w.elsevier.com/locate/tsf

Thin Solid Films 515 (

Diamond chemical vapour deposition on seeded cemented

tungsten carbide substrates

Gil Cabral *, J.C. Madaleno, E. Titus, N. Ali, J. Gracio

Centre for Mechanical Technology and Automation, Department of Mechanical Engineering, University of Aveiro, 3810-193 Aveiro, Portugal

Available online 23 January 2006

Abstract

Diamond particles were deposited onto seeded cemented tungsten carbide (WC–Co) substrates using conventional hot-filament chemical

vapour deposition (HFCVD) and time-modulated CVD (TMCVD) processes. The substrates were pre-seeded ultrasonically with diamond

particles of different grit sizes. In this investigation, we employ timed methane (CH4) gas modulations, which are an integral part of our TMCVD

process in order to enhance diamond nucleation density. During diamond deposition using the conventional HFCVD process, methane gas flow

was maintained constant. The total hydrogen flow into the reactor during TMCVD process was higher than in the HFCVD process. Hydrogen

etching can be expectedly more prominent in the TMCVD process than in HFCVD of diamond particles. Scanning electron microscopy (SEM)

and atomic force microscopy (AFM) results showed that a proper selection of the diamond grit size for seeding using ultrasounds can lead to

enhancement in the nucleation density values of about two orders of magnitude (107 to 109 cm�2). The TMCVD process using the different

seeded substrates can result in high nucleation density values of up to 1010 cm�2.

D 2005 Elsevier B.V. All rights reserved.

Keywords: Diamond; Nucleation; Cemented carbides; Hot-filament

1. Introduction

One of the continuing challenges facing CVD diamond films

onWC–Co substrates is to obtain smooth and adherent diamond

coatings. There is on-going research and development in

attempting to improve the coating performance. It is well known

that, in CVD process, discrete particles of diamond crystals

sparsely appear on the substrate during the early stages of CVD

diamond deposition and coalesces subsequently to form a

continuous film. The nucleation density of diamond particles

plays an important role in defining the adhesive properties of the

resulting films [1,2]. The presence of cobalt binder inWC–Co is

problematic in diamond deposition since it solubilise carbon at

high temperatures and catalyses the formation of graphite in the

conditions typical for diamond growth. The high carbon

solubility and diffusivity in cobalt retard diamond nucleation.

This induces nucleation from graphite instead of tungsten

carbide and results in a poor adherence. Thus, the success in

obtaining good adherence and high diamond nucleation depends

on the ability to suppress carbon–cobalt interactions. A number

0040-6090/$ - see front matter D 2005 Elsevier B.V. All rights reserved.

doi:10.1016/j.tsf.2005.12.148

* Corresponding author.

E-mail address: gcabral@mec.ua.pt (G. Cabral).

of surface treatments can be used to overcome these interac-

tions including chemical etching, ion implantation, interface

coating and bias treatment. Among these the widely used

method is chemical etching [2–4]. Peters and Cummings [5]

introduced a widely recognized two step etching procedure

with effectiveness in improving the overall coating adhesion.

The most commonly used method to enhance nucleation on

non-diamond substrates is by damaging the surface of the

substrate either mechanically (polishing) or by ultrasound

abrasion using diamond or other abrasives [6–8]. Ihara et al.

[9] correlated the diamond nucleation density to the amount of

diamond dust (seed) left on the substrate surface after

prolonged ultrasonication in a diamond powder suspension.

They noticed changes in the nucleation density values from 106

to 1010 nuclei cm�2 on mirror polished silicon substrates

treated in various suspensions consisting of different sized

diamond abrasive particles.

In this paper, we ultrasonically seed WC–Co substrates with

different sized diamond particles and subsequently use

TMCVD and HFCVD processes to deposit diamond particles

onto their surfaces. The current investigation will correlate the

as-obtained nucleation densities, using TMCVD and HFCVD

processes, with the size of the diamond particles used for

2006) 158 – 163

ww

Table 1

Information about the samples prepared in this investigation together with the pre-treatments used and the growth process employed

Sample Murakami’s etching (min) Acid etching (s) Diamond seeding (Am) CVD process Number of samples

TMCVD<0.25 6 10 <0.25 TMCVD 2

TMCVD 3–5 6 10 3–5 TMCVD 2

TMCVD 30–50 6 10 30–50 TMCVD 2

HFCVD<0.25 6 10 <0.25 HFCVD 2

HFCVD 3–5 6 10 3–5 HFCVD 2

HFCVD 30–50 6 10 30–50 HFCVD 2

Fig. 1. The flow of CH4/H2 gas used during diamond deposition using the

HFCVD and TMCVD processes.

