Tranformations of faulted halves of the DAS structure on quenched Si(111)

Surface Science 423 (1999) L291-L298



Wataru Shimada a, Hiroshi Tochihara a *, Tomoshige Sato b and Masashi Iwatsuki b

a Department of Molecular and Material Sciences, Kyushu University at Kasuga Kasuga, 816-8580, Japan
b JEOL Ltd., 3-1-2 Musashino, Akishima, Tokyo 196-8558, Japan

* Corresponding author. E-mail: tochihar@mm.kyushu-u.ac.jp


Abstract

We have observed peculiar faulted (F) - halves of the dimer-adatom-stacking fault (DAS) structure by scanning tunneling microscopy on quenched Si(111) surfaces at 380C, and denoted them irregular F-halves (F'). In the F'-halves, the arrangement of adatoms is partially different from that of corresponding regular F-halves and one adatom is missing. We observed single 5x5-F', 7x7-F', 9x9-F', 11x11-F' and 13x13-F' in unreconstructed regions, and proposed their structures. Fluctuations in the size and structure of the faulted halves have been observed in real time by rapid imaging at 380C. It is found that single odd-sized faulted-halves appearing in the size changes are always the irregular ones.

Keywords: Surface structure, morphology, roughness and topography; Scanning tunneling microscopy (STM); Silicon


The Si(111)7x7 surface is established as the dimer-adatom-stacking fault (DAS) structure [1], and continues to attract us for investigating interesting surface phenomena occurring on this surface, such as diffusion of Si atoms, homo- and hetero-epitaxial growth, chemical reaction, etc. The formation process of the 7x7 reconstruction at the atomic level is also unresolved interesting issue. To date, the process was studied by various methods. By reflection electron microscopy [2], low-energy electron microscopy (LEEM) [3] high temperature (HT) STM [4,5] and secondary emission microscopy [6], it had been demonstrated that the 7x7 reconstruction, starting from each upper step edge upon cooling the surface to 860C after flashing to 1250C for cleaning, grows across the terrace until it reaches the neighboring step. Even with HT STM it was not possible to observe the formation of the 7x7 unit cell at the atomic level, because the growth is extremely rapid at 860C. Ogawa and coworkers [7] pointed out wrong assignment of the direction for the 7x7 domain growth in the previous studies. LEEM revealed that triangular 7x7 domains are formed on terraces when rapidly quenched [3]. Yang and Williams [8] confirmed this on a quenched surface with very wide terraces by room temperature (RT) STM and revealed that small unreconstructed regions, randomly decorated by adatoms (denoted "1x1", in which quotation mark means a disordered surface), remain on terraces. They discovered small triangular 9x9 and 11x11 domains as well as large 7x7 domains, and claimed that the formation of the 9x9 and 11x11 domains is due to a high atomic-density in the 1x1 region. Ohdomari and coworkers [9-13] carried out in situ STM observation of the "1x1" regions at 440~700C prepared by quenching. They observed small 5x5 domain in the "1x1" region as well as the 7x7 and 9x9 domain [9,10]. The atomic density of the 5x5 domain is lower than that of the 7x7 one.

Ohdomari et al. [9] showed that the formation of the faulted (F) - half of the 7x7 unit cell leads to the growth of the 7x7 domain through in situ STM observation of the "1x1" region. [Hereafter we denote the F-half of the 7x7 unit cell as the 7x7-F.] Recently, we have demonstrated by STM [14] that an odd-sized F-half changes its size dynamically in a sequence 5x5-F <-> 7x7-F <-> 9x9-F in the "1x1" region at about 350C after rapid quenching, and proposed the dynamic DAS size conversion model for the formation of the 7x7 unit cell and the 7x7 domain. We claim [14] that the formation of domains of various sizes of the DAS structure in the "1x1" region is caused by the formation-mechanism itself of single F-half. Very recently, we have discovered even-sized F-halves, e.g., 6x6-F, 8x8-F, 10x10-F and 12x12-F in the "1x1" region at 380C after rapid quenching [15]. The even-sized F-halves appear transiently as intermediates during size-changes between odd-sized F-halves.

