Growth of GP arrays occurred on electrode-grade graphite owing to its high electrical/thermal conductivity. The sample size was approximately 10 mm × 5 mm in area with a thickness of 2 mm. A catalyst-free petal growth process was performed in a SEKI AX5200S microwave plasma chemical vapor deposition system utilizing H₂ and CH₄ feed gases in conditions described previously (Bhuvana et al., 2010; McMullen, 2010; Vander Laan, 2011). The sample was placed atop a 2.5 cm ceramic post in order to promote coupling of the plasma to the sample. The plasma was initiated in 50 sccm of H₂ at 10 Torr pressure. No substrate heat was applied, but exposure to the plasma raised the substrate temperature to near 600 K with the sample temperature presumably being higher than that due to a more direct exposure to the plasma. The initial microwave generator power was 300 W. The pressure and generator power were increased stepwise, with pressure increases preceding generator power increases. The pressure was first raised to 20 Torr, and then power to 500 W; then the pressure was taken to 30 Torr, and finally the power was set to 700 W. This stepwise increase ensures that the power is not too great for a given chamber pressure to cause plasma instabilities. Once the final growth conditions were established (30 Torr pressure, 700 W generator power), a flow of 10 sccm of CH₄ was added to the H₂, and petal growth occurred for 30 min.

Boron nitride modification of the GP arrays was performed following a previously reported procedure utilizing a microwave heat-assisted chemical treatment in a homogenous water solution of boric acid (H₃BO₃) and urea (CON₂H₂) (Paul et al., 2012). Afterward, the sample was vacuum treated for 12 h to remove water content. Finally, the sample was annealed in a N₂ environment at 1,180 K for 12 h. SEM images of the C_x(BN) petals are provided in Figure 1.

Raman spectra of the GP and C_x(BN) petal samples are provided in Figure 2. Peaks at approximately 1,360, 1,560, and 2,720 cm⁻¹ correspond to the D, G, and 2D peaks, respectively. The G peak indicates sp² bonding, while the D peak indicates breaks in sp² bonds and is commonly referred to as the defect peak (Ferrari, 2007). The 2D peak is the second order harmonic of the D peak. GP arrays have a relative peak intensity I_D/I_G = 0.67, while the C_x(BN) petal sample exhibits I_D/I_G = 0.28, suggesting that the boron nitride modification preferentially takes place at the defect sites of the GP arrays.

Given that potassium reacts readily with water and oxygen, sample loading for potassium intercalation was performed in an inert Ar environment. An Ar-filled glove box was used to load the sample into a sealed glass tube with a pure potassium source. The glass tube was then loaded into a N₂-filled tube furnace (Lindberg/Blue M) where a two-step intercalation process was performed (McMullen, 2010; Westover et al., 2010; Vander Laan, 2011). The vessel was first heated to 540 K for 48 h, followed by a second heating process at 340 K for 15 h. The first process vaporizes the potassium source, coating the sample, while the second stage occurs just above the melting temperature of potassium to allow for the physisorbed potassium to diffuse into the petal lattice. Once cooled, the intercalated sample was loaded into a different Ar-filled glass tube without a potassium source prior to testing. The sample was unloaded from the Ar-filled glass tube prior to loading into the vacuum chamber for testing, resulting in exposure to air of several seconds, including additional exposure within the chamber itself prior to air being pumped out of the chamber. All told, several minutes of exposure to air is typical.

Chemical composition of the samples was investigated by XPS using a Kratos Axis Ultra DLD spectrometer with monochromatic Al K α radiation (hν = 1,486.58 eV). Survey and high-resolution spectra were collected from a 700 µm × 400 µm spot size in the normal direction with respect to the sample surface. Fixed analyzer pass energies of 160 and 20 eV were used for the survey and high-resolution spectra, respectively. XPS data were analyzed with commercially available CasaXPS software (www.casaxps.com), and individual peaks were fitted to a Gaussian–Lorentzian function. The atomic concentrations of the chemical elements on the near-surface region were estimated after subtraction of the Shirley-type background, taking into account the corresponding Scofield atomic sensitivity factors and inelastic mean free paths of photoelectrons using standard procedures in the CasaXPS software. A charge neutralizer was used for C_x(BN) petals because of their low electrical conductivity, and the resulting spectra were then corrected on the binding energy scale using the position of the main C 1s component at 284.5 eV as a reference for graphitic carbon (Paul et al., 2012).

