Formation of nanoclusters on silicon from carbon deposition

Applied Surface Science 226 (2004) 191–196

Formation of nanoclusters on silicon from carbon deposition V. Palermo*, D. Jones ISOF, Istituto per la Sintesi Organica e Fotoreattivita’, CNR, via Gobetti 101, I-40129 Bologna, Italy

Abstract Changes in the structure of silicon surfaces can be induced by adsorption of carbon-containing molecules followed by thermal treatments. Clean Si(111) surfaces, prepared in vacuum and exposed to different adsorbants such as methanol or carbon monoxide, change their structures with the formation of self-organised nanostructures (15–50 nm diameter) after suitable UHV annealing procedures. Evolution of the size and density per unit area over different heating periods indicates that the structures are nucleated by carbon atoms present on the surface while their growth derives from mobile surface silicon atoms during the annealing process. Methanol adsorbs dissociatively on silicon at room temperature thus leading to a high density of nucleation centres, but when the process is applied to partially oxide-masked silicon surfaces using CO as adsorbant the nanostructures form preferentially at the Si/SiO2 interface around the mask border thus offering the possibility to grow more ordered selforganised nanoscale patterns. Monte Carlo simulations of this process correlate well with STM measurements. # 2003 Elsevier B.V. All rights reserved. PACS: 68.35.-p; 61.16.Ch; 81.07.-b; 81.16.Rf Keywords: STM; Nanostructures; Silicon; Adsorption

1. Introduction The continuous drive for extreme miniaturisation and the vicinity of the physical limits of classical lithographic techniques in the microelectronics industry has led to increased efforts to find novel fabrication techniques. The use of self-organisation phenomena is one of the most promising, simple and economic methods for producing nanometric structures on surfaces. Small lines and ordered structures can be selfassembled on surfaces and, in particular, on silicon [1,2]. It has been previously shown how contaminant molecules can adsorb onto silicon surfaces and nucle-

*

Corresponding author. Tel.: þ39-051-639-8336; fax: þ39-051-639-8349. E-mail address: [email protected] (V. Palermo).

ate the growth of small, protruding nanostructures (‘‘nanoislands’’) on the otherwise flat silicon surface during UHV annealing [3]. When surface oxide layers are removed from silicon by UHV thermal annealing at 700–800 8C, the process begins through the creation of small oxide-free areas called ‘‘voids’’ which enlarge radially until the whole surface is oxide free [4–7]. The above-mentioned nanoislands grow at void nucleation points with concomitant radial enlargement of the voids during annealing. Carbon-containing compounds present in UHV systems as well as extrinsic surface impurities have been cited as the source of this contamination [3,8,9]. The nanoislands form over the whole surface and their formation mechanism is still not clear. This paper compares the results obtained by adsorbing carbon monoxide and methanol on silicon at room temperature to nucleate nanoisland growth. Many

0169-4332/$ – see front matter # 2003 Elsevier B.V. All rights reserved. doi:10.1016/j.apsusc.2003.11.021

192

V. Palermo, D. Jones / Applied Surface Science 226 (2004) 191–196

different molecules were tested, but only the most significant results are reported here. The molecules are adsorbed on clean silicon surfaces prepared in vacuum, and the nanoisland growth process is followed using STM. Furthermore, samples partially masked by an oxide layer are used to have a selective growth of the nanoislands into voids of the oxide layer. Finally, Monte Carlo simulations are used to reproduce the island growth process.

temperature but, after degassing overnight at 150 8C, heated for 10 min at 800 8C to create the voids [3]. These samples were then directly exposed to CO in the vacuum system. Further annealing at 800 8C for 10 min led to the radial expansion of the voids with loss of the oxide and the creation of the first series of nanoislands around the original void perimeter. Repeating this process led to the formation of a second, larger concentric circular arrangement of nanoislands.

2. Experimental 3. Results and discussion Samples of Si(111) (p-doped, 0.015 O cm) were degassed in vacuum for 5 h at 650 8C and then flashed repeatedly at 1200 8C to clean the surface. Surface morphology was observed in situ between heating cycles with an Omicron UHV-STM operating at a base pressure of 1011 Torr. Typical measurement parameters were sample bias þ1.2 V, tunnelling current 50 pA. The adsorbant gases were introduced into the vacuum system using a leak valve and the pressure controlled with an ion gauge. The molecules tested were oxygen, water, methanol, carbon monoxide, carbon dioxide and methane. During gas adsorption the surface was observed periodically with STM. Between the measurements, the STM tip was retracted to avoid shadowing effects. After gas adsorption, the silicon was resistively heated and nanoisland growth monitored periodically with the STM. For the experiments with oxide-free voids in the native oxide layer the samples were not flashed at high

