Si nanocrystal synthesis in HfO2/SiO/HfO2 multilayer ...

Home

Search

Collections

Journals

About

Contact us

My IOPscience

Si nanocrystal synthesis in HfO2/SiO/HfO2 multilayer structures

This content has been downloaded from IOPscience. Please scroll down to see the full text. 2010 Nanotechnology 21 055606 (http://iopscience.iop.org/0957-4484/21/5/055606) View the table of contents for this issue, or go to the journal homepage for more

Download details: IP Address: 193.205.155.112 This content was downloaded on 26/09/2014 at 09:39

Please note that terms and conditions apply.

IOP PUBLISHING

NANOTECHNOLOGY

Nanotechnology 21 (2010) 055606 (7pp)

doi:10.1088/0957-4484/21/5/055606

Si nanocrystal synthesis in HfO2/SiO/HfO2 multilayer structures M Perego1,4 , G Seguini1 , C Wiemer1 , M Fanciulli1,2 , P-E Coulon3 and C Bonafos3 1 Laboratorio Nazionale MDM CNR-INFM, Via C. Olivetti 2, I-20041 Agrate Brianza (MI), Italy 2 Dipartimento di Scienza dei Materiali, Universit`a degli Studi di Milano-Bicocca, I-20126 Milano, Italy 3 nMat Group, CEMES-CNRS, rue J. Marvig 29, Toulouse FR-31055, France

E-mail: [email protected]

Received 14 October 2009, in final form 1 December 2009 Published 24 December 2009 Online at stacks.iop.org/Nano/21/055606 Abstract The synthesis of two-dimensional arrays of Si nanocrystals in an HfO2 matrix has been achieved by deposition of HfO2 /SiO/HfO2 multilayer structures followed by high temperature (1100 ◦ C) thermal treatment in nitrogen atmosphere. Silicon out-diffusion from the SiO layer through the HfO2 films has been shown to be the limiting factor in the formation of the Si nanocrystals. Suitable strategies have been identified in order to overcome this limitation. Si nanocrystal formation has been achieved by properly adjusting the thickness of the SiO layer. (Some figures in this article are in colour only in the electronic version)

challenging than in SiO2 [11]. The supersaturation conditions that are necessary to induce nanocrystal nucleation and growth are difficult to achieve in high-k oxides since silicon diffusivity in these materials is usually much higher than in SiO2 [12]. Moreover, moisture absorption as well as oxygen diffusion in the high-k dielectrics represent potential sources of oxidizing agents that could induce a strong reduction of the Si supersaturation in the SiO layer, limiting in this way the formation of Si nanocrystals in the HfO2 matrix [11]. In this work we investigate the possibility to form Si nanocrystals in an HfO2 matrix by deposition of HfO2 /SiO/HfO2 multilayer structures and subsequent high temperature thermal treatment. The main material science issues related to the synthesis of silicon nanocrystals in an HfO2 matrix are faced. We demonstrate that Si out-diffusion is the limiting factor that prevents Si nanocrystal formation in the HfO2 matrix. The goal of Si nanocrystal synthesis has been achieved by introducing proper compensation mechanisms to sustain Si supersaturation.

1. Introduction High dielectric constant materials (high-k ) have been suggested as a suitable solution to overcome the fundamental trade-off between programming speed and data retention characteristics in Si nanocrystal-based memory devices [1]. In principle a dielectric with a higher k than SiO2 allows us to use a thicker tunnel oxide, reducing leakage currents and increasing charge retention. At the same time, due to the smaller conduction band off-set between the high-k material and the silicon substrate, the goal of a low voltage high speed non-volatile memory device can be achieved [2–4]. 2D layers of Si nanocrystals embedded in SiO2 have been synthesized by using different techniques, such as ion beam synthesis [5, 6], chemical vapor deposition [7], molecular beam epitaxy [8], SiOx /SiO2 multilayer deposition followed by high temperature thermal treatment [9, 10]. The last approach is based on thermal decomposition of SiOx into silicon and silicon dioxide [9]. This methodology is extremely appealing since an independent control of the dimension and density of the nanocrystals can be achieved by properly tuning the thickness and the stoichiometry of the SiOx film [9]. The synthesis of Si nanocrystals in high-k materials is more

