- Email: [email protected]
Synthesis of mesoporous amorphous silica by Kr and Xe ion implantation: Transmission electron microscopy study of induced nanostructures
Microporous and Mesoporous Materials 132 (2010) 163–173
Contents lists available at ScienceDirect
Microporous and Mesoporous Materials journal homepage: www.elsevier.com/locate/micromeso
Synthesis of mesoporous amorphous silica by Kr and Xe ion implantation: Transmission electron microscopy study of induced nanostructures E. Oliviero a,*, M.-O. Ruault a, B. Décamps a, F. Fotuna a, E. Ntsoenzok b, O. Kaïtasov a, S. Collin a a b
CSNSM, UMR 8609, CNRS/IN2P3-Univ-Paris-Sud, Bâtiment 108, 91405 Orsay Campus, France CEMHTI-CNRS, 3A, rue de la férollerie, 45071 Orléans, France
a r t i c l e
i n f o
Article history: Received 13 November 2009 Received in revised form 18 February 2010 Accepted 20 February 2010 Available online 24 February 2010 Keywords: Mesoporous silica Noble gas implantation Bubbles Cavities Structural investigation
a b s t r a c t Thermally grown amorphous SiO2 was implanted at room temperature with heavy noble gases Kr and Xe in order to create cavities in the oxide and increase its porosity. The implantation energies were chosen in order to have the same implantation depth for both ions. Although both ions induce bubbles in amorphous SiO2, bubble size and spatial distribution depend upon the ion mass. Moreover, Xe implantation leads to the additional formation of ‘‘nanoclusters”. Thermal stability of bubbles/cavities depends on the implanted ion. The nucleation of bubbles and nanoclusters in amorphous SiO2 is discussed in terms of ion mobility, gas–defect interactions, and chemical interaction. Bubble growth is shown to occur by a migration and coalescence process. Ó 2010 Elsevier Inc. All rights reserved.
1. Introduction With the constant miniaturization of electronic devices and consequently the continual increase in the density of multilevel interconnections, material with lower dielectric constant (k) value is required to overcome some of the resulting problems such as transmission delay, power consumption and cross-talk noise. Until now, SiO2 having a dielectric constant of 3.9 has been used as a dielectric material for the interconnections in integrated circuit devices. The International Technology Roadmap for Semiconductors (ITRS) has projected that the dielectric constant of the material should fall below 2 by 2010. There are two possible approaches to achieve this. The obvious one is the use of new materials with lower k, but care has to be taken on their compatibility with microelectronics (interface with silicon, mechanical strength, thermal and etching properties, etc.). The second approach is to decrease the effective dielectric constant of SiO2. This can be achieved by incorporating pores into the material. Air has a dielectric constant of roughly 1.00059, thus increasing the SiO2 layer porosity leads to the reduction of the dielectric constant. The pore structure and its distribution have distinct influences on the basic physical properties of the material. While the reduction in k is desirable, there are also adverse effects such as deterioration of the mechanical properties as porosity increases. There is also the problem of mois-
* Corresponding author. E-mail address: [email protected] (E. Oliviero). 1387-1811/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.micromeso.2010.02.015
ture adsorption in the presence of pores at the surface. Therefore, an ideal porous material would consist of a network of small pores embedded into material, with a regular size distribution (uniformly-sized and distributed pores). An appropriate technique to realize such a pores network within SiO2 is rare gas ion implantation. The formation of bubbles/cavities resulting from the agglomeration of noble gas after ion implantation has been extensively studied over several decades in crystalline materials such as metals. Blistering, swelling and embrittlement are unwanted effects in these materials, which are used in fusion or fission reactors [1]. Bubble/cavity formation, blistering and swelling were also studied in semiconductors as smart tools for applications such as impurity gettering [2], strain relaxation [3], thin layer separation [4] and defect engineering [5]. Bubble formation and growth in such materials is a complex phenomenon involving point defect production/ recombination, mobility and agglomeration. Because of their extremely low solubility, noble gases have a strong tendency to segregate into gas–vacancy complexes which may further coarsen and grow forming stable gas-filled cavities. The final defect morphology resulting from noble gas implantation and post-annealing is intricate. It consists of point and extended defects of both vacancy and interstitial types. The vacancy-type defects may contain gas. The main factors influencing the eventual defect morphology depends on the rate at which damage is introduced, the thermal and radiation-enhanced mobility of both gas and defects, and most importantly the ability of the gas to permeate from cavities back into the matrix.
164
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
The extension of these studies from crystalline to amorphous materials such as SiO2 could give insights into the damage accumulation processes and a better understanding of the bubble formation mechanisms. Indeed, the formation and growth mechanisms of cavities are relatively well understood in metals [6] and semiconductors [7]. In the case of SiO2, the nucleation and growth mechanisms of metallic/semiconductor nanoprecipitates has been extensively studied but these mechanisms are not yet well understood for bubbles/cavities [8,9] and require detailed analyses of bubble characteristics including (i) the diffusivity of the gas atoms in SiO2 and (ii) the nature of defects created by implantation. The aim of the work presented in this paper is to understand the processes that could play a role in the nucleation and growth of the bubbles formed by Kr or Xe implantation in amorphous SiO2 (aSiO2). Thus, a detailed TEM study of the induced nanostructures (involving bubbles) was performed after two implantation fluences (3.5 and 5 1016/cm2) and after isochronal annealing up to 1370 K. The as-implanted results are discussed in terms of gas mobility and the interaction between gas atoms and vacancy-type defects or negatively charged defects. The structural evolutions during annealing are explained using the migration and coalescence mechanism. 2. Experimental details Thermal amorphous SiO2 layers were grown by heating n-type silicon wafers at 1370 K in ambient air. Samples of two different layer thicknesses (220 nm and 2.3 lm) were used in this study. The samples were implanted, at room temperature, with 220 keV Kr ions or 300 keV Xe ions at the same fluences (3.5 and 5 1016 ions/cm2). The implantation energies were chosen, according to SRIM (Stopping and Range of Ions in Matter) simulations [10], in order to have the same projected range (Rp) of 120 nm for both ions. Ion flux was kept as low as 1– 1.5 1012 ion/cm2 s to avoid any temperature increase during implantation. After implantation, samples were annealed at temperatures ranging from 670 to 1370 K for 1 h under nitrogen atmosphere. The as-implanted and post-annealed samples were characterised by cross-sectional transmission electron microscopy (XTEM) and compared with the ion distributions obtained previously [9] by Rutherford Backscattering Spectrometry (RBS) in specimens implanted under the same experimental conditions. The measurements were performed using a FEI CM12 TEM and a Tecnai G2 20 TEM. The cross-section samples for XTEM were mechanically thinned by means of a tripod. For a better comprehension of the XTEM results, a few preliminary remarks are needed before a detailed presentation of the results. (1) Amorphization of the Si substrate occurs when implanting in the 220 nm thick SiO2 layer. In fact, the ion range is up to about 250 nm for both implanted ions. Thus, a Si band just behind the interface is amorphized due to the induced mixing phenomena at the SiO2/Si interface and to the induced irradiation defects. During post-annealing P800 K, the amorphous band recrystallizes leaving defects such as dislocations or rod-like defects.
(2) The thickness of the 220 nm thick SiO2 layer was controlled before and after implantation by RBS technique. No swelling of the layer due to noble gas implantation was observed in the bulk specimens. However, a widening of the SiO2 layer was observed by TEM (Fig. 1). This enlargement occurs during the mechanical thinning process and is due to an increase of the SiO2 ductility. The change in ductility is induced by the implantation/irradiation. In the following, the deformation of the SiO2 layer will be taken into account. (3) For non-spherical nanostructures, the size will be given as the equivalent diameter for spherical shape having the same area. (4) The density depth distribution of nanostructures is obtained by counting them in 10–20 nm width bands only if the size of these nanostructures is small enough (i.e. at least half of the band width). In our study nanostructure sizes are mostly quite large so that it is not possible to give the bubble density distribution as a function of depth. The same is true for the distribution of the bubble centres as a function of depth. It seems more accurate to evaluate the projected area of the nanostructures. A normalization is done by dividing the results by the total projected area of all the nanostructures present in the layer. It should be noted that, due to the amorphous structure of SiO2, it is difficult to assess the thickness of the XTEM specimen in the studied zones. In fact, the XTEM specimen thickness can be evaluated only by the electron transparency of the sample (see for example Fig. 1). The accuracy of such an evaluation is not better than 30–50 nm. However, each projected area depth distribution is obtained on a sufficiently small region, so that the sample thickness variation can be estimated at about 20% inside an individual depth distribution. 3. Results Both Kr and Xe implanted samples clearly show the formation of nanostructures in the implanted region. A detailed description of the observed nanostructures is given in the following. 3.1. Xe implantation 3.1.1. As-implanted Figs. 2 and 3 show the nanostructures induced in amorphous SiO2 by 300 keV Xe implantation at room temperature for fluences of 3.5 and 5 1016 Xe/cm2. Two types of nanostructures are observed: (i) Nanostructures without Fresnel fringes (Fresnel fringes are characteristic of bubbles/cavities [11]). These nanostructures exhibit a black contrast in overfocus conditions, which tends to vanish in the underfocus condition (Fig. 3a–d). These nanostructures could not therefore be identified as bubbles or cavities. From this point on, these nanostructures will be termed as ‘‘nanoclusters”. (ii) Nanostructures identified as bubbles, i.e. exhibiting Fresnel fringes at the interface whose contrast reverse from black to white from under to overfocus conditions [11]
Fig. 1. An assembly of brightfield image showing a TEM cross-section of an as-implanted sample. The white bars illustrate the size of the SiO2 layer before implantation (220 nm) along the cross-section. The thickness of the TEM sample increases from the left to the right.
