Applied Surface Science 226 (2004) 52 56 Structural and optical properties of Cu:silica nanocomposite films prepared by co-sputtering deposition G. Battaglin a, E. Cattaruzza a, F. Gonella a,*, R. Polloni a, B.F. Scremin a, G. Mattei b, P. Mazzoldi b, C. Sada b a INFM and Dipt. di Chimica Fisica, Università di Venezia, Dorsoduro 2137, 30123 Venezia, Italy b INFM and Dip. di Fisica, Università di Padova, via Marzolo 8, 35131 Padova, Italy Abstract Copper-containing silica films were synthesized by radiofrequency (rf) co-sputtering deposition technique, and then heattreated in different annealing atmospheres, i.e. either oxidizing or reducing, with the aim to develop suitable preparation methodologies for controlling the composite structure. Characterization of the samples along the various preparation steps was performed by Rutherford backscattering spectrometry (RBS), transmission electron microscopy and optical absorption spectroscopy. The nonlinear optical coefficient n 2 of the nanocomposite films was estimated by the Z-scan technique. Experimental observations showed that copper migration and aggregation depend critically on the annealing conditions, giving rise to quite different stable structures. In particular, for samples heat-treated first in air and then in a H 2 Ar gas mixture, the oxidizing atmosphere drives copper towards the surface while the reducing one promotes the subsequent clusterization in a well defined region. # 2003 Elsevier B.V. All rights reserved. PACS: 61.46.þw; 81.05.Kf Keywords: Nanoclusters; Glass composites 1. Introduction Metal nanoclusters embedded in fused silica exhibit peculiar optical properties that have made them attractive in several application fields. In particular, metal nanocluster composite glasses (MNCGs) are expected to exhibit features that can be exploited in integratedoptical devices [1], owing to their transparency in the optical fibers transmission window. Prescribed Cu:silica composite features require the control of the * Corresponding author. Tel.: þ39-041-234-8595; fax: þ39-041-234-8594. E-mail address: gonella@unive.it (F. Gonella). cluster formation and growth, and therefore the definition of effective preparation protocols. Sputtering deposition technique is particularly suitable for codepositing the matrix together with more than one dopant element, also possibly creating core-shell or alloy aggregates [2 6]. Moreover, by sputtering deposition one can prepare homogeneous films several microns thick, so operating in the field of light waveguiding structures. In this work, copper-containing silica films were synthesised by radiofrequency (rf) co-sputtering deposition technique, and then heattreated (in some cases sequentially) in different annealing atmospheres, i.e. either oxidizing or reducing. 0169-4332/$ see front matter # 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2003.11.030
G. Battaglin et al. / Applied Surface Science 226 (2004) 52 56 53 2. Experimental Samples were synthesized at room temperature by radiofrequency magnetron multitarget co-sputtering deposition of silica and copper, in pure Ar atmosphere at a working pressure of 35 10 4 mbar, using two different radiofrequency sources for silica and copper, respectively. Substrate were fused silica slides. The rfpower to the 2 00 diameter targets was 250 W for silica and 12 W for copper. Co-deposition time was 40 min, preceeded and followed by a 5 min deposition of sole silica. The final composite film thickness was 0.5 mm. After the deposition, different thermal treatments were performed on the sample, namely: 2 h at 900 8CinH 2 Ar (10 90%) reducing atmosphere (samples labeled as H ); 5 h at 700 8C in air (samples labeled as A ); sequential annealing first in oxidizing then in reducing atmosphere at the above respective conditions (samples labeled as AH ). Characterization of the samples at the various preparation steps was performed by several techniques. Rutherford backscattering spectrometry (RBS) measurements were performed at the INFM-INFN Laboratories in Legnaro (Padova, Italy) using a 2.2 MeV 4 He þ beam. The incident direction was normal to the sample surface, and scattered particles were detected at the angle of 1608. Transmission electron microscopy (TEM) was performed using a Philips Tecnai F20 microscope operated at 200 kv, equipped to provide also X-ray energy dispersion spectroscopy (EDS) and selected-area electron diffraction (SAED). Measurements were done at the CNR-IMM centre in Bologna, Italy. Optical absorption spectroscopy was performed by a double-beam UV-Vis-NIR Cary 5 spectrophotometer. The determination of the nonlinear optical coefficient n 2 of the nanocomposite films, related to the electronic thirdorder optical nonlinearity, was realized by means of the Z-scan method [7], using a doubled-frequency Nd:glass source (l ¼ 527 nm) at 1 Hz of repetition rate, with pulse duration t ¼ 6 ps and light intensity up to about 1 GW/cm 2. 