Plasma assisted deposition of nanocrystalline BCN thin films and ...

Plasma assisted deposition of nanocrystalline BCN thin films and property characterization Z. X. Caoa) and L. M. Liu National Laboratory for Surface Physics, Institute of Physics, P.O. Box 603, 100080 Beijing, China

H. Oechsner Institut fu¨r Oberfla¨chen-und Schichtanalytik, Universita¨t Kaiserslautern, Postfach 3049, 67653 Kaiserslautern, Germany

共Received 7 March 2002; accepted 9 September 2002兲 Electron-cyclotron-wave-resonance plasma assisted deposition is an effective technique for preparing superhard materials. In this work, nanocrystalline BCN thin films were grown on Si共001兲 wafers and Corning glass substrates, where the growing surface was bombarded with nitrogen plasma at energies between 60 and 180 eV. Energy-dispersive x-ray analysis revealed the formation of very clean, homogeneous films with a bulk composition around B42C33N25 , which changes only slightly with ion energy. Under an atomic force microscope the films displayed a morphology composed of crystallites of about 200 nm in lateral size in cubic habits. Both the x-ray photoelectron spectroscopy and infrared absorption indicated that the deposits are ternary BCN compounds. The films are highly transparent and hard; the Vicker’s hardness scatters in the range of 26 –28 GPa. Strong photoluminescence peaked at 430 nm was detected on the as-deposited specimens at room temperature. The band gap for the deposits was estimated to be circa 3.0 eV. © 2002 American Vacuum Society. 关DOI: 10.1116/1.1518973兴

I. INTRODUCTION The structural and electronic similarities between carbon dimer and BN molecule have motivated the synthesis of some ternary BCN structures. It is believed that the ternary BCN compounds, depending on the structure and composition, may manifest tunable properties between the extremes for elemental carbon and boron nitride, i.e., the properties are capable of tailoring.1,2 In the past two decades, intensive efforts have been devoted to the theoretical simulation and experimental investigation on this family of materials, and some remarkable progress has been made, though only very slowly. Liu et al. studied the various bonding configurations for the hexagonal BCN, and Lambrecht et al. calculated the anomalous band-gap behavior and phase stability for the diamond-cBN alloy.3,4 A variety of deposition techniques have been employed for the synthesis of BCN materials, in particular the cubic ones. These include the direct conversion under high-temperature high-pressure conditions,2,5 reactive magnetron sputtering,6 – 8 chemical vapor deposition 共CVD兲, and pulsed-laser deposition 共PLD兲,9–11 etc. The primary interest in the cubic BCN is the combination of their superhigh hardness, excellent transparency to visible light and even to x rays, and other relevant superior properties, which renders these materials prospective protective coatings.12–14 It was reported that the cubic BC2 N has a hardness and an elastic modulus of values higher than those of cBN. 1 Montasser et al. have prepared BCN films transparent to photons in the wavelength range of 20–1000 nm using the plasma assisted CVD method.10 Now it has been gradually recognized that both the attaina兲

Author to whom correspondence should be addressed; electronic mail: [email protected]

2275

J. Vac. Sci. Technol. B 20„6…, NovÕDec 2002

ment of a true ternary compound and the unambiguous confirmation of a unique ternary phase for the BCN materials are a troublesome enterprise. These problems have their origin precisely in the structural and electronic similarity between the constituent units. The barely contrasted scattering capability, bonding strength, and mass among the constituent atoms and relevant bonded entities can incapacitate some conventional analyzing tools, which rely on the electronic characters and/or the mass, for resolving the local structural features. It is then very difficult for these tools to distinguish a true ternary BCN compound from a microcrystalline mixture of carbon and boron nitride or something alike. Segregation of diamond and cBN phases takes place routinely in specimens prepared at high temperature. It is hypothesized that, even for a BC2 N of strict stoichiometry, there exist a few competitive structures which may emerge in a specimen prepared with energetic ions.14,15 Fortunately, however, there is no practical impediment to the application of the cubic BCN materials since phase separation can be hindered by a large kinetic barrier, as pointed out by Lambrecht et al.4 Electron-cylcotron-wave-resonance 共ECWR兲 plasma assisted deposition is an effective low temperature method which has been successfully applied to the synthesis of superhard materials such as cBN and nanocrystalline SiCN.16,17 The flexibility in tuning the beam composition and the exact control of ion energy and current in a separate way allow a careful investigation of the film features’ dependence on the processing parameters. For the deposition of ternary materials, the equally energetic ions of all precursors are expected to enhance the homogeneous mixing, favoring the formation of a true ternary phase. In this work, we report the synthesis of nanocrystalline BCN thin films using the ECWR plasma assisted deposition. Analyzing tools such as atomic