2. Experimental details SiO layers of different thickness (ranging between 4 and 10 nm) have been deposited and incorporated in a multilayer

4 Author to whom any correspondence should be addressed.

0957-4484/10/055606+07$30.00

1

© 2010 IOP Publishing Ltd Printed in the UK

Nanotechnology 21 (2010) 055606

M Perego et al

Table 1. List of the samples and description of their structure in terms of material composition and thickness of the SiO layer. The thickness of the tunnel oxide and control oxide is fixed for all the samples (4 nm) as well as the thickness of the SiO2 capping layer (10 nm).

Label

Tunnel oxide

SiO thickness (nm)

Control oxide

Capping layer

M13 M14 M15 M16 M17 M18

HfO2 HfO2 HfO2 HfO2 SiO2 HfO2

4 7 10 4 4 4

HfO2 HfO2 HfO2 — — HfO2

SiO2 SiO2 SiO2 SiO2 SiO2 —

structure with a 4 nm thick HfO2 film as a tunnel oxide on top of an Si(100) substrate. Before evaporation, substrates have been cleaned using conventional Radio Corporation of America (RCA) processing followed by a 30 s long dip in a diluted HF solution at room temperature. Each cleaning step has been followed by a 30 s long rinse in deionized water. The samples have been capped in situ with a 10 nm thick SiO2 film in order to prevent moisture absorption during air exposure and oxygen penetration during the thermal treatment. Reference samples with a 4 nm thick SiO layer have been prepared by replacing the HfO2 tunnel oxide and/or control oxide with SiO2 layers of the same thickness. A list of samples with the indication of the tunnel and control oxide and of the thickness of the SiO layer is summarized in table 1. HfO2 , SiO and SiO2 depositions have been performed by means of an electron beam evaporation system equipped with a four-crucible e-beam gun (5 kV power supply) working in the high vacuum regime (base pressure 3 × 10−7 mbar). The samples have been annealed in a conventional tubular furnace at different temperatures (600– 1100 ◦ C) for 60 min in N2 flux at atmospheric pressure. Several characterization techniques have been carried out in order to check the nanocrystal formation. Photoluminescence (PL) spectra have been acquired at room temperature by exciting the samples with a 10 mW He–Ne laser at 632.8 nm. Luminescence has been detected by means of a Renishaw 2000 spectrometer equipped with a thermoelectrically cooled charge-coupled device. Time-of-flight secondary ion mass spectrometry (ToF-SIMS) measurements have been performed in a dual-beam IONTOF IV system using Cs+ ions at 1 keV for sputtering and Ga+ ions at 25 keV for analysis. X-ray reflectivity (XRR) and x-ray diffraction (XRD) measurements were acquired using a conventional laboratory Cu Kα x-ray source equipped with a parabolic multilayer monochromator and a position-sensitive detector. Specimens transparent to electrons are prepared for cross-sectional (XTEM) observations using the standard procedure involving mechanical polishing and Ar+ ion milling. High resolution transmission electron microscopy (HREM) and energy-filtered TEM (EFTEM) are performed on a field emission TEM, FEI Tecnai™ F20 microscope operating at 200 kV, equipped with a corrector for spherical aberration and the Gatan Imaging Filter (GIF) TRIDIEM. For EFTEM, an image is formed with the electrons that are selected by a slit placed in the energy dispersive plane

Figure 1. (a) Photoluminescence spectra of SiO films of different thicknesses embedded in HfO2 after the 1100 ◦ C thermal treatment. The photoluminescence spectrum of the as-deposited HfO2 /4 nm thick SiO/HfO2 sample (M13) is reported as a reference. (b) Photoluminescence spectra of the SiO2 /4 nm thick SiO/SiO2 sample (M17) before and after thermal treatment.

of the spectrometer, centered at 17 eV with a width of ±2 eV for the detection of the plasmon signal associated with the Si nanoparticles.

