Spectroscopic insight of low energy electron emission from diamond surfaces

Low work function materials are desirable in many applications such as electron emission and photo-catalysis. We have studied low energy electron emission from low work function hydrogen terminated diamond surfaces via electron spectroscopy to gain insight into the mechanisms involved during electron excitation and emission. Electron emission was found to be dominated by electrons within the band gap energy region, allowed due to the negative electron af ﬁ nity diamond surface, while sub-bandgap illumination was able to signi ﬁ cantly increase emission current. Substantial upward surface band bending greater than 2 eV was observed for the diamond samples, which affect electron accumulation at the surfaces. Intra-bandgap states are shown to strongly in ﬂ uence electron emission behavior, which can have great implications for various energy conversion devices. © 2021 Published by Elsevier Ltd. This is an open access article under the CC BY license (http:// creativecommons.org/licenses/by/4.0/).


Introduction
Thermionic emission involves the removal of electrons from within a material after absorbing energies greater than the work function (WF), sharing similarities to that of secondary electron emission or photoemission, but simply using thermal energy to overcome the surface potential barrier [1]. This simplicity in generating free electrons allows it to form the basis of a variety of devices such as electron guns, X-ray sources, magnetrons and energy converters [2] [e] [6]. Improvements to the thermionic emission will increase energy efficiency and device performance, while minimizing fatigue over time with reduced operating temperatures.
During the thermionic emission process, the electron thermal energy distribution allows a portion of electrons to occupy surface electronic states above the vacuum level, which can then escape through the surface. Thermionic emission current is thus dependent on the density of states (DOS), temperature and WF of the material, and can be approximated for bulk metals using the Richardson-Dushman equation [1].
where J is the current density, A is a material dependent constant, T is the temperature and F is the WF. With an exponential dependence on the WF, the emission current can be significantly increased using lower WF materials. Although the equation does not take into account the true DOS of the material, by assuming it is free-electron-like, it has generally provided good approximation for a material's WF and thermionic emission profiles. However, it may be less likely to predict thermionic emission behavior from semiconductors due to the existence of a band gap. Diamond is a wide-bandgap semiconductor that can exhibit low WF when nitrogen or phosphorus doped and terminated with hydrogen to produce a negative electron affinity (NEA) surface, such that the vacuum level lies below the surface conduction band minimum (CBM) [7]. When considered with other properties such as vacuum stability and extremely high thermal conductivity, diamond is a strong candidate for thermionic emitter material. However, nitrogen and phosphorus doped diamond (NDD and PDD) have a high activation energy unlike boron doped diamond (BDD) acceptor levels [8,9]. This leads to reduced electrical conductivity until relatively high temperatures, as well as an increased barrier for electron emission. While the temperature-dependent current density from Equation (1) provides quantitative information on the thermionic emission capabilities, it lacks insight on the origins of electrons as well as their energy distribution. Here we have used electron spectroscopy to study the mechanisms involved in thermionic emission from diamond surfaces, which allows for detailed investigation of electronic states and their occupancy. By observing electron emission as a function of electron kinetic energy rather than the total emission current, effects on the energetic emission intensities arising from the WF, band gap, band bending and electron affinity can be visualized.

Methods and materials
Phosphorus, nitrogen and boron doped polycrystalline diamond were grown on 10 mm Â 10 mm nano-diamond seeded silicon substrates in a microwave-enhanced plasma chemical vapor deposition (CVD) reactor. The PDD was prepared at Hasselt University by microwave plasma enhanced chemical vapor deposition (MWPECVD) in a 2.45 GHz ASTeX PDS17 Microwave plasmaassisted CVD reactor with a water cooled molybdenum substrate holder [10]. The total flow rate was 500 SCCM with 0.5% CH 4 in H 2 , and 1% PH 3 in CH 4 . A constant forward microwave power of 2500 W and pressure of 70 Torr was used for a growth duration of 3 h.
The NDD and BDD were grown at the University of Bristol, using an ASTeX-type 2.45 GHz MWPECVD reactor. The total gas flow rate was 312 SCCM with 4% CH 4 in H 2 , where 2.5% N 2 and 100 ppm B 2 H 6 was used respectively in CH 4 for the NDD and BDD. A forward microwave power of 1500 W and pressure of 150 Torr was used, for a growth duration of 30 min. All samples were then hydrogen terminated using the MWPECVD reactor using 300 SCCM of H 2 at 800 C and 500 C for 2 min each. The hydrogen termination was confirmed using contact angle measurements [11,12], showing angles of about 100 in Fig. S1.
