Ablation of CVD diamond with nanosecond laser pulses of UV IR rangek

Transcription

1 Diamond and Related Materials 7 (1998) Ablation of CVD diamond with nanosecond laser pulses of UV IR rangek T.V. Kononenko, V.G. Ralchenko *, I.I. Vlasov, S.V. Garnov, V.I. Konov General Physics Institute, Russian Academy of Sciences, ul. Vavilova 38, Moscow, Russia Received 5 August 1997; accepted 28 April 1998 Abstract Etch rates of CVD diamond upon irradiation by nanosecond (5 9 ns) pulses at three different wavelengths 1078, 539 and 270 nm at laser fluences in the range J/cm2 were measured. A Nd:YAP laser system operated at first, second and fourth harmonics was used in the ablation experiments. Both shallow (<15 microns) and through holes were etched in a 95-mm thick free-standing diamond film grown by microwave plasma CVD. The ablation rate was found to be wavelength-independent, this result being ascribed to surface blackening caused by amorphization/graphitization as confirmed by Raman analysis. The maximum etch rate approached 600 nm/pulse. The etch rate depended on the crater depth, which was ascribed to the effect of laser plasma interaction inside the deep channel. The possibility of cutting trenches of high aspect ratio has been demonstrated. In a separate experiment, a batch of thin diamond films differing in thermal conductivity (k=2 5 W/cmK ) was ablated with a KrF excimer laser (l=248 nm). No dependence of ablation rate on film quality was observed, which could be explained assuming grain boundaries to be the main source of thermal resistance Elsevier Science S.A. All rights reserved. Keywords: Ablation rate; CVD diamond; Laser 1. Introduction ultraviolet to infrared, the influence of wavelength on etch rate may be significantly reduced. Lasers can be effectively used for diamond processing, Comparison of data reported by different authors in particular, cutting, drilling, patterning and polishing revealed [1, 3] a tendency of ablation fluence threshold of diamond films have been realized (for a review see to increase with laser wavelength increase from UV Ref. [1] and references there). The pulsed laser radiation (193 nm) to IR (10.6 mm). However, for more correct being absorbed in a diamond surface layer can result in analysis ablation experiments are desirable which are material ejection from the beam spot. Both the area performed at similar irradiation parameters such as laser treated and the thickness of the removed layer could be pulse width, beam spot size and radial intensity distribu- precisely controlled by appropriate selection of laser tion. This condition can be best fulfilled using one laser parameters. For a particular application, a choice of with variable wavelength. In the present paper we report optimum laser system can be made from a broad variety on measurements of ablation rates of CVD diamond of laser types available today. That choice should be films upon irradiation by nanosecond (5 9 ns) pulses at based on comparative experimental data on laser dia- three different wavelengths of 1078, 539 and 270 nm in mond interaction. From a general point of view the a broad range of fluences of J/cm2 using the lasers operated at shorter wavelengths, especially those same laser system. In addition, a KrF excimer laser emitting in the spectral region below fundamental operated at 248 nm wavelength was used for etching of absorption band of diamond at 225 nm, such as ArF diamond films with different values of thermal excimer laser (l=193 nm), are preferable due to higher conductivity. optical absorption. On the other hand, since the ablation process proceeds via a surface graphitization [2], which provides stronger absorption in a broad range from 2. Experimental * Corresponding author. Fax: k Presented at the Diamond 97 Conference, Edinburgh, Scotland, August 3 8, An original Nd:YAP laser system consisting of a master oscillator and two-passed amplifier was used in /98/$ see front matter 1998 Elsevier Science S.A. All rights reserved. PII S (98)