In this study, a Newport solar simulator (model 69907 power supply and model 67005 lamp box) with a xenon lamp (model 6255 bulb) and an AM1.5 global filter (model 81094) provides a source of electron excitation for the samples studied. By knowing the wavelength-dependent spectrum of light provided by the solar simulator and utilizing the proper photoemission theory, one can calculate approximate work functions for the materials under study. Figure 3 provides a basic schematic of the photoemission process performed in this report. A solar flux illuminates the sample surface with a spectrum of energy-dependent photons. Photons with energies above the work function (φ) can promote electron emission depending on the direction of emission, while photons below φ will merely heat the surface. An energy analyzer sweeps through specific electron energies such that only emitted electrons within a small energy window are recorded at a particular instant. Given the directionality of electron emission, both low-energy and high-energy photons can contribute to low-energy electron emission, while only high-energy photons can contribute to high-energy electron emission. Therefore, more photons in the solar flux can contribute to the emission of an electron whose final energy state is close to the Fermi energy plus work function (E_F + φ) than an electron with final energy state greater than that. A sharp peak in the PEED near E_F + φ is generated as a result, with a nearly exponential decay in emission at higher electron energies.

As indicated in Figure 3, the characteristic shape of the PEED work function reflects a sharp rise in electron emission followed by a smooth exponential-like decaying trailing edge. The energy at which this sharp rise occurs is a reliable measure of the sample work function. The overall shape of the PEED at higher energies generally reflects the decrease in illuminated photons capable of inducing electron emission. With the complex materials developed in this study, multiple rises in electron emission across the PEED and/or a broadening of emission peaks may occur. This behavior is a result of multiple emission sources being present on the surface of the sample, thereby demonstrating several work functions present in the material.

Extraction of one or more work functions from the geometrically complicated and chemically diverse surfaces in this study can be accurately performed via least squares fits to the PEED data of an appropriate photoemission model. The photoemission model below is fit to the experimental PEEDs and best estimates are taken to determine whether a sample exhibits a single emission peak, or multiple emission peaks with the fitting procedure adjusted accordingly. The resulting work functions are considered accurate within ±0.2 eV, accounting for inaccuracies in experimental measurements and data fitting.

Knowledge of thermionic emission theory is critical to developing the photoemission model utilized here. In thermionic theory, electrons are assumed to behave as a free electron gas, and the material is assumed to have a single parabolic conduction band, such that (McMullen, 2010; Westover et al., 2010): (1)ITEED(E) dE=4πmeh3(E−EF−ϕ)H(E−EF−ϕ)1+exp(E−EFkBT)dE where E is electron energy, I_TEED is the thermionic emission intensity, m_e is the rest mass of an electron, h is Planck’s constant (4.14 × 10⁻¹⁵ eV⋅s), E_F is the Fermi level, ϕ is the work function, H is the Heaviside step function, k_B is the Boltzmann constant (8.62 × 10⁻⁵ eV/K), and T is temperature.

Photoemission theory can be complicated, but a simplified three-step model is often utilized (Berglund and Spicer, 1964; Spicer and Herreragomez, 1993). Photoemission occurs by: (1) photon penetration and absorption, (2) electron excitation and transport to the surface, and (3) electron emission over the potential barrier (Jensen, 2007). While this is an effective model for many photoemission processes, it requires an extensive knowledge of the material characteristics (e.g., surface geometry, electronic band structure) that are often unknown when studying photoemission properties of newly developed materials.