In the present study, we have investigated irregular structures of odd-sized F-halves formed in the "1x1" region upon quenching to 380C by STM. It is found that odd-sized F-halves appearing in the dynamic size-changes mentioned above are always the irregular type.

Experiments were conducted by using HT STM (JEOL JSTM-4500XT) in ultrahigh vacuum (2x10-8 Pa). A n-type (0.1 cm) Si(111) sample of size 1x7x0.3 mm3 was cleaned by flashing at 1250C using direct heating, and cooled down to 380C within 1-2 s. The temperature was measured by an infrared pyrometer. The sample bias was estimated to be in a range of 0.5 ~ 1.0 V. After maintaining temperature at 380C for 2~10 min, the "1x1" regions were continuously recorded in the constant-current mode on a video-tape. Scanning speed is 1.7 s per frame with 128x128 pixels otherwise noted.



Figure 1

Fig. 1. STM images of the 1x1 region prepared by rapid quenching to 380C.
(a) Single isolated 9x9-F' is formed as indicated by a short arrow. This image is about 1/4 of one frame (18 s per frame with 512x512 pixels). This was taken 760.0 s after the start of measurement as shown at the right-bottom. Hereafter, we show time at the right-bottom of each image taken for this isolated faulted-half. Two 7x7-F's are also seen as outlined. (b) Single 5x5-F' at the neighbor of a 9x9 DAS domain is outlined. This was taken in a different experiment at 524.0 s after the start of measurement as shown at the left-bottom. For this half, we show time at the left-bottom in each image. This was taken with a scanning speed of 5.0 s per frame (256x256 pixels).

Using the procedure above we can produce small 1x1 regions as shown in Fig. 1(a). As indicated in Ref. [14], the small 1x1 regions eventually convert to 7x7 domains leaving three zigzag lines as domain boundaries. Therefore, the 1x1 regions are living place as for the creation of unit cells of the DAS structure. In fact, a single F-half is created there during in situ observation as described below. Here, we note three features of the 1x1 regions. (i) Atom-resolved images of adatoms are not usually achieved at 380C except for adatoms at the peripheral T4 sites around F-halves [9,14]. (ii) The regions are always surrounded by F-halves [7,14]. (iii) There are many unknown species appearing as small clouds, whose images are strongly dependent on the bias.

A F-half indicated by a short arrow in Fig. 1(a) is magnified in Fig. 2(c) and assigned to be a kind of the 9x9-F. The arrangement of adatoms is partially different from that of the regular 9x9-F, and one adatom is missing [16]. We denote it the irregular F-half (F'-half for simplicity). Typical STM images of the 5x5-F', 7x7-F' and 9x9-F' are shown in Figs. 2(a), (b), (c), respectively. [The 11x11-F' and 13x13-F' are also observed.] In each panel, the F'-half is outlined along dimer-rows. Since the dimers are not recognized in STM images, the outlining of the boundary between the F'-half and the 1x1 region is carried out by using the fact that the regular F-half is always surrounded by adatoms at the peripheral T4 sites of the 1x1 region at 380C [9,14]. Remember that adatoms in the 1x1 region do not exhibit atomic resolution except for these special adatoms at 380C.

Structural models of the 5x5-F', 7x7-F' and 9x9-F' are proposed in Figs. 2(d), (e) and (f), respectively, as follows. First of all, we paid attention to adatoms just outside of the F'-halves, sitting at the peripheral T4 sites. In Fig. 2(c), for example, there are 4, 4 and 5 adatoms along three sides of the triangle (9x9-F'), which indicates that there are 4, 4 and 5 dimers in the three sides, respectively. This assignment is based on the fact that a T4 site just in front of a dimer is always occupied by an adatom in the 1x1 region [9,14,17]. Thus, we naturally suggest a structural model for the 9x9-F' as in Fig. 2(f). Similarly, models for the 5x5-F' and 7x7-F' are obtained in Figs. 2(d) and (e), respectively. In each model, the arrangement of adatoms is in good agreement with the image.



