Chapter 6

Ce/Si{111}

Abstract

The deposition and the subsequent anneal of submonolayer amounts of cerium on Si{111}7x7 surfaces produces silicide structures. A Si{111}√3x√3-30-Ce phase can be formed. It is probably a surface-stabilized polymorph of CeSi₂ in the AlB₂ structure, as opposed to either the ThSi₂ or the GdSi₂ bulk structure. A Si{111}2x1-Ce phase can be formed, but it probably arises from either a ThSi₂-, a GdSi₂-, or a CaSi₂-polymorph of CeSi₂. In both cases, √3x√3-30 and 2x1, the diffraction patterns are not associated with adsorbate superstructures. Instead, we believe they derive from silicide structures.

6.1 Introduction

Cerium and other rare-earth metals have been deposited on Si surfaces for use as oxidation enhancers [113-117]. The high reactivity of Ce can cause order-of-magnitude increases in the oxidation rate of Si surfaces, even at relatively low temperatures. Fujimori et.al.[113] noted the presence of distinct Si{111}√3x√3-30-Ce and 2x2-Ce surface structures. These surfaces were created by heating (350 C) 1/3 and 1/4 ml of Ce, respectively, deposited on clean Si{111}7x7 substrates. The √3 phase was attributed to a H₃-site structure. The 2x2 phase was attributed to a buckled-chain structure.

The intent of the current research was to create a Si{111}√3x√3-30-Ce surface, collect the LEED spectra, and analyze the structure. However, a purely √3x√3-30 diffraction-pattern of such a structure was never observed in the present work. Even though √3 LEED beams were seen, these beams were always of similar intensity as the Si{111} integral-order beams, in the presence of 2x1 beams, and additional minor beam-clusters around the integer-order beams were present. Owing to the complexity of these patterns, LEED I(V) spectra could not be collected. The Ce/Si{111} surface structures formed were probably those of a cerium silicide. The next sections detail the formation of such surfaces and describe their possible structure.

6.2 Experiment

The experiment was done with a similar sample and techniques as those used in the previous Au and Mg experiments. The prime difference in the Ce experiment was the evaporation of Ce metal.

Cerium is a very reactive metal and quickly tarnishes in air. It has a relatively low vapor pressure (3x10^-11 Torr) at its melting point (798 C). For this reason, Ce would have to be evaporated from the liquid state at approximately 1100 C with a vapor pressure of 2x10^-7 Torr. However, since liquid Ce is denser than the very stable oxide, ceria, it was necessary to use oxide-free source material, otherwise the ceria could float above the liquid cerium and hinder any evaporated Ce from reaching the Si sample surface.

High purity Ce was obtained from Ames Laboratory, Iowa State University, sealed in glass ampules. A semi-cylindrical (6 mm radius, 10 mm length) slug was loaded into a Ta tubular evaporation-boat, which was previously outgassed. The boat was mounted in the vacuum system 10-cm away from the sample surface and inclined at a 30 degree angle.

The temperature of evaporation was monitored using an infra-red detector on the side the tubular Ta-boat. Initially, the Ta boat required 60 Amperes to reach roughly 800 C and melt the Ce. At this point there was not a large burst of gas. Instead, the melted Ce uniformly filled the bottom of the Ta boat and effectively shorted-out the source. This was clearly visible in the form of a `silhouette' of Ce in the tube. The decreased resistance of the Ta boat, in the volume filled with Ce, created a sharp drop in temperature at the Ce/vacuum interface. For example, the unfilled part of the boat would be at 1000 C (bright red) and the Ce-filled part of the boat was at 800 C (dull red). From then on, larger current would be necessary to heat the source and overcome the reduced electrical resistance of the source. Typically, more than 100 Amperes would be required to heat the Ce source material to 1100 C. Such high currents caused large initial outgassing (which subsided), but never caused any measurable sample substrate-heating.

The cleanliness of the deposition was monitored using AES. Initial contaminants included S, Cl, C, and O, but oxygen was the only contaminant to remain after sustained outgassing and prolonged use of the Ce source. Typical values of R_(O/Ce) (the ratio of the O_(510 eV) and the Ce_(82 eV) Auger peaks) were 1/50, indicating a proportion of less than one oxygen atom per cerium atom [37]. Interestingly, several AES peaks were observed which are absent in the standard AES spectra [37], at approximately 902, 884, 291, 207, and 109 eV. These peaks probably correspond to transitions involving cerium's M_IV 3d_3/2, M_V 3d_5/2, N_I 4s, N_III 4p_3/2, and N_IV 4d_3/2 electron energy levels and some interfacial states near the Fermi energy.