G. Cabral et al. / Thin Solid Films 515 (2006) 158–163 159

seeding. The quality of the diamond particles deposited will be

assessed using Raman spectroscopy. Finally, we discuss our

findings in relation to the type of CVD process used to produce

the diamond deposits.

2. Experimental

2.1. Substrate pre-treatment

The WC–Co samples used in this investigation were of

commercial grade, as supplied by Palbit (Portugal) and

consisted of 6-wt.% cobalt content with an average WC grain

size of 6 Am. The samples were cylindrically shaped and were

ground using D126/C75 and D20/C75 diamond wheels in

order to obtain final dimensions of 6T0.05 mm diameter

and 0.4T0.05 mm height. To remove the defects induced

during grinding, the sample heights were reduced 0.03T0.01mm by mechanical polishing using diamond paste of 7 and

3 Am. After polishing, the samples were ultrasonically cleaned

in pure acetone for 5 min. The substrates were chemically

etched with Murakami’s solution (10 g K3(Fe(CN)6)+10 g

KOH+100 ml H2O) for 6 min at room temperature in

ultrasounds followed by a rinse in de-ionised water. Subse-

quently, the surface cobalt was removed by acid treatment (1 ml

46-wt.% HNO3+9 ml 30% m/v H2O2) for 10 s, followed by a

rinse in de-ionised water.

Prior to diamond deposition, the samples were ultrasonically

abraded for 20 min in diamond suspension. The suspensions

were prepared using three different grit sizes, of commercial

diamond; (i) <0.25 Am, (ii) 3–5 Am and (iii) 30–50 Am. In all

cases, 1 carat of diamond powder was mixed with 20 ml of

pure methanol. Table 1 gives details of the substrates, surface

pre-treatment, and the CVD process employed to produce the

diamond deposits.

2.2. HFCVD and TMCVD processes

Diamond deposition was carried out in a custom built

stainless steel HFCVD system. Deposition pressure and

filament power values were set to 5 kPa and 200 W per

filament, respectively. The substrate surface temperature was

measured using a K-type thermocouple, which was located

directly underneath and touching the bottom of the substrate

during deposition. In the HFCVD process, the substrate

temperature was 700T8 -C whereas in TMCVD, the substrate

temperature was in the range 680T5 to 700T8 -C. The

temperature fluctuation in TMCVD was due to the timed CH4

modulations with respect to deposition time. In HFCVD

process, the total gas flow was kept constant at 200 sccm.

The CH4 /H2 volume ratio was fixed at 1% as per calculation

from Eq. (1), where t, n and A represents the time of the

modulation, the number of modulation in deposition cycle and

the total amount of CH4 in each modulation, respectively. The

indices 1 and 2 refer to the high and low volume of CH4 in the

cycle in accordance with the graphical representation shown in

Fig. 1. In both cases, the deposition time was set to 32 min. For

each condition, the tests were carried out twice in order to

check the reproducibility and repeatability of the results.

CH;

4 ¼t1n1A1 þ t2n1A1

n1t1 þ n2t2ð1Þ

2.3. Characterization techniques

Surface morphology of the as-grown diamond deposits was

characterized by scanning electron microscopy (SEM/EDS,

Hitachi 4100) and atomic force microscope (AFM: Nano-

scope03-Digital Instrument). The nucleation density values

were determined by digitalizing SEM images (total area of

46 Am2 each at three different locations) using AUTOCAD

software. The quality of diamond nuclei was assessed by

Raman spectroscopy at room temperature using a ISA JOBIN

YVON-SPEX T6400 spectrometer equipped with an argon ion

laser (k =514.5 nm). The data was smoothened and then fitted

Fig. 2. SEM micrographs showing the surface morphologies of WC-6-wt.% Co substrates after 6 min etching with Murakami’s reagent (a) followed by 10 s etching

in HNO3 and H2O2 acid solution (b).