3. Results and discussion Optical absorption spectroscopy is the first indication of the possible formation of copper nanoclusters. Optical density 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 AH A H as deposited 0 300 400 500 600 700 800 Wavelength (nm) Surface plasmon resonance (copper nanoparticles) λ=560 nm Fig. 1. Optical absorption spectra for the as-deposited sample and for the samples annealed in air (A), H 2 Ar (H), and sequentially in both atmospheres (AH). In Fig. 1, the optical spectra are shown for the various heat-treated samples, compared with the spectrum for the as-deposited sample. The peak at about 560 nm is related to the surface plasmon resonance (SPR) of copper metallic nanoclusters, and is evident in the two cases of H 2 -annealing. Fig. 2 shows the RBS spectra for the same samples. From the observation of these spectra, it is evident how the uniform distribution of copper inside the as-deposited sample is strongly Normalized Yield 12 10 8 6 4 2 Energy (MeV) 1.4 1.6 1.8 as deposited A AH H 0 300 350 400 450 Channel Fig. 2. RBS spectra for the as-deposited sample and for the samples annealed in air (A), H 2 Ar (H), and sequentially in both atmospheres (AH).
54 G. Battaglin et al. / Applied Surface Science 226 (2004) 52 56 modified by the annealing processes: annealing in air at 700 8C causes the diffusion of copper towards the surface, where it accumulates in a layer about 40 nm thick. This is confirmed by TEM analysis (not reported here), that points out the formation of relatively big clusters in the near-surface region. On the other hand, the annealing at 900 8C in a reducing environment forces copper atoms to move into the sample, giving rise to a redistribution which is evident in both the sample labeled as H and AH. The oxygen-tosilicon ratio in the as-deposited sample has a value of about 2.3. For the sample annealed in air, the oxygen excess is roughly equal to the copper content, with nearly the same depth distribution. In the samples annealed in the reducing atmosphere, a ratio O=Si ¼ 2 was determined, except for the Cu-rich surface layer of the previously air-annealed specimen, where we still detect an oxygen excess corresponding to about one-half of the copper content in the same region. We then expect that copper is present in its reduced form in hydrogen-annealed samples, except for the surface region of the already air-annealed sample, where both oxidized and reduced species should coexist. A confirmation of this picture is given by TEM analysis and electron diffraction compositional analysis: they showed that in the air-annealed samples oxide clusters are present only at the surface, while in the hydrogenannealed samples both copper oxide forms (CuO and Cu 2 O) are present, being oxidation of copper less effective with the depth, with clusters of average size of 5:9 2:3 nm. It is in general evident that several factors come into play in determining the cluster formation process. In particular, chemical driving forces due to the different annealing atmospheres determine the behavior of copper in the silica matrix in terms of migration and aggregation and growth processes. Predictions on the formation of either metallic or oxide clusters could be made on the basis of the respective Gibbs free energy values for the compounds formation, but a comprehensive picture accounting for the process is anyhow still lacking [8,9]. Z-scan technique has demonstrated to be particularly suitable for the determination of the optical nonlinear behavior of thin MNCG films [10,11]. The intensity-dependent refractive index is usually defined by n ¼ n 0 þ n 2 I, where n 0 and n 2 (proportional to the real part of the third-order optical susceptibility, w (3) ) are the linear and nonlinear refractive indices, respectively, and I is the intensity of the incident radiation. For ultrafast all-optical switching applications, third-order nonlinear materials must satisfy some figures of merit (FOMs). The most quoted FOMs are related to important features that a material must accomplish for application in ultrafast all-optical switching devices. For example, the possibility to process light signals without needing to convert them to electronic form should allow alloptical devices to operate in a frequency range inaccessible to electronics. However, high switching Normalized intensity (a) Normalized intensity (b) 1.02 1.01 1 0.99 0.98 1.02 1.01 1 0.99 0.98 sample H Error bar -20-15 -10-5 0 5 10 15 20 Z (mm) sample AH Error bar -20-15 -10-5 0 5 10 15 20 Z (mm) Fig. 3. Z-scan figures collected for the H and AH samples, together with the corresponding theoretical curves used to determine the numerical results. Experimental parameters are l ¼ 527 nm, pulse duration t ¼ 6 ps, repetition rate 1 Hz, light intensity about 1 GW/cm 2.