2 1624 T.V. Kononenko et al. / Diamond and Related Materials 7 (1998) Table 1 using a DC plasma CVD reactor [5]. The methane Laser parameters concentration varied from 3 to 9% to produce films of Parameter Value different grade with thermal conductivity ranging from 4.6 to 1.9 W/cmK as measured by the photothermal l, nm deflection technique ( mirage effect ) [5]. The freet,ns standing samples were irradiated with a KrF excimer E, mj laser emitting 20-ns pulses at 248 nm wavelength to form through holes of mm2 size. The irradia- tion with both lasers was performed in an ambient atmosphere. l is wavelength, t is pulse duration, E is pulse output energy. ablation experiments. The laser produced single pulse shots with a repetition rate of 1 10 Hz at 1078 nm basic wavelength and a maximum output energy of 7 mj. The 3. Results pulse-to-pulse deviation in pulse energy was less than 20%. The laser beam spatial profile was nearly Gaussian, 3.1. Nd:YAP laser the pulse width being 9 ns at l=1078 nm wavelength. A KTP crystal of 1 cm length was used as the second Fig. 1 shows dependence of ablation rate on laser harmonics converter, and a 2 cm length KDP crystal fluence for three different wavelengths used, 1078, 539 was employed for further wavelength conversion to and 270 nm, as measured for shallow craters and fourth harmonics in the UV region. The laser parameters through holes. The most striking result is that the data are summarized in Table 1. for all laser harmonics fall in a single curve, so the For ablation rate measurements the laser beam was ablation rate appears to be practically insensitive to focused with a lense of F=21 cm focal length into a wavelength. The ablation rate increases with fluence spot of mm diameter on the sample surface. The nonlinearly in the whole fluence range over three orders ablation rate was determined in two ways. First, shallow of magnitude of fluence ( J/cm2). The threshold craters with aspect ratio (depth-to-diameter) of less than fluence for the multipulsed ablation process is 0.5 were produced, and the depth measured with the E #2J/cm2 for shallow craters, while it increases by a th optical microscope was divided by the number of pulses factor of 8 for through holes. In the latter case the high received by the sample. The accuracy of this procedure value of the threshold means that no through channel was about 20%. Second, through holes were etched in can be produced at E<17 J/cm2, even upon prolonged the film, and the ratio of the known film thickness to irradiation, although a clear crater forms. the number of pulses was taken as an average ablation At fluences well above the threshold the ablation rates rate upon drilling of deep channels. for craters of small depth are higher by a factor of up A diamond film of 95 mm thick film was grown to three compared with those for through holes. The on mirror-polished Si substrate of 2.25 inches diameter possible reason for the reduced ablation from deep channels is an enhanced absorption in dense plasma by microwave plasma-enhanced CVD using a CH H O gas mixture as previously described [4] Then the substrate was removed in HF H SO acid to 2 4 obtain free-standing film. The film appeared greyish in colour but was still translucent. Only a minor amount of amorphous carbon inclusions was contained in this material as deduced from Raman spectra. The diamond peak at 1332 cm 1 had a width of 4 cm 1 when the spectrum was taken on a well-faceted growth side, and 5cm 1 on a smooth fine-grained substrate side. The substrate side was irradiated in all experiments to ensure well-defined laser imprints on the surface. Based on our previous measurements of other diamond films of similar quality we estimate the thermal conductivity of the sample to be k=6 10 W/cmK in the direction perpendicular to the film plane. The samples were characterized with SEM and microraman spectroscopy. Raman spectra were taken at 488 nm excitation wavelength by focusing an Ar+ laser beam into a spot of 2 mm diameter. In another experiment several 9 14-mm thick diamond films were grown on Mo substrates in CH H mixtures 4 2 Fig. 1. Ablation rate vs fluence at three wavelengths of Nd:YAP laser: 1078 nm (circles), 539 nm (squares) and 270 nm (triangles). Open symbols correspond to shallow craters produced, closed symbols refer to drilling of through holes in 95-mm thick film.

3 T.V. Kononenko et al. / Diamond and Related Materials 7 (1998) Fig. 3. Crater depth vs number of pulses at l=1078 nm at laser fluence E=5.4 J/cm2 (circles) and E=43.2 J/cm2 (squares). Note the absence of ablation for the first 50 pulses at lower fluence. estimate the thickness of this layer to be of the order of 100 nm or less. When the spectrum was recorded using a confocal optical scheme (the depth of the probed volume is restricted by 3 5 microns in this case, the signal from the underlying diamond being significantly weaker), only the broad asymmetric band of the a C phase was observed. The ejected carbon forming a ring of redeposited material around the crater displayed the presence of disordered graphite only with two well-resolved peaks at 1350 cm 1 and 1580 cm 1 (D and G bands, respec- tively). Similar spectra are frequently observed for dia- mond surfaces treated with a scanning laser beam [6, 7] but at lower fluences than in the present work. This may mean, at least at certain irradiation conditions, that for the beam scanning