Methanol is known to interact strongly with the Si(111) 7  7 reconstructed surface dissociating upon adsorption [10,11]. The adsorption of methanol can, in fact, be followed by the corresponding darkening of dangling bonds in the STM image. The well-known 7  7 unit cell has many highly reactive dangling bonds; in particular each cell has 12 symmetric dangling bonds on so-called ‘‘adatoms’’ and six on ‘‘rest’’ atoms (see Fig. 1a). The CH3OH molecule reacts with an adatom of the 7  7 unit cell forming Si–OCH3 [10]. The remaining hydrogen atom reacts with a Si rest atom. Since there are 12 adatoms and six rest atoms in the unit cell, only half of the adatoms are darkened to STM observations, and the rest remain visible (Fig. 1b). After half of the adatoms have formed SiOCH3 the surface saturates and no more molecules adsorb. The surface reaches saturation after only a few langmuirs; even after more than 30 L the

Fig. 1. (a) STM image of the clean Si(1 1 1) surface before methanol adsorption, 8 nm  8 nm. An half unit cell is highlighted, with adatoms (circles) and rest atoms (crosses). Rest atoms are not visible. (b) The same surface after 30 l exposure to methanol, 20 nm  20 nm. (c) Zoom out, 80 nm  80 nm.

V. Palermo, D. Jones / Applied Surface Science 226 (2004) 191–196

193

Fig. 2. (a) Nanoislands grown after methanol adsorption and heating 5 s at 800 8C, 80 nm  80 nm. The typical silicon 7  7 pattern has been restored. (b) After 1 min at 8008, 200 nm  200 nm. (c) After 18 min at 8008, 200 nm  200 nm. All images have been gradient-filtered.

Although their positions are quite random, there is a preponderance of nucleated structures along step edges and domain boundaries (Fig. 2). This indicates that most of the carbon present on the surface desorbs during the heating, but carbon atoms adsorbed at defect sites (i.e. steps or domain boundaries) are more stable and nucleate nanoisland growth. The same adsorption procedure was repeated using carbon monoxide instead of methanol. CO is known to adsorb onto and desorb from silicon surfaces in its neutral molecular state without reacting chemically with the surface. STM images of the silicon surface, even after 180 L of CO exposure did not, in fact, even show the presence of molecular adsorption at room temperature. This can be due to selective preferential adsorption of CO at surface defect sites which were too few to be observed with certainty, or to the elevated mobility of CO molecules on the surface, 2500

160 140 100

1500

80 1000

60 40

Volume

20

Islands /um 2

2000

120 V (nm 3)

STM images like the one in Fig. 1b remain exactly the same. This leads to a very uniform coverage equivalent to 9:6  1013 carbon atoms/cm2. The process has been described in detail in [10]. Observing the surface on a larger scale (Fig. 1c), it is possible to note that the adsorption is quite uniform on the whole surface. The only irregularities are observed at the border between different 7  7 domains, already present on the clean surface, which remain visible as dark lines on the contaminated one. Following methanol adsorption the samples were heated to 800 8C, allowed to cool and then examined again with the STM. Although the heating cycle restored the crystalline surface with the 7  7 reconstruction, a large number of nanoislands was observed over the whole surface (Fig. 2a). The island density is 1500  60 mm2. It is known that annealing alkyl monolayers on silicon surfaces at 500 8C leads to the disappearance of C–H and Si–H vibrational modes and the appearance of a Si–C vibrational mode [12]. Although photoemission results [11] indicate a uniform dispersion of SiC over the surface, the real space STM images in Fig. 2 show a quite clean 7  7 surface with the contaminating atoms concentrated into the nanoislands. Heating for longer times the number of islands per unit area remains the same but their dimensions increase (see Fig. 2b and c and Fig. 3). During the heating no more carbon contaminants are provided to the surface, so island growth must be due to silicon atoms diffusing on the surface and feeding the islands.

500

Isl. Density

0

0 90

150

330

450

1050

Heating time (s)

Fig. 3. Evolution of nanoisland volume and density with heating time.

