Structural and electrical properties of carbon-ion-implanted ultrananocrystalline diamond films

We investigate the structural and electrical properties of carbon-ion-implanted ultrananocrystalline diamond (UNCD) films. Impedance spectroscopy measurements show that the impedance of diamond grains is relatively stable, while that of grain boundaries (GBs) (Rb) significantly increases after the C+ implantation, and decreases with the increase in the annealing temperature (Ta) from 650 °C to 1000 °C. This implies that the C+ implantation has a more significant impact on the conductivity of GBs. Conductive atomic force microscopy demonstrates that the number of conductive sites increases in GB regions at Ta above 900 °C, owing to the formation of a nanographitic phase confirmed by high-resolution transmission electronic microscopy. Visible-light Raman spectra show that resistive trans-polyacetylene oligomers desorb from GBs at Ta above 900 °C, which leads to lower Rb of samples annealed at 900 and 1000 °C. With the increase in Ta to 1000 °C, diamond grains become smaller with longer GBs modified by a more ordered nanographitic phase, supplying more conductive sites and leading to a lower Rb.


Introduction
Diamond films have excellent properties such as high hardness, good chemical stability, and high thermal conductivity. [1][2][3] Ultrananocrystalline diamond (UNCD) film is a unique form of carbon, in which diamond grains with sizes of 3-5 nm are surrounded by an amorphous carbon matrix. UNCD films have applications where high conductivities are required, including electrochemical electrodes, [4][5][6] field emission, [7] and heterostructures. [8,9] Extensive studies have been performed to improve the conductivities of UNCD films. It has been reported that nitrogen-doped UNCD films exhibit n-type conduction with an enhanced electrical conductivity, which is attributed to the increased number of sp 2 bonds at grain boundaries (GBs). [10,11] Phosphorous and oxygen ion implantations were used to increase the conductivity of nanocrystalline diamond grains and amorphous-carbon GBs, which improved the n-type Hall mobility of the UNCD film. [12,13] A hybrid structure of UNCD and graphene nanoribbons promoted a high mobility and n-type conductivity in an oxygen-ion-implanted UNCD film. [14] Lithium ion implantation induced the formation of electron trap centers inside diamond grains, whereas post-annealing healed the defects and converted the amorphous carbon phase into nanographite, forming conduction channels for an effective transport of electrons. [7] Amorphous carbon in GBs is a common component in UNCD films, whose electrical conductivities were reported to be not sufficiently high, thus limiting the applications of UNCD films. [15] It is believed that the conversion of amorphous carbon to a conducting graphitic phase along GBs enhances the efficiency for electron transport on the surface, thus improving the conductivity of the UNCD film. [16,17] It was also reported that copper [18] and gold [19] ion implantations and annealing promoted electrical properties of UNCD films by decreasing the amorphous carbon content and even by inducing the nanographitic phase in GBs. This suggests that the electrical properties can be effectively improved by modifying the structure and component of amorphous-carbon GBs. Carbon ions are not dopants for diamond grains, but preferentially lead to more ordered amorphous-carbon GBs to enhance the conduction mobility. A high-dose C + -ion implantation (10 15 cm −2 ) and annealing lead to crystallization of the graphite phase that was not transformed from amorphous carbon in the UNCD film and degradation of the field-emission properties. [20] In this study, we implanted a lower dose of C + to regulate the microstructures of the UNCD films particularly in GBs to improve their electrical properties. Highresolution transmission electronic microscopy (HRTEM) and visible-light Raman spectroscopy were used to characterize the microstructures of the films after ion implantation and annealing. Conductive atomic force microscopy (AFM) was used to acquire images of conductive capacities at different parts of the UNCD films. Impedance spectroscopy (IS) was employed to distinguish the conductivity contributions from diamond grains and GBs.
The results show that the impedance resistance of GBs significantly increases after the C + implantation; it is effectively reduced with the increase in the annealing temperature (T a ) in the range of 650-1000 • C. The impedance of diamond grains is relatively stable. This suggests that the C + implantation has a larger impact on the impedance of GBs than on diamond grains. When T a is above 900 • C, the number of conductive sites significantly increases in the GB region of the UNCD film owing to the formation of a nanographitic phase, which is responsible for the reduced impedance resistance of the GBs.

