Skip to content

Electrophysical properties of diamond films deposited on silicon substrates by hfcvd method

FieldValue
JournalPhysical Sciences and Technology
Published2026-06-19
AuthorsА. С. Саидов, А. Кутлимратов, U. Kh. Rakhmonov, Otabek Turgunov, D. V. Saparov
DOI10.26577/phst20261313

Conductive diamond films 2.2 to 2.8 μm thick were grown on n-type Si(100) by HFCVD without preliminary hydrogen etching. Raman spectroscopy showed a strong diamond peak at 1333 cm⁻¹, and Hall measurements from 300 to 500 K confirmed semiconducting transport. Carrier concentration rose with temperature, while mobility decreased and resistivity fell overall. The process produced a growth rate of 0.2 to 0.3 μm/h and left a weak graphite band near 1561 cm⁻¹.

The study used n-type monocrystalline Si(100) wafers, 300 μm thick, with a resistivity of 0.1 Ω·cm. Methanol supplied carbon in a hydrogen-rich gas mixture, and ammonia supplied nitrogen. Unlike the authors’ earlier silicon growth work, this process omitted hydrogen etching of the substrate before deposition.

ParameterSetting
Growth methodHFCVD
Surface preparationNo preliminary hydrogen etching
FilamentTungsten at 2250 to 2300 °C
Substrate temperature870 to 900 °C
Chamber pressure75 to 80 Torr
Total flow of CH₃OH, H₂, and NH₃65 to 70 cm³/min
Methanol concentration0.5 to 0.8% of the total gas mixture
Deposition time8 to 10 h
Film thickness2.2 to 2.8 μm
Average growth rate0.2 to 0.3 μm/h

About 50% of samples produced under identical conditions had a polycrystalline structure. The authors associated this variability with possible changes in nucleation density, local temperature gradients, and active hydrogen concentration. The selected temperatures were higher than those in several earlier studies. Their stated purpose was to increase the concentration and activity of atomic hydrogen, which preferentially removes sp²-bonded carbon. This selective removal improved phase quality but reduced growth rate. The resulting rate was below the roughly 1 μm/h value reported as typical for HFCVD in the paper’s introduction.

Close to the tungsten filament, methanol decomposition produced CO and H₂, while hydrogen dissociation supplied atomic hydrogen. The paper describes ammonia decomposition as a source of NH radicals and hydrogen. These reactive species reached the silicon surface, where carbon first formed a thin SiC layer. Diamond nuclei then formed on that interface while sp² and sp³ carbon arrived together.

Atomic hydrogen preferentially etched the sp² phase because the paper describes its bonding as weaker than sp³ bonding. When the etching rate exceeded the formation rate of non-diamond carbon, the film quality improved at the expense of deposition speed.

The authors cite nitrogen concentrations of 0.1 to 0.6% from prior work. They state that this range can increase growth rate and stabilize the surface. They also state that nitrogen creates donor levels about 1.7 eV below the conduction band. The paper does not report a measured nitrogen concentration for these films. The cited concentration range therefore describes prior work, not the doping level verified in this study.

Raman spectra were recorded at 300 K with a 514 nm laser. The main diamond line appeared near 1333 cm⁻¹, close to the cited reference position of 1332 cm⁻¹. Its full width at half maximum was about 10 to 12 cm⁻¹. The authors interpret this width as relatively good crystalline quality with minor defects and residual sp² carbon.

A weaker graphite G band appeared near 1561 cm⁻¹. Its intensity was low relative to the diamond line, supporting a film dominated by sp³-bonded carbon. The data do not indicate phase-pure diamond, because the paper explicitly reports incomplete etching of non-diamond carbon.

Figure 1a also lacks the 15R-SiC interlayer feature found in earlier films. A sample prepared with hydrogen pre-etching produced no detectable diamond peak. It instead showed a graphite-like phase and a 15R-SiC feature. Under the tested conditions, this comparison supports the authors’ conclusion that omitting pre-etching improved the observed carbon phase structure.

Transport data came from Van der Pauw measurements with an HMS-7000 Hall system. Four silver contacts were thermally evaporated onto the sample edges in vacuum at about 10⁻⁵ Torr. Contact deposition heated the sample to about 450 °C, and the contacted sample was then annealed for 10 minutes.

The system acquired resistivity, Hall mobility, and carrier concentration over 300 to 500 K. The reported uncertainty for all three quantities was about 2%. The transport figures refer to the film whose Raman spectrum appears in Figure 1a. The authors used the resistivity relation ρ = 1/[q(μₙn + μₚp)], linking resistivity to electron and hole concentrations and their mobilities.