G. Cabral et al. / Thin Solid Films 515 (2006) 158–163160

with mixed Gaussian–Lorentzian functions [10] using PeakFit

software.

3. Results and discussion

3.1. Chemical pre-treatments

The average substrate surface roughness (Ra) was measured

using a surface profilometer (Hommel Tester T1000 profil-

ometer). After Murakami’s and acid etching pre-treatments the

surface roughness was found to be 0.07T0.01 Am. Fig. 2a)

shows the SEM image of the roughened surface after etching

the sample by Murakami’s reagent for 6 min. The presence of

intergranular cobalt is apparent from the SEM image. The

presence of cobalt was evident in EDS microanalysis since it

was not attacked by the alkaline solution of potassium

ferricyanide. Fig. 2b) shows the surface morphology of the

same sample after 10 s etching with acid, HNO3+H2O2,

solution. This extended acid treatment dissolves any binder

residues remaining behind after Murakami’s etching.

3.2. Diamond seeding

SEM results showed very good seeding uniformity in the

substrates and high reproducibility. From Fig. 3a) it is evident

that residual diamond particles with sizes between 8 and 40 nm

cover the WC grains. The decrease in the number of seeding

sites with larger sized diamond particles was apparent from

SEM examinations (see Fig. 3b). It is expected that smaller

sized diamond particles are likely to get trapped and become

embedded in the WC–Co substrate cavities, whereas, for larger

Fig. 3. SEM micrographs showing (a) diamond seeds as obtained after ultrasonically

an average size between 10 and 30 nm) and (b) diamond seeding using diamond p

sized particles the process of particle entrapment in the cavities

is more difficult. Therefore, by using smaller sized powders,

this will result in the seeding of greater number small sized

particles compared to larger powder sizes, which are the active

sites for subsequent diamond nucleation using CVD.

The correlation between the as-obtained diamond nucleation

densities and the diamond grit size used during ultrasonic

seeding is shown in Fig. 4. Under our standard experimental

conditions, the diamond nucleation density for 0.25 Am was

more than two orders of magnitude higher than for 30–50 Amsize diamond powder. The comparison of nucleation density

values attained by the two CVD processes are shown in Fig. 5a

and b. Using 0.25 Am size diamond powder in seeding the

WC–Co substrates, we observed an increase in the nucleation

density from ¨4�109 to 1.5�1010 which is approximately an

order of magnitude. Detailed SEM studies showed that with

TMCVD the grain size distribution of the diamond nuclei was

more homogeneous, the grains were in the nano-range and the

deposits displayed better substrate surface area coverage. This

is in agreement with our earlier reports [11,13,14], which

demonstrated that the TMCVD process promoted secondary

diamond nucleation during the timed methane modulations. It

is remarkable, the high uniformity and reproducibility is

obtained in all the samples, as can be seen at low magnification

micrographs, Fig. 5c and d. Both SEM (Fig. 5a) and AFM

(Fig. 6a and b) micrographs show the formation of diamond

grains of size ranging from 50 to 110 nm. The size distribution

of the particles produced by the TMCVD process, as shown in

Fig. 7 suggests a Gaussian distribution with an average crystal

size of 82.5 nm. Similar distribution was found for the HFCVD

process, which produced diamond particles with an average

seeding for 20 min using diamond grit size lower than 0.25 Am (the seeds are of

owder of grain size between 3 and 5 Am.

Fig. 4. Graph showing the relationship between diamond nucleation densities

and the size of diamond particles used for seeding for both HFCVD and

TMCVD processes.

G. Cabral et al. / Thin Solid Films 515 (2006) 158–163 161

size of 110.5 nm. From these observations, it is indicative that

the rate of crystal growth on the diamond seeds is practically

almost the same, which is in some agreement with the kinetic

model proposed by Molinari et al. [15].

Several researchers [16–18] reported that in HFCVD

reactors, an additional source of hydrogen due to the

heterogeneous dissociations of H2 on the filament is also

included along with the experimentally observed H-atom [19]

concentration. When the level of carbon in the gas phase

becomes greater than a critical value, depending on the

filament temperature, deposition of carbon onto the filament

occurs. This deposition is commonly referred to as filament

‘‘poisoning’’ or ‘‘sooting’’. Filament poisoning results in a

reduced ability to transform H2 to H-atom [17,21,22]. A

subsequent decrease in the substrate temperature arising from

Fig. 5. SEM images showing the diamond deposits formed on the surfaces of th

b and d) processes.

the increased methane concentration is the result of the lower

level of H-atom production in the HFCVD reactor [23].

