The structure and electronic properties of tetrahedrally bonded hydrogenated amorphous carbon

Dense forms of hydrogen-free amorphous carbon (a-C), including tetrahedrally bonded amorphous carbon (ta-C),¹ offer high hardness, high Young's modulus, optical transparency, low friction coefficient, and biocompatibility.^2,3 The properties of these materials make them desirable for a range of applications, including neuromorphic materials for computing and bio-sensing.^4–7 The maximum reported fraction of sp³ bonds in ta-C is ∼80% with the remainder sp² bonded.^2,8 Further reducing the sp² content could lead to a dense a-C material with increased visible transparency and improved carrier mobility by removing sp² bonded atoms that introduce states in the bandgap.^1–3 One way of reducing the fraction of sp² states is to create polymer-like structures containing entirely sp³ bonded carbon, since up to ∼30 at. % hydrogen can be included in a-C.⁹ However, most polymeric materials are of low density and are not sufficiently robust for many applications. In order to retain high density while introducing hydrogen, it is necessary to employ energetic deposition technology to create compressive stress conditions that encourage densification. The question remains as to whether it is possible to synthesize a high density form of hydrogenated a-C (that is, hydrogenated tetrahedral amorphous carbon, ta-C:H), where the carbon atoms are present in a tetrahedrally bonded network with hydrogen passivation of sp² defects, analogous to the passivation of dangling bonds by hydrogen in hydrogenated amorphous silicon (a-Si:H).

The highest density hydrogenated a-C so far reported has been prepared as thin films by deposition of acetylene or methane ions using an ion beam.^10,11 The structure of this form of amorphous carbon was studied using electron diffraction.¹¹ The first neighbor C–C bond lengths were in the range 1.46–1.50 Å between the diamond value of 1.54 Å and the graphite value of 1.42 Å. The material was found to have a C–C–C bond angle in the range 113°–118°, significantly higher than the tetrahedral bond angle of 109.5° and approaching that of graphite (120°). A neutron diffraction study of a deuterated form of hydrogenated a-C prepared using material prepared by a fast atom plasma source with an impact energy of 950 eV showed that this material also did not have a high content of tetrahedrally bonded carbon atoms, since the first neighbor C–C bond length occurred at 1.45 Å.¹² Hydrogenated a-C deposited from a plasma beam source using methane or acetylene exhibited an optical bandgap of ∼2.5 eV, only marginally higher than a ta-C of ∼2.3 eV, indicating only partial hydrogenation or loss of tetrahedral bonding. Therefore, it appears that fully tetrahedrally bonded, wide bandgap, hydrogenated a-C (henceforth referred to as ta-C:H) has not yet been synthesized. Here, we describe the synthesis of ta-C:H using an ion energy and hydrogen level that simultaneously maximize the density, tetrahedral carbon content, and bandgap.

Hydrogenated and deuterated samples were prepared at room temperature using a double-bend filtered cathodic vacuum arc (FCVA) system capable of a base pressure of <10⁻⁷Pa with a 99.9% purity graphite cathode with an arc current of 60 A. Deuterated samples were included to enable the use of neutron diffraction to carry out radial distribution analysis of the material. Films were synthesized on n-type (100) silicon, 300 nm-thick SiO₂/Si, A-plane (11–20) sapphire and glassy carbon substrates subjected to cleaning in an ultrasonic bath with acetone, ethanol, and de-ionized water. Hydrogen or deuterium gas was bled into the sample chamber at an operating pressure of 0.13 Pa, and the substrate bias was varied in the range 0–650 V. Direct electrical bias to the substrate holder was used for conducting substrates while in the case of insulating substrates, a conducting grid placed in front of the substrate was used instead as described previously.¹³ 