Experimental methods
UNCD films were deposited using a hot-filament chemical vapor deposition (HFCVD) system on single-crystal silicon (111) wafers. The carbon source was acetone, the volume ratio between acetone and hydrogen was 1%-2% with a total gas flow of 200 ml/min, the working pressure in the reaction chamber was 0.5-1.4 kPa, and the temperature of the substrate was approximately 850 • C. The thickness of the as-deposited UNCD films was approximately 5 µm, obtained at a growth rate of 1.5 µm/h. C + implantation was performed on as-deposited UNCD films with an implantation energy of 90 keV and dose of 10 12 cm −2 . For roomtemperature implantations, there is a universal critical damage density (10 22 vacancies/cm 3 ), which, when exceeded, leads to complete graphitization of heavily damaged diamond crystals upon annealing. [21] It was shown that the C + implantation level of 10 12 cm −2 at room temperature was significantly below the critical dose of (2-3)×10 15 cm −2 . [22] The implantation profile was determined by the Stopping and Range of Ions in Matter (SRIM-2013) software. C ions were implanted into diamond with an average ion range of 121 nm, creating 10 19 vacancies/cm 3 in the meantime. The implanted samples were annealed at 500, 650, 725, 800, 900, and 1000 • C for 30 min at a pressure of 4000 Pa (80% N 2 and 20% O 2 ), denoted as C12500, C12650, C12725, C12800, C12900, and C121000, respectively. The annealing at 900 • C for 30 min was performed on an as-deposited UNCD film (sample 900-A) to distinguish the effects of C + implantation and annealing on the electrical properties of the UNCD films.
HRTEM was used to characterize the microstructures of the films using a Tecnai G2 F30 S-Twin microscope, produced by Philips-FEI. Visible-light Raman spectroscopy measurements (JobinYvon, Labor Raman HR-800) of the samples were performed with an excitation wavelength of 514 nm at room temperature. Conductive AFM and scanning Kelvin microscopy (SKM) were performed in a scanning probe microscope (SPM) (Solver P47, NT-MDT, Russia) under ambient conditions.
Ti (500 nm)/Au (300 nm) contacts were fabricated using an SPC-350 magnetron sputtering system with a metal mask covering on the UNCD films set before the deposition. The metal mask was fixed by a high-temperature tape with the UNCD film to ensure its stability during the deposition. The contacts were then annealed under a vacuum at 450 • C for 20 min. The diameter of the electrode was 2 mm. I-V measurements of the samples at room temperature showed that ohmic contacts were formed. The impedance properties of the samples were determined using an Autolab potentiostatgalvanostat connected with a four-probe station in the frequency range of 0.01 Hz to 10 MHz. IS was employed to study the electrical properties of single-crystal diamond and polycrystalline diamond. This technique involves a measurement of the real and imaginary components of the film impedance as a function of both frequency and temperature. The impedance as a complex number can be defined as where the real part refers to the resistance contribution, while the imaginary part to the capacitance contribution. In a typical impedance spectroscopic analysis, the impedance is measured as a function of frequency, and the real component is plotted against the imaginary component with the change of the frequency. These so-called "Cole-Cole" plots may reveal different semicircular responses for each resistance-capacitance (RC) component of the film. [23] Theoretically, a double RC parallel model in series could be used to simulate the electrical conduction of diamond films attributed to uniformly distributed diamond grains and GBs, as illustrated in Fig. 1. The resistance between the electrode and diamond films is considered during the measurements, denoted as R e . In the figure, R g , R b , C g , and C b represent the resistances of diamond grains and boundaries, and capacitances of diamond grains and boundaries, respectively. Each parallel RC equivalent circuit model accurately fits the corresponding Cole-Cole semicircle. The employed fitting procedure is the same as that described by Kleitz and Kennedy, which enables us to determine the resistance and relaxation frequencies with a good precision. [24] 096104-2 The complex impedance can be expressed as In the complex plane, the plot of Z against Z has the form of two semicircles, where the contributions of diamond grains and GBs are easily identified. Therefore, the electrical conduction paths of the bulk material can be studied separately. Equivalent circuit for the electrical response of a UNCD sample with contributions from diamond grains (denoted as "g"), grain boundaries (denoted as "b"), and electrode interface (denoted as "e").