The samples were then transferred into the ultra-high vacuum (UHV) chambers at the Bristol NanoESCA Facility, where they were annealed at 300 C for 1 h before analysis and remained under UHV for the remainder of the experiment. The thermal annealing under vacuum cleans the diamond surface and removes any surface conductivity resulting from the negative electron affinity and surface adsorbates [13].
Electron spectroscopy was performed using a hemispherical analyzer with a pass energy of 6 eV, for an analyzer energy resolution of about 300 meV. A bias of 20 V was placed between the sample and analyzer during analysis to overcome the analyzer work function. Heating of the sample during measurements was done via a resistive heating filament behind the sample, where the current-dependent sample temperature was pre-calibrated. A monochromated Al-Ka X-ray source and white LED lights were also used as excitation light sources for some measurements.

Results and discussion
Scanning electron microscopy (SEM) of the two samples are shown in Fig. 1, showing a micro-crystalline diamond structure. The images show no specific crystallographic orientation, meaning that emissions observed are an average of all the present surface orientations. While WF variations commonly exist between different surface orientations [14], their contributions are insignificant for the studies done here. The surface morphology difference between the PDD and NDD are insignificant and thus any emission differences are attributed to the different dopants. It should also be noted that for other electron emission mechanisms such as field emission, the diamond surface morphology and grain boundaries can play a crucial role [15,16], thus the Raman spectra for the two samples are shown in Fig. S2 for reference. Fig. S3a also shows an electron emission microscope of a diamond surface during thermionic emission, indicating that thermionic emission was generally uniform on the microscopic scale, in agreement with other works in the literature [17].
As shown from beta-radiation enhanced emissions, the energy distribution of excited electrons also play an important role towards the emission current [18]. Even at thermal equilibrium, determining the electron energy distribution can also be extremely beneficial towards understanding electron emission from low WF semiconducting surfaces. The spectroscopy of the thermionic emission is therefore performed on the low WF polycrystalline PDD and NDD from Fig. 1.
Electron emission and thermal peak broadening were studied using sample temperatures between 300 and 700 C. Emission at lower temperatures were not tested due to the reduced electrical conductivity of the PDD and NDD, and the insignificant amount of thermionic emission current below 300 C; while thermal desorption of hydrogen would likely begin at temperatures around 700 C [19] [e] [21], which would lead to changes in the electron affinity, WF and surface band bending [22] [e] [25].
X-ray photoelectron spectroscopy (XPS) was performed on the samples at elevated temperatures, where the C 1s core level is shown in Fig. 2. The C 1s binding energies for the hydrogen terminated PDD and NDD are approximately 286.5 eV, about 2 eV higher than typically observed for BDD. This can be explained by the slight n-type doping induced on the PDD and NDD, where the Fermi level (E F ) is closer to the CBM in the bulk, in contrast to the ptype doping in BDD with the E F near the VBM. The binding energy is the energy difference between E F and a core-level, thus a higher E F position will shift the core-level spectra to higher binding energies. The wide band gap in diamond allows for large changes in E F position and thus binding energies with chemical doping.
With the higher E F of PDD and NDD, the shift to higher binding energies for their C 1s peaks relative to a BDD is to be expected [26], though the magnitude of the shift is significantly less than suggested from simple comparison between their bulk donor level positions in diamond. With the E F at 0.37 eV above the VBM for BDD [27], 1.7 and 0.56 eV below the CBM for NDD and PDD respectively [8,28], a C 1s binding energy shift of 3.4 and 4.6 eV is expected. This indicates a downward band bending for the p-type and significant upward band bending for n-type diamond surfaces.
Surface band bending is difficult to precisely determine using XPS due to contributions from different depth within the material [29], and the theoretical calculations needed for such comparisons. The elevated temperatures used also affect the carrier concentration and thus band bending amount, while also broadening spectral peaks. However, the relative band bending between a hydrogen and bare termination can be observed. The unterminated samples were prepared by thermal desorption of the hydrogen terminated surfaces at 900 C within the UHV chambers after completion of other measurements. The dotted lines in Fig. 2 show the C 1s emission spectra of bare diamond surfaces, indicating an increase in upward band bending by approximately 0.5 eV. This is consistent with previous studies where hydrogen termination was shown to induce a downward surface band bending on diamond [22].