G. Battaglin et al. / Applied Surface Science 226 (2004) 52 56 55 Table 1 Results from the analysis of the Z-scan data for the samples annealed in reducing atmosphere, together with the estimation of the figures of merit Sample n 2 (cm 2 /W) a 0 (mm 1 ) b (cm/w) w ð3þ (esu) W T H (5.0 0.8) 10 11 2.0 0.1 (1.7 0.2) 10 10 (4.7 1.0) 10 9 0.05 0 AH (3.5 0.8) 10 11 2.0 0.1 ( 1.0 0.3) 10 10 (3.2 1.5) 10 9 0.07 0 speeds must be coupled with low power switching threshold, to get high packing densities in optoelectronic or photonic devices. The nonlinear material must exhibit a large n 2 value to operate at watt peak power in centimetre long devices. At the same time, it must have a low optical absorption (a nonlinear phase shift of 2p should be possible over one attenuation distance for reasonable device throughput), a high intensity threshold for laser-induced damage, a low thermo-optic contribution to n 2, a very short response þ recovery time (i.e. a high recycling frequency). Moreover, thermal stability and wavelength tunability are important peculiarities for promising nonlinear materials. Depending on the main absorption mechanism of a given material, i.e. linear or nonlinear, two figures of merit have to be satisfied for a 2p phase shift: j W ¼ n 2jI max > 1 (1) a 0 l T ¼ bl jn 2 j < 1 (2) where n 2 is the fast (electronic) nonlinear refractive index, a 0 and b are the one- and the two-photon absorption coefficients, respectively (for an optical Kerr medium, the absorption coefficient a can be written as a ¼ a 0 þ bi, where I is the light intensity), l the light wavelength, and I max is the maximum permitted value of the light intensity. Significantly high nonlinearity was measured for the hydrogen-annealed samples: Fig. 3a and b shows the Z-scan figures collected for the H and AH samples, respectively, together with the corresponding theoretical curves used to determine the numerical results. A well defined valley-peak behavior is evident for both samples, related to a positive nonlinear optical response of the copper nanoclusters. In Table 1 are summarized the results from the analysis of the Z-scan experiments, together with the estimation of the figures of merit. In particular, the coefficients were determined by analyzing the Z-scan figures in both farand near-field configurations. Most common units are given for the w ð3þ values to make easy the comparison with the various data reported in the literature. A different sign for the b coefficient can be noticed for the two systems, indicating a saturated nonlinear absorption for the AH sample. This is ascribed to a different cluster size distribution, due to the different aggregation process. The optical properties of MNCGs are still insufficient at least for the W figure for the application of these materials in all-optical switching technology, but the challenge for an improvement of MNCGs features could be faced in two ways. The first one is to explore very large values of the cluster volume fraction, p.itis well-known that at low metal concentration the w (3) of a MNCG is linearly proportional to p [12]. In Au:SiO 2 composite films, working at the SPR, Liao et al. [13] recently foundthat the linear relation is valid up to p ¼ 0:15, while beyond this value and below the percolation limit (around 0.5), w (3) becomes roughly proportional to the third power of p. Moreover, the ratio the w ð3þ =a 0 also follows a power law, with an exponent of about 2.7. Also reducing particles size down to about 1 nm would increase the nonlinear response, since a 0 drops abruptly for this size range [1]. The second way to improve the MNCG optical behavior could be to explore glass matrices different from silica. At SPR, the ratio w ð3þ =a 0 is proportional to the fifth power of the linear refractive index, n 0 [14]. Therefore, host glass matrices as zirconia (n 0 ¼ 2:0) or titania (n 0 ¼ 2:7) should give more favourable W figure values than silica (n 0 ¼ 1:45). 4. Conclusions The radiofrequency magnetron co-sputtering deposition, combined with suitable subsequent ther-