Results and discussion
I-V measurements were performed before the IS measurements to ensure that ohmic contacts were formed between the electrodes and UNCD films. Figure 2 shows the Cole-Cole plots of the samples and Si substrate; as the Cole-Cole plot of the Si substrate is too small to be clearly observed, it is enlarged in the inset of Fig. 2(a). One semicircle is observed in the Cole-Cole plot of the Si substrate. The impedance is lower than 1 kΩ, which is 1-2 orders of magnitude lower than those of the annealed samples. This implies that the Cole-Cole plots of the samples mainly represent the electrical properties of the UNCD films. The Cole-Cole plots of all samples can be fitted by two semicircles, which indicates that two components contribute to the conductivity. The total impedance of the as-deposited sample is the lowest among the analyzed samples, approximately 28 kΩ. It increases to 62 kΩ for sample C12500 and 121 kΩ for sample C12650. It is worth noting that the semicircular response of sample C12725 is accompanied by an additional semicircle extending to low frequencies, in which the low-frequency response is incomplete and measured as an arc owing to the limited measuring range. Nevertheless, the total impedance of sample C12725 is estimated to be higher than that of sample C12500, but lower than that of sample C12650. Figure 2(b) shows the Cole-Cole plots of samples 900-A, C12800, C12900, and C121000. The total impedance of sample 900-A is approximately 8 kΩ, which increases to 47 kΩ for sample C12800. With the increase in T a above 800 • C, it decreases to 29 kΩ and 32 kΩ for samples C12900 and C121000, respectively. The impedance of sample 900-A is significantly lower than that of sample C12900, which indicates that the IS measurements effectively reveal the effect of the C + implantation on the electrical properties of the UNCD films, although the ion-implanted layer is thin, compared to the UNCD films. The total impedance of sample 900-A is lower than that of the as-deposited sample, which suggests that the annealing decreases the impedance of the UNCD film.  The resistance and capacitance values were obtained according to the equivalent circuit model and presented in Table 1 to quantitatively understand the responses. The semicircular response at low frequencies corresponds to a capacitance value in the range of 0.2-9 nF, whereas the capacitance at the high-frequency region is 70-110 pF. It is well known that picofarad-scale capacitances for polycrystalline materials originate from the contribution of crystal grains, while nanofarad-scale responses are attributed to GBs. [25] Therefore, both nanocrystalline diamond grains and GBs in the UNCD films contribute to the measured electrical characteristics.
The R e values presented in Table 1 are in the range of 1.4-1.7 kΩ, except that of sample C12650 (as the electrodes for sample C12650 were deposited by the vacuum evaporation method), which implies that the ohmic contacts are stable in these samples and that good conditions for IS measurements are provided. The R g value of the as-deposited sample is 25.9 kΩ; it significantly decreases to 4.5 kΩ for sample 900-A. This implies that the annealing at 900 • C decreases the resistance of diamond grains in the UNCD films. It is worth  noting that the R g of sample C12900 is 19.5 kΩ, significantly higher than that of sample 900-A, but lower than that of the asdeposited sample. In the C + -implanted samples, the R g value decreases from 33.4 kΩ to 18.9 kΩ with the increase in T a from 500 • C to 725 • C. When T a is above 800 • C, the R g values of samples C12800, C12900, and C121000 are 27.3, 19.5, and 23.6 kΩ, respectively. The C + implantation increased the R g of the UNCD films, without significant dependence on the annealing temperature. The R b value of sample 900-A is 2.2 kΩ, slightly higher than that of the as-deposited sample of 1.4 kΩ. The R b value of sample C12500 is 26 kΩ; it significantly increases to 97.4 kΩ for sample C12650. It starts to decrease when T a further increases, reaching a value of 6.6 kΩ for sample C121000. These R b values are higher than that of the as-deposited sample, which implies that the C + implantation increases the impedance of GBs even though it is effectively reduced by the annealing.
The IS results show that the contribution of diamond grains to the conductivity is increased after the C + implantation and annealing. The significantly increased R b values of the C + -implanted samples decrease with the increase in T a in the range of 650-1000 • C, while the R g values remain stable, which indicates that the C + implantation has a larger impact on the impedance of GBs. The lower R b values of samples C12900 and C121000 show that their GB regions are more conductive than those of the C + -implanted samples annealed at lower T a .