Figure 2

Fig. 2. (a)-(c) STM images of odd-sized irregular faulted-halves of 5x5-F', 7x7-F' and 9x9-F', respectively, at 380C. (a) and (b) are magnified from Figs. 1(a) and (b), respectively. (d)-(f) Proposed structural models (top view) for 5x5-F', 7x7-F' and 9x9-F', respectively. Circles represent Si atoms; the closer these atoms are to the surface, the larger the diameters of the circles. The largest (dotted) circles represent adatoms and the smallest closed ones the fourth layer atoms. The dangling bond at the corner holes is indicated. In (d), a line connecting two arrows is a mirror plane of the structure.

The proposed model of the 7x7-F' was confirmed by another images taken at RT after rapid quenching in Figs. 3. [Note that atom-resolved images of adatoms in the 1x1 region (upper side of Fig. 3(a)) are obtained at RT.] The 7x7-F' is outlined in Fig. 3(a), and its topographic images at various biases are shown in Figs. 3(b)-(d). This F'-half is located at the edge of a 7x7 domain. At negative sample-bias (cf. Figs. 3(c),(d)), the arrangement of the restatoms reflects well the proposed model (cf. Fig. 2(e)).



Figure 3

Fig. 3. (a) Topographic image of Si(111) at RT after rapid quenching from 1250C. Sample bias is 2.0 V. The 7x7-F' is outlined. (b)-(d) Topographic images of the 7x7-F at sample bias of 1.0, 0.5 and 2.0 V, respectively.

It is very interesting to note where and how the F'-halves are formed on Si(111). We have never found them inside of DAS domains of various sizes. They do not form their own domain. Single F'-halves are observed in the living place, i.e., in the 1x1 region or at the edge of DAS domains. We show two sites where F'-halves are created. First, the isolated F'-half shown in Fig. 1(a) by the short arrow is described. Changes of its size (1 corresponds to 1x1) are plotted as a function of time at 380C in Fig. 4(a). At 200 s from the start of the observation, a 6x6-F is created (not shown). As seen in Fig. 4(a) a stable size of the faulted half is 9x9, but its size fluctuates around 9x9. Even-sized F-halves, such as 6x6, 8x8 and 10x10 [15], are created for short time (about 10 s or less). At 1170 s, the faulted half terminates quickly with decreasing its size. A thin vertical bar at 1015 s in Fig. 4(a) means that the faulted half at that time is the regular type. Almost all odd-sized faulted-halves appearing in its whole life (not shown) are irregular ones, which may be related to the fact that the faulted half changes its size frequently in this case. Soon we know this is true.



Figure 4

Fig. 4. (a) Size changes of single isolated faulted-half (indicated by arrow in Fig. 1(a)) as a function of time. See text for a thin vertical-bar. (b) Size changes of single faulted-half (outlined in Fig. 1(b)) at the neighbor of a 9x9 domain as a function of time. See text.

Second, we observed at 380C continuously a single faulted-half (outlined in Fig. 1(b)) at the neighbor of a 9x9 domain sharing one corner hole and its magnified image was shown in Fig. 2(a)). Its size-changes are plotted in Fig. 4(b). At 509 s from the start of the measurement, a 5x5-F' is born. In this measurement the size of 7x7 is stable. In Fig. 4(b), we emphasize following three important findings: regular odd-sized F-halves are formed much more frequently than in the isolated case, as indicated by many thin vertical-bars; the size changes do not occur frequently in comparison with the isolated case; the faulted half always takes the irregular type before and after the size-changes irrespective of its size.

We confirm that the third finding mentioned above is also realized in Fig. 4(a). Therefore, nascent odd-sized faulted-halves are always the irregular type. This implies that the faulted-halves are irregular type at almost all the time in the case where frequent size-changes occur, which is actually realized in Fig. 4(a). If the size change does not take place frequently, the chance of the irregular-regular conversion increases. As a result, in that case regular faulted-halves are observed frequently as in Fig. 4(b). Therefore, the first and second findings mentioned above are strongly correlated each other through the third one.