1071-1023Õ2002Õ20„6…Õ2275Õ6Õ$19.00

©2002 American Vacuum Society

2275

2276

Cao, Liu, and Oechsner: Plasma assisted deposition of nanocrystalline BCN

2276

force microscopy 共AFM兲, Fourier-transform infrared spectroscopy 共FTIR兲, x-ray photoelectron spectroscopy 共XPS兲, and energy-dispersive x-ray spectroscopy 共EDX兲 were exploited for the characterization of the film morphology, bonding states, and bulk composition. For the investigation of the mechanical and optical properties of the deposits, nanoindentation was performed to determine the Vicker’s hardness and light absorption was measured in the wavelength range of 300– 800 nm. Also the photoluminescence 共PL兲 at room temperature was scrutinized. II. EXPERIMENT The films were grown on n-type Si 共001兲 wafers and on Corning glass 共7059兲 substrates using a custom-designed ECWR plasma source. In this experiment, the ECWR plasma of nitrogen was rf biased to sputter the high-purity hBN/graphite target and to bombard the growing surface simultaneously. The plasma of a nitrogen/argon mixture was applied to deposit a few samples for comparison. Depending on the plasma parameter and biasing power, the energy for the ions striking the growing surface was regulated between 60 and 180 eV. A detailed description of the experimental setup can be found in our early publications.18 The glass substrate can be characterized by three material parameters: a softening point at 844 °C, a thermal expansion 共0–300 °C兲 of 4.60⫻10⫺6 /°C, and a refractive index 共589.3 nm兲 of 1.5333. The base pressure in the chamber was better than 4.0 ⫻10⫺7 Pa, and the substrate temperature was maintained at about 450 °C during film growth ordinarily for half an hour. The typical film thickness is 1.0 ␮m as measured using a stylus method 共Dektak 3030兲. The deposits were first inspected with EDX analysis performed on an Oxford 6566 spectrometer to determine their bulk compositions. Both XPS 共PHI 5700兲 and FTIR 共Nicolet MX-I兲 absorption were employed to investigate the bonding states in the samples. The XPS measurement was performed using the Mg K ␣ line of photon energy h ␯ ⫽1253.6 eV and the spectra were registered with a pass energy of 10 eV for better resolution. Before XPS analysis, the samples were sputter cleaned for a short while using the 1 keV Ar⫹ ions. AFM height images were obtained in the contact mode on a Park instrument to reveal the surface morphology. Hardness of the deposits was read from the load versus displacement curve in a complete load/unload cycle obtained on a nanoindenter 共CSEM兲; the load was restricted below 3.0 mN so as to limit the maximum indent depth below 120 nm. The transparency of the deposits to the visible light was verified on a Spectrapro’ 500I spectrometer 共Roper Scientific兲, and the PL spectrum was measured at room temperature on a florescence system 共PTI 710兲 using the spectral lines from a Xe lamp. III. RESULTS AND DISCUSSION Some 40 samples in total were prepared on the silicon and Corning glass substrates. The deposits on silicon substrates appear dark blue or violet, while those on glass substrates are transparent with a light brown color. Improved adhesion was achieved even on the glass substrate due to the low energy J. Vac. Sci. Technol. B, Vol. 20, No. 6, NovÕDec 2002

FIG. 1. EDX spectrum for a BCN film grown on Si substrate at 141 eV.

plasma bombardment of the growing surface. No delamination could have been observed on thick films even after having been exposed to the ambient for half a year. A. Composition and surface morphology

A typical EDX spectrum is presented in Fig. 1, showing that the oxygen was the only detectable extrinsic atomic species. The concentration of oxygen from this spectrum amounts to merely 0.8 at. %. Therefore we can say that the deposits are very clean, taking into account the fact that the contaminants reside primarily on the sample surface as provoked during sample handling in ambient. A bulk composition of B42C33N25 was determined for the sample in Fig. 1, and it manifests a fluctuation within 2.0 at. % for each element across the applied energy window. The films possess essentially a homogeneous integration; to the resolving power of EDX operated at 10 keV no variation of composition across the sample was confirmed. Under AFM the film surfaces revealed densely packed, isometrically distributed crystallites in cubic habits. Though it may be a weak argument, this indicates the evolution of a cubic phase material in the deposits. Figures 2共a兲–2共c兲 display the images for samples grown with pure nitrogen plasma, and for Fig. 2共d兲 a plasma of nitrogen/argon mixture was applied. It can be seen that, when nitrogen plasma was applied, the crystallites developed well in the applied energy window. With the addition of argon, however, the effective sputtering and particularly the forward momentum transfer initiated by Ar ions could considerably modify the surface morphology. The crystallites were reduced in size and the boundaries became diffuse, therefore the surface was much smoother. Figure 3 displays the surface roughness versus applied ion energy. It shows a valley at about 112 eV where the changeover to the internal growth mode characteristic of this technique is supposed to take place. The behavior of this curve can be readily understood by the nature of sputter growth of crystalline materials. Initially the surface becomes rougher with increasing ion energy when crystallites emerge,