An alternative model was developed by Westover and McCarthy (Westover et al., 2010; McCarthy et al., 2014) as an extension of the theory developed by Fowler and DuBridge (Fowler and Nordheim, 1928; DuBridge, 1933) and offers a generalized approach for predicting photoemission with regard to specific material properties. As with the thermionic theory embodied in Eq. 1, this model assumes a free electron gas with a single parabolic energy band. Additional assumptions are that photon absorption is independent of photon energy, and that all photon energy is manifested by electron momentum normal to the surface. Given these assumptions, the electron population for a three-dimensional material, described by the Fermi-Dirac function, is simply increased by the photon energy (Vander Laan, 2011): (2)IPEED(E) dE=4πmeh3(E−EF−ϕ)H(E−EF−ϕ)1+exp(E−EF−ℏωkBT)dE where I_PEED is the photoemission intensity, and ħω is the photon energy. Equation 2 is termed the normal energy model. Westover and McCarthy further improved this model by weighting the number of electrons available for emission based on the angle of electron emission, for which the derivation is provided elsewhere (Westover et al., 2010; McCarthy et al., 2014). The angle of photonic illumination is referenced to the surface normal. The random energy model is given by: (3)IPEED (E) dE=∑n=1NNavail,ΔθP(θp)ΔθPNavail,totalIPEED,ΔθP (E) dE where (4)Navail,Δθp(θp)=[∫EF+ϕ−ℏωcos(θp)∞kBT2meWme(meh)3ln(exp(EF−WkBT)+1) dW×[(cos (θp) −Δθp2)−(cos(θp)+Δθp2)]] (5)IPEED,ΔθP (E) dE=4πmeh3(E−EF−ϕ)H(E−EF−ϕ)1+exp(E−EF−ℏωcos(θP)kBT)dE

N_avail,_ΔθP is the number of electrons available for emission at a given angle of photonic illumination, W is the energy associated with momentum in the normal direction, and θ_P is the angle of photonic illumination. A final modification of Eq. 3 is performed in order to account for electron scattering and “smearing” due to the energy analyzer. As these scattering events are not discussed extensively in this report, the details of the scattering theory and additional data related to scattering are provided in Supplementary Material.

By fitting the foregoing photoemission theory to the resulting temperature-dependent PEEDs, electron emission sources from the samples can be isolated and identified as either basic materials (e.g., potassium, graphite, boron-nitride) or newly developed emission sources resulting from potassium intercalation and/or boron nitride modification.

Photoemission characterization is performed utilizing a SPECS-Phoibos 100 SCD hemispherical energy analyzer (HEA) built into a high-vacuum system. The sample stage has a heater beneath it, with the HEA entrance located 40 mm above the sample. Substrate temperature is monitored by a thermocouple placed beneath the stage. Additionally, at temperatures above 850 K, a pyrometer is used as a secondary temperature measurement. The HEA system records the number of electrons collected per second for a given electron energy. Pressure within the chamber was approximately 2.0 × 10⁻⁸ Torr for room temperature studies, but could rise as high as 3.0 × 10⁻⁷ Torr for substrate temperatures of 1,180 K. A Newport solar simulator (Model 69907 power supply and Model 67005 lamp box) with a xenon lamp (Model 6255 bulb) and an AM1.5 global filter (Model 81094) provides photonic illumination of the samples. The substrate was supplied with a bias of −4 eV to avoid contributions from the detector near 0 eV. The supplied bias was subtracted from all of the data reported here. The chosen bias will impact the intensity of the recorded PEEDs and as such should be kept the same for all studies.

Potassium intercalated within graphite tends to exhibit a high mobility that is temperature dependent. This mobility causes deintercalation of the sample, thereby reducing the effectiveness of the sample as an electron emitter. It is, therefore, important to evaluate emission intensity and sample stability with respect to temperature. Shifts in emission peaks, changes in PEED shapes, and multiple work function peaks often develop over multiple heating cycles. The first heating cycle in particular could produce large changes in PEEDs with temperature, as oxides formed on the sample surface during pretreatment and transfer to the vacuum chamber presumably bake off. In order to investigate the sample work functions and temperature limits of the samples in this study, a systematic heat treatment schedule was developed. Four heating cycles of the samples were performed over 4 days (one per day). A cycle consisted of heating from room temperature to a selected maximum temperature, at which point the sample was allowed to cool to room temperature overnight. The next cycle would begin the following day. Each day, PEEDs were recorded at room temperature (310 K) and then at 100 K intervals starting at 380 K. The selected maximum temperatures for cycles 1–4 were 580, 780, 980, and 1,180 K, respectively. For example, cycle 1 was performed on day 1 with recordings taken up to a maximum temperature of 580 K. Cycle 2 was performed on day 2 with testing up to a maximum temperature of 780 K, and so on. Figure 4 provides the temperatures at which measurements were performed each day.