Figure 1 shows the relationship between the intrinsic film stress and the substrate bias measured from the change in curvature of the substrate wafer before and after film deposition using Stoney's equation.¹⁴ The trends of the curves with substrate bias for the hydrogenated and deuterated samples were similar. A maximum intrinsic film stress was observed at ∼80 V bias, as has been found for carbon deposition in the same system in the absence of H or D.¹⁵ The stress dropped sharply for films deposited at higher bias. This trend follows the correlation observed in pure carbon films prepared at room temperature, which have a maximum film stress, density, and sp³ fraction at the same bias value.⁸ These stress curves can be understood in terms of a thermal spike model, which is relevant when growing films using energetic deposition. At low bias (and energy), the depositing flux does not penetrate the surface to generate stress, while at high bias (and energies), the heated volume (the “thermal spike”) is sufficiently large and long lasting to allow stress relief.^15,16 This model could explain why the deuterated curve has a higher maximum stress than the hydrogenated curve since the deuterium is more efficient at exchanging its forward momentum in collision with carbon atoms. The same mechanism also explains why the deuterated material shows generally lower stress than the hydrogenated material at higher biases. Analysis of the films using transmission electron microscopy (JEOL 2100F operating at 200 kV) showed that regardless of bias or whether hydrogen or deuterium was used, the structure was amorphous (the inset in Fig. 1).

Figure 2(a) compares XPS C 1s spectra (Kratos Analytics AXIS Supra using a monochromated Al 1437 eV K-α source) from hydrogenated and deuterated films deposited at 80 V bias. The spectrum from the deuterated sample is almost identical to that of the hydrogenated sample, confirming that both have essentially the same bonding configuration. This result confirms previous work, which showed that deuterium can be substituted for hydrogen without significantly affecting the bonding.¹⁷ The spectrum was fitted with CasaXPS with a single symmetric Gaussian–Lorentzian profile at ∼285.2 eV, indicating the samples produced at 80 V bias contain predominantly sp³ bonding. The diamond-like nature of this sample is consistent with its synthesis near the peak in stress where the maximum sp³ bonding is found in pure carbon (corresponding to ta-C).^12, Figure 2(b) shows the results for the films produced at 650 V, which again shows little difference between deuterated and hydrogenated forms. In this case, the spectrum was fitted with an asymmetric peak at 284.4 eV, consistent with predominantly sp² bonding and synthesis under lower stress conditions.¹⁸  Figure 2(c) shows the XPS spectrum from a carbon-only film (ta-C) deposited at 80 V, which exhibits mixed sp² and sp³ bonding. The reduction in sp² bonding in the deuterated/hydrogenated films prepared at the same bias [Fig. 2(a)] demonstrates that the included hydrogen/deuterium forms mainly sp³ bonds with carbon.

In order to investigate the atomic structure of the deuterated material produced at 80 V, neutron diffraction was performed at the spallation neutrons and pressure (SNAP) diffractometer of the spallation neutron source (SNS) with the beamline setup for large Q-range measurements (0.7–30 Å⁻¹). Deuterated material was used since hydrogen has a large incoherent scattering cross section, increasing the noise background. A ∼5 g powder sample was prepared by collecting material delaminated during repeated depositions onto glassy carbon and placed into a thin glass capillary. Figure 3 (black solid line) shows the reduced density function G(r), a type of radial distribution function, obtained by Fourier transformation of the background-subtracted and normalized scattering intensity S(Q) (refer to Ref. 19 for details of the method). G(r) is defined as 4πr(ρ(r)-ρ_average), where ρ(r) is the density of atom centers at a distance r from an average atom and ρ_average is the average density of atom centers in the sample. The main peaks have been labeled, and the bond lengths and angles were summarized in Table I. In comparison to ta-C,²⁰ the nearest neighbor C–C bond distance, the C–C second nearest neighbor distance, and the C–C–C bond angle of the deuterated material are the same within experimental uncertainty. Therefore, the carbon network structure in our material has the same local environment of carbon atoms as in ta-C, confirming that we have formed a material dominated by tetrahedral carbon–carbon bonding. The C–D bond is clearly visible at ∼1.1 Å, confirming the material contains a significant fraction of deuterium. The deuterium content of this sample was measured to be 24% using elastic recoil detection (ERD). The microscopic density of the material was determined from the plasmon peak position measured using electron energy loss spectroscopy (EELS).²¹ The microscopic density was measured to be 2.7 ± 0.1 g/cm³, consistent with previous measurements of dense amorphous carbons.^11,21

The G(r) from the material synthesized by depositing acetylene ions in an ion beam¹² is shown for comparison in Fig. 3 and has a shorter C–C distance (Table I) and a larger bond angle of ∼115° both indicating a lower density and the presence of significant sp² hybridized carbon. Also shown for comparison is the G(r) for a model with a density of 2.9 g/cm³ containing 52 carbon atoms and 12 hydrogen atoms that were produced using ab initio molecular dynamics.²² The G(r) of our ta-C:H sample and this model are in reasonable agreement, including C–C bond distances and C–C–C angles shown in Table I. The lack of an obvious peak at the predicted D–C–D peak, a 1.8 Å in our experiment indicates that most of the deuterium is in the monohydride configuration with one deuterium atom bonded to each carbon atom as in the model of Ref. 22.