The low kinetic energy region of the electron emission can also be used to determine the band alignments, though with a significantly greater surface sensitivity than the energy region used for the XPS in Fig. 2 [30]. A slight charging of up to 0.2 eV was observed for the NDD due to insufficient conductivity at 300 C, where an increase in emission current lead to a greater voltage drop across the sample. Contributions to the electron emission in this region can include thermionic emission, photoemission from low-energy light sources and secondary electron emission [31] [e] [35].
We begin with the emission spectra from the PDD and NDD surface at 300 C in Fig. 3, where a single peak between 2 and 2.5 eV above E F is observed. Without any charging, the low-energy cutoff indicates the position of the vacuum level, and hence the WF of the surface. Emission with the additional excitation from white LED lights are also shown in dotted lines, where an increase in emission intensity was observed, showing that even low energy photons can excite electrons from doped diamond into the vacuum, or at least promote carrier transport to the surface. This is significant for photon enhanced thermionic emission and diamond photocatalytic devices [4] [e] [6,36], and helps explain observation of enhanced photocatalysis when using sub-bandgap radiation [37]. To ensure the intensity increase in emission is not from localized spots during LED illumination, microscopy of the thermionic emission was done using an energy-filtered photoemission electron microscope setup in Fig. S3, where the emission increase appears uniform over the sample surface. The increased emission from LED lights are also observed at higher temperatures, as shown in Fig. S4 for the samples at 400 C, where the featured emission peaks can potentially be due to defect-induced states [5,6].
To further investigate the enhanced electron emission from subbandgap illumination on diamond, two laser sources were also used on the PDD. A red laser diode showed no noticeable difference in the thermionic emission behavior, while a 5 mW 532 nm (2.33 eV) green laser showed enhancements, indicating a lower limit for the photon energy required. The emission spectra using the green laser on the PDD at 300 and 400 C are shown in Fig. 4. Other than increases in the overall emission intensity, the emission line shape also shows slight differences, which can be further investigated using tunable photon sources. The interactions between the initial and final states in photo-assisted thermionic emission can also be predicted with computational theory [38,39].
Sub-bandgap excitation and emission using low energy light sources can have significant implications for a wide range of devices and applications. For example in diamond-based thermionic energy converters where significant advances towards obtaining ultra-low WF materials have been achieved [40], the phosphorus or nitrogen doped diamond can have a further enhancement of performance under optical illumination. This would be convenient for solar-based converters, and may allow for lower operating temperatures as well. Other possibilities include taking advantage of the photoexcitation process for semiconducting purposes, where doped diamond substrates can be used as a photo-conductive semiconducting switch [41]. Understanding and optimizing the electron excitation process may lead to breakthroughs in diamond  semiconducting devices, where the n-type conductivity has often been a limiting factor.
The emission intensity for the NDD increased significantly more than on the PDD after the additional LED excitation, which can be due to different doping concentrations of nitrogen and phosphorus in the diamond. As phosphorus doping efficiency into the diamond structure is significantly lower than nitrogen [10,28], a greater density of states at the donor level in NDD can allow for greater absorption efficiency of the LED light. However, with the shallower phosphorus donor level in the PDD, a greater bulk conductivity can be achieved at 300 C, leading to a reduced shift of the peak to lower energies from charging. The broader NDD peak is also likely due to differential charging across the sample surface. This shows the importance of both doping concentration and the dopant thermal activation energies.
Emission spectra from a hydrogen terminated BDD sample is shown in Fig. 5 for comparison, where no emission was observed without using X-ray illumination. The WF is noticeably higher at approximately 3.5 eV, which would explain its lack of thermionic emission and photoemission from LED lights in contrast to the PDD and NDD. To provide the extra energy needed for emission, X-rays were used to generate pure secondary electrons without direct photoemission from occupied states. Under X-ray illumination, emission was observed where electrons thermalize into unoccupied states and are emitted as secondary electrons. A peak with a broad shoulder towards the low-energy cutoff is observed on the BDD, which was previously studied and attributed to the CBM emission at the peak, and intra-bandgap emission at the shoulder [7,35]. Similar to the XPS results in Fig. 2, while a significant WF difference of over 1.2 eV is observed between the BDD and the ntype samples, it is significantly less than the difference between their bulk E F positions. The downward band bending on the BDD decreases its WF while also allowing for electron accumulation in the CBM at the surface, giving rise to the large peak around 5 eV [29]. A small amount of electron emission was still observed for several minutes after the X-ray illumination as shown in orange in Fig. 5, likely also owing to the downward band bending as well as the indirect diamond band gap contributing to the longer lifetime of excited electrons. The n-type PDD and NDD however exhibit upward band bending which draw surface electrons at the CBM into the bulk, in turn causing secondary electrons to accumulate less near the CBM. This highlights the dominating effect of band bending on low-energy surface emission.