Three main samples were tested via XPS. The first sample consisted of GP arrays grown on electrode graphite. A C_x(BN) petal sample was tested second. The last sample was also C_x(BN) petals but was intercalated with potassium [C_x(BN)-K]. Figures 5 and 6 display X-ray photoelectron survey spectra performed on the GP, C_x(BN) petal, and C_x(BN)-K petal samples. Table 1 contains the elemental compositions derived from XPS analysis of all three samples. High concentrations of boron and nitrogen are detected in the C_x(BN) petal samples, but as a byproduct of the acid treatment, a significant amount of oxygen is also present. Oxygen content is potentially detrimental to sample performance, partially because of the rapid oxidation of potassium. The increase in oxygen content of the C_x(BN)-K petal sample relative to the C_x(BN) petal sample is likely a result of oxidized potassium on the surface.

Figure 6 contains the fitted C 1s spectra for graphitic petals, C_x(BN) petals, and C_x(BN)-K petals together with the K 2p spectrum for C_x(BN)-K petals. Negligible amounts (≈1 at %) of fluorine and silicon were detected in the XPS analysis of the C_x(BN)-K petals that were most likely introduced during sample handling in the glove box.

A GP array grown on electrode graphite was intercalated with potassium (GP-K) and was studied in this effort along with a potassium-intercalated boron nitride modified petal sample [C_x(BN)-K]. Results from the first heating cycle (day 1) are presented in Supplementary Material. Data resulting from the first heating cycle are not discussed in detail here as significant variation in PEEDs occurred owing to oxidized surface product bake off from the samples. This oxidation is largely removed during the first heating cycle, allowing for more consistent emission. Figure 7 contains PEEDs from potassium-intercalated samples illuminated by the solar simulator, and associated theoretical fits corresponding to PEED theory, from the second heating cycle (day 2: up to 780 K substrate temperature). The second heating cycle promotes annealing of the petal surfaces, the benefits of which are demonstrated during later heating cycles. Detailed results of the fitted data are provided in Supplementary Material. Plots are normalized for a given set of samples with respect to the PEED with the highest emission peak. Multiple work functions and shifts in peak locations are prevalent with changes in temperature. The maximum emission intensity for each sample occurred between 480 and 580 K. Above 580 K, large reductions in emission were evident with the GP-K sample. For the C_x(BN)-K petal sample, a second high-energy peak emerged above 580 K, coinciding with a significant reduction in low-energy electron emission.

Three distinct emission peaks occurred during these studies. The first work function is 2.3 ± 0.2 eV, the second is 2.8 ± 0.2 eV, and the third is 3.3 ± 0.2 eV. Given that the work function of bulk potassium is approximately 2.3 eV (Osterlund et al., 1999), the dominant emission peak near room temperature during the first two cycles/days of testing likely results from potassium adsorbed on the surface. Osterlund et al. (1999) measured the change in graphite work function with respect to potassium adsorption thickness and showed that a reduction in the graphite work function from 4.7 to 2.6 eV occurred for 0.4-monolayer coverage of potassium. A 1.0-monolayer coverage decreased the work function to 2.3 eV, verifying that ultrathin layers of coverage generate significant reductions in sample work function. The GP-K sample exhibits a photoemission spectrum attributed solely to surface adsorbed potassium at 310 K. Near 580 K, some potassium apparently remains on the surface, but a dominant PEED edge emerges near 2.8 eV, providing good evidence of intercalation. After annealing at 780 K, nearly all potassium has desorbed from the surface with the dominant edge near 2.8 eV indicating that potassium on the petal surfaces has essentially been incorporated (intercalated) into the GPs.