regime the Raman spectra charac- terize a redeposited carbon rather than the underlying adherent transformed layer, since in this case the fresh ablated area must be covered by a redeposit from the next (shifted) laser spots. Fig. 2. Raman spectra of 95-mm thick film: (a) virgin surface; (b) inside laser produced crater; (c) redeposited carbon around crater; and (d) inside the crater, the spectrum being recorded with confocal optical scheme. restricted by the walls of the channel. The beam defocusing upon progressive deepening of a hole seems to be negligible as the length of the focal region (#600 mm) is much larger than the film thickness. Laser ablation occurs via surface graphitization [2,6,7]. In the ablation process a part of the previously It is necessary to emphasize that the multipulse ablation graphitized layer is removed, but simultaneously a threshold determined from Fig. 1 does not coincide deeper layer of diamond is converted to graphite, so the with the damage threshold for a single pulse. The most laser etching occurs as a pulse-by-pulse penetration of pronounced effect of delay in surface damage at low the graphitic piston into diamond [2]. The formation fluences (E<20 J/cm2) was observed for IR radiation. of the modified layer which is opaque in the whole Fig. 3 shows the dependence of crater depth on the UV IR range could explain the observed constant abla- number of pulses at l=1078 nm, for which the optical tion rate at different wavelengths. Fig. 2 shows Raman absorption of virgin diamond is minimal. A few initial spectra taken from the virgin film surface, from the pulses did not cause any visible craters, for instance, at bottom of a laser-induced crater, and from a ring of E=5.4 J/cm2 it took 50 shots to start the ablation redeposited ablated carbon around the crater. The inten- process. This initiation stage ( priming ) is believed to sity of the diamond Raman peak at 1332 cm 1 frequency be necessary to accumulate a concentration of defects strongly decreases inside the crater by two orders of high enough to result in surface heating followed by magnitude, the peak width of #5cm 1 remaining graphitization. The priming period becomes shorter with unchanged. A broad peak at 1550 cm 1 corresponding fluence, for instance, surface damage was detected after to amorphous carbon is slightly increased relative to the 10 and 5 pulses at fluences of 12 and 23 J/cm2, respec- diamond signal. This means that a transformed (graphi- tively, while at E=43 J/cm2 the crater was created by tized) layer did form, but of minimal thickness, since it the first pulse (see Fig. 3). was partially transparent to the blue light of the Raman To illustrate the capabilities of the Nd:YAP laser for laser source. Based on the fact of the transparency we practical applications, a series of trenches were produced

4 1626 T.V. Kononenko et al. / Diamond and Related Materials 7 (1998) Fig. 5. Ablation rate vs fluence of KrF excimer laser (l=248 nm) for four thin diamond films of different thermal conductivity: k= 4.6 W/cmK (circles), 3.9 W/cmK (squares), 2.6 W/cmK (d) and 1.9 W/cmK (c). Fig. 4. SEM picture of cuts in 110-mm thick diamond film made by the scanning beam of a Nd:YAP laser at 270 nm wavelength at different fluences (left to right): 210, 70, 25 and 10 J/cm2. Each groove is produced by 10 scans, by sending 5 pulses per one step of 2.5 mm. in 110 mm-thick film at 270 nm wavelength, the fluence being varied in the range of J/cm2. The cut depth increases with fluence as shown in Fig. 4, and a through cut was performed at E=210 J/cm2 after 10 scans (each point along this cut received 300 pulses). The cut width is only 20 mm, giving an aspect ratio of about KrF excimer laser 200 nm/pulse at E=11 J/cm2. The drilling process is very fast, e.g. upon an etch rate of 100 nm/pulse and pulse repetition rate of 50 Hz which is quite common for excimer lasers, the drilling speed is 300 mm/min, that is by 4 5 orders of magnitude higher than typical etch rates of diamond by oxygen plasma [ 10,11]. To explain why the ablation was insensitive to thermal conductivity, one has to take into account the real grain structure of the films. Our samples exhibit crystallite size d=4 10 mm on the growth side. The thermal wave originated by laser energy absorbed in a thin graphitized layer travels into depth for a distance l =(Dt)1/2 t #2 mm, where D= cm2/s is thermal diffusivity of the films, and t=20 ns is laser pulse duration. We adopt a one-dimensional heat dissipation model, so the heat flows perpendicularly to film plane. Upon layerby-layer material removal the thermal wave meets grain boundaries relatively rarely, since l <d (columnar grain t shape results in the same