Figure 5

Fig. 5. Sudden size-change of a faulted-half during line scans. In each image, the time from the start of measurement is indicated at bottom-right (cf. Fig. 4(a)). See text in detail. White lines at bottom are used for determining absolute position of the faulted-half.

It is shown in the above two cases that the faulted half takes the irregular type before and after its size-changes. More direct evidence for this was obtained in continuous images (see Figs. 5). The time is indicated in each panel (cf. Fig. 4(a)). In Fig. 5(a), a 9x9-F' is present in the 1x1 region, as outlined. In the next image it changes to a 8x8-F suddenly during a line scan at arrows in Fig. 5(b). The formed 8x8-F is seen in Fig. 5(c). Therefore, the above sudden change clearly demonstrates that the faulted-half just before the conversion to the 8x8-F is really the irregular type. We observed another sudden change from the 9x9-F' to the 10x10-F during a line scan. Thus, we conclude that the faulted half takes the irregular type before and after its size-changes.

Here, we focus on the proposed models of the F'-halves (cf. Figs. 2(d)-(f)). There are two prominent features: there exists only one corner hole; the number of the dimer in a side of the F'-half is larger than that of the corresponding regular type by one. Here, we discuss the models in detail. First, we should note that the proposed structures have a mirror symmetry (e.g. along a line between arrows in Fig. 2(d)). Therefore, there are two equivalent and one inequivalent sides in triangles of the F'-halves (see models). We examined movement directions of the F'-halves after size-changes. It is found that the displacement of the sides occurs only in the two equivalent sides. No displacement is found in other inequivalent side. Next, the number of dangling bonds of the F'-halves is the same as that of the corresponding regular ones (dangling bonds at corner holes are included): 7, 12 and 19 for the 5x5-F', 7x7-F' and 9x9-F', respectively. Third, we note that it is impossible for the proposed structures to form domains of the F'-halves. In fact, the F'-halves are formed isolatedly or at the neighbor of DAS domains in the 1x1 region and in the edge of DAS domains as mentioned above. In addition, single F'-half has never been observed inside of DAS domains, which is also consistent with the models. It is concluded that the F'-halves are intermediate species appearing only in the living place.

Recently, we have proposed the dynamic DAS size conversion model as the mechanism of the formation of the 7x7 structure [14]. In this model, the F-half is created or annihilated in a sequence 1x1 <-> 5x5-F <-> 7x7-F <-> 9x9-F <-> 11x11-F <-> 13x13-F. More recent study has exhibited that even-sized F-halves are transiently formed between odd-sized ones [15]. The sequence above has been replaced by the following one: 1x1 <-> 5x5-F <-> 6x6-F <-> 7x7-F <-> 8x8-F <-> 9x9-F <-> 10x10-F <-> 11x11-F <-> 12x12-F <-> 13x13-F. The present study provide a new finding into the sequence. That is, the odd-sized faulted-halves are always irregular type in the sequence. Therefore, the sequence is again rewritten as 1x1 <-> 5x5-F' <-> 6x6-F <-> 7x7-F' <-> 8x8-F <-> 9x9-F' <-> 10x10-F <-> 11x11-F' <-> 12x12-F <-> 13x13-F'. Now we believe that the process of the size-changes of the single faulted-half is established.

Then, how do domains of various sizes of the DAS structure grow in the 1x1 region? As indicated earlier [14], single F-half is stabilized when another neighboring F-half of the same size is formed during its lifetime, sharing two corner-holes. Thus, a domain of one size of the DAS structure is grown. Of course, there is a chance for other sizes of DAS domains to grow. It should be noted here that in the domain growth irregular F-halves convert to regular ones. In situ observation of such domain-growth has been successfully carried out and will be described elsewhere [18]. We conclude that excessive atomic-density in the 1x1 region mentioned at the introduction [8] is not the primary condition for the formation of the 9x9 and 11x11 domains. Formation of small domains of various DAS sizes is caused by the microscopic formation-mechanism itself of the faulted half.