3. Results and discussion In figure 1(a) we report the PL spectra of samples with SiO films of different thickness embedded in HfO2 after the 1100 ◦ C thermal treatment. The PL spectrum of a not-annealed sample is reported as a reference. The PL spectrum of the annealed sample with a 4 nm thick SiO film (M13) is equivalent to the spectrum of the as-deposited one, indicating that no phase separation and, consequently, no nanocrystal formation occurs in this sample after thermal treatment. In contrast, a clear PL signal can be observed, after thermal treatment, for the samples with 7 nm and 10 nm thick SiO layers. Figure 1(b) shows the PL spectra of a 4 nm thick SiO layer embedded in an SiO2 matrix (M17) before and after the 1100 ◦ C thermal treatment in nitrogen atmosphere. In this case, as previously observed for the samples with 7 and 10 nm thick SiO films embedded in HfO2 , the PL peak after annealing exhibits a maximum in the range between 1.5 and 1.7 eV, which is the position corresponding to the emission from Si nanocrystals [9, 13]. The presence of this PL emission in these samples after annealing is undoubtedly related to the formation of silicon nanoparticles within the dielectric matrix. Transmission electron microscopy observations confirm the PL results. Indeed, while no Si nanocrystals have been detected in the HfO2 /4 nm thick SiO/HfO2 stack, clear evidence of the presence of Si nanocrystals has been obtained in the HfO2 /7 nm thick SiO/HfO2 and HfO2 /10 nm thick SiO/HfO2 multilayers (respectively M13, M14 and M15 samples). In figure 2(a) we report an EFTEM image obtained by filtering around the Si plasmon (17 eV) of the 1100 ◦ C 2


M Perego et al

Figure 2. (a) EFTEM image of 1100 ◦ C annealed sample with a 10 nm thick SiO film embedded in the HfO2 matrix, showing the presence of Si nanoparticles in the SiO layer; (b) HREM image of the same sample showing that the nanoparticles are crystalline.

annealed sample with 10 nm thick SiO films embedded in the HfO2 matrix (M15). White dots are clearly observed in the SiO layer after annealing, showing the presence of Si nanoparticles. Figure 2(b) is the associated HREM image. The measured thickness of the SiO film is slightly larger than the nominal one, in agreement with XRR results (data not shown). This is true for all the analyzed samples. In addition, it is important to notice that no interfacial layer is observed between the HfO2 tunnel layer and the Si substrate. Finally, crystalline planes characteristic of (111) Si are observed in the SiO layer, showing that the Si nanoparticles are crystalline. After calibration of the scale in the silicon substrate, their inter˚ further confirming planar distance has been found to be 3.14 A, that the nanocrystals are silicon-made. The same morphology is observed for the HfO2 /7 nm thick SiO/HfO2 stack but no nanoparticles are observed when the SiO layer is only 4 nm thick. These results clearly indicate that the formation of Si nanoparticles in HfO2 /SiO/HfO2 multilayer structures is inhibited for SiO films with nominal thicknesses lower than 7 nm. In figure 3 we report the ToF-SIMS depth profiles of the as-deposited and 1100 ◦ C annealed samples with 4, 7 and 10 nm thick SiO films embedded in the HfO2 matrix (M13, M14 and M15, respectively). Analysis of ToF-SIMS signals confirms the results obtained from PL measurements and TEM observations. In SiO2 /SiO/SiO2 multilayer structures, the Si− n signals usually increase after thermal treatment [10]. This intensity variation of the Si− n signals is unambiguously associated with the formation of Si nanocrystals in the SiO2 matrix [13, 14]. In the HfO2 /SiO/HfO2 multilayer structures, this effect can be observed only for the samples with 7 and

Figure 3. ToF-SIMS depth profiles of the as-deposited (on the left) and 1100 ◦ C annealed (on the right) samples with 4 (a), 7 (b) and 10 (c) nm thick SiO films embedded in the HfO2 matrix. All the films have been capped in situ with a 10 nm thick SiO2 layer.