The two n-type samples were then heated up to 400 C for further analysis, where a significant increase in emission intensity was observed (Fig. 6), as predicted by the Richardson equation in Equation (1). Other notable differences include the shape of the low-energy peak, where a double peak can now be observed, with a separation of approximately 0.1 eV. A double peak is generally expected for hydrogen terminated diamond due to its NEA providing emission peaks at the CBM and vacuum level [7], though the separation is typically in the region of 1 eV [25]. Given the energy  position of the observed peaks, they likely originate from within the band gap, indicating spectral features at the surface near the vacuum level. Angle-resolved measurements of secondary electrons on hydrogen terminated BDD (100) and (111) surfaces at room temperature however show no observable surface states in this region [35]. The difference in emission behavior is therefore either a temperature or band bending effect. Potential explanations include existence of surface states outside the free-electron parabola horizon, which are disallowed emissions due to the lack of momentum matching with vacuum states [35,42], and therefore not observed with secondary electron emission. However, while under sufficient thermal excitation, electrons in these states may be inelastically scattered and emitted into the vacuum. The effect was likely not observed in Fig. 3 due to the significantly lower population of the state at 300 C. Another possibility is that the surface dipole induced by a hydrogen termination may have additional effects on the electric field near the surface [21], producing localized states where electrons can be excited from, similar to those seen with image potentials [43].
The inset in Fig. 6 shows a magnified view of the spectra at around 3 eV where the intensity increases again after tailing off from the first peak near 2 eV. The intensity of this peak is substantially lower than the initial peak, which can be expected due to the thermal occupancy of electronic states in accordance with the Boltzmann distribution. The reduction in intensity between 2.2 and 2.8 eV therefore indicates a lower density-of-states than at 3 eV. Separation between the two observed emission peaks is roughly 1 eV, which is consistent with the expected NEA of hydrogen terminated diamond surfaces. These observations further confirm that the low-energy emission cutoff is at the vacuum level and lies within the band gap, whereas the emission near 3 eV is at the surface CBM position.
Thermal energy at 400 C is shown to be sufficient to populate and emit a small proportion of electrons from the surface CBM, which lies at about 2.8 eV above E F for both the PDD and NDD. A noticeable degree of Fermi-level pinning is therefore shown to exist for diamond, though the pinning amount is not entirely constant. As seen with the BDD in Fig. 5, the CBM position to be  approximately 4.8 eV above E F , showing that the surface Fermi level position can be altered with sufficient doping. The difference in measured band bending between the BDD and n-type diamond determined in this low-energy region can differ from XPS in Fig. 2, due to the greater bulk contributions when using XPS [29,30]. Comparing to the bulk Fermi level positions of the PDD and NDD samples, a significant amount of upward band bending of 2.2 and 1.1 eV respectively is obtained. This resulted in the emission being almost completely dominated by electron emission within the band gap, particularly at lower temperatures. For thermionic emission devices, surface states at the band gap and a large NEA value are therefore critical to performance. The PDD and NDD were then annealed to 900 C within the same UHV chamber to desorb hydrogen from the surface, while minimizing surface graphitization. This leaves a predominantly unterminated bare diamond surface, which has been reported to exhibit a slight positive electron affinity (EA) and affect the surface band bending [44]. The increased EA and upward surface band bending would therefore result in an increased WF and reduction in electron emission, which was characterized by further spectroscopy near the work function threshold.
The emission spectra for unterminated PDD and NDD at different temperatures are shown in Fig. 7 with X-ray excitation, which was necessary for emission below 600 C. The low energy threshold for emission up to 600 C is approximately 4.5 eV for the unterminated samples, indicating a significantly increased CBM position from the 2.8 eV of the hydrogen terminated surface. The exact CBM position for the unterminated samples may be lower than the emission threshold due to the slight PEA of unterminated diamond surfaces (0.38 eV) [44].