2277


2277

FIG. 2. AFM height images in an area of 4⫻4 ␮ m obtained on the specimens prepared at 共a兲 177 eV, 共b兲 141 eV, 共c兲 112 eV, and 共d兲 140 eV. The sample for 共d兲 was grown with a plasma of nitrogen/argon mixture for comparison. The averaged crystallite size, from 共a兲 to 共d兲, is about 200, 210, 140, and 65 nm, respectively.

the changeover into the internal growth mode at about 100 eV has then effectively reduced the surface roughness. However, the crystallites continually grow larger at higher ion energy, consequently the surface becomes rougher once again until at about 170 eV when the heavy sputtering eventually overobliterates the crystallite evolution.

FIG. 3. Root-mean-square roughness vs ion energy for the film surfaces obtained with pure nitrogen plasma. The solid line is drawn to guide the eye. The two half-filled circles represent the values for samples deposited when Ar was added at a flow ratio Ar/N2 ⫽1:4. JVST B - Microelectronics and Nanometer Structures

B. Bonding states

Bonding states are an important feature for the confirmation of true ternary BCN compounds. Although a definitive confirmation of the ternary phases can be achieved with techniques such as x-ray absorption near edge spectroscopy 共XANES兲 extracted from the K 1s near edge fine structure sensitive to the local order arrangement,19 however, such methods are not always readily accessible. The conventional methods such as XPS and IR spectroscopy can still provide helpful hints to the structure when interpreted with care. On the XPS spectra for ‘‘dirty’’ samples, a strong O 1s line always appears at a binding energy of 533 eV, which can be indubitably ascribed to the B–O bond in B2 O3 . As in BN deposits, here the B atoms at the surface have captured O2 and H2 O molecules in ambient. The binding energy for the 1s lines of B, C, and N, as shown in Fig. 4, is 190.2, 285.8, and 399.4 eV, and the corresponding full width at half maximum 共FWHM兲 is 1.84, 1.56, and 2.33 eV, respectively. Even though the binding energies cannot unequivocally point to a ternary BCN bonding configuration, the narrow, nearly Gaussian profiles indicate that the material is most likely chemically homogenous. For comparison, the spectral lines from Fig. 3 in Ref. 11, where the BCN sample was prepared in a plasma-assisted pulsed-laser deposition system, are much broader; the FWHM for them is 2.3, 2.8, and 2.4 eV, respectively. The comparable width for the N 1s line excludes a significant difference in the apparatus’ resolving

2278


2278

FIG. 5. FTIR spectrum for a BCN film deposited at 177 eV.

FIG. 4. XPS narrow scan profiles of the 共a兲 B 1s line, 共b兲 C 1s line, and 共c兲 N 1s line.

power. Thus at the current stage we can say that phase separation in our specimens, if there is any, is not serious. In Fig. 5 is shown a typical FTIR spectrum for the BCN deposits. Four distinct absorption peaks and a wide band were identified at 615, 798, 1107, 1404, and 3500 cm⫺1, respectively. This spectrum bears resemblance to a superposition of the spectra for diamond and sputter deposited BN films when both hBN and cBN phases are available. The band at 1404 cm⫺1 corresponds to the stretching mode of the B–N in-plane bond of hBN 共around 1383 cm⫺1 in hBN) and the one at 798 cm⫺1 refers to the bending mode of the B–N–B out-of-plane bond 共around 770 cm⫺1 in hBN). The band at 1107 cm⫺1 arises from the B–N bond of sp 3 type 共around 1065 cm⫺1 in cBN). Nevertheless, this band does J. Vac. Sci. Technol. B, Vol. 20, No. 6, NovÕDec 2002