10 nm thick SiO films. In contrast, in the HfO2 /4 nm thick SiO/HfO2 sample, the Si− n signals (for the sake of simplicity in figure 3 we report only the Si− 2 signal) decrease after thermal treatment while the SiO− 3 signal increases. This experimental evidence indicates that, in this sample, after high temperature annealing, the Si atoms in the SiO layer are more oxidized than in the as-implanted sample [15]. That means that the formation of Si nanocrystals in HfO2 /4 nm thick SiO/HfO2 multilayer structures is inhibited by a mechanism that reduces the Si supersaturation in the SiO layer, preventing in this way the SiO phase separation and the formation of Si nanocrystals. The mechanisms that can be effective in reducing the silicon supersaturation in the multilayer structures are related either to the presence of a reservoir that provides oxygen reacting with the Si atoms in the SiO layer or to silicon out-diffusion through the HfO2 layers. The presence of an oxygen reservoir could be ascribed to oxygen absorption during the thermal treatment or to an oxygen excess already present in the as-deposited multilayer structure. In our case oxygen absorption during the annealing in a tubular furnace is very unlikely considering that all the samples have been capped with a 10 nm thick SiO2 film [10]. The presence 3


M Perego et al

Figure 4. ToF-SIMS profile of the HfO2 /4 nm thick SiO/HfO2 multilayer structure without SiO2 capping layer before (a) and after thermal treatment at 600 ◦ C (b) and 900 ◦ C (c) in N2 atmosphere for 60 min. The intensity of the Si− 2 signal at the surface increases with the annealing temperature.

of this capping layer has been shown to be effective in preventing oxygen absorption during the thermal treatment in SiO2 /SiO/SiO2 multilayer structures as revealed by the analysis of the SiO2 /4 nm thick SiO/SiO2 multilayer structure (M17) after annealing. The presence of an excess of atomic or molecular oxygen in the multilayer structure can be ruled out since e-beam-evaporated oxides are usually oxygen-deficient. The hypothesis of moisture absorbed in the HfO2 film is more reasonable considering that HfO2 is a very hygroscopic material. In a recent paper Ferrari et al showed that moisture trapped in an HfO2 film grown by atomic layer deposition (ALD) is responsible for the formation of an SiO2 interfacial layer between the HfO2 film and the Si substrate during the early phase of annealing [16]. In contrast, in our samples, there is no interfacial layer and no increase of the SiO2 related signals at the HfO2 /Si interface is observed after annealing in ToF-SIMS profiles reported in figure 3, suggesting negligible moisture concentration in the HfO2 films. In principle, the presence of moisture in our HfO2 films should be naturally limited because evaporations have been performed in a high vacuum regime. Finally the in situ deposition of the 10 nm thick SiO2 capping layer should prevent moisture absorption during sample exposure to air. Therefore the mechanism reducing the Si supersaturation during the thermal treatment should be related to Si out-diffusion. In order to verify this hypothesis we perform ToF-SIMS measurements on an HfO2 /4 nm thick SiO/HfO2 multilayer structure (M18) without any SiO2 capping layer. The ToF-SIMS profiles of this sample before and after thermal treatments at 600 and 900 ◦ C are reported in figure 4. The absence of the 10 nm thick SiO2 capping layer allows us to detect the silicon diffusion process occurring during the − annealing process. The Si− 2 and SiO3 signals are not ◦ substantially modified by the 600 C thermal treatment. Only a slight increase of the silicon-related signals is observed at