Above 600 C, the emission spectra for both samples show a lower energy emission peak beginning at 3 eV that is unaffected by the X-ray excitation. This shows that secondary electrons at the CBM are extremely inefficient at populating the surface states within the band gap, whereas thermally excited electrons can be directly emitted from those states. For the NDD, the peak position relative to the CBM (~1 eV below) fits well with previous spectra of the hydrogen terminated surfaces, indicating that the emissions were likely from residual hydrogen terminated sites after the thermal desorption by annealing. The small amount of hydrogen did not appear to induce any changes in the local surface band bending, seen from its position relative to the CBM peak, indicating insignificant changes to the surface state occupancy. Meanwhile, significant broadening of peaks were observed for the unterminated PDD, likely a result of inhomogeneous and strong upward band bending across the surface from its low dopant density and lack of hydrogen termination.
The PDD sample was later exposed to atmosphere when partially hydrogen terminated before being tested again, significantly reducing its emission intensity. Contact angle measurements of a hydrogen desorbed sample after air exposure shows contact angles of about 65 , indicating a significantly decreased hydrophobicity after removal of hydrogen [11,12], compared with the contact angle of hydrogen terminated diamond in Fig. S1. Fig. 8 shows the emission spectra from 300 to 500 C with and without LED excitation, where the two distinct peaks observed earlier in Fig. 6 are observed and have varying height dependence on the thermal and LED excitation. The two peaks at around 3 eV correspond to the double peaks observed on the hydrogen terminated diamond, while the higher energy region from the conduction band was also observed when using X-ray excitation (not shown). The conduction band peaks while not explicitly shown and magnified in Fig. 8a, can be barely observed at precisely 1 eV higher energy than the taller LED peak in red. The constant energy spacing between those two peaks even as the WF and bend bending changes with temperature strongly suggest the LED illumination peak is from a band feature.
Focusing back onto the two main features seen in Fig. 8a, it is clear that thermionic emission only contributes to the lower energy peak at lower temperatures, while the LED lights promote emission from the second peak without any noticeable effect on the first peak. This is a particularly interesting observation as the LED is a low-energy non-monochromatic source, which is successfully exciting a narrow distinct peak without affecting emission at a lower energy. The possibility of sub-bandgap radiation absorption from intermediate states has also been discussed for laser irradiated diamond substrates [5,6]. The two peaks from each LED illumination spectra have been decomposed into two Gaussian components, and a small shift in relative energy position is observed between different temperatures, unlike mentioned above for the second peak and the conduction band. This is likely due to small WF changes on the sample surface with temperature, as the lowest energy peak position is determined by the WF cutoff of electron emission. Therefore, the lower energy peak of the two is unlikely to originate from a band feature. The energy range of a typical LED extends to approximately 2.7 eV, meaning that any direct electronic excitation into the state at 3.2 eV above E F in Fig. 8 must have an initial state above E F . This indicates that there are sufficiently occupied states within the band gap and above the Fermi level at 300 C, while the states at 3.2 eV are initially unoccupied without LED excitation. This is in agreement with the proposal of intra-bandgap surface states mentioned above and potentially of electron photoexcitation from imagepotential like states [43], but into a final state rather than directly into the vacuum due to the narrow peak width. The lack of decay into the lower energy first peak also suggests a long lifetime of the surface state at 3.2 eV, or no direct decay pathway. LED excitation of the second peak also increases with temperature, which is consistent with the theory of initial states being above E F , as they increase in occupation with temperature. The thermally excited first peak is likely to be from surface emissions at the vacuum level rather than a surface state [35,45], which could explain its lack of response towards photoexcitation. For applications in air or under solution, the additional states induced by surface adsorbates may further enhance photon capture.

Conclusion
Spectroscopic measurements of thermionically emitted electrons provide insight into mid-bandgap energy levels in diamond, showing interesting properties of hydrogen terminated diamond. Detailed investigation shows a double peak exists near the vacuum level, where the higher energy component may indicate existence of a surface state. At lower temperatures, electron emission from below the CBM essentially contributed the entire thermionic emission current. Fermi-level pinning was found to be significant for both the nitrogen and phosphorus doped diamond, leading to upward band bending of up to 2.2 eV, therefore higher doping concentration is likely more important than lower activation energy dopants for electron emission devices. While perhaps most interestingly, low energy excitation from LED light was found to significantly increase emission current at low temperatures, which can prove very useful for devices in energy conversion or photocatalysis, and encourages further investigation to exploit a potential extremely efficient photoemitter.

Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.