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
165
Fig. 2. Evolution during TEM observations (i.e. 120 keV electron beam) of the nanostructures created in a 2 lm thick SiO2 with 300 keV Xe ions up to 5 1016 Xe/cm2 in underfocus conditions. The observation time (increasing from left to right) is reported on the micrographs.
Fig. 3. XTEM bright field images of as-implanted 220 nm-SiO2 with 300 keV Xe ions up to 3.5 1016 Xe/cm2 in (a) overfocus and (b) underfocus conditions, and up to 5 1016 Xe/cm2 in (c) overfocus and (d) underfocus conditions. (e and f) Enlarged area for detailed examination of nanostructures: same experimental conditions as in (c and d) showing the nanostructure band (between the black lines) observed in overfocus conditions vanishes in underfocus conditions. Note also that, among three dark nanostructures circled in underfocus conditions, only two nanostructures exhibit the inverse characteristic contrast (white) for nanobubble/cavity at underfocus condition, the contrast of the last vanishes.
(Fig. 3a–d). The contrast inside the large bubbles is mainly grey in overfocus conditions and becomes lighter – but not white – in the underfocus condition. This behaviour is probably due to the heavy mass of Xe present inside the bubbles. Some of the large bubbles exhibit a hole contrast (white con-
trast in under and overfocus conditions) due to specimen cutting during sample thinning. It should be noted that those few holes/cavities are not stable under the TEM electron beam as it was shown in a time evolution of the nanostructures under the electron beam (Fig. 2). This effect is not
166
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
yet fully understood and needs further extensive study to elucidate on this. In this work, all the size studies were done on pictures obtained within 1 or 2 min to avoid any bubble evolution under the electron beam. For both fluences (Fig. 3a–d) a band of large bubbles is surrounded by smaller nanostructures mainly located towards the Si interface. The diameter of the nanostructures ranges between 2 and 10 nm with an average diameter of 6 nm. Even if a few small nanostructures (diameter < 3 nm) exhibit the reversal contrast of nanobubbles – black to white for over to underfocus conditions – most of them are identified as nanoclusters. Fig. 3e and f show an example of the study done for the TEM determination of the small nanostructures whose diameter was 63 nm. It is acknowledged that their size (especially for the largest ones) might be overestimated because of superimposition and their black contrast. As a consequence, the nanocluster concentration might be underestimated. The bubbles exhibit a random size distribution within the bubble band (e.g. Fig. 3b and d) and their average diameter does not significantly change (21 ± 2 nm) with increasing fluence (Fig. 4a and b). However, the shape of their distribution does change. A few larger bubbles (50–90 nm in diameter) appear at 5 1016 Xe/cm2. For this fluence, the peak of the distribution (at 14 ± 2 nm) is narrower than at 3.5 1016 Xe/cm2. In fact, bubbles of 20–25 nm in diameter disappear at the expense of the larger bubbles possibly due to bubble coalescence. This point seems to be confirmed by the fact that a non-negligible number of elongated bubbles are observed. Fig. 5a–d shows, for both implantation fluences (3.5 and 5 1016/cm2), the normalized1 depth distribution of the bubbles and the normalized depth distribution of the nanoclusters. We would like to point out that bubble concentration is about five times greater than the nanocluster concentration at a fluence of 3.5 1016/ cm2 and of an order of magnitude greater at 5 1016/cm2. Thus the proportion of bubbles increases with Xe concentration in a-SiO2. At 3.5 1016/cm2, the bubble depth distribution follows the simulated Xe profile and is centred at 110–130 nm (corresponding to the Rp of the Xe calculated with SRIM) and the nanoclusters stand at the tail-end of the Xe distribution (Fig. 5a). At 5 1016/ cm2, the bubble/cavity depth distribution is shifted towards the free surface and the peak (80–120 nm) corresponds to the maximum of the simulated damage distribution (Fig. 5b). This is coherent with Xe depth profiles, previously obtained by RBS [9]. Indeed, at 3.5 1016/cm2, the Xe RBS profile exhibited a wide peak extended from 100 to 140 nm (not shown here). By increasing the fluence to 5 1016/cm2, the Xe RBS profile shifts towards a lower depth and becomes narrower (from 70 to 90 nm). Moreover, a shoulder peak appeared at a depth of 110–140 nm (reported in Ref. [9]). The latter could be associated to the shoulder observed in the bubble area distribution (Fig. 5d). In Fig. 5d, focusing on the total nanostructures profile, it can be noticed that the shoulder is only weakly increased when adding the nanoclusters. Moreover, the nanocluster distribution peak (170–200 nm) does not shift with increasing fluence. The shoulder is thus ascribed to bubbles.
3.1.2. Thermal annealing One of the main effects is the disappearance of the nanoclusters from annealing temperatures P870 K. The other striking effect is that for annealing at 1370 K, cavities remain in the sample while the Xe has already exodiffused [9]. These cavities exhibit quasispherical shapes (Figs. 6e and 7d). 1
Normalisation is carried out by dividing by the total nanostructure area.
Fig. 4. Normalized size distribution of the bubbles created by 300 keV Xe implantation in silicon dioxide for a fluence of (a) 3.5 1016/cm2 and (b) 5 1016/cm2.
Although the general feature of the nanostructure evolution during post-annealing is fairly similar between 3.5 and 5 1016 Xe/cm2, the detailed statistic analysis shows some differences (Figs. 6–9). At a fluence of 3.5 1016 Xe/cm2 the maximum bubble diameter observed is 60 nm. Both the bubble/cavity shape and the size distribution do not significantly change for annealing up to 1370 K. The mean diameter remains constant. The bubble depth distribution does not significantly change up to 1020 K. After 1370 K annealing, the cavities located near the SiO2/Si interface progressively disappear leading to a sharp distribution centred at about 100 nm (Fig. 8). At a fluence of 5 1016 Xe/cm2 the bubble size distribution and the bubble mean diameter do not significantly change up to 1370 K. However, the number of elongated bubbles slightly increases (e.g. Fig. 7c). The bubble depth distribution does not significantly change up to 1020 K. In particular, the shoulder remains (Fig. 9) despite the disappearance of the nanocluster peak. After a 1370 K anneal, the depth distribution tightens between 50 and 150 nm and the cavities become predominantly spherical. The narrowing of the depth distribution is however not as significant as for the lower fluence (see Figs. 8 and 9). 3.2. Kr implantation 3.2.1. As-implanted As already mentioned in the case of Xe implantation, TEM study of Kr implantation in a-SiO2 also faces the problem of bubble evolution under the electron beam. An example of this evolution is shown in Fig. 10. Primarily, when the electron beam is applied, the bubbles appear to grow (see Fig. 10 after 2 and 4 min of observation under the electron beam) and a few bubbles exhibit deformed shapes. Between 10 and 15 min of observation (Fig. 10), the bubbles return to a spherical shape and their contrast faints. This effect was observed whichever electron beam voltage was used (from 80 to 200 kV) and will be studied in more detail outside the remit of this paper. Therefore, exposure to the electron beam was limited to less than 4 min during the observation of the bubble
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
167
Fig. 5. Normalized area profile for bubbles (bars) and ‘‘nanoclusters” (full-line) [Units on the left] compared with SRIM simulation of Xe distribution profile and target displacements [Units on the right] created by 300 keV Xe implantation in 220 nm-SiO2 for a fluence of (a) 3.5 1016/cm2 and (b) 5 1016/cm2. In order to clearly point out the shape of the area distributions we have reported on the right the distributions with a line between points for the distributions of bubbles and total nanostructrures (bubbles + nanoclusters), respectively, for fluences: (c) 3.5 1016/cm2 and (d) 5 1016/cm2.
layer. The Kr implantation-induced nanostructures form on a band near the free surface. The width of the nanostructure band fluctuates throughout the implanted area with an average width of around 150–180 nm. The SiO2 layer thickness has no effect on the band width (Fig. 11a and b) neither does the Kr fluence (Fig. 11a and c). For both implanted fluences (3.5 and 5 1016 Kr/cm2) the nanostructures were identified as bubbles as they exhibit the contrast characteristic of bubbles/cavities (Fig. 11c and d) and Kr is confirmed by RBS to be trapped in the sample. No nanoclusters are observed. The size repartition is quasi-homogeneous within the band (Fig. 11a–d) and the bubble diameter ranges from 4 to 30 nm (Fig. 12a–c), with an average diameter of 10–12 nm. This is independent of the fluence and of the SiO2 layer thickness. The depth distribution also does not significantly change for the different fluences or the different SiO2 layer thicknesses (Fig. 13a and b): it clearly displays a plateau shape. This differs from the Kr distribution predicted by SRIM [10] which exhibits a Gaussian profile with a peak at 120–130 nm (Fig. 13a–c). The bubbles extend preferentially on the deeper side of the theoretical Kr profile predicted by SRIM (Fig. 13a and b). For the 220 nm thick SiO2 layer (Fig. 13c), the bubble layer is formed very close to the SiO2/Si interface. If the SiO2/Si interface was absent (as in the 2.3 lm-thick SiO2) the bubble layer would extend deeper up to 240 nm (Fig. 13b). However, bubbles are not observed either at the 220 nm-SiO2/Si interface or in the Si substrate (Figs. 11b and 13c). Instead, the bubbles seem to extend towards the free surface corresponding to the damaged region. It illustrates the ease for the Kr to diffuse and agglomerate in the SiO2 compared to silicon.