The plotted carrier concentration increased from 6.39843 × 10¹⁷ cm⁻³ at 300 K to 4.72784 × 10¹⁸ cm⁻³ at 500 K. The rise was gradual up to about 480 to 485 K. Above roughly 490 K, the increase became stronger. The authors associated the high-temperature change with activation of deeper impurity or defect states, possibly residual sp² carbon.

Mobility moved in the opposite direction and decreased as temperature rose. The decline was slow at first and steeper at higher temperatures. The authors interpret this as a change from impurity-limited scattering at low temperature to phonon-limited scattering at high temperature. They cite acoustic and optical phonons as the elevated-temperature scattering source and use Matthiessen’s rule to combine scattering contributions.

Resistivity decreased overall as carrier concentration increased. A slight rise occurred from 300 to 320 K, which the authors associate with possible carrier localization or compensation. Figure 4 uses axes from 0 to 250 cm²/(V·s) for mobility and 0.032 to 0.052 Ω·cm for resistivity. The authors interpret these trends through impurity and phonon scattering rather than reporting a fitted microscopic transport model.

Earlier films from the same group were predominantly polycrystalline. They had 2 to 3 μm grains, noticeable graphite phases, and a 15R-SiC interlayer. That work also omitted temperature-dependent carrier, mobility, and resistivity measurements. The present films instead produced a pronounced diamond Raman line and enabled Hall measurements from 300 to 500 K.

The direct pre-etch comparison strengthens the process result. The pre-etched comparison sample showed no diamond Raman peak. The no-pre-etch sample gave the 1333 cm⁻¹ line and a weak 1561 cm⁻¹ graphite band. The improvement applies to the tested gas chemistry and temperature range. It does not establish that pre-etching is harmful under every HFCVD condition.

The structural gain came with a process penalty. The measured growth rate was 0.2 to 0.3 μm/h. This was well below the roughly 1 μm/h cited as typical for HFCVD. The authors attributed this result to stronger atomic-hydrogen etching of non-diamond carbon.

Engineering implications and unresolved issues

Section titled “Engineering implications and unresolved issues”

The process offers a direct route for depositing conductive diamond on n-type Si(100) without a hydrogen pre-etch. The Raman result shows predominantly sp³-bonded carbon, and the measured transport trends show semiconducting behavior over 300 to 500 K. The authors propose potential electronic and optoelectronic applications, but they did not fabricate or test a device.

Several issues remain. Roughly half of identically processed samples were polycrystalline, which limits consistency between nominally identical samples. Residual sp² carbon and minor structural defects remain in the Raman data. The transport interval ends at 500 K, so behavior at operating temperatures outside that range is unknown. Nitrogen incorporation is discussed, but no film nitrogen concentration or fitted activation energy is reported. The abstract and methods give thicknesses of 2.2 to 2.8 μm, while the conclusion gives 2.2 to 2.5 μm. The paper also notes filament degradation as a known HFCVD limitation.

For material selection, direct growth on silicon is the main process benefit. Measured semiconducting transport is documented only over 300 to 500 K. Device integration must account for the silver-contact anneal at about 450 °C. It must also tolerate residual non-diamond carbon and variability in nucleation and structure. The demonstrated result is semiconducting transport over the measured range, not device-level performance.

View Original Abstract

Electrically conductive diamond films were deposited on n-type Si(100) substrates by the hot filament chemical vapor deposition (HFCVD) method using a hydrogen-methanol gas mixture with ammonia addition. In contrast to conventional approaches, film growth was performed without preliminary hydrogen etching of the silicon surface, which simplified the technological process. The deposition conditions were optimized to suppress the formation of non-diamond carbon phases and improve film quality. Raman spectroscopy revealed a pronounced diamond-related peak at 1333 cm⁻¹ with a relatively small contribution of sp²-bonded carbon, indicating the predominance of the diamond phase in the deposited films. The films had thicknesses of 2.2-2.8 μm and were grown at a rate of 0.2-0.3 μm h⁻¹. The electrophysical properties were investigated in the temperature range of 300-500 K using Hall-effect measurements. The carrier concentration increased exponentially with temperature, while the carrier mobility decreased due to enhanced phonon scattering, confirming the semiconducting nature of the deposited films. The observed temperature dependences of carrier concentration, mobility, and resistivity were found to be consistent with the behavior expected for semiconductor materials. The results demonstrate that high-temperature HFCVD growth without preliminary hydrogen etching can produce diamond films with satisfactory structural quality and stable electrical characteristics, making them promising for potential electronic and optoelectronic applications. Keywords: CVD technology, diamond film, silicon substrate, Raman spectra, charge carrier mobility, temperature dependence.