For comparison purposes, Fig. 3 also includes the G(r) from a-Si:D,²³ scaled so that the first Si–Si peak position aligns to the first C–C peak in our sample. The close match (except for the Si–D peak position, which will not scale correctly) confirms that the underlying C/Si networks are similar, and our material is a dense, fully tetrahedrally bonded hydrogenated solid. A small discrepancy occurs in the position of the second scaled Si–Si peak compared to the corresponding C–C peak consistent with the increase in the average C–C–C bonding angle (Table I) resulting from a small residual of the sp² component. In a comparison of our G(r) with that of tetrahedrally bonded polymers including polyethylene, we found a poor match due to incorrect peak positions and intensities. Instead, our results indicate a structure that consists of disordered diamond-like fragments surrounded by tetrahedrally bonded hydrogen zones. The neopentane molecule, although having a much higher level of hydrogenation (70%), has similar structural characteristics in that it contains a diamond-like core surrounded by hydrogen terminations. The G(r) of neopentane shown in Fig. 3 compares well to our material.²⁴ 

The confirmation that we have produced dense, ta-C:H material motivated an investigation of its electronic properties. Figure 4(a) compares the valence band spectra [ultra-violet photoelectron spectroscopy or UPS measured using a Kratos AXIS Supra and a He (II) (40.8 eV) emission line] for the dense hydrogenated material produced at 80 V with carbon-only (ta-C) material at the same bias. Regions have been labeled that correspond to different bonding states.²⁷ Comparing the non-hydrogenated and hydrogenated material, the signal in the region corresponding to π bonding decreased in intensity as a result of the hydrogenation. This observation is consistent with the C 1s XPS results (Fig. 2). The valence band maximum (VBM, determined by fitting the secondary electron cut-off²⁸) was determined to increase from ∼1.1 to ∼1.8 eV with the incorporation of hydrogen. This increase in VBM and the reduction of signal due to π bonding indicates that the addition of hydrogen has effectively increased the bandgap and removed defect states from within the bandgap. The ta-C:H could, therefore, have improved electronic properties compared to ta-C since the states in the bandgap likely act as traps, affecting carrier lifetime and mobility.²⁹ 

Scanning tunneling spectroscopy (STS) analysis supported the UPS results. STS was conducted by collecting I–V sweeps on our ta-C:H and ta-C samples deposited onto substrates of Pt coated SiO₂ and probed with a solid Pt/Ir tip using an atomic force microscope (Asylum Research MFP-3D). Figure 4(b) shows the dI/dV plot corresponding to the electronic density of states of ta-C:H and ta-C and clearly demonstrates that hydrogenation removes states from the bandgap region. Linear fits were applied to the band edges of each curve, yielding the conduction and valence band energies. The onset of the valence band was determined to be 2.0 eV for the ta-C:H material, in close agreement with the UPS measurement of 1.9 eV. The energy difference between the valence and conduction bands represents the electronic gap, which was measured to be E_g = 4.3 eV for ta-C:H compared to E_g = 2.7 eV for ta-C. The value was found to compare well with measurements of the optical bandgap of 4.5 eV determined using the Tauc method from transmittance values of the material deposited onto A-plane (⁠ $11 2 ¯ 0$⁠) sapphire measured using UV-Vis spectroscopy (Agilent Technologies Cary 60 spectrophotometer). Using the values of VBM and E_g, the conduction band offset was calculated to be 2.5 eV. Comparing this offset with the energy of the VBM indicates that the material is weakly p-type as reported for ta-C.³⁰ 