In order to observe the conductive behavior at the microscale, conductive-AFM and SKM were employed. A Ptcoated Si cantilever with a tip curvature radius of ∼ 35 nm was used to obtain the surface height relief in the tapping mode. Local conductance was scanned in the contact mode across a previously detected surface-height-relief map (with a sampleprobe gap of ∆ = 0). At each point, a bias voltage is applied between the sample and grounded probe to obtain the surface conductivity. Therefore, the local conductance represents the conductive behavior of the ion-implanted layer. A sampleprobe gap of ∆ = 100 nm was used during the SKM measurements. For each sample, the measurements were repeated at 5 arbitrarily chosen 3 × 3 µm 2 regions. Figure 3 shows the surface height-relief and local conductance maps of samples C12725, C12900, and C121000. The surface height-relief maps show that the most probable relief heights of the samples increase with T a , which indicates that the exposed area of nanoclusters becomes more significant at higher T a . The local conductance maps show that the samples become more conductive with the increase in T a . The most conductive areas in sample C12725 are scattered in the film; they become continuous in samples C12900 and C121000, which indicates that more conductive sites appear in these two samples. Combined with the IS results, we revealed that the R g values of these three samples increase with T a , while the R b values decrease, which indicates that the conductive sites observed at the high T a are mainly located at the GB regions. Figure 4(a) shows the conductance histograms of samples C12725, C12900, and C121000. They show that the annealings at 900 • C and 1000 • C lead to ∼ 10 times higher conductances in the range of 100-500 nS compared to that of the sample annealed at 725 • C, which confirms the increased number of conductive sites. Figure 4(b) shows the contact-potential histograms of samples C12725, C12900, and C121000. The contact potentials of the samples increase with T a , which implies loss of negative electron affinity at higher T a . [26] This could be attributed to the absence of Htermination of nanocrystalline diamond grains when T a is above 800 • C. [27][28][29] The increased numbers of conductive sites in samples C12900 and C121000 indicate that the microstructure changes in the GB regions. In order to understand the origin of these highly conductive GBs, we performed HRTEM measurements to characterize the microstructures of the as-deposited sample and samples C12900 and C121000. During the TEM specimen preparation, first, the Si substrate was removed and then the UNCD film close to the Si side was removed by argon-ion milling, providing surface area (ion implantation region) for observation. Figures 5(a)-5(c) show low-magnification TEM images of the as-deposited sample and samples C12900 and C121000, and corresponding selected-area electron diffraction (SAED) patterns (insets), respectively. The SAED pattern of the as-deposited sample contains diffraction rings corresponding to the (111), (200), and (311) lattice planes of diamond, which indicates that diamond grains are the main component in the film. Many bright spots are observed on the diffraction rings, which implies that some large diamond grains exist in the film. The SAED patterns of samples C12900 and C121000 also contain diffraction rings corresponding to the (111), (200), and (311) lattice planes of the diamond. However, these rings are smooth without bright spots, which implies that the diamond grains in these samples are smaller and uniformly distributed. The grain sizes in the as-deposited sample are in the range of approximately 10-15 nm. They decrease to ∼ 5-10 nm in sample C12900 and become even smaller in sample C121000, providing more GB paths along the diamond grains.  096104-5 Figure 5(d) shows high-magnification TEM images of the as-deposited sample with Fourier-transform (FT) diffractograms corresponding to the entire image (FT 0 ), area 1 (FT 1 ), and area 2 (FT 2 ). The FT 0 image in Fig. 5(d) shows a ring with many spots and a circular bright spot in the center, which implies that amorphous carbon exists in the GBs of the asdeposited sample. The presences of diamond and amorphous carbon are demonstrated by images FT 1 and FT 2 corresponding to area 1 and area 2, respectively. The circular bright spot in the FT 1 pattern of area 1 is inversely transformed to provide insights into the GB microstructure, as shown in Fig. 5(d 1 ). The carbon clusters randomly intertwine in the GB area (representing the amorphous carbon). The FT 2 pattern is also inversely transformed in Fig. 5(d 2 ), which shows a lattice structure with a spacing of ∼ 0.21 nm, reflecting the ordering of the (111) planes of the diamond. Figure 5(e) shows high-magnification TEM images of sample C12900 with FT diffractograms corresponding to the entire image (FT 0 ), area 1 (FT 1 ), and area 2 (FT 2 ). The FT 0 image in Fig. 5(e) shows a ring with many spots and a central diffuse ring, which represent the diamond grains and GBs in sample C12900. The central diffuse ring in the FT 1 pattern of area 1 is inversely transformed in Fig. 5(e 1 ), which shows a different GB microstructure from that of the as-deposited sample. Some chain fringes are observed in the GBs. The spacing between the curved fringes is ∼ 0.35 nm in Fig. 5(e 1 ), corresponding to the (0002) planes of the graphitic structure, which indicates that a nanographitic phase forms in the GBs of sample C12900. The FT 2 diffraction pattern is inversely transformed in Fig. 5(e 2 ), which shows a lattice structure with a spacing of ∼ 0.21 nm, reflecting the ordering of the (111) planes of the diamond. Therefore, the spots in the diamond lattice ring of the FT 0 image represent nanocrystalline diamond grains in different orientations. The nanographitic phase in the GBs is less ordered than the diamond phase as no spots are observed in the corresponding graphite ring. The spots slightly deviate from the lattice ring, probably due to the diamond lattice damage by the ion implantation.  and 5(f 2 ) show the orderings of the graphite (0002) and diamond (111) planes, respectively. The center area of the FT 2 image in Fig. 5(f) is significantly darker than those in Figs. 5(d) and 5(e), which implies that the amount of amorphous carbon is significantly lower in sample C121000. These HRTEM results confirm that diamond grains are the main component in the films, and that the GBs of the as-deposited sample mainly consist of amorphous carbon. In contrast, a nanographitic phase is formed in the GBs of samples C12900 and C121000, induced by the ion implantation and annealing. A more ordered nanographitic phase and more GB paths along smaller diamond grains exist in sample C121000 compared to those in sample C12900, in agreement with the lower R b value of sample C121000. It can be concluded that the conductive sites observed in Figs. 3(e) and 3(f) are mainly formed by the nanographitic phase, attributed to the low-impedance GBs.