These results are in agreement with the RBS results [8] at 5 1016 Kr cm 2 reported in Fig. 13c which show that the Kr profile exhibits a plateau instead of the peak predicted by SRIM calculations (see Fig. 13c). The plateau shape corresponds to Kr diffusion within the implanted region. SRIM calculations show that, without diffusion, damaged region is very close in depth to the Kr distribution, so it would be useful to separate these two profiles, in order to study the influence of the defects on the observed plateau distribution. Future work entails a detailed study of a possible enhanced diffusion by defects due to irradiation. This could be achieved by increasing either the Kr energy or the damage rate (dpa) in a dual-beam experiment. 3.2.2. Thermal annealing After annealing at 670 K, XTEM observations in 2.3 lm SiO2 show the presence of bubbles/cavities at the same depth region as after Kr implantation (Fig. 14a). The average bubble/cavity diameter remains at about 12 ± 2 nm and the size distribution does not significantly change (Figs. 15 and 16). Only a few larger bubbles/cavities are observed (less than 5%) (Fig. 15c). They have an elongated shape (Fig. 14b) which indicate that they probably result from a coalescence process. These results are coherent with previous RBS measurements [8] which show that after an anneal at 670 K, the Kr distribution is still a plateau although a strong decrease in the Kr concentration (by a factor of about 3) is observed. After a 970 K annealing, the bubbles/cavities are no longer present inside the SiO2 layer. This disappearance is in agreement with the exodiffusion of Kr observed by RBS above 870 K [8]. Thus, no cavities are detected within the material in the absence of Kr.
168
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
Fig. 6. XTEM images showing the evolution of bubbles/cavities obtained by Xe implantation in 220 nm-SiO2 at a fluence of 3.5 1016/cm2 after 1 h annealing at (a) 670 K, (b) 970 K, (c) 1020 K, (d) 1170 K and (e) 1370 K.
4. Discussion 4.1. Nucleation and growth mechanisms of bubbles The nucleation of bubbles in metals [6] or in semiconductors [12] is attributed to the interaction of gas atom with vacancies. In amorphous SiO2 the concept of vacancy is difficult to address but bubble nucleation can be seen as the interaction of gas atom with some free volume. We have previously [9] suggested from RBS and PAS measurements that the interaction of Xe atoms with vacancy-type defects (i.e. oxygen vacancies – oxygen deficiency centres, or defects with large open volume) leads to complexes that could be precursors of bubbles. This interpretation is coherent with the fact that the bubble distribution and the Xe profile measured by RBS (§ III.A1) closely follow the simulated implantation-induced damage distribution for the highest fluence. For the lowest fluence, both the Xe profile obtained by RBS and the bubble distribution are closer to the simulated Xe profile. In fact, for the implantation energy used, the simulated Xe profile and the induced damage profiles overlap in the region where bubbles are formed. Bubbles are thus formed where the ratio Xe/defect is maximum. The fact that,
at the highest fluence the bubble and the Xe profile shift towards the damage profile shows that above a threshold concentration, the induced damage act as sinks for Xe atoms leading to the bubble formation centred at the maximum of the simulated damage distribution. The interaction of implanted gas atoms with vacancytype defects is also coherent in the case of Kr, since, even if a plateau shaped distribution is observed, bubbles are formed at about the depth of the simulated ion distribution (Fig. 13). These results could be related to the mobility of Xe and Kr in a-SiO2. Actually, Kr is qualitatively more mobile than Xe in a-SiO2 [13]. However, Kr does not follow the simulated implantation damage profile. In fact, the damage concentration (dpa) for Kr at the highest fluence is equivalent to that of the Xe implantation at the lowest fluence where no diffusion occurs. So the induced damage is below the threshold concentration, and do not act as sink for Kr, but can enhanced its diffusion. Nevertheless, when the fluence increases, a small shift of the bubble distribution occurs towards the implantation damage profile, indicating that damage concentration is getting closer to the threshold. Implanting to a higher dose (>5 1016 Kr/cm2) should lead to the formation of a bubble band centred on the simulated damage peak. The formation of bubbles
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
169
Fig. 7. XTEM images showing the evolution of bubbles/cavities obtained by Xe implantation in 220 nm-SiO2 at a fluence of 5 1016/cm2 after 1 h annealing at (a) 670 K, (b) 1020 K, (c) 1170 K, (d) 1370 K.
Fig. 8. Normalized area profiles for bubbles (bars) and ‘‘nanoclusters” (dotted-line) after a 300 keV Xe implantation at 3.5 1016 Xe/cm2: as-implanted, and after an isochronal anneal of one hour between 670 and 1370 K.
Fig. 9. Normalized area profile for bubbles (bars) and ‘‘nanoclusters” (dotted-line) after a 300 keV Xe implantation at 5 1016 Xe/cm2: as-implanted, and after an isochronal annealing of one hour between 670 and 1370 K.
all along the Kr profile and even beyond, where very small amounts of Kr and damage are expected, is proof that the complexes Kr-vacancy-type defect are mobile and are able to spread resulting in the observed plateau distribution.
4.1.1. Ostwald ripening versus migration-coalescence Two different mechanisms, Ostwald ripening (OR) and migration-coalescence (MC) are generally mentioned to explain the growth of bubbles. In this part of the discussion we will compare both mechanisms and determine which is responsible for bubble growth in amorphous SiO2.
170
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
Fig. 10. The evolution during TEM observations (i.e. 200 keV electron beam) of the nanostructures created in a 2 lm thick SiO2 with 300 keV Xe ions up to 5 1016 Kr/cm2 in underfocus conditions. The observation time (increasing from left to right) is reported on the micrographs. Examples of bubbles which exhibit deformed shapes are highlighted with an arrow.
Fig. 11. XTEM bright field images of as-implanted SiO2 with 220 keV Kr ions: (a) fluence of 3.5 1016/cm2 in 2.3 lm thick SiO2, the insert shows an enlargement of the bubble band/(overfocus condition), (b) fluence of 5 1016/cm2 in 220 nm thick SiO2 (underfocus condition), (c and d) 2.3 lm thick SiO2, respectively, in overfocus and underfocus conditions.
The formation of bubbles in amorphous material seems to be quite different than in crystalline materials. It is well known that noble gas implantation into crystalline materials such as Si [14], SiC [15], GaN [16] leads to the formation of tiny bubbles of 1– 3 nm in size. Recently it was shown that He implantation in amorphous silicon leads to the formation of significantly larger bubbles compared to those observed in crystalline Si under identical implantation conditions [17]. In a-SiO2, Kr implantation leads to the formation of bubbles with a mean diameter of 10 nm and the Xe implantation leads to the formation of bubbles with a mean diameter of 20 nm. Moreover, during thermal annealing we do not observed any significant growth of the bubbles. This confirms that growth already occurs during ion implantation. From our results, we can state that bubble mobility in a-SiO2 almost certainly results from migration and coalescence mechanism limited by a surface migration process. It can then be assumed, qualitatively, that the surface diffusivity in amorphous SiO2 is relatively low. This allows bubble coarsening directly during implantation explaining the observation of large bubbles directly after Kr and Xe implanta-
tion into a-SiO2. Small bubbles could be mobile at room temperature (and thus during the implantation) and could grow by thermally or radiation-induced induced motion and coalescence. In addition, this would explain the rather non-spherical aspect of the bubbles. This would be the result of a recent coalescence of spherical bubbles as stated in helium-implanted silicon [7]. Moreover, the bubbles formed during implantation are not stable. If an excitation is provided such as electron irradiation, the morphology of the bubbles can change thus illustrating again the relatively ease for migration and coalescence. Through thermal annealing, the bubbles formed by Kr implantation are not stable when Kr totally exodiffuses. However, in the case of Xe, cavities are found to be stable even after 1370 K where Xe is no longer present. Two factors seem to influence the formation of cavity. Firstly, the bubble size is significantly smaller for as-implanted Kr than for as-implanted Xe. This point suggests a possible threshold size for which the cavities are stable through thermal annealing. Secondly, some chemical mechanisms are necessary to lead to cavity formation. Indeed, it was shown by the
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
171
Fig. 14. XTEM bright field in underfocus condition of a 2.3 lm SiO2 layer implanted with 220 keV Kr ions up to a fluence of 5 1016/cm2 and post-annealed 1 h at 670 K. (a) View of the whole band in a thick part of the sample and (b) example of a bubble which is probable result of a coalescence process (indicated by a white arrow).
Fig. 12. Size profiles of bubbles created by 220 keV Kr, (a) at 3.5 1016/cm2 for 2.3 lm thick SiO2, (b) at 5 1016 /cm2 for 2.3 lm thick SiO2 and (c) at 5 1016/ cm2 for 220 nm thick SiO2.
work of Marstein et al. [18] that the presence of the gas is not required to create cavities. They report the formation of cavities in aSiO2 after 1270 K post-annealing of samples containing Ge nanoprecipitates formed by ion implantation although, in this case, the cavity precursors are complexes Ge-vacancy-type defect. They assumed that during annealing, firstly, the Ge precipitates grow and then secondly, the cavities are formed by the out diffusion of Ge from the crystalline Ge nanoprecipitates. So in this case, no gas is required to form cavities in a-SiO2 after annealing. It shows again that above a critical size, vacancy-type defect agglomeration
Fig. 13. Normalized depth distribution of bubbles area, after 220 keV Kr ion implantation in 2.3 lm SiO2 up to (a) 3.5 1016/cm2, (b) 5 1016/cm2 and (c) in 220 nm-SiO2 up to 5 1016/cm2. (c) Also shows the Kr concentration previously obtained by RBS [8]. These distributions are compared to the ion range distribution and the ‘‘defects” distribution in SiO2 obtained by SRIM simulation.