Photoconductivity provides a useful assessment of carrier mobility and lifetime. An interdigitated test device [the inset in Fig. 4(c)] was fabricated that consisted of a 30 nm layer of our ta-C:H with Cr top contacts. Figure 4(c) shows the conductivity of the ta-C:H in the dark and when illuminated with a mercury lamp, plotted as a function of the inverse temperature. These measurements were conducted in the voltage range of −40 to 40 V. The curves show little difference at high temperatures but separate at temperatures below 300 K indicating the presence of photoconductivity. Photoconductivity is more apparent at low temperatures where the background of thermally activated conductivity is greatly reduced, and thermal quenching of photoconduction is minimized. Similar behavior has been reported for a-Si:H³¹ as well as disordered carbons with a range of sp³ fractions.^10,32 The conductivity in a-C materials in the high temperature range has been attributed to localized sp² “defect states,” which give rise to hopping conductivity in which carriers pass from one localized state to another.^2,11 Our value of conductivity at room temperature of 10⁻⁸ S/cm is much lower than that of a-C with lower sp³ fraction (10⁻²–10⁻³ S/cm) and lower than that of ta-C (10⁻⁶–10⁻⁷ S/cm),^2,11 indicating that hydrogen passivation of these defect states has occurred.

We propose a model of the structure of the ta-C:H material, where regions of tetrahedrally bonded carbon are separated by zones of hydrogenation by analogy with the molecular structure of neopentane. This model predicts regions where electron states are diamond-like within the tetrahedral region, separated by barriers of hydrogenation. Conductivity would then take place by thermally assisted tunneling or hopping of carriers between diamond-like regions with limited mobility. This structural model for ta-C:H is necessitated by the bond angle rigidity of the sp³ hybridized carbon, which discourages the incorporation of hydrogen “defects” in the tetrahedral network. In contrast, the amorphous silicon network can tolerate defects by its greater tetrahedral bond angle flexibility as demonstrated by ab initio simulation of the silicon and carbon tetrahedral network structures, prepared using liquid quench methods.^25,26 The top right inset of Fig. 3 demonstrates that the tetrahedral bond angle distribution in ta-C is significantly narrower than that in amorphous silicon. These findings on bond angle distribution are also evident in the G(r) results from hydrogenated materials from Fig. 3. The width of the second scaled Si-Si peak is wider than the corresponding second C–C peak at ∼2.5 Å, indicating that the bond angle distribution is larger in a-Si:H than in our ta-C:H material. Large deviations in bond angles in sp³ bonded carbon are not favored energetically and instead the network reconfigures to allow sp² bonding. To overcome the constraint of bond angle rigidity in carbon tetrahedral networks, it may be possible to inject hydrogen into the structure under pressure, encouraging a more uniform distribution of hydrogenation, enabling extended state rather than hopping conduction. Such an experiment could be done by compressing hydrogenated material in a diamond anvil cell while measuring the temperature dependent conductivity.

In summary, a tetrahedrally bonded hydrogenated amorphous carbon, with a density of 2.7 ± 0.1 g/cm³, has been synthesized using energetic deposition in a filtered cathodic vacuum arc. The atomic structure of this material was investigated using neutron diffraction and found to be related to the fully tetrahedrally bonded molecule neopentane. Therefore, the structure of ta-C:H and a-Si:H is different in that the hydrogen distribution is likely to be inhomogeneous leading to tetrahedrally bonded carbon regions surrounded by hydrogen rich regions. This structure leads to poor carrier mobility and weak photoconductivity compared to a-Si:H. While hydrogenation removes states from the bandgap, hopping conduction remains the dominant conduction mechanism in this material.

The authors would like to thank B. A. Cook, S. Wong, and T. Shiell for assistance with the synthesis of the hydrogenated materials. Electron microscopy was conducted at the RMIT Microscopy and Microanalysis Facility (RMMF). ERDA analysis was performed at the Australian Facility for ion-implantation Research (AFAiiR), a node of the NCRIS Heavy-Ion Accelerator capability. This research used resources at the Spallation Neutron Source, a DOE Office of Science User Facility operated by the Oak Ridge National Laboratory.

This manuscript has been authored by UT-Battelle, LLC under Contract No. DE-AC05-00OR22725 with the U.S. Department of Energy. The United States Government retains and the publisher, by accepting the article for publication, acknowledges that the United States Government retains a non-exclusive, paid-up, irrevocable, world-wide license to publish or reproduce the published form of this manuscript, or allow others to do so, for United States Government purposes. The Department of Energy will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).