It is known that atomic hydrogen etches both diamond and

graphite; however, under typical CVD diamond deposition

conditions, the rate of diamond growth exceeds the etch rate

whilst for other forms of carbon, such as graphite the case is

reverse. In our case, we explain our findings by giving two

hypothesis. First, during the high timed methane modulations

in TMCVD, the relative monohydrogen concentration in the

reactor is relatively (to HFCVD) lower, therefore, less

hydrogen etching of the diamond seeds can be expected. In

our case, we propose that as the hydrogen etches the seeds, thus

reducing the surface density of diamond growth centres. We

hypothesise that less of the seeds will be etched out during

TMCVD compared to HFCVD of diamond deposits. In

HFCVD process, the hydrogen concentration in the reactor is

constant and the overall H-concentration is higher than in the

TMCVD process. Therefore, more of the seeds are expectedly

etched out leaving behind inactive sites for any possible

diamond growth by CVD. Our SEM results showed that after

the HFCVD and TMCVD processes, the density of the

diamond seeds on the WC grains, as was clearly observed

after ultrasonic seeding and prior to diamond deposition (see

Fig. 3), had decreased significantly. We postulate by attributing

the disappearance of these seeds to the hydrogen etching effect

during diamond CVD.

Second, we can also expect some heterogeneous nucleation

taking place in correspondence of surface defects or sites

produced by the impacts of diamond particles during the

seeding treatment. The larger CH4 concentration pulse would

favour a more abundant heterogeneous nucleation process [24].

Therefore, we expect the heterogeneous nucleation rate to be

higher in TMCVD than in the HFCVD process, mainly due to

e WC–Co substrates using HFCVD (panels a and c) and TMCVD (panels

Fig. 6. AFM images showing (a) two dimensional representation at high

magnification; and b) three dimensional representation at lower magnification

of the surface morphologies of diamond deposits produced by the TMVCVD

process using <0.25 Am seeding powder.

Fig. 7. Graph showing particle size distribution for the sample TMCVD <0.25

after 36 min of deposition.

Fig. 8. Raman spectra of the diamond deposits produced by the TMCVD (a

and HFCVD (b) processes. The seeding powder used was of grit size<0.25 Am

G. Cabral et al. / Thin Solid Films 515 (2006) 158–163162

the utilisation of high/low timed methane gas flow cycles

during deposition.

Secondary nucleation could also take place on defects on the

surface of growing diamond crystallites. Again, we speculate

that in this case the high methane bursts in TMCVD could have

potentially increased the overall surface density of growing

diamond crystals.

3.3. Raman analysis

Fig. 8 shows the Raman spectra of diamond deposits

produced on seeded WC–Co substrates with diamond powder

using TMCVD (a) and HFCVD (b) processes. In this case, the

diamond powder size used for pre-seeding the substrates was

0.25 Am in size. Besides the diamond peak (1334 cm�1), five

other main components contributed from the carbonaceous

)

.

G. Cabral et al. / Thin Solid Films 515 (2006) 158–163 163

Raman background were noticed. A band centred around

1140 cm�1 assigned to transpolyacetylene lying at the grain

boundaries [24], a band around 1490 m�1 from diamond

precursors [25], the band around 1250 cm�1 is generally

accepted as arising from scattering in diamond, due to

relaxation of the wave vector selection rule [26] and two broad

bands at 1350 and 1580 cm�1, attributed to D- and G-bands of

graphite [27]. The full width at half maximum (FWHM) values

for the Raman diamond lines were 8.0 for both the samples

shown in Fig. 8. The variation in the FWHM values reflects the

amount of disorder and defect densities [28]. The constant

FWHM shows the uniformity in the crystallite size of diamond

particles found in the TMCVD and HFCVD deposits. The data

also confirm our previous findings, i.e., that the Raman quality

of the diamond grown by TMCVD is only slightly lower than

that obtained by conventional HFCVD.