Finally, we mention the tip-induced effects in the present observations. Ichimiya and co-workers [19] have observed a strong effect of the tip during the observation of decay of hillocks and crater on Si(111) 7x7. However, the effect is much more likely at steps than at terraces where our observation is carried out. In fact, no meaningful change in the process of the size-conversions of F-halves has been observed when we changed the interval of the scanning. Therefore, we conclude that there is little tip-induced effects in the present observations.

In summary, we have investigated faulted halves of the DAS structure having peculiar arrangement of adatoms at 380C on unreconstructed (1x1) regions formed by rapid quenching. They are single faulted-halves of 5x5, 7x7, 9x9, 11x11 and 13x13. We have denoted them the irregular faulted-halves (F'-half), and proposed their structural models. Not the regular odd-sized faulted-halves but the irregular ones appear during their size-changes. It is concluded that single faulted-half changes its size in the 1x1 region in a sequence 1x1 <-> 5x5-F' <-> 6x6-F <-> 7x7-F' <-> 8x8-F <-> 9x9-F' <-> 10x10-F <-> 11x11-F' <-> 12x12-F <-> 13x13-F'


This work is supported by the Grand-in Aid for Creative Basic Research (09NP1201).



References

[1] K. Takayanagi, Y. Tanishiro, M. Takahashi, S. Takahashi, J. Vac. Sci. Technol. B 4 (1985) 1079. ; K. Takayanagi, Y. Tanishiro, S. Takahashi, M. Takahashi, Surf. Sci. 164 (1985) 367.
[2] N. Osakabe, Y. Tanishiro, K. Yagi, G. Honjo, Surf. Sci. 109 (1981) 353.
[3] W. Telieps, E. Bauer, Surf. Sci. 162 (1985) 163; M. Mundschau, E. Bauer, W. Telieps, Phil. Mag. A, 61 (1990) 257.
[4] S. Kitamura, T. Sato, M. Iwatsuki, Nature 351 (1991) 215.
[5] K. Miki, Y. Morita, H. Tokumoto, T. Sato, M. Iwatsuki, M. Suzuki, T. Fukuda, Ultramicroscopy 42-44 (1992) 851.
[6] N. Aizawa, Y. Homma, Surf. Sci. 340 (1995) 101.
[7] M. Hoshino, Y. Shigeta, K. Ogawa, Y. Homma, Surf. Sci. 365 (1996) 29.
[8] Y. -N. Yang, E. D. Williams, Phys. Rev. Lett. 72 (1994) 1862.
[9] T. Hoshino, K. Kumamoto, K. Kokubun, T. Ishimaru, I. Ohdomari, Phys. Rev. B 51 (1995) 14594.
[10] T. Hoshino, K. Kokubun, K. Kumamoto, T. Ishimaru, I. Ohdomari, Jpn. J. Appl. Phys. 34 (1995) 3346.
[11] T. Hoshino, K. Kokubun, H. Fujikawa, K. Kumamoto, T. Ishimaru, I. Ohdomari, Phys. Rev. Lett. 75 (1995) 2372.
[12] K. Kumamoto, T. Hoshino, K. Kokubun, T. Ishimaru, I. Ohdomari, Phys. Rev. B 52 (1995) 10784.
[13] K. Kumamoto, T. Hoshino, K. Kokubun, T. Ishimaru, I. Ohdomari, Phys. Rev. B 53 (1996) 12907.
[14] H. Tochihara, W. Shimada, H. Yamamoto, M. Taniguchi, A. Yamagishi, J. Phys. Soc. Jpn., 67 (1998) 1513.
[15] W. Shimada, H. Tochihara, T. Sato, M. Iwatsuki, submitted to Phys. Rev. Lett..
[16] This peculiar structure observed by STM was reported by M. Sakaki, K. Ogawa on The 52th annual meeting of the Physical Society of Japan, 31aT2, March 1997.
[17] M. C. Payne, J. Phys. C, 20 (1987) L983.
[18] W. Shimada, H. Tochihara, T. Sato, M. Iwatsuki, unpublished.
[19] A. Ichimiya, Y. Tanaka, K. Ishiyama, Phys. Rev. Lett. 76 (1996) 4721; A. Ichimiya, Y. Tanaka, K, Hayashi, Surf. Sci. 386 (1997) 182.