In order to further investigate the microstructure changes of GBs in the C + -implanted samples, visible-light Raman spectra in the range of 800-2000 cm −1 for the as-deposited sample, sample 900-A, and C + -implanted films are shown in Fig. 6(a). The Raman spectra of the UNCD films can be well fitted with six Gaussian-Lorentzian peaks. The peaks at 1550-1570 cm −1 and 1340 cm −1 correspond to the G-and Dbands, respectively. [30] The peak at 1332 cm −1 originates from diamond, [31] while that at 1200 cm −1 is related to amorphous sp 3 carbon. [32] In addition, two peaks are observed at 1140-1190 cm −1 (ω 1 peak) and 1480 cm −1 (ω 3 peak) attributed to bending and stretching vibrations of trans-polyacetylene (TPA) at the GBs of the UNCD films, respectively. [33] 1000 1200 1400 1600 1800   Figure 6(b) shows the I D /I G values (intensity ratio between the D and G bands) as a function of T a , obtained from the fitted peaks in Fig. 6(a). The I D /I G value of the asdeposited sample is the highest (1.07), which decreases after the annealing at 900 • C (0.96). It becomes lower in the T a range of 500-800 • C (0.91-0.93) and then increases at higher T a (∼ 1.05) for the C + -implanted samples. This indicates that the size of the sp 2 -bonded carbon located at GBs decreases at 500-800 • C and then increases at 900-1000 • C. [34,35] This is consistent with the result that the nanographitic phase forms conductive sites in samples C12900 and C121000. Figure 6(c) shows the dependences of the full-width-at-half-maximum (FWHM) values of the ω 1 and ω 3 peaks on T a . They are stable when T a is lower than 900 • C and increase at larger  temperatures. The I TPA /I sum values of the samples annealed at 500-725 • C (15.0%-15.5%) are slightly higher than that of the as-deposited sample (14.75%). The value is lower for sample C12800 (14.6%), while those of samples C12900 (11.8%) and C121000 (9.6%) are significantly decreased, as shown in Fig. 6(d). This implies that TPA oligomers in GBs start to desorb when T a is higher than 800 • C, in agreement with the contact-potential result. TPA chains, as one of the πconjugated polymers, are non-conductive without dopants. [36] It was reported that more TPA dangling bonds in UNCD films led to a high sheet resistance. [37] After the annealing at 1000 • C, the FWHM values of the ω 1 and ω 3 peaks increase, while the I TPA /I sum ratio significantly decreases, which is consistent with the lower R b of sample C121000 than that of sample C12900.
IS results show that the R g is relatively stable, while the R b significantly increases after the C + implantation, which indicates that the contribution of diamond grains is increased after the ion implantation. The R b gradually decreases with the increase in T a in the range of 650-1000 • C, which implies that the C + implantation has a larger impact on GBs than on diamond grains. The IS and AFM results demonstrate that the number of conductive sites in the GBs increases with T a in the range from 725 • C to 1000 • C. The HRTEM and AFM results demonstrate that the number of conductive sites in GBs increases at T a higher than 900 • C owing to the formation of the nanographitic phase. Visible-light Raman spectra and HRTEM results show that the nanographitic phase is more ordered in sample C121000 than that in sample C12900, which provides more conductive sites and leads to the lower R b of sample C121000.

Conclusion
We showed that the C + implantation significantly increased the impedance resistance of GBs, which was effectively reduced with the increase in T a in the range of 650-1000 • C, while the impedance resistance of diamond grains was relatively stable in UNCD films. This suggested that the C + implantation had a larger impact on the impedance of GBs than on that of diamond grains. Conductive-AFM results demonstrated that the number of conductive sites in GBs significantly increased when T a was above 900 • C owing to the formation of a nanographitic phase, confirmed by HRTEM. Visible-light Raman spectra and HRTEM results showed that the nanographitic phase was more ordered in sample C121000 than in sample C12900, which further decreased the R b of sample C121000.