and are defined as follows:⁴³ 

where φ is emitter work function (eV) and β is field enhancement factor. Here, the work function is assumed to be 4.05 eV for Si.^44,45 This value is used only to analyze the data as the work function before UV exposure will be higher because of the water vapor on the emitter surface, but after exposure or high temperature bake, it is expected that the work function will be close to that of silicon.

The values of ln(a_FN) and b_FN were extracted from the slope and intercept of the F–N plot shown in Fig. 3(b), and β (cm⁻¹) was calculated from Eq. (3) using the silicon work function. Table I shows the extracted values for the plots before and after UV exposure along with the current density and the maximum test voltage for the array. Also shown are the extracted values from the temperature testing from our previous work.²⁸ The F–N data come from Fig. 5 in that work. From the table, it can be seen that b_FN, after UV exposure, increases by more than 10% indicating either a decrease in the field enhancement factor, β, or an increase in the work function. An increase in the work function is inconsistent with water vapor desorption, which should cause a decrease in the work function. Using the same work function value, the array β decreases slightly. However, an increase in the emission current and in the magnitude of the ln(a_FN) from −13.975 to −10.71 is also measured, indicating the proposed reduction in work function as expected from water desorption. A similar result is seen for the high temperature results from the prior work²⁸ with similar GFEAs, where the magnitude of ln(a_FN) increased from −10.76 to −8.97 and the b_FN again increased from a magnitude of 147.7 to 227. The resulting β also decreased. Hence, the bake to 400 °C gives similar trends to the F–N characteristics compared to the UV exposure.

To explain these results, first consider that all of the emitter tips have approximately the same work function before they are initially tested. The emission of the gated emitter depends also on the diameter and proximity of the gate to the emitter tip. There is a distribution of gate diameters and gate-to-tip distances likely well described as a Gaussian distribution.⁴⁶ These values are assumed fixed for our experiments and do not change with testing, UV exposure, or temperature unless there is a tip arc damaging the emitter. The tip sharpness also determines emission and is included in β. While it is possible that contaminants and adsorbates on the tip surface might reduce the effective tip sharpness as measured, this seems to be a small effect as will be evidenced by the results. As with the gate diameter, the sharpness of the tips³⁴ is also described by a Gaussian distribution. Our hypothesis is that during initial testing, the sharpest emitters with the closest and smallest gate openings are the first to emit. Therefore, some fraction of the distribution of tips emits initially. This emission can cause desorption and some level of cleaning of those emitters as evidenced by the burn-in or conditioning effect where emission current increases while at a fixed voltage during initial testing. Once fully conditioned, those emitting tips are now stable and may have a somewhat lower work function, due to this burn-in, but the remaining nonemitting tips are unchanged, and these are the less sharp parts of the tip population. Now UV exposure or high temperature is applied desorbing the water vapor and other adsorbates. Upon conclusion, all tips now have approximately the same lower work function (near that of silicon). However, because the work function decreases, emitter tips that were either not emitting because they were not sharp enough (nonemitting part of the distribution) or were only emitting slightly before treatment now emit.

Hence, the fraction of tips emitting increases, likely significantly in our experiment, so that a much larger fraction of the tip sharpness distribution is now emitting current. This increase in the fraction of emitting tips accounts for much of the ten times increase in emission current.

This effect results in a decrease in the effective array sharpness (β) measured for the array. In other words, the lower work function results in duller tips emitting, which then skews the array β lower and increases the resulting b_FN,

I-V characterization of Si-GFEA with 1000 × 1000 tip arrays were carried out before and after UV light exposure. It was found that the gate current reduced by more than ten times and the emission current improved by more than ten times after 60 min of UV exposure. These results were replicated for several arrays. An RGA was used to capture the outgassing from the devices during the UV irradiation. These experiments show very clear water desorption during the hour-long exposure. If the devices were kept in atmosphere for three days, the collector and gate currents went back to their previous values, but subsequent UV exposure recovers the gate and collector currents to their enhanced levels. These results strongly suggest that desorption from the emitter surface and from the dielectric between the gate and emitters is affecting field emission array performance as previously demonstrated using high temperature. F–N plots of the collector I-V curves show a straighter line after UV exposure, which indicates a cleanser emitter surface. This result is backed by fitting the data to the F–N equation and finding that a lower work function is required to fit the data along with a lower array tip sharpness. This lower effective array sharpness is hypothesized to be due to a much larger fraction of tips now emitting because of the reduced work function; these duller tips were not emitting prior to UV, and they skew the resulting array β lower. A very similar result is seen from temperature testing of a different array. Hence, UV irradiation shows a much less complex and simpler method of water desorption for cleaning Si-GFEAs. This technique can be used in testing and in vacuum packaging of devices where high temperature sealing is not desired. Future work will study different emitter materials and the effects of adsorption of different introduced gases.

Material support for this research was provided by the Air Force Office of Scientific Research under Award No. FA9550-18-1-0436. The authors would also like to acknowledge the support from Mason Cannon, Gerardo Herrera, Patrick Epperson, Patrick Ward, John McClarin, Liz Gaffney, Robby Davis, and Jessica Carlson.

The data that support the findings of this study are available from the corresponding author upon reasonable request.