56 G. Battaglin et al. / Applied Surface Science 226 (2004) 52 56 mal treatments, provides effective MNCG preparation protocols. The hydrogen-annealed films formed by copper nanoparticles embedded in silica exhibit very large values of the nonlinear refractive index, n 2, thus making them interesting as composite materials for photonic applications. The optical properties of the films annealed in reducing atmosphere are promising compared to those of similar copper MNCGs described in literature, but yet insufficient for an immediate application in alloptical switching technology. The improvement of the figures of merit could be achieved by further modifying other parameters that are known to affect the optical nonlinear performances, that is, by further reducing the nanoclusters size, that should be not larger than 1 nm in size, by exploring relatively high values of the cluster-volume ratio and/or by using other suitable matrices. Experimental observations have shown the complexity of the copper behaviour during the composite formation: copper migration and aggregation depend critically on the annealing conditions, in particular on the atmosphere, and quite different stable structures actually result. In the samples heat-treated first in air and then in a H 2 Ar gas mixture, the oxidizing atmosphere drives copper towards the surface while the reducing one promotes the subsequent clusterization in a well defined region. Generally speaking, the combined preparation methodology has demonstrated to provide further degrees of freedom in the control of the final composite structure. References [1] F. Gonella, P. Mazzoldi, Metal nanocluster composite glasses, in: H.S. Nalwa (Ed.), Handbook of Nanostructured Materials and Nanotechnology, vol. 4, Academic Press, San Diego, 2000, pp. 81 158. [2] S. Padovani, F. D Acapito, E. Cattaruzza, A. De Lorenzi, F. Gonella, G. Mattei, C. Maurizio, P. Mazzoldi, M. Montagna, S. Ronchin, C. Tosello, M. Ferrari, Eur. Phys. J. B 25 (2002) 11. [3] I. Tanahashi, M. Yoshida, Y. Manabe, T. Tohda, J. Mater. Res. 10 (1995) 362. [4] T. Akai, H. Yamanaka, H. Wakabayashi, J. Am. Ceram. Soc. 79 (1996) 859. [5] I. Tanahashi, Y. Manabe, T. Tohda, S. Sasaki, A. Nakamura, J. Appl. Phys. 79 (1996) 1244. [6] C. Tosello, S. Ronchin, E. Moser, M. Montagna, P. Mazzoldi, F. Gonella, M. Ferrari, C. Duverger, R. Belli, G. Battaglin, Phil. Mag. B 79 (1999) 2103. [7] J. Wang, M. Sheik-Bahae, A.A. Said, D.J. Hagen, E.W. Van Stryland, J. Opt. Soc. Am. B 11 (1994) 1009. [8] G. Battaglin, M. Catalano, E. Cattaruzza, F. D Acapito, C. de Julián Fernández, G. De Marchi, F. Gonella, G. Mattei, C. Maurizio, P. Mazzoldi, A. Miotello, C. Sada, Nucl. Instrum. Meth. B 178 (2001) 176. [9] E. Cattaruzza, Nucl. Instrum. Meth. B 169 (2000) 141. [10] G. Battaglin, P. Calvelli, E. Cattaruzza, F. Gonella, R. Polloni, G. Mattei, P. Mazzoldi, Appl. Phys. Lett. 78 (2001) 3953. [11] E. Cattaruzza, G. Battaglin, F. Gonella, R. Polloni, G. Mattei, C. Maurizio, P. Mazzoldi, C. Sada, M. Montagna, C. Tosello, M. Ferrari, Phil. Mag. B 82 (2002) 735. [12] D. Ricard, P. Roussignol, C. Flytzanis, Opt. Lett. 10 (1985) 511. [13] H.B. Liao, R.F. Xiao, J.S. Fu, P. Yu, G.K.L. Wong, P. Sheng, Appl. Phys. Lett. 70 (1997) 1. [14] O. Maruyama, Y. Senda, S. Omi, J. Non-Cryst. Solids 259 (1999) 100.