the surface of the HfO2 top layer, suggesting limited silicon diffusion towards the surface through the HfO2 film. At the same time XRD measurements indicate that the HfO2 layers − have been completely crystallized. The Si− 2 and SiO3 signals ◦ after 900 C annealing clearly indicate a silicon accumulation at the surface of the upper HfO2 layer and an increase of the silicon oxidation state in the SiO layer corresponding to a reduction of the silicon content due to the out-diffusion process. The formation of a top silicate layer at the surface of an HfO2 film grown on a silicon substrate during high temperature annealing has already been reported in the literature and is commonly attributed to silicon outdiffusion [17]. According to theoretical models, HfO2 was predicted to be thermodynamically stable in contact with Si up to 1200 ◦ C [18]. However, thermodynamic stability analysis considering only reactions between two stoichiometric solids appears to be insufficient for predicting stable heterointerfaces. According to data available in the literature an Si diffusion mechanism is expected to be activated at the SiO2 /HfO2 and at the HfO2 /Si interfaces during high temperature thermal treatment [19–21]. To our knowledge, no information is available in the literature on the stability of the HfO2 /SiO interface. In our system different interfaces are involved in this diffusion process. Si out-diffusion at SiO/HfO2 interfaces could be compensated by a diffusion mechanism activated at SiO2 /HfO2 and HfO2 /Si interfaces. According to our results the net balance of the diffusion processes occurring in the two HfO2 films determines a decrease of the Si concentration in the SiO film below the supersaturation threshold. Nevertheless the contributions of the diffusion processes occurring in the two HfO2 layers can have different impacts on the nucleation and growth of the Si nanocrystals. It is worth noting here that, for temperatures above 500 ◦ C, crystallization processes are 4


M Perego et al

Figure 6. Tof-SIMS profiles of the as-deposited (a) and 1100 ◦ C annealed (b) SiO2 /4 nm thick SiO/HfO2 sample (M16). The intensity of the Si− 2 signal after thermal treatment indicates the presence of Si nanocrystals in the annealed sample.

In order to discriminate among the various contributions to the diffusion processes occurring in the two HfO2 layers, we grew a new sample (M16) without the upper HfO2 layer, i.e. by depositing the SiO2 capping layer directly on top of the 4 nm thick SiO film. Due to the low Si diffusivity in SiO2 , no Si diffusion is expected to occur at the SiO2 /SiO interface [12, 26]. The intensity of the Si− n signals in the ToFSIMS profile after thermal treatment indicates the presence of Si nanocrystals in the annealed sample (figure 6) as confirmed by EFTEM observations reported in figure 7(a). This result suggests that Si diffusion through the HfO2 film is negligible or that Si diffusion at the HfO2 /Si interface compensates the Si out-diffusion at the SiO2 interface. The XRD pattern of this sample is reported in figure 7(b). No signature of the orthorhombic phase is detected, further confirming the previous hypothesis that the HfO2 tunnel oxide is mainly composed by HfO2 in the monoclinic phase. Our experiment clearly indicates that the loss of Si during the thermal treatment in HfO2 /SiO/HfO2 multilayer structures occurs essentially through the upper HfO2 film whereas no significant loss of Si occurs through the lower HfO2 layer. The origin of this difference is probably related to the different crystallographic phases of the two HfO2 layers. Actually, crystallization of the upper layer in the orthorhombic phase occurs already at 600 ◦ C, before significant Si diffusion is detected. As already shown, the increase of the SiO film thickness can compensate this phenomenon by preserving the Si supersaturation during the thermal treatment, leading to the formation of Si nanocrystals. Moreover, the higher the SiO thickness, the lower the orthorhombic component of HfO2 . The reason for the formation of orthorhombic phase in the 4 nm thick HfO2 top layers grown on SiO deserves further investigation. The main drawback of increasing the SiO thickness is the limited control on the Si nanocrystal size and positioning; a thicker SiO film leads to a wider size distribution and to a broader region where nanocrystals are located. To overcome these strong limitations, different strategies have to be envisioned. Two main approaches can be suggested in order to reach the goal of a 2D matrix of Si nanocrystals with a limited dispersion in size. One possible solution is