194

V. Palermo, D. Jones / Applied Surface Science 226 (2004) 191–196

which excludes their detection with the STM tip. However, some CO adsorption must have occurred because nanoisland nucleation and growth was indeed observed after the subsequent heating step (10 min at 800 8C). At 230 L CO exposure of silicon, Chamberlain et al. did, in fact, observe on silicon the presence of small amounts of CO in their TPD spectra [13]. The island distribution density was only 360  30 mm2 (25% of the density obtained with methanol) and approximate island dimensions, given the limitations of such measurements with STM, much larger with respect to methanol-nucleated islands (510  60 nm3 versus 120  20 nm3). Similar treatment of blank samples with no exposure to CO gave no significant production of surface nanoislands. Analogous experiments using other molecules, like oxygen, water, carbon dioxide or methane, also did not show any nanoisland growth. Given that CO did, in some way, react at some particular sites on the surface, we exposed partially oxide-masked silicon surfaces to the gas with the aim of presenting active sites to the adsorbant. These surfaces, prepared in UHV [3,14] contain surface oxide voids of 50–100 nm diameter with naked silicon delimited by an exposed Si/SiO2 interface (see Fig. 4a). At the centre of these voids there is a nanostructure produced during the original radial enlargement of the void. After cooling, these samples were exposed to 180 L of CO at room temperature and then heated again to 800 8C for 10 min. During this second heating cycle, with consequent further radial enlargement of the voids with oxide decomposition,

new series of nanostructures were formed in a circular arrangement corresponding to the original perimeter of the oxide void with the Si/SiO2 interface during CO exposure (see Fig. 4b). When the same samples were heated without having been exposed to CO, there was simply void enlargement without the formation of secondary nanostructures around the void perimeter. This suggests that, although the whole surface (oxide and naked silicon), is exposed to CO, the gas reacts chemically only at the Si/SiO2 interface around the oxide-free void perimeter. These reaction sites thus become nucleation sites, leading to growth of the circular arrangements of nanoislands during subsequent annealing. A similar phenomenon of preferential CO decomposition at the perimeter of oxygen islands has been already observed [15]. On repeating the CO exposure/heating cycle, a second circle of nanoislands formed around the newly expanded surface void perimeter (Fig. 4c). This process can, in principle, be repeated to obtain a series of concentric circular arrangements of nanostructures, similar to a nanoscopic ‘‘dartboard’’, the only limit being the minimum distance between neighbouring voids. The island circles are very irregular and their dimension quite variable. However, the most interesting prospect which this site-dependant chemical reaction, leading to the nucleation and growth of nanostructures, brings to light its possible use for producing surface nanostructures by exploiting this ‘‘edge effect’’. The use of selective adsorption along the edge of an oxide pattern followed by the formation of nanostructures only

Fig. 4. (a) Nanoscopic void in the surface oxide layer, 120 nm  120 nm. (b) Nanoisland circle created after CO adsorption and heating at 800 8C, 400 nm  400 nm. (c) After another cycle of CO adsorption and heating, 590 nm  590 nm.

V. Palermo, D. Jones / Applied Surface Science 226 (2004) 191–196 18

195

400

(a)

16

(b)

350

14

300

12

250 10

200 8

150 6

100

4

50

2

0

0 0

200

400

600

800

1000

0

200

400

600

800

1000

Fig. 5. (a) Real STM profile of islands grown in a void. (b) Simulated profile of island growth, arbitrary units.

along that edge through self-organisation processes would allow a significant reduction in feature dimensions whilst still using classical processing techniques. The exploitation of this ‘‘reactive adsorption edge effect’’ to produce nanoscale structures is analogous to the ‘‘light propagation edge effect’’ exploited by Li et al. [16]. For nanoisland growth, silicon atoms have to move uphill on the surface. Similar mound growth on various surfaces has been explained as a consequence of the Ehrlich–Schwo¨ bel (ES) barrier effect where a mobile surface atom when approaching a step from above finds a potential barrier at the step and is reflected. This leads to a net uphill flux of atoms on crystal steps, thus being a major cause of roughening [17,18]. The presence of the ES barrier effect on silicon surfaces is controversial: STM measurements during epitaxial growth suggested that the barrier is weak or even absent on Si(111) [19], but thermal relaxation measurements of hillocks and craters at high temperature showed the presence of a significant ES barrier effect [20]. Thus, it would seem that the ES barrier is a sufficient but not a necessary condition [21] for mound growth on silicon surfaces. In order to assess the effect of the ES barrier on nanoisland growth in our system, Monte Carlo simulations of surface roughening were carried out. A simple, mono-dimensional system was used where the probability of each atom of diffusing depends upon the number of nearest neighbours and

upon an interaction energy Es [22]. Details of the algorithm and the parameters used will be reported elsewhere [23]. The starting configuration was a flat surface with a central peak, like the one shown in Fig. 4a, and simulation temperature was set to 800 8C. Fig. 5 compares a real STM profile taken from a void with the simulated one. Best results were obtained using an Es interaction energy of 0.4 eV. We can see that, even though the complexity and the dimensionality of the simulated system is much simpler than the real one, in both cases sharp mounds develop on the otherwise flat surface. This indicates that the presence of an ES barrier is sufficient to produce structures similar to those observed experimentally, although further studies, using more complex, 2-D systems, are necessary.