172
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
cess. Thus, there must be an alternative mechanism leading to cavity formation. This mechanism has to involve chemical interaction between vacancy-type defects and implanted atoms. Through this interaction, the implanted atom stabilizes the vacancy-type defects. In order to better understand these effects, in situ TEM during ion implantation experiments are necessary in an attempt to detect any bubble motion during implantation. Moreover, the stability of bubbles under electron/ion beam will be investigated. This will be done on the dual-beam TEM of JANNuS facility at Orsay [19]. By assimilating cavities to pores, the porosity created by rare gases implantation can be estimated from the TEM results. The porosity is found to increases directly after implantation to about 2%. After thermal annealing, the porosity of the SiO2 layer is found higher than 10%. It is worthwhile to underline that this porosity is relative to the thickness of the SiO2 layer and that a higher level of porosity can be reached by achieving multi-energy implantation to create a larger cavity layer. C–V measurements [20] performed on Metal Oxide Semiconductor (MOS) show that the dielectric constant k is reduced in both cases from 3.9 for the unimplanted SiO2 to about 1.5 for the as-implanted samples. The decrease of k is attributed both to the increase of porosity and to the reduction of SiO2 polarizability (modification of the dipoles and/or breaking of polarizable Si–O bonds). Indeed, after thermal annealing, the dielectric constant increase slightly while the porosity is found to increase. Fig. 15. 2.3 lm thick SiO2 implanted with 220 keV Kr ions up to a fluence of 5 1016/cm2. Comparison of the size profile (a) as-implanted and (b) post-annealed for 1 h at 670 K. In (c), the tail of the distribution (for size P 20 nm) is enlarged (i.e. the normalized ratio scale was a magnitude lower).
Fig. 16. 2.3 lm thick SiO2 implanted with 220 keV Kr ions up to a fluence of 5 1016/cm2. Comparison of the normalized depth distribution of bubbles area: (a) as-implanted and (b) post-annealed for 1 h at 670 K.
can be stable. The formation of cavities after annealing is, however, not only related to the vacancy-type defect concentration created in the matrix, since we recently showed that a Bi implantation at same conditions of fluence and penetration depth in the matrix does not lead to any cavity formation after a 1270 K annealing pro-
4.2. The nature of the nanoclusters In the case of Xe implantation, nanocluster formation was observed. The first question that arose was what is the possible composition of nanoclusters? To tentatively determine directly their composition, Electron Energy Loss Spectroscopy (EELS) measurements were carried out. However, it was not entirely successful due to the very small diameter (few nm) of these nanoclusters in comparison with the probe size limit (P10 nm). However, from the following points, a probable composition of the nanocluster can be deduced. (1) We have shown above that the proportion of bubbles versus nanoclusters increases with the Xe concentration and that the Xe shoulder at the end of the distribution is clearly more associated to the bubble distribution than to the nanoclusters. Thus, the nanoclusters are not directly related to the Xe concentration. If Xe is involved in the nanoclusters, it is a weak contribution. (2) It was shown by Positon Annihilation Spectrometry measurements [9] that after Xe implantation, the ‘‘void” volume (i.e. oxygen vacancies or defects with large open volume) increases in the bubble region. This ‘‘void” volume increase concurrently with the fluence. Meanwhile, negatively charged defects (Si–O , Si–O–O , O2 ) were observed at the tail-end of Xe depth distribution. No effect of the fluence was observed on their density. The formation of negatively charged defects is due to the ejection of interstitial oxygen (primary knock-on atom) from the implanted region. If this ejected oxygen has enough energy, it can break Si–O–Si bonds in the matrix leading to negatively charged defects. The inward orientation of the incoming atom (Xe) causes a momentum transfer in the atomic collision events with a preferential inward component. The ejected oxygen has thus an inward momentum component and, as a consequence, the negatively charged defects are formed at a deeper depth (i.e. at the tail of the Xe depth distribution). The same behaviour was observed by Raman measurements in SiO2
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
implanted with boron[21]. Thus the nanoclusters are probably related to the presence of those negatively charged defects. (3) Recently we have pointed out that Bi implantation under similar implantation condition (fluence, Rp) does not lead to either bubbles or nanoclusters even if the presence of open volume defects and negatively charged defects has been detected (not published). It should be note that the Bi is heavier than Xe and thus induced a higher concentration of defects for the same implanted fluence. Again, it shows that a chemical interaction between the implanted ion and the implantation-induced defects must take place to stabilize the defects and to allow their agglomeration. From these results, we can assume that Xe atoms, which are not involved in bubble formation (i.e. no interaction with vacancy-type defects), can interact with negatively charged defects and stabilize them. If the concentration of negative defects-Xe cluster is high enough, it leads to the agglomeration and the formation of the observed nanoclusters. This interpretation tends to show that the nanoclusters are probably complexes of Xe with negatively charged defects. In this case, their chemical composition is nearly the same as the matrix, and the observed contrast could result from a strained region. The stability of the nanoclusters is weak. They disappeared for annealing temperature P870 K while Xe atoms have not yet desorbed and bubbles are still present. Open-volume-Xe atom complexes are energetically more stable than negatively charged defects-Xe complexes. The absence of nanoclusters after Kr implantation suggests that a threshold density for negatively charged defects is required in order to induce a visible nanocluster in TEM or/and Kr is too mobile to induced a precipitation of negatively charged defects. This point seems to be confirmed by recent observations in samples with Ar ions implanted in same conditions which show bubbles in the whole defected region without any nanoclusters. Our interpretation is also in agreement with our previous results for Ne and He ions implantations [8,9] where neither bubbles or nanoclusters are observed. In these cases, Ne and He are very mobile and cannot be sinks for vacancy-type defects to form bubble precursors.
5. Summary This study provides important data concerning nanostructures (bubbles and nanoclusters) and defects induced by noble gas implantation in amorphous SiO2 thermally grown on Si single crystal. TEM studies show that both Kr and Xe form bubbles in amorphous SiO2 while nanoclusters are formed only with Xe atoms. The size distribution of the bubbles depends the nature of the implanted ion. Bubble formation was shown to result from the interaction of vacancy-type defects and the gas. For Kr implantation, upon thermal annealing, the bubbles are stable only when Kr is still present, whereas Xe desorption at 1370 K results in the formation of stable cavities. Two factors were pointed out for the cavity for-
173
mation: the initial bubble size and the chemical interaction between the implanted ion and the vacancy-type defect. It was demonstrated that, both during implantation and during post-annealing, the bubble growth is mainly due to a coalescence process. For Xe implantation, nanoclusters are induced in a region free of bubbles, at the tail of the Xe profile. It was shown that these nanocluster consists probably of an agglomeration of Xe and negatively defects created by primary knock-on atoms during implantation. The dielectric constant of the as-implanted SiO2 is lowered in both cases from 3.9 to about 1.5. This significant reduction of the dielectric constant makes irradiated/implanted SiO2 a promising candidate for low-k applications in silicon semiconductor technology. Acknowledgments The authors are very grateful to K.J. Abrams for proof reading this paper and for her critical comments. The authors would like to thank H. Assaf-Reda for her contribution in initiating the noble gas implantation study in silicon oxide in the frame of her PhD thesis. This work was granted by ANR French organization through the contract Nanocafon NT05-2_42001. References [1] Materials Challenges for Advanced Nuclear Energy Systems, MRS Bull. 34 (2009) 1–53. [2] D.M. Follstaedt, S.M. Myers, C.A. Petersen, J.W. Medernach, J. Electron. Mater. 25 (1996) 157–164. [3] H. Trinkaus, B. Hollander, S. Rongen, S. Mantl, H.J. Herzog, J. Kuchenbecker, T. Hackbarth, Appl. Phys. Lett. 76 (2000) 3552–3554. [4] M. Bruel, Mater. Res. Innovations 3 (1999) 9–13. [5] E. Oliviero, M.L. David, P.F.P. Fichtner, Phys. Status Solidi C 6 (2009) 1969– 1973. [6] S.E. Donnelly, J.H. Evans, Fundamental Aspects of Inert Gases in Solids, NATO ASI Series B: Physics, vol. 279, Plenum, New York, 1991. [7] J.H. Evans, Nucl. Instr. Meth. B 196 (2002) 125–134. [8] H. Assaf-Reda, Ph.D. Thesis, Orléans University, France, 2006. [9] H. Assaf, E. Ntsoenzok, M.-F. Barthe, M.-O. Ruault, T. Sauvage, S. Ashok, Nucl. Instr. Meth. B 253 (2006) 222–226. [10] J.P. Biersack, J.F. Ziegler, The Stopping and Ranges of Ions in Solids, Lulu Press Co, Morisville, USA, 2008, see also http://www.srim.org for code description. [11] M.H. Loretto, R.E. Smallman, Defect Analysis in Electron Microscopy, Taylor & Francis, Inc. Publishers, London, 1975. [12] V. Raineri, M. Saggio, E. Rimini, J. Mater. Res. 15 (2000) 1449–1477. [13] K. Roselieb, W. Rammensee, H. Büttner, M. Rosenhauer, Chem. Geol. 120 (1995) 1–13. [14] V. Raineri, P.G. Fallica, G. Percolla, A. Battaglia, M. Barbagallo, S.U. Campisano, J. Appl. Phys. 78 (1995) 3727–3735. [15] E. Oliviero, M.L. David, M.F. Beaufort, J. Nomgaudyte, L. Pranevicius, A. Declemy, J.F. Barbot, J. Appl. Phys. 91 (2002) 1179–1186. [16] J.F. Barbot, F. Pailloux, M.L. David, L. Pizzagalli, E. Oliviero, G. Lucas, J. Appl. Phys. 104 (2008) 043526. [17] M.F. Beaufort, L. Pizzagalli, A.S. Gandy, E. Oliviero, D. Eyidi, S.E. Donnelly, J. Appl. Phys. 104 (2008) 094905. [18] E. Marstein, A.E. Gunnaes, U. Serincan, S. Jorgensen, A. Olsen, R. Turan, T.G. Finstad, Nucl. Instr. Meth. B 207 (2003) 424–433. [19] http://jannus.in2p3.fr. [20] H. Asaf, E. Ntsoenzok, M.O. Ruault, O. Kaïtasov, Nucl. Instr. Meth. B, Solid State Phenom 108–109 (2005) 291–296. [21] H. Hosono, N. Matsunami, Nucl. Instr. Meth. B 141 (1998) 566–574.