not imply the presence of cBN aggregates in deposits, since a difference as large as 42 cm⫺1 cannot be accounted for by internal stress induced upward shift;20 besides, the internal stress in BCN is surely less than in cBN prepared under comparable conditions. In fact, no noticeable shift was observed for this peak in the BCN deposits across the applied energy window. A most probable origin for this absorption is the TO mode in a cBN lattice with its sites partially occupied by carbon atoms. The peak of feeble intensity at around 615 cm⫺1 was detected in B4 C. 15 The appearance of this peak is reasonable since the deposits have an N-deficit stoichiometry. We have also performed x-ray diffraction 共XRD兲 and transmission electron microscopic 共TEM兲 analysis in order to achieve a direct verification of the crystal structure. The XRD was carried out using the Cu K ␣ line on a Rikagu 2400D/X diffractometer and the TEM analysis on a Philips CM200 FEG microscope. The nanocrystalline nature of BCN materials disfavors the determination of their crystal structure by the x-ray scattering method. Due to the very low scattering capability of BCN atoms for hard x rays, the XRD spectrum, not shown here, revealed only a weak broad peak at about 2 ␪ ⬇43° where the 共111兲 reflection for cubic BCN is believed to appear. The TEM analysis was impeded by sample preparation using ion milling, whereby the sample lattice structure was decisively destroyed by the energetic ion impact. C. Mechanical and optical properties

The microhardness for the films was determined, following Oliver and Pharr,21 from the load versus displacement curve shown in Fig. 6. Assuming a Poisson ratio of 0.096 for our deposits, the Vicker’s hardness was then calculated to be about 28 GPa and the Young’s modulus was measured to be about 240 GPa. This hardness is much lower than the values on c-BC2 N from direct conversion, but agrees well with the results obtained on samples prepared with magnetron sputtering.13 To our surprise, though the crystallite size spans

2279


FIG. 6. Load vs displacement curve for the measurement of microhardness. Loading rate: 5 mN/min.

a large range from 65 to 210 nm, the hardness is not sensitive to the crystallite size; it fluctuates within 26 –28 GPa. This is contradictory to our observation on the SiCN system prepared under comparable conditions where a remarkable reduction in the hardness was recognized on samples with large crystallites. The nanocrystalline cubic BCN is anticipated to be transparent to visible light. Figure 7 represents the transmittance for a 1.0 ␮m thick sample in the wavelength range of 300– 800 nm. It is seen that the averaged transmittance within 500– 800 nm amounts to over 85%. Taking into account the fact that the light scattering was mainly caused by the defected microstructures, a much larger transmittance is anticipated for the crystalline cubic BCN samples. Light emission of the wide gap materials may significantly extend the range of prospective applications. The PL spectra for BCN deposits were stimulated using the 380 nm 共3.26 eV兲 irradiation corresponding to the resonant absorp-

2279

FIG. 8. Excitation spectrum and the photoluminescence spectrum stimulated with photons of ␭⫽380 nm for a BCN prepared at 141 eV. The photoluminescence spectra can be decomposed into three Gaussian profiles centered at 410, 435, and 451 nm, respectively.

tion, see Fig. 8. A strong PL emission peaked at 430 nm 共2.88 eV兲 was detected at room temperature. This is clearly a band-edge emission. By considering the peak positions of the light absorption and PL spectra, the band gap for this material was estimated to be about 3.0 eV. Such a wide band gap suggests a cubic phase for the deposits, since the hexagonal BCN structures are unlikely to have a band gap over 1.6 eV.3 The FWHM of nearly 0.4 eV indicates that there may be multiple channels for the spectral feature of PL on BCN. Briefly, the PL spectra for samples grown at different ion energies displayed the same profile, but those grown at energy less than 120 eV showed a considerably reduced intensity. We hypothesize that the internal growth mode results in an improved film quality leading to a considerable reduction of nonradiative recombination for the electron–hole pairs; therefore the sample grown at higher energies shows much more intense photoluminescence. Moreover, due to the large surface-to-volume ratio, defect sites at the surface can contribute a great part to the nonradiative recombination; consequently the films that grow at higher energies 共thus composed of larger crystallites, cf. Fig. 2兲 manifest a more efficient PL emission. IV. SUMMARY

FIG. 7. Light transmittance in the wavelength range of 300– 800 nm. The deceptive abrupt reduction in the transmittance from 500 nm on is due to the strong absorption by glass substrate. JVST B - Microelectronics and Nanometer Structures