4. Conclusions Carbon-containing molecules which react with the silicon surface nucleate the growth of nanoislands during UHV annealing. Island dimension and number depends strongly on molecule–surface interaction. Methanol, which adsorbs dissociatively over the whole silicon surface, leads to a large number of structures whereas CO reacts preferentially at Si/ SiO2 interfaces. The presence of carbon in the nucleating molecule is a condition necessary but not sufficient to nucleate

196

V. Palermo, D. Jones / Applied Surface Science 226 (2004) 191–196

island growth. Molecules without carbon (oxygen, water) or interacting weakly with silicon (methane, CO2), do not nucleate island growth. The carbon adsorbed on the surface through the contaminating molecules is a main component of the central island core, but further growth is due to a flux of silicon atoms from the surrounding surface. The site-specific reactivity of CO offers the possibility of producing ordered arrangements of nanostructures by exploiting this surface adsorption edge effect. The initial results from Monte Carlo simulations indicate that the presence of an Ehrlich–Schwo¨ bel barrier can cause the growth of well-defined, protruding nanoislands as those observed experimentally.

Acknowledgements A Ph.D. grant (V.P.) from the CNR in Rome is gratefully acknowledged. Thanks are also due to Mr. G. Bragaglia for technical assistance with the STM instrumentation.

References [1] T. Ogino, Surf. Sci. 386 (1997) 137. [2] T. Ogino, H. Hibino, Y. Homma, Y. Kobayashi, K. Prabhakaran, K. Sumimoto, H. Omi, Acc. Chem. Res. 32 (1999) 447.

[3] D. Jones, V. Palermo, Appl. Phys. Lett. 80 (2002) 673. [4] R. Tromp, G.W. Rubloff, P. Balk, F.K. LeGoues, E.J. Van Loenen, Phys. Rev. Lett. 55 (1985) 2332. [5] E.A. Gulbransen, S.A. Jansson, Oxid. Met. 4 (1972) 181. [6] R.E. Walkup, S.I. Raider, Appl. Phys. Lett. 53 (1988) 888. [7] M. Liehr, J.E. Lewis, G.W. Rubloff, J. Vac. Sci. Technol. A 5 (1987) 1559. [8] M. Liehr, H. Dallaporta, J.E. Lewis, Appl. Phys. Lett. 53 (1988) 589. [9] H. Yamada, J. Vac. Sci. Technol. A 19 (2001) 627. [10] Z.-X. Xie, Y. Uematsu, X. Lu, K. Tanaka, Phys. Rev. B 66 (2002) 125306. [11] M. Carbone, R. Zanoni, M.N. Piancastelli, G. Comtet, G. Dujardin, L. Hellner, Surf. Sci. 352/354 (1996) 391. [12] M.M. Sung, J. Kluth, O.W. Yauw, R. Maboudian, Langmiur 13 (1997) 6164. [13] J.P. Chamberlain, J.L. Clemons, A.J. Pounds, H.P. Gillis, Surf. Sci. 301 (1994) 105. [14] V. Palermo, D. Jones, Mater. Sci. Eng. B88 (2002) 220. [15] J. Wintterlin, S. Vo¨ lkening, T.V.W. Janssens, T. Zambelli, G. Ertl, Science 278 (1997) 1931. [16] S.P. Li, A. Lebib, D. Peyrade, M. Natali, Y. Chen, W.S. Lew, J.A.C. Bland, J. Appl. Phys. 90 (2001) 521. [17] R.L. Schwo¨ bel, J. Appl. Phys. 40 (1969) 614. [18] S. Das Sarma, P. Punyindu, Z. Toroczkai, Surf. Sci. 457 (2000) L369. [19] C.J. Lanczycki, R. Kotlyar, E. Fu, Y.-N. Yang, E.D. Williams, S. Das Sarma, Phys. Rev. B 57 (1998) 13132. [20] S. Hildebrandt, A. Kraus, R. Kulla, G. Wilhelmi, M. Hanbu¨ cken, H. Neddermeyer, Surf. Sci. 486 (2001) 24. [21] S. Das Sarma, P. Punyindu, Z. Toroczkai, Surf. Sci. 457 (2000) L369. [22] C.J. Lanczycky, S. Das Sarma, Phys. Rev. Lett. 76 (1996) 780. [23] V. Palermo, Philos. Mag. Lett., accepted for publication.