Contents lists available at ScienceDirect
Microporous and Mesoporous Materials journal homepage: www.elsevier.com/locate/micromeso
Synthesis of mesoporous amorphous silica by Kr and Xe ion implantation: Transmission electron microscopy study of induced nanostructures E. Oliviero a,*, M.-O. Ruault a, B. Décamps a, F. Fotuna a, E. Ntsoenzok b, O. Kaïtasov a, S. Collin a a b
CSNSM, UMR 8609, CNRS/IN2P3-Univ-Paris-Sud, Bâtiment 108, 91405 Orsay Campus, France CEMHTI-CNRS, 3A, rue de la férollerie, 45071 Orléans, France
a r t i c l e
i n f o
Article history: Received 13 November 2009 Received in revised form 18 February 2010 Accepted 20 February 2010 Available online 24 February 2010 Keywords: Mesoporous silica Noble gas implantation Bubbles Cavities Structural investigation
a b s t r a c t Thermally grown amorphous SiO2 was implanted at room temperature with heavy noble gases Kr and Xe in order to create cavities in the oxide and increase its porosity. The implantation energies were chosen in order to have the same implantation depth for both ions. Although both ions induce bubbles in amorphous SiO2, bubble size and spatial distribution depend upon the ion mass. Moreover, Xe implantation leads to the additional formation of ‘‘nanoclusters”. Thermal stability of bubbles/cavities depends on the implanted ion. The nucleation of bubbles and nanoclusters in amorphous SiO2 is discussed in terms of ion mobility, gas–defect interactions, and chemical interaction. Bubble growth is shown to occur by a migration and coalescence process. Ó 2010 Elsevier Inc. All rights reserved.
1. Introduction With the constant miniaturization of electronic devices and consequently the continual increase in the density of multilevel interconnections, material with lower dielectric constant (k) value is required to overcome some of the resulting problems such as transmission delay, power consumption and cross-talk noise. Until now, SiO2 having a dielectric constant of 3.9 has been used as a dielectric material for the interconnections in integrated circuit devices. The International Technology Roadmap for Semiconductors (ITRS) has projected that the dielectric constant of the material should fall below 2 by 2010. There are two possible approaches to achieve this. The obvious one is the use of new materials with lower k, but care has to be taken on their compatibility with microelectronics (interface with silicon, mechanical strength, thermal and etching properties, etc.). The second approach is to decrease the effective dielectric constant of SiO2. This can be achieved by incorporating pores into the material. Air has a dielectric constant of roughly 1.00059, thus increasing the SiO2 layer porosity leads to the reduction of the dielectric constant. The pore structure and its distribution have distinct influences on the basic physical properties of the material. While the reduction in k is desirable, there are also adverse effects such as deterioration of the mechanical properties as porosity increases. There is also the problem of mois-
* Corresponding author. E-mail address: [email protected] (E. Oliviero). 1387-1811/$ - see front matter Ó 2010 Elsevier Inc. All rights reserved. doi:10.1016/j.micromeso.2010.02.015
ture adsorption in the presence of pores at the surface. Therefore, an ideal porous material would consist of a network of small pores embedded into material, with a regular size distribution (uniformly-sized and distributed pores). An appropriate technique to realize such a pores network within SiO2 is rare gas ion implantation. The formation of bubbles/cavities resulting from the agglomeration of noble gas after ion implantation has been extensively studied over several decades in crystalline materials such as metals. Blistering, swelling and embrittlement are unwanted effects in these materials, which are used in fusion or fission reactors [1]. Bubble/cavity formation, blistering and swelling were also studied in semiconductors as smart tools for applications such as impurity gettering [2], strain relaxation [3], thin layer separation [4] and defect engineering [5]. Bubble formation and growth in such materials is a complex phenomenon involving point defect production/ recombination, mobility and agglomeration. Because of their extremely low solubility, noble gases have a strong tendency to segregate into gas–vacancy complexes which may further coarsen and grow forming stable gas-filled cavities. The final defect morphology resulting from noble gas implantation and post-annealing is intricate. It consists of point and extended defects of both vacancy and interstitial types. The vacancy-type defects may contain gas. The main factors influencing the eventual defect morphology depends on the rate at which damage is introduced, the thermal and radiation-enhanced mobility of both gas and defects, and most importantly the ability of the gas to permeate from cavities back into the matrix.
164
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
The extension of these studies from crystalline to amorphous materials such as SiO2 could give insights into the damage accumulation processes and a better understanding of the bubble formation mechanisms. Indeed, the formation and growth mechanisms of cavities are relatively well understood in metals [6] and semiconductors [7]. In the case of SiO2, the nucleation and growth mechanisms of metallic/semiconductor nanoprecipitates has been extensively studied but these mechanisms are not yet well understood for bubbles/cavities [8,9] and require detailed analyses of bubble characteristics including (i) the diffusivity of the gas atoms in SiO2 and (ii) the nature of defects created by implantation. The aim of the work presented in this paper is to understand the processes that could play a role in the nucleation and growth of the bubbles formed by Kr or Xe implantation in amorphous SiO2 (aSiO2). Thus, a detailed TEM study of the induced nanostructures (involving bubbles) was performed after two implantation fluences (3.5 and 5 1016/cm2) and after isochronal annealing up to 1370 K. The as-implanted results are discussed in terms of gas mobility and the interaction between gas atoms and vacancy-type defects or negatively charged defects. The structural evolutions during annealing are explained using the migration and coalescence mechanism. 2. Experimental details Thermal amorphous SiO2 layers were grown by heating n-type silicon wafers at 1370 K in ambient air. Samples of two different layer thicknesses (220 nm and 2.3 lm) were used in this study. The samples were implanted, at room temperature, with 220 keV Kr ions or 300 keV Xe ions at the same fluences (3.5 and 5 1016 ions/cm2). The implantation energies were chosen, according to SRIM (Stopping and Range of Ions in Matter) simulations [10], in order to have the same projected range (Rp) of 120 nm for both ions. Ion flux was kept as low as 1– 1.5 1012 ion/cm2 s to avoid any temperature increase during implantation. After implantation, samples were annealed at temperatures ranging from 670 to 1370 K for 1 h under nitrogen atmosphere. The as-implanted and post-annealed samples were characterised by cross-sectional transmission electron microscopy (XTEM) and compared with the ion distributions obtained previously [9] by Rutherford Backscattering Spectrometry (RBS) in specimens implanted under the same experimental conditions. The measurements were performed using a FEI CM12 TEM and a Tecnai G2 20 TEM. The cross-section samples for XTEM were mechanically thinned by means of a tripod. For a better comprehension of the XTEM results, a few preliminary remarks are needed before a detailed presentation of the results. (1) Amorphization of the Si substrate occurs when implanting in the 220 nm thick SiO2 layer. In fact, the ion range is up to about 250 nm for both implanted ions. Thus, a Si band just behind the interface is amorphized due to the induced mixing phenomena at the SiO2/Si interface and to the induced irradiation defects. During post-annealing P800 K, the amorphous band recrystallizes leaving defects such as dislocations or rod-like defects.
(2) The thickness of the 220 nm thick SiO2 layer was controlled before and after implantation by RBS technique. No swelling of the layer due to noble gas implantation was observed in the bulk specimens. However, a widening of the SiO2 layer was observed by TEM (Fig. 1). This enlargement occurs during the mechanical thinning process and is due to an increase of the SiO2 ductility. The change in ductility is induced by the implantation/irradiation. In the following, the deformation of the SiO2 layer will be taken into account. (3) For non-spherical nanostructures, the size will be given as the equivalent diameter for spherical shape having the same area. (4) The density depth distribution of nanostructures is obtained by counting them in 10–20 nm width bands only if the size of these nanostructures is small enough (i.e. at least half of the band width). In our study nanostructure sizes are mostly quite large so that it is not possible to give the bubble density distribution as a function of depth. The same is true for the distribution of the bubble centres as a function of depth. It seems more accurate to evaluate the projected area of the nanostructures. A normalization is done by dividing the results by the total projected area of all the nanostructures present in the layer. It should be noted that, due to the amorphous structure of SiO2, it is difficult to assess the thickness of the XTEM specimen in the studied zones. In fact, the XTEM specimen thickness can be evaluated only by the electron transparency of the sample (see for example Fig. 1). The accuracy of such an evaluation is not better than 30–50 nm. However, each projected area depth distribution is obtained on a sufficiently small region, so that the sample thickness variation can be estimated at about 20% inside an individual depth distribution. 3. Results Both Kr and Xe implanted samples clearly show the formation of nanostructures in the implanted region. A detailed description of the observed nanostructures is given in the following. 3.1. Xe implantation 3.1.1. As-implanted Figs. 2 and 3 show the nanostructures induced in amorphous SiO2 by 300 keV Xe implantation at room temperature for fluences of 3.5 and 5 1016 Xe/cm2. Two types of nanostructures are observed: (i) Nanostructures without Fresnel fringes (Fresnel fringes are characteristic of bubbles/cavities [11]). These nanostructures exhibit a black contrast in overfocus conditions, which tends to vanish in the underfocus condition (Fig. 3a–d). These nanostructures could not therefore be identified as bubbles or cavities. From this point on, these nanostructures will be termed as ‘‘nanoclusters”. (ii) Nanostructures identified as bubbles, i.e. exhibiting Fresnel fringes at the interface whose contrast reverse from black to white from under to overfocus conditions [11]
Fig. 1. An assembly of brightfield image showing a TEM cross-section of an as-implanted sample. The white bars illustrate the size of the SiO2 layer before implantation (220 nm) along the cross-section. The thickness of the TEM sample increases from the left to the right.
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
165
Fig. 2. Evolution during TEM observations (i.e. 120 keV electron beam) of the nanostructures created in a 2 lm thick SiO2 with 300 keV Xe ions up to 5 1016 Xe/cm2 in underfocus conditions. The observation time (increasing from left to right) is reported on the micrographs.