At 310 K, the C_x(BN)-K petal sample exhibits a PEED that is dominated by surface adsorbed potassium, similar to that observed for the GP-K sample. Additionally, presence of a lower energy shoulder suggests that regions of the sample have work functions below 2 eV. As this energy shoulder is far less pronounced in the GP-K sample, its presence is likely attributed to boron nitride incorporated into the petals. Identifying the geometric origin of these low work functions, high-value active surfaces (perhaps petal edges) could be critical to increasing emission intensity. As the temperature increases, the low work function areas (below 2 eV) remain until disappearing near 780 K. The work function for most of the surface shifts to 3.3 eV, rather than the 2.8 eV work function of the GP-K sample. This work function is most likely a result of potassium intercalation, and the increase in work function relative to the GP-K sample can again be attributed to the boron nitride modification. While some charging of C_x(BN) may be present, resulting in a slight shift of the measured work function of C_x(BN), the potassium intercalation of the C_x(BN)-K petal sample should ensure a high electrical conductivity, thereby minimizing charging. Given that the measured work functions arise from potassium, potassium intercalated graphite, and potassium intercalated C_x(BN), the effects of charging were neglected in this study and any inaccuracies were assumed to fall within the ±0.2 eV uncertainty assigned to the work functions reported.

Both samples exhibited repeatable performance after heating to 780 K, but prior to heating above 980 K. Data from the third cycle/day of testing are, therefore, not presented, as those results closely match the PEEDs recorded during the fourth cycle/day of testing. Figure 8 shows solar-induced PEEDs from the potassium-intercalated samples tested in the fourth heating cycle (up to 1,180 K) along with corresponding fits from theory. Figure 8 shows the stability of the GP-K sample after thorough annealing produced by the first three heating cycles. The GP-K sample is very stable during the fourth heating cycle for temperatures below 680 K with a nearly constant work function of 2.8 eV. Emission intensity increases dramatically upon heating to 980 K, with little to no change in work function and PEED shape. This increased emission might be due to the decomposition of potassium oxide (K₂O), potassium peroxide (K₂O₂), and potassium superoxide (KO₂) that likely form overnight in a vacuum of 10⁻⁸ Torr between testing cycles, but further studies would be required to verify. Upon heating above 980 K, the emission intensity decreases rapidly and the work function increases as potassium deintercalates.

The C_x(BN)-K petal sample shows an unstable and evolving PEED in the fourth heating cycle, indicative of a highly dynamic surface. The C_x(BN)-K petal sample exhibits low-intensity, high work function emission (approximately 3.5 eV) at room temperature. At 680 K, the PEED shows emission from low work function sites (about 2 eV), a dramatic rise in emission from 2.5 eV work function sites (similar to the activation in the GP-K sample at 980 K), and sites still having a 3.5 eV work function. For temperatures above 680 K, a gradual reduction in emission occurs and similar to the plain GP-K sample, the C_x(BN)-K petal sample rapidly deintercalates above 980 K. Despite this, the C_x(BN)-K petal sample retains multiple work function peaks, even up to 1,180 K where the PEED shows emission from work function sites of 3.3 eV and 4.4 eV, approaching the work function of pristine GPs.

The contrasting PEEDs of the GP-K and C_x(BN)-K petal samples clearly demonstrate that the incorporation of boron nitride in the GPs significantly alters the intercalation of potassium. One possible explanation is that the binding energy of potassium atoms at the carbon and boron nitride interfaces may be higher than its binding energy in non-modified GPs, causing potassium atoms to be pinned. Pinning of potassium atoms to these interfaces may result in increased retention of surface adsorbed potassium on the C_x(BN)-K petal sample relative to the GP-K sample and, in turn, the retention of multiple work functions exhibited in the C_x(BN)-K petal sample PEEDs. Further discussion regarding the origins of these work samples can be found in Supplementary Material.