In summary, nanocrystalline BCN thin films were grown using the ECWR plasma assisted deposition method. In the applied energy window, the films prepared with nitrogen plasma showed a smooth surface composed of crystallites of about 200 nm in size, whereas much smaller crystallites came out when Ar was introduced into the working gas. The surface demonstrates a polyhedral morphology with a rootmean-square roughness less than 3 nm. With supporting data from AFM, FTIR, XRD, and also the optical measurements, the deposits are conjectured to be of cubic phase. The deposits are hard and transparent, hence they are competitive ma-

2280


terials in the application as a protective coating for optical components. Moreover, strong PL peaked at 430 nm was detected on the as-deposited materials at room temperature. This implies that the cubic BCN compounds may also find applications in the fabrication of violet light-emitting diodes and laser diodes. At present, there has been uncertainty regarding the nature and the degree of order in the samples. To explore the possibility of the formation of a true cubic BCN phase, it is of great importance to obtain a sharp TEM lattice image or diffraction pattern, which will be essential for a deeper discussion on the microstructure of the deposits. A wellcrystallized sample with crystallites of a few microns in dimension will favor the analysis using XRD and XANES. To this end, we are now undertaking the film 共approaching a 1:1 BN stoichiometry兲 growth at higher substrate temperatures and looking for the possibility of preparing the TEM specimens with a nondestructive technique. ACKNOWLEDGMENT This work was supported by the Natural Science Foundation of China 共NSFC兲 Grant No. 19974065, by the Volkswagen Foundation, Germany, Grant No. I74701, and by the Chinese Academy of Sciences. 1

C. Popov, K. Saito, B. Ivano, Y. Koga, Sh. Fujiwara, and V. Shanov, Thin Solid Films 312, 99 共1998兲. 2 V. L. Solozhenko, S. N. Dub, and N. V. Novikov, Diamond Relat. Mater. 10, 2228 共2001兲.

J. Vac. Sci. Technol. B, Vol. 20, No. 6, NovÕDec 2002

3

2280

A. Y. Liu, R. M. Wenrcovitch, and M. L. Cohen, Phys. Rev. B 39, 1760 共1989兲. 4 W. R. L. Lambrecht and B. Segall, Phys. Rev. B 47, 9298 共1993兲. 5 S. Nakano, M. Akaishi, T. Sasaki, and Sh. Yamaoka, Mater. Sci. Eng., A 209, 26 共1996兲. 6 M. K. Lei, Q. Li, Z. F. Zhou, I. Bello, C. S. Lee, and S. T. Lee, Thin Solid Films 389, 194 共2001兲. 7 E. Martinez, A. Lousa, and J. Esteve, Diamond Relat. Mater. 10, 1892 共2001兲. 8 R. Gago, I. Jimenez, T. Sajavaara, and J. M. Albella, Diamond Relat. Mater. 10, 1165 共2001兲. 9 M. Mieno and T. Satoh, J. Mater. Sci. 36, 3925 共2001兲. 10 K. Montasser, S. Hattori, and S. Morita, Thin Solid Films 117, 311 共1984兲. 11 Y. K. Yap, Y. Wada, M. Yamaoka, M. Yoshimura, Y. Mori, and T. Sasaki, Diamond Relat. Mater. 10, 1137 共2001兲. 12 S. Ulrich, T. Thel, J. Schwan, and H. Ehrhardt, Surf. Coat. Technol. 97, 45 共1997兲. 13 Y. Tateyama, T. Ogitsu, K. Kusakabe, and S. Tsuneyuki, Phys. Rev. B 55, R10161 共1997兲. 14 M. Nesla´dek, M. Vanı`e`ek, K. Meykens, K. Hanen, J. Manca, L. De Schepper, E. Pace, A. Pini, G. Verona, C. Kimura, Y. Etou, and T. Sugino, Phys. Status Solidi A 185, 107 共2001兲. 15 F. L. Freire, D. C. Reigada, and R. Prioli, Phys. Status Solidi A 187, 1 共2001兲. 16 Z. X. Cao, Diamond Relat. Mater. 9, 1881 共2000兲. 17 Z. X. Cao, J. Vac. Sci. Technol. A 19, 485 共2001兲. 18 H. Oechsner, in Plasma Processing of Semiconductors 共Kluwer Academic, Dordrecht, Netherlands, 1997兲, p. 529. 19 R. Gago, I. Jimenez, and J. M. Albella, Appl. Phys. Lett. 78, 3430 共2001兲. 20 D. L. Medlin, T. A. Friedmann, P. B. Mirkarimi, P. Rez, M. J. Mills, and K. F. McCarty, J. Appl. Phys. 76, 295 共1994兲. 21 W. C. Oliver and G. M. Pharr, J. Mater. Res. 7, 1564 共1992兲.