Fig. 3. XTEM bright field images of as-implanted 220 nm-SiO2 with 300 keV Xe ions up to 3.5 1016 Xe/cm2 in (a) overfocus and (b) underfocus conditions, and up to 5 1016 Xe/cm2 in (c) overfocus and (d) underfocus conditions. (e and f) Enlarged area for detailed examination of nanostructures: same experimental conditions as in (c and d) showing the nanostructure band (between the black lines) observed in overfocus conditions vanishes in underfocus conditions. Note also that, among three dark nanostructures circled in underfocus conditions, only two nanostructures exhibit the inverse characteristic contrast (white) for nanobubble/cavity at underfocus condition, the contrast of the last vanishes.
(Fig. 3a–d). The contrast inside the large bubbles is mainly grey in overfocus conditions and becomes lighter – but not white – in the underfocus condition. This behaviour is probably due to the heavy mass of Xe present inside the bubbles. Some of the large bubbles exhibit a hole contrast (white con-
trast in under and overfocus conditions) due to specimen cutting during sample thinning. It should be noted that those few holes/cavities are not stable under the TEM electron beam as it was shown in a time evolution of the nanostructures under the electron beam (Fig. 2). This effect is not
166
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
yet fully understood and needs further extensive study to elucidate on this. In this work, all the size studies were done on pictures obtained within 1 or 2 min to avoid any bubble evolution under the electron beam. For both fluences (Fig. 3a–d) a band of large bubbles is surrounded by smaller nanostructures mainly located towards the Si interface. The diameter of the nanostructures ranges between 2 and 10 nm with an average diameter of 6 nm. Even if a few small nanostructures (diameter < 3 nm) exhibit the reversal contrast of nanobubbles – black to white for over to underfocus conditions – most of them are identified as nanoclusters. Fig. 3e and f show an example of the study done for the TEM determination of the small nanostructures whose diameter was 63 nm. It is acknowledged that their size (especially for the largest ones) might be overestimated because of superimposition and their black contrast. As a consequence, the nanocluster concentration might be underestimated. The bubbles exhibit a random size distribution within the bubble band (e.g. Fig. 3b and d) and their average diameter does not significantly change (21 ± 2 nm) with increasing fluence (Fig. 4a and b). However, the shape of their distribution does change. A few larger bubbles (50–90 nm in diameter) appear at 5 1016 Xe/cm2. For this fluence, the peak of the distribution (at 14 ± 2 nm) is narrower than at 3.5 1016 Xe/cm2. In fact, bubbles of 20–25 nm in diameter disappear at the expense of the larger bubbles possibly due to bubble coalescence. This point seems to be confirmed by the fact that a non-negligible number of elongated bubbles are observed. Fig. 5a–d shows, for both implantation fluences (3.5 and 5 1016/cm2), the normalized1 depth distribution of the bubbles and the normalized depth distribution of the nanoclusters. We would like to point out that bubble concentration is about five times greater than the nanocluster concentration at a fluence of 3.5 1016/ cm2 and of an order of magnitude greater at 5 1016/cm2. Thus the proportion of bubbles increases with Xe concentration in a-SiO2. At 3.5 1016/cm2, the bubble depth distribution follows the simulated Xe profile and is centred at 110–130 nm (corresponding to the Rp of the Xe calculated with SRIM) and the nanoclusters stand at the tail-end of the Xe distribution (Fig. 5a). At 5 1016/ cm2, the bubble/cavity depth distribution is shifted towards the free surface and the peak (80–120 nm) corresponds to the maximum of the simulated damage distribution (Fig. 5b). This is coherent with Xe depth profiles, previously obtained by RBS [9]. Indeed, at 3.5 1016/cm2, the Xe RBS profile exhibited a wide peak extended from 100 to 140 nm (not shown here). By increasing the fluence to 5 1016/cm2, the Xe RBS profile shifts towards a lower depth and becomes narrower (from 70 to 90 nm). Moreover, a shoulder peak appeared at a depth of 110–140 nm (reported in Ref. [9]). The latter could be associated to the shoulder observed in the bubble area distribution (Fig. 5d). In Fig. 5d, focusing on the total nanostructures profile, it can be noticed that the shoulder is only weakly increased when adding the nanoclusters. Moreover, the nanocluster distribution peak (170–200 nm) does not shift with increasing fluence. The shoulder is thus ascribed to bubbles.
3.1.2. Thermal annealing One of the main effects is the disappearance of the nanoclusters from annealing temperatures P870 K. The other striking effect is that for annealing at 1370 K, cavities remain in the sample while the Xe has already exodiffused [9]. These cavities exhibit quasispherical shapes (Figs. 6e and 7d). 1
Normalisation is carried out by dividing by the total nanostructure area.
Fig. 4. Normalized size distribution of the bubbles created by 300 keV Xe implantation in silicon dioxide for a fluence of (a) 3.5 1016/cm2 and (b) 5 1016/cm2.
Although the general feature of the nanostructure evolution during post-annealing is fairly similar between 3.5 and 5 1016 Xe/cm2, the detailed statistic analysis shows some differences (Figs. 6–9). At a fluence of 3.5 1016 Xe/cm2 the maximum bubble diameter observed is 60 nm. Both the bubble/cavity shape and the size distribution do not significantly change for annealing up to 1370 K. The mean diameter remains constant. The bubble depth distribution does not significantly change up to 1020 K. After 1370 K annealing, the cavities located near the SiO2/Si interface progressively disappear leading to a sharp distribution centred at about 100 nm (Fig. 8). At a fluence of 5 1016 Xe/cm2 the bubble size distribution and the bubble mean diameter do not significantly change up to 1370 K. However, the number of elongated bubbles slightly increases (e.g. Fig. 7c). The bubble depth distribution does not significantly change up to 1020 K. In particular, the shoulder remains (Fig. 9) despite the disappearance of the nanocluster peak. After a 1370 K anneal, the depth distribution tightens between 50 and 150 nm and the cavities become predominantly spherical. The narrowing of the depth distribution is however not as significant as for the lower fluence (see Figs. 8 and 9). 3.2. Kr implantation 3.2.1. As-implanted As already mentioned in the case of Xe implantation, TEM study of Kr implantation in a-SiO2 also faces the problem of bubble evolution under the electron beam. An example of this evolution is shown in Fig. 10. Primarily, when the electron beam is applied, the bubbles appear to grow (see Fig. 10 after 2 and 4 min of observation under the electron beam) and a few bubbles exhibit deformed shapes. Between 10 and 15 min of observation (Fig. 10), the bubbles return to a spherical shape and their contrast faints. This effect was observed whichever electron beam voltage was used (from 80 to 200 kV) and will be studied in more detail outside the remit of this paper. Therefore, exposure to the electron beam was limited to less than 4 min during the observation of the bubble
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
167
Fig. 5. Normalized area profile for bubbles (bars) and ‘‘nanoclusters” (full-line) [Units on the left] compared with SRIM simulation of Xe distribution profile and target displacements [Units on the right] created by 300 keV Xe implantation in 220 nm-SiO2 for a fluence of (a) 3.5 1016/cm2 and (b) 5 1016/cm2. In order to clearly point out the shape of the area distributions we have reported on the right the distributions with a line between points for the distributions of bubbles and total nanostructrures (bubbles + nanoclusters), respectively, for fluences: (c) 3.5 1016/cm2 and (d) 5 1016/cm2.
layer. The Kr implantation-induced nanostructures form on a band near the free surface. The width of the nanostructure band fluctuates throughout the implanted area with an average width of around 150–180 nm. The SiO2 layer thickness has no effect on the band width (Fig. 11a and b) neither does the Kr fluence (Fig. 11a and c). For both implanted fluences (3.5 and 5 1016 Kr/cm2) the nanostructures were identified as bubbles as they exhibit the contrast characteristic of bubbles/cavities (Fig. 11c and d) and Kr is confirmed by RBS to be trapped in the sample. No nanoclusters are observed. The size repartition is quasi-homogeneous within the band (Fig. 11a–d) and the bubble diameter ranges from 4 to 30 nm (Fig. 12a–c), with an average diameter of 10–12 nm. This is independent of the fluence and of the SiO2 layer thickness. The depth distribution also does not significantly change for the different fluences or the different SiO2 layer thicknesses (Fig. 13a and b): it clearly displays a plateau shape. This differs from the Kr distribution predicted by SRIM [10] which exhibits a Gaussian profile with a peak at 120–130 nm (Fig. 13a–c). The bubbles extend preferentially on the deeper side of the theoretical Kr profile predicted by SRIM (Fig. 13a and b). For the 220 nm thick SiO2 layer (Fig. 13c), the bubble layer is formed very close to the SiO2/Si interface. If the SiO2/Si interface was absent (as in the 2.3 lm-thick SiO2) the bubble layer would extend deeper up to 240 nm (Fig. 13b). However, bubbles are not observed either at the 220 nm-SiO2/Si interface or in the Si substrate (Figs. 11b and 13c). Instead, the bubbles seem to extend towards the free surface corresponding to the damaged region. It illustrates the ease for the Kr to diffuse and agglomerate in the SiO2 compared to silicon.