conclusion). This means that the thermal resistance caused by grain boundaries where amorphous carbon is concentrated, does not influence the temperature distribution within the grain very much, while the thermal conductivity of the grain bulk seems to be more important. The Raman diamond peak width was nearly constant, Dn#8cm 1 for these samples, so we assume the grains to have a similar quality, but to contain different amounts of a C phase at the grain boundaries. Since the thermal conductivity was evaluated from mirage-effect measurements [ 5], which give an in-plane thermal conductivity value, k,affected by grain boundaries, the local (within the grain) and average thermal properties may not be strictly related. Therefore, the etch rate might display no dependence on the k value as we observed. However, in the case when the diamond grain size is very small, d<l, many t Generally, the materials with lower thermal conductivity, k, are heated and vaporized by laser radiation easier than those with high k. As the low optical absorption of diamond correlates to improved thermal conductivity it is not surprising that an increase of ablation threshold fluence and/or decrease of etch rate for diamond films of better quality were reported [1,2,6,8,9]. We tested a number of diamond films with different thermal conductivity to investigate how this parameter influences the ablation process. The ablation rate for a KrF excimer laser was determined from a number of pulses required to drill through diamond films of about 10-mm thickness. The aspect ratio of the holes was very low, <0.05, thus the hole s wall effects on plasma run away was negligible. The plot of ablation rate against laser fluence for four samples with thermal conductivity varying from 1.9 to 4.6 W/cmK is shown in Fig. 5. No dependence of etch grain boundaries would be encountered on the journey rate on film quality is observed, the threshold ablation of the thermal wave induced by each laser pulse. Then fluence E #2J/cm2 being the same for all the films. the etch rate would depend on an averaged value of k, th The ablation rate increases with E reaching approx. rather than on its local value. Thus, a difference in

5 T.V. Kononenko et al. / Diamond and Related Materials 7 (1998) quality of the films would manifest itself in the etch rate References difference, as was observed by Johnston et al. [8] for diamond films with grain size <1 mm. [1] V.G. Ralchenko, S.M. Pimenov, Diamond Films Technol. 7 (1997) 15. [2] M. Rothschild, C. Arnone, D.J. Ehrlich, J. Vac. Sci. Technol. B 4 (1986) Conclusions [3] C.A. Klein, Laser-induced damage in optical materials: 1994, SPIE Proc (1994) 517. Etch rates of CVD diamond film were determined [4] V.G. Ralchenko, A.A. Smolin, V.I. Konov, K.F. Sergeichev, I.A. upon irradiation by nanosecond (5 9 ns) pulses of a Sychov, I.I. Vlasov, V.V. Migulin, S.V. Voronina, A.V. Khomich, Nd:YAP laser operated at first, second and fourth har- Diamond Relat. Mater. 6 (1997) 417. monics (wavelengths 1078, 539 and 270 nm). Ablation [5] M. Bertolotti, G.L. Liakhou, A. Ferrari, V.G. Ralchenko, A.A. Smolin, E.D. Obraztsova, K.G. Korotoushenko, S.M. Pimenov, threshold fluence (E#2J/cm2) and etch rates under V.I. Konov, J. Appl. Phys. 75 (1994) multipulse exposure coincide for the wavelengths used, [6] V.G. Ralchenko, S.M. Pimenov, T.V. Kononenko, K.G. presumably due to enhanced surface absorption caused Korotushenko, A.A. Smolin, E.D. Obraztsova, V.I. Konov, in: by the graphitization effect. The maximum etch rate A. reached 600 nm/pulse. Narrow cuts with a high aspect Feldman et al. (Eds.), Applied Diamond Conference. NIST ratio can be produced with this laser system. The ablation Spec. Publ. 885, 1995, p [7] S.M. Pimenov, V.I. Konov, E.D. Obraztsova, U. Bogli, P. Tosin, rate of thin diamond films of different quality by KrF E.N. Loubnin, Diamond Films Technol. 7 (1997) 61. excimer laser irradiation was found to be insensitive to [8] C. Johnston, P.R. Chalker, I.M. Buckley-Golder, P.J. Marsden, thermal conductivity, which was explained in terms of S.W. Williams, Diamond Relat. Mater. 1 (1992) 829. grain boundary-dominated thermal resistance. [9] A. Blatter, U. Boegly, L.L. Bouilov, N.I. Chapliev, V.I. Konov, S.M. Pimenov, A.A. Smolin, B.V. Spitsyn, in: Proceedings of the 2nd International Symposium on Diamond Materials, Washington D.C., 5 10 May 1991, The Electrochemical Society, Penning- Acknowledgement ton, 1992, p [10] G.S. Sandhu, W.K. Chu, Appl. Phys. Lett. 55 (1989) 437. This work was supported in a part by International [11] O. Dorsch, M. Werner, E. Obermeier, R.E. Harper, C. Johnston, Science and Technology Center, grant No I.M. Buckley-Golder, Diamond Relat. Mater. 1 (1992) 277.