4. Conclusions

Substrate pre-treatments and deposition conditions strongly

influence the nucleation behaviour of diamond deposited onto

seeded cemented WC–Co substrates. The size of the diamond

seeding powder proved to be an influential factor in governing

the diamond nucleation density.

The diamond deposits produced using conventional

HFCVD or TMCVD process were of similar Raman quality,

in terms of both crystallite size and carbon phase purity.

However, the TMCVD process enabled the enhancement of

diamond nucleation density by about one order of magnitude

compared to the HFCVD process.

We explained our findings by proposing that the larger time-

averaged CH4 concentration in the TMCVD process could

have reduced the H-etching of diamond seeds, and enhanced

both the heterogeneous and secondary nucleation processes.

Acknowledgements

The authors wish to thank the Foundation for Science and

Technology in Portugal for financial support of the project

POCTI/EME/61676/2004. The authors also wish to thank

Palbit, Hard MetalTools Solutions—Portugal for supplying

the substrates.

References

[1] K. Mallika, R. Komanduri, Wear (1999) 245.

[2] R. Haubner, B. Lux, J. Phys. C 5 (1989) 169.

[3] B.S. Park, Y.J. Baik, K.R. Lee, K.Y. Eun, H.D. Kim, Diamond Relat.

Mater. 2 (1993) 910.

[4] Y. Saito, K. Sato, S. Matsuda, H. Koinuma, Mater. Sci. 26 (1991) 2937.

[5] M.G. Peters, R.H. Cummings, European Patent 0519587 A1 (1992).

[6] O. Ternyak, R. Akhvlediani, A. Hoffman, Diamond Relat. Mater. 14

(2005) 323.

[7] F. Arezzo, N. Zacchetti, W. Zhu, J. Appl. Phys. 75 (1994) 5375.

[8] Shoji Kamiya, et al., Surf. Coat. Technol. 142–144 (2001) 738.

[9] Manabu Ihara, Hiroshi Komiyama, Tatsuya, Appl. Phys. Lett. 65

(1994) 1192.

[10] W. Bambynek, et al., Rev. Mod. Phys. 44 (1972) 716.

[11] N. Ali, V.F. Neto, Y. Kousar, G. Cabral, J. Gracio, Mater. Sci. Technol.

19 (2003).

[13] N. Ali, G. Cabral, A.B. Lopes, J. Gracio, Diamond Relat. Mater.

(2004) 495.

[14] N. Ali, V.F. Neto, Sen Mei, G. Cabral, Y. Kousar, E. Titus, A.A. Ogwu,

D.S. Misra, J. Gracio, Thin Solid Films 469 (2004) 154.

[15] E. Molinari, R. Polini, M. Tomellini, J. Mater. Res. 8 (1993) 798.

[16] D.G. Goodwin, G.G. Gavillet, J. Appl. Phys. 68 (1990) 6393.

[17] M.C. McMaster, W.L. Hsu, M.E. Coltrin, D.S. Dandy, J. Appl. Phys. 68

(1990) 2424.

[18] D.S. Dandy, M.E. Coltrin, J. Appl. Phys. 76 (1994) 3102.

[19] K.H. Chen, M.C. Chuang, C.M. Penney, W.F. Banholzer, J. Appl. Phys.

71 (1992) 1485.

[21] D. Morel, W. Hanni, Diamond Relat. Mater. 7 (1998) 826.

[22] M. Sommer, F.W. Smith, J. Mater. Res. 5 (1990) 2433.

[23] F. Jansen, M.A. Machonkin, D.E. Kuhman, J. Vac. Sci. Technol. A8

(1990) 3785.

[24] A.C. Ferrari, J. Robertson, Phys. Rev., B 63 (2001) 121405.

[25] Nemanich, J.T. Glass, G. Lucovsky, R.E. Schroder, J. Vac. Sci. Technol.

A6 (1988) 1783.

[26] Lopez-Rıos, E. Sandre, S. Leclercq, E. Sauvain, Phys. Rev. Lett. 76

(1996) 4935.

[27] P.K. Bachmann, D.U. Wiechert, Diamond Relat. Mater. 1 (1992) 422.

[28] P. Ascarelli, E. Cappelli, G. Mattei, F. Pinzari, S. Martelli, Diamond Relat.

Mater. 4 (1995) 464.