These results are in agreement with the RBS results [8] at 5 1016 Kr cm 2 reported in Fig. 13c which show that the Kr profile exhibits a plateau instead of the peak predicted by SRIM calculations (see Fig. 13c). The plateau shape corresponds to Kr diffusion within the implanted region. SRIM calculations show that, without diffusion, damaged region is very close in depth to the Kr distribution, so it would be useful to separate these two profiles, in order to study the influence of the defects on the observed plateau distribution. Future work entails a detailed study of a possible enhanced diffusion by defects due to irradiation. This could be achieved by increasing either the Kr energy or the damage rate (dpa) in a dual-beam experiment. 3.2.2. Thermal annealing After annealing at 670 K, XTEM observations in 2.3 lm SiO2 show the presence of bubbles/cavities at the same depth region as after Kr implantation (Fig. 14a). The average bubble/cavity diameter remains at about 12 ± 2 nm and the size distribution does not significantly change (Figs. 15 and 16). Only a few larger bubbles/cavities are observed (less than 5%) (Fig. 15c). They have an elongated shape (Fig. 14b) which indicate that they probably result from a coalescence process. These results are coherent with previous RBS measurements [8] which show that after an anneal at 670 K, the Kr distribution is still a plateau although a strong decrease in the Kr concentration (by a factor of about 3) is observed. After a 970 K annealing, the bubbles/cavities are no longer present inside the SiO2 layer. This disappearance is in agreement with the exodiffusion of Kr observed by RBS above 870 K [8]. Thus, no cavities are detected within the material in the absence of Kr.
168
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
Fig. 6. XTEM images showing the evolution of bubbles/cavities obtained by Xe implantation in 220 nm-SiO2 at a fluence of 3.5 1016/cm2 after 1 h annealing at (a) 670 K, (b) 970 K, (c) 1020 K, (d) 1170 K and (e) 1370 K.
4. Discussion 4.1. Nucleation and growth mechanisms of bubbles The nucleation of bubbles in metals [6] or in semiconductors [12] is attributed to the interaction of gas atom with vacancies. In amorphous SiO2 the concept of vacancy is difficult to address but bubble nucleation can be seen as the interaction of gas atom with some free volume. We have previously [9] suggested from RBS and PAS measurements that the interaction of Xe atoms with vacancy-type defects (i.e. oxygen vacancies – oxygen deficiency centres, or defects with large open volume) leads to complexes that could be precursors of bubbles. This interpretation is coherent with the fact that the bubble distribution and the Xe profile measured by RBS (§ III.A1) closely follow the simulated implantation-induced damage distribution for the highest fluence. For the lowest fluence, both the Xe profile obtained by RBS and the bubble distribution are closer to the simulated Xe profile. In fact, for the implantation energy used, the simulated Xe profile and the induced damage profiles overlap in the region where bubbles are formed. Bubbles are thus formed where the ratio Xe/defect is maximum. The fact that,
at the highest fluence the bubble and the Xe profile shift towards the damage profile shows that above a threshold concentration, the induced damage act as sinks for Xe atoms leading to the bubble formation centred at the maximum of the simulated damage distribution. The interaction of implanted gas atoms with vacancytype defects is also coherent in the case of Kr, since, even if a plateau shaped distribution is observed, bubbles are formed at about the depth of the simulated ion distribution (Fig. 13). These results could be related to the mobility of Xe and Kr in a-SiO2. Actually, Kr is qualitatively more mobile than Xe in a-SiO2 [13]. However, Kr does not follow the simulated implantation damage profile. In fact, the damage concentration (dpa) for Kr at the highest fluence is equivalent to that of the Xe implantation at the lowest fluence where no diffusion occurs. So the induced damage is below the threshold concentration, and do not act as sink for Kr, but can enhanced its diffusion. Nevertheless, when the fluence increases, a small shift of the bubble distribution occurs towards the implantation damage profile, indicating that damage concentration is getting closer to the threshold. Implanting to a higher dose (>5 1016 Kr/cm2) should lead to the formation of a bubble band centred on the simulated damage peak. The formation of bubbles
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
169
Fig. 7. XTEM images showing the evolution of bubbles/cavities obtained by Xe implantation in 220 nm-SiO2 at a fluence of 5 1016/cm2 after 1 h annealing at (a) 670 K, (b) 1020 K, (c) 1170 K, (d) 1370 K.
Fig. 8. Normalized area profiles for bubbles (bars) and ‘‘nanoclusters” (dotted-line) after a 300 keV Xe implantation at 3.5 1016 Xe/cm2: as-implanted, and after an isochronal anneal of one hour between 670 and 1370 K.
Fig. 9. Normalized area profile for bubbles (bars) and ‘‘nanoclusters” (dotted-line) after a 300 keV Xe implantation at 5 1016 Xe/cm2: as-implanted, and after an isochronal annealing of one hour between 670 and 1370 K.
all along the Kr profile and even beyond, where very small amounts of Kr and damage are expected, is proof that the complexes Kr-vacancy-type defect are mobile and are able to spread resulting in the observed plateau distribution.
4.1.1. Ostwald ripening versus migration-coalescence Two different mechanisms, Ostwald ripening (OR) and migration-coalescence (MC) are generally mentioned to explain the growth of bubbles. In this part of the discussion we will compare both mechanisms and determine which is responsible for bubble growth in amorphous SiO2.
170
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
Fig. 10. The evolution during TEM observations (i.e. 200 keV electron beam) of the nanostructures created in a 2 lm thick SiO2 with 300 keV Xe ions up to 5 1016 Kr/cm2 in underfocus conditions. The observation time (increasing from left to right) is reported on the micrographs. Examples of bubbles which exhibit deformed shapes are highlighted with an arrow.
Fig. 11. XTEM bright field images of as-implanted SiO2 with 220 keV Kr ions: (a) fluence of 3.5 1016/cm2 in 2.3 lm thick SiO2, the insert shows an enlargement of the bubble band/(overfocus condition), (b) fluence of 5 1016/cm2 in 220 nm thick SiO2 (underfocus condition), (c and d) 2.3 lm thick SiO2, respectively, in overfocus and underfocus conditions.
The formation of bubbles in amorphous material seems to be quite different than in crystalline materials. It is well known that noble gas implantation into crystalline materials such as Si [14], SiC [15], GaN [16] leads to the formation of tiny bubbles of 1– 3 nm in size. Recently it was shown that He implantation in amorphous silicon leads to the formation of significantly larger bubbles compared to those observed in crystalline Si under identical implantation conditions [17]. In a-SiO2, Kr implantation leads to the formation of bubbles with a mean diameter of 10 nm and the Xe implantation leads to the formation of bubbles with a mean diameter of 20 nm. Moreover, during thermal annealing we do not observed any significant growth of the bubbles. This confirms that growth already occurs during ion implantation. From our results, we can state that bubble mobility in a-SiO2 almost certainly results from migration and coalescence mechanism limited by a surface migration process. It can then be assumed, qualitatively, that the surface diffusivity in amorphous SiO2 is relatively low. This allows bubble coarsening directly during implantation explaining the observation of large bubbles directly after Kr and Xe implanta-
tion into a-SiO2. Small bubbles could be mobile at room temperature (and thus during the implantation) and could grow by thermally or radiation-induced induced motion and coalescence. In addition, this would explain the rather non-spherical aspect of the bubbles. This would be the result of a recent coalescence of spherical bubbles as stated in helium-implanted silicon [7]. Moreover, the bubbles formed during implantation are not stable. If an excitation is provided such as electron irradiation, the morphology of the bubbles can change thus illustrating again the relatively ease for migration and coalescence. Through thermal annealing, the bubbles formed by Kr implantation are not stable when Kr totally exodiffuses. However, in the case of Xe, cavities are found to be stable even after 1370 K where Xe is no longer present. Two factors seem to influence the formation of cavity. Firstly, the bubble size is significantly smaller for as-implanted Kr than for as-implanted Xe. This point suggests a possible threshold size for which the cavities are stable through thermal annealing. Secondly, some chemical mechanisms are necessary to lead to cavity formation. Indeed, it was shown by the
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
171
Fig. 14. XTEM bright field in underfocus condition of a 2.3 lm SiO2 layer implanted with 220 keV Kr ions up to a fluence of 5 1016/cm2 and post-annealed 1 h at 670 K. (a) View of the whole band in a thick part of the sample and (b) example of a bubble which is probable result of a coalescence process (indicated by a white arrow).
Fig. 12. Size profiles of bubbles created by 220 keV Kr, (a) at 3.5 1016/cm2 for 2.3 lm thick SiO2, (b) at 5 1016 /cm2 for 2.3 lm thick SiO2 and (c) at 5 1016/ cm2 for 220 nm thick SiO2.
work of Marstein et al. [18] that the presence of the gas is not required to create cavities. They report the formation of cavities in aSiO2 after 1270 K post-annealing of samples containing Ge nanoprecipitates formed by ion implantation although, in this case, the cavity precursors are complexes Ge-vacancy-type defect. They assumed that during annealing, firstly, the Ge precipitates grow and then secondly, the cavities are formed by the out diffusion of Ge from the crystalline Ge nanoprecipitates. So in this case, no gas is required to form cavities in a-SiO2 after annealing. It shows again that above a critical size, vacancy-type defect agglomeration
Fig. 13. Normalized depth distribution of bubbles area, after 220 keV Kr ion implantation in 2.3 lm SiO2 up to (a) 3.5 1016/cm2, (b) 5 1016/cm2 and (c) in 220 nm-SiO2 up to 5 1016/cm2. (c) Also shows the Kr concentration previously obtained by RBS [8]. These distributions are compared to the ion range distribution and the ‘‘defects” distribution in SiO2 obtained by SRIM simulation.