Figure 5. XRD patterns of the 1100 ◦ C annealed samples with 4, 7 and 10 nm thick SiO films embedded in the HfO2 . On increasing the thickness of the SiO layer, a decrease of the orthorhombic component at 30◦ is observed.

known to occur in HfO2 layers during thermal treatment [22]. Differences in crystallographic phase can have a huge effect on the diffusion mechanism. Moreover the grain boundary diffusion coefficient in polycrystalline materials can be orders of magnitude higher than in bulk [23]. XRD analysis of the SiO2 capped HfO2 /SiO/HfO2 multilayer structures revealed the presence of two different HfO2 phases after thermal treatment, as shown in figure 5; a first component at 28◦ can be unambiguously identified as monoclinic HfO2 , while a second component at 30◦ is attributed to the presence of orthorhombic HfO2 [22]. It is worth noting that the identification of the latter phase is not straightforward due to the superposition of the diffracted peaks of the orthorhombic phase with those of the cubic or tetragonal phases. Nevertheless the orthorhombic phase is observed to decrease by increasing the thickness of the SiO layer. Moreover, according to XRR measurements, the electronic density of the upper HfO2 layers is systematically higher than in the lower HfO2 films. These data suggest that the upper HfO2 film is characterized by a higher fraction of orthorhombic HfO2 than the lower HfO2 film. According to the phase diagram the stable crystallographic phase for HfO2 at the temperature of interest for this work is monoclinic [24]. Nevertheless different crystallographic phases can be obtained in ultrathin films by changing the film thickness [25]. XRR analysis of our samples indicates that the thickness of the HfO2 films is the same (4.0 ± 0.1 nm) for all the samples. The variations of crystallographic phase observed in our samples are not induced by variation of the HfO2 film thickness. The origin of this evolution of the crystal structure is somehow related to the variation of the SiO thickness. 5


M Perego et al

Figure 7. EFTEM image (a) and XRD pattern (b) of the SiO2 /4 nm thick SiO/HfO2 sample (M16) after 1100 ◦ C thermal treatment in N2 atmosphere.

to compensate the Si out-diffusion process during the thermal treatment by increasing the Si concentration, i.e. using a SiOx film with x lower than one. In this way we could try to keep the Si concentration above the supersaturation threshold during the annealing. Another possibility to reach this goal is the introduction of an additional SiO layer, on top of the upper HfO2 film, acting as a silicon reservoir. The Si out-diffusion process from the two SiO layers in contact with the HfO2 film should compensate each other. These hypotheses are currently under investigation.