172
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
cess. Thus, there must be an alternative mechanism leading to cavity formation. This mechanism has to involve chemical interaction between vacancy-type defects and implanted atoms. Through this interaction, the implanted atom stabilizes the vacancy-type defects. In order to better understand these effects, in situ TEM during ion implantation experiments are necessary in an attempt to detect any bubble motion during implantation. Moreover, the stability of bubbles under electron/ion beam will be investigated. This will be done on the dual-beam TEM of JANNuS facility at Orsay [19]. By assimilating cavities to pores, the porosity created by rare gases implantation can be estimated from the TEM results. The porosity is found to increases directly after implantation to about 2%. After thermal annealing, the porosity of the SiO2 layer is found higher than 10%. It is worthwhile to underline that this porosity is relative to the thickness of the SiO2 layer and that a higher level of porosity can be reached by achieving multi-energy implantation to create a larger cavity layer. C–V measurements [20] performed on Metal Oxide Semiconductor (MOS) show that the dielectric constant k is reduced in both cases from 3.9 for the unimplanted SiO2 to about 1.5 for the as-implanted samples. The decrease of k is attributed both to the increase of porosity and to the reduction of SiO2 polarizability (modification of the dipoles and/or breaking of polarizable Si–O bonds). Indeed, after thermal annealing, the dielectric constant increase slightly while the porosity is found to increase. Fig. 15. 2.3 lm thick SiO2 implanted with 220 keV Kr ions up to a fluence of 5 1016/cm2. Comparison of the size profile (a) as-implanted and (b) post-annealed for 1 h at 670 K. In (c), the tail of the distribution (for size P 20 nm) is enlarged (i.e. the normalized ratio scale was a magnitude lower).
Fig. 16. 2.3 lm thick SiO2 implanted with 220 keV Kr ions up to a fluence of 5 1016/cm2. Comparison of the normalized depth distribution of bubbles area: (a) as-implanted and (b) post-annealed for 1 h at 670 K.
can be stable. The formation of cavities after annealing is, however, not only related to the vacancy-type defect concentration created in the matrix, since we recently showed that a Bi implantation at same conditions of fluence and penetration depth in the matrix does not lead to any cavity formation after a 1270 K annealing pro-
4.2. The nature of the nanoclusters In the case of Xe implantation, nanocluster formation was observed. The first question that arose was what is the possible composition of nanoclusters? To tentatively determine directly their composition, Electron Energy Loss Spectroscopy (EELS) measurements were carried out. However, it was not entirely successful due to the very small diameter (few nm) of these nanoclusters in comparison with the probe size limit (P10 nm). However, from the following points, a probable composition of the nanocluster can be deduced. (1) We have shown above that the proportion of bubbles versus nanoclusters increases with the Xe concentration and that the Xe shoulder at the end of the distribution is clearly more associated to the bubble distribution than to the nanoclusters. Thus, the nanoclusters are not directly related to the Xe concentration. If Xe is involved in the nanoclusters, it is a weak contribution. (2) It was shown by Positon Annihilation Spectrometry measurements [9] that after Xe implantation, the ‘‘void” volume (i.e. oxygen vacancies or defects with large open volume) increases in the bubble region. This ‘‘void” volume increase concurrently with the fluence. Meanwhile, negatively charged defects (Si–O , Si–O–O , O2 ) were observed at the tail-end of Xe depth distribution. No effect of the fluence was observed on their density. The formation of negatively charged defects is due to the ejection of interstitial oxygen (primary knock-on atom) from the implanted region. If this ejected oxygen has enough energy, it can break Si–O–Si bonds in the matrix leading to negatively charged defects. The inward orientation of the incoming atom (Xe) causes a momentum transfer in the atomic collision events with a preferential inward component. The ejected oxygen has thus an inward momentum component and, as a consequence, the negatively charged defects are formed at a deeper depth (i.e. at the tail of the Xe depth distribution). The same behaviour was observed by Raman measurements in SiO2
E. Oliviero et al. / Microporous and Mesoporous Materials 132 (2010) 163–173
implanted with boron[21]. Thus the nanoclusters are probably related to the presence of those negatively charged defects. (3) Recently we have pointed out that Bi implantation under similar implantation condition (fluence, Rp) does not lead to either bubbles or nanoclusters even if the presence of open volume defects and negatively charged defects has been detected (not published). It should be note that the Bi is heavier than Xe and thus induced a higher concentration of defects for the same implanted fluence. Again, it shows that a chemical interaction between the implanted ion and the implantation-induced defects must take place to stabilize the defects and to allow their agglomeration. From these results, we can assume that Xe atoms, which are not involved in bubble formation (i.e. no interaction with vacancy-type defects), can interact with negatively charged defects and stabilize them. If the concentration of negative defects-Xe cluster is high enough, it leads to the agglomeration and the formation of the observed nanoclusters. This interpretation tends to show that the nanoclusters are probably complexes of Xe with negatively charged defects. In this case, their chemical composition is nearly the same as the matrix, and the observed contrast could result from a strained region. The stability of the nanoclusters is weak. They disappeared for annealing temperature P870 K while Xe atoms have not yet desorbed and bubbles are still present. Open-volume-Xe atom complexes are energetically more stable than negatively charged defects-Xe complexes. The absence of nanoclusters after Kr implantation suggests that a threshold density for negatively charged defects is required in order to induce a visible nanocluster in TEM or/and Kr is too mobile to induced a precipitation of negatively charged defects. This point seems to be confirmed by recent observations in samples with Ar ions implanted in same conditions which show bubbles in the whole defected region without any nanoclusters. Our interpretation is also in agreement with our previous results for Ne and He ions implantations [8,9] where neither bubbles or nanoclusters are observed. In these cases, Ne and He are very mobile and cannot be sinks for vacancy-type defects to form bubble precursors.
5. Summary This study provides important data concerning nanostructures (bubbles and nanoclusters) and defects induced by noble gas implantation in amorphous SiO2 thermally grown on Si single crystal. TEM studies show that both Kr and Xe form bubbles in amorphous SiO2 while nanoclusters are formed only with Xe atoms. The size distribution of the bubbles depends the nature of the implanted ion. Bubble formation was shown to result from the interaction of vacancy-type defects and the gas. For Kr implantation, upon thermal annealing, the bubbles are stable only when Kr is still present, whereas Xe desorption at 1370 K results in the formation of stable cavities. Two factors were pointed out for the cavity for-
173
mation: the initial bubble size and the chemical interaction between the implanted ion and the vacancy-type defect. It was demonstrated that, both during implantation and during post-annealing, the bubble growth is mainly due to a coalescence process. For Xe implantation, nanoclusters are induced in a region free of bubbles, at the tail of the Xe profile. It was shown that these nanocluster consists probably of an agglomeration of Xe and negatively defects created by primary knock-on atoms during implantation. The dielectric constant of the as-implanted SiO2 is lowered in both cases from 3.9 to about 1.5. This significant reduction of the dielectric constant makes irradiated/implanted SiO2 a promising candidate for low-k applications in silicon semiconductor technology. Acknowledgments The authors are very grateful to K.J. Abrams for proof reading this paper and for her critical comments. The authors would like to thank H. Assaf-Reda for her contribution in initiating the noble gas implantation study in silicon oxide in the frame of her PhD thesis. This work was granted by ANR French organization through the contract Nanocafon NT05-2_42001. References [1] Materials Challenges for Advanced Nuclear Energy Systems, MRS Bull. 34 (2009) 1–53. [2] D.M. Follstaedt, S.M. Myers, C.A. Petersen, J.W. Medernach, J. Electron. Mater. 25 (1996) 157–164. [3] H. Trinkaus, B. Hollander, S. Rongen, S. Mantl, H.J. Herzog, J. Kuchenbecker, T. Hackbarth, Appl. Phys. Lett. 76 (2000) 3552–3554. [4] M. Bruel, Mater. Res. Innovations 3 (1999) 9–13. [5] E. Oliviero, M.L. David, P.F.P. Fichtner, Phys. Status Solidi C 6 (2009) 1969– 1973. [6] S.E. Donnelly, J.H. Evans, Fundamental Aspects of Inert Gases in Solids, NATO ASI Series B: Physics, vol. 279, Plenum, New York, 1991. [7] J.H. Evans, Nucl. Instr. Meth. B 196 (2002) 125–134. [8] H. Assaf-Reda, Ph.D. Thesis, Orléans University, France, 2006. [9] H. Assaf, E. Ntsoenzok, M.-F. Barthe, M.-O. Ruault, T. Sauvage, S. Ashok, Nucl. Instr. Meth. B 253 (2006) 222–226. [10] J.P. Biersack, J.F. Ziegler, The Stopping and Ranges of Ions in Solids, Lulu Press Co, Morisville, USA, 2008, see also http://www.srim.org for code description. [11] M.H. Loretto, R.E. Smallman, Defect Analysis in Electron Microscopy, Taylor & Francis, Inc. Publishers, London, 1975. [12] V. Raineri, M. Saggio, E. Rimini, J. Mater. Res. 15 (2000) 1449–1477. [13] K. Roselieb, W. Rammensee, H. Büttner, M. Rosenhauer, Chem. Geol. 120 (1995) 1–13. [14] V. Raineri, P.G. Fallica, G. Percolla, A. Battaglia, M. Barbagallo, S.U. Campisano, J. Appl. Phys. 78 (1995) 3727–3735. [15] E. Oliviero, M.L. David, M.F. Beaufort, J. Nomgaudyte, L. Pranevicius, A. Declemy, J.F. Barbot, J. Appl. Phys. 91 (2002) 1179–1186. [16] J.F. Barbot, F. Pailloux, M.L. David, L. Pizzagalli, E. Oliviero, G. Lucas, J. Appl. Phys. 104 (2008) 043526. [17] M.F. Beaufort, L. Pizzagalli, A.S. Gandy, E. Oliviero, D. Eyidi, S.E. Donnelly, J. Appl. Phys. 104 (2008) 094905. [18] E. Marstein, A.E. Gunnaes, U. Serincan, S. Jorgensen, A. Olsen, R. Turan, T.G. Finstad, Nucl. Instr. Meth. B 207 (2003) 424–433. [19] http://jannus.in2p3.fr. [20] H. Asaf, E. Ntsoenzok, M.O. Ruault, O. Kaïtasov, Nucl. Instr. Meth. B, Solid State Phenom 108–109 (2005) 291–296. [21] H. Hosono, N. Matsunami, Nucl. Instr. Meth. B 141 (1998) 566–574.