References [1] Kim D-W, Kim T and Banerjee S K 2003 IEEE Trans. Electron Devices 50 1823 [2] Chen J H, Wang Y Q, Yoo W Y, Yeo Y C, Samudra G, Chan D S H, Du A Y and Kwong D-L 2004 IEEE Trans. Electron Devices 51 1840 [3] Monzio Compagnoni C, Ielmini D, Spinelli A S and Lacaita A L 2005 IEEE Trans. Electron Devices 52 569 [4] Monzio Compagnoni C, Ielmini D, Spinelli A S and Lacaita A L 2005 IEEE Trans. Electron Devices 52 2473 [5] Kapetanakis E, Normand P, Tsoukalas D and Beltios K 2002 Appl. Phys. Lett. 80 2794 [6] Bonafos C et al 2004 J. Appl. Phys. 95 5696 [7] Nakajima A, Sugita Y, Kawamura K, Tomita H and Yokoyama N 1996 J. Appl. Phys. 80 4006 [8] Kanjilal A, Lundsgaard Hansen J, Gaiduk P, Nylandsted Larsen A, Cherkashin N, Claverie A, Normand P, Kapelanakis E, Skarlatos D and Tsoukalas D 2003 Appl. Phys. Lett. 82 1212 [9] Zacharias M, Heitmann J, Scholz R, Kahler U, Schmidt M and Blasing J 2002 Appl. Phys. Lett. 80 661 [10] Perego M, Fanciulli M, Spiga S, Bonafos C and Cherkashin N 2006 Mater. Sci. Eng. C 26 835 [11] Fanciulli M, Perego M, Bonafos C, Mouti A, Schamm S and Benassayag G 2006 Adv. Sci. Technol. 51 156 [12] Mathiot D, Schunck J P, Perego M, Fanciulli M, Normand P and Tsoukalas D 2003 J. Appl. Phys. 94 2136 [13] Perego M, Ferrari S, Spiga S, Bonera E and Fanciulli M 2003 Appl. Phys Lett. 82 121 [14] Perego M, Ferrari S, Fanciulli M, Ben Assayag G, Bonafos C, Carrada M and Claverie A 2004 J. Appl. Phys. 95 257 [15] Perego M, Ferrari S, Fanciulli M, Ben Assayag G, Bonafos C, Carrada M and Claverie A 2004 Appl. Surf. Sci. 231/232C 813 [16] Ferrari S and Fanciulli M 2006 J. Phys. Chem. B 110 14905 [17] Chang H S, Hwang H, Cho M, Kim H H and Moon D W 2004 J. Vac. Sci. Technol. A 22 165 [18] Shin D and Liu Z-K 2007 Scr. Mater. 57 201 [19] Hoshino Y, Kido Y, Yamamoto K, Hayashi S and Niwa M 2002 Appl. Phys Lett. 81 2650 [20] Liu C-L, Stoker M, Hegde R I, Rai R S and Tobin P J 2002 MRS Symp. Proc. 731 W5.2

4. Conclusions In conclusion we demonstrate that the synthesis of a 2D array of Si nanocrystals embedded in HfO2 by high temperature annealing of HfO2 /SiO/HfO2 multilayer structures is strongly inhibited by Si out-diffusion phenomena occurring at the SiO/HfO2 interfaces. Due to the loss of silicon, the supersaturation conditions that are necessary to induce nanocrystal nucleation and growth are difficult to be achieved. To overcome this problem a compensation mechanism for the reduction of Si concentration in the SiO layer during high temperature annealing is required. By increasing the thickness of the SiO layer Si supersaturation can be preserved and Si nanocrystal synthesis achieved. Other strategies to overcome the reduction in Si concentration during the thermal treatment have been suggested.

Acknowledgments The authors would like to thank Sylvie Schamm (CEMES) for very fruitful discussions. This work has been supported by the Project MATRIX ‘2-D arrays of metallic and semiconducting nanocrystals for microelectronic applications’ funded by Fondazione CARIPLO. 6


M Perego et al

[21] Lopez C M and Irene E A 2007 J. Appl. Phys. 99 024101 [22] Wiemer C, Ferrari S, Fanciulli M, Pavia G and Lutterotti L 2004 Thin Solid Film 450 134 [23] Stemmer S 2004 J. Vac. Sci. Technol. B 2 791 [24] Massalski T B (ed) 1986 Binary Alloy Phase Diagrams (Metals Park, OH: American Society for Metals)

[25] Alessandri M, Piagge R, Caniatti M, Del Vitto A, Wiemer C, Pavia G, Alberici S, Bellandi E and Nale A 2006 ECS Trans. 3 183 [26] Takahashi T, Fukatsu S, Itoh K M, Uematsu M, Fujiwara A, Kageshima H, Takahashi Y and Shiraishi K 2003 J. Appl. Phys. 93 3674

7