Cerium thickness was also monitored using AES. The ratio R_Ce, see Table 2.1, was compared to calculated values in the same way as for Mg and Au, where R_inf = I^inf_Ce(82) / I^inf_Si(92)=0.38 and lambda_82 = 4.3 Ang. were assumed [37, 38]. Using this method the deposition rate was determined to be 3.0 Ang./min. This number could not be converted to LE/min., since the structure and the thickness of a layer were not known. Also, it should be noted that the thicknesses were based on the assumption of non-interdiffusion; this was most likely not a correct assumption.

Ce was deposited onto a room-temperature Si{111}7x7 surface several times. Increasing thicknesses of Ce masked the fractional-order 7x7 LEED beams, created higher background, and eventually masked the integral-order beams. When visible, the integral-order beams were always sharp and there was never any indication of a room-temperature ordered-phase of Ce forming on the substrate surface.

The sample was also heated progressively from 100 C to 1100 C with various Ce thicknesses, and was then allowed to cool to room temperature. LEED observations at no time revealed the presence of a separate and distinct Si{111}√3x√3-30-Ce structure. Instead, when 1/2 Ang. of Ce was deposited the pattern remained 7x7 with an increase in the background. Anneal at 200, 250, ........, 500, 550 C for 2 min. caused no apparent change of the LEED pattern. Anneal at 600 C caused the appearance of single-domain 2x1 streaks in the LEED pattern. Further progressive anneal only caused the streaks to sharpen, but the streaks never completely transformed into individual beams. AES measurements indicated a halving of the surface Ce content as a consequence of the annealing treatment.

Deposition of 1.5 Ang. of Ce onto a clean substrate caused the almost complete disappearance of (1/7)-order beams into the increasing LEED background, but the integral-order beams remained sharp. No new ordered-phases were seen. Therefore, the surface was covered by an amorphous Ce film, which increased the amount of incoherent scattering. Anneal caused no change until 400 C, when the integral-order beams broadened and inter-integral-order streaking became visible. With increasing anneal temperature, up to 500 C, the streaking became more pronounced and formed into a 2x1 pattern. However, the pattern was not uniform across the sample. √3 beams became visible after anneal to 600 C. These beams were of equal intensity as the integral-order beams, and the ratio of intensity between the 1x1 and √3x√3-30 beams did not change appreciably across the sample. Additional minor beam-clusters located around the integral-order beams were visible. It should be noted that the R_Ce ratio had not changed, and the distribution of 2x1 and √3x√3-30 domains on the sample was not uniform. Anneal to 850 C (in 50 C steps) sharpened the pattern, anneal to 1000 C caused a uniform distribution of 2x1 and √3 beams. These beams had intensity comparable to the original Si{111}1x1 beams. The satellite clusters were more distinct, and there was a reduction in the Ce thickness by a factor of four.

Higher thicknesses (up to 12 Ang.) of Ce on Si{111}7x7 caused a complete obliteration of the LEED pattern into a very high background. Anneal of such surfaces yielded a similar progression of results: the appearance of broad 1x1 beams, inter-integral-order streaks, 2x1 beams, √3 beams, and minor beam-clusters. Owing to the complexity of these phases, LEED I(V) spectra were not collected.

Once these structures were formed, it was not straightforward to return to the clean Si{111}7x7 structure. Anneal alone to 1100-1200 C did not remove the Ce. Although slight ion-bombardment removed the Ce atoms, anneal brought back the above mentioned Ce/Si{111} structures. The sample often needed to be bombarded for more than 5 hours in order to completely remove the subsurface Ce atoms. After this process, anneal to 1100 C produced a clean Si{111}7x7 surface.

6.3 Discussion

6.3.1 Si{111}2x1-Ce

In contrast to Fujimori et.al.'s[113] report of a low-coverage 2x2-Ce phase, the present observations show that the surface unit-mesh was actually 2x1. Although a single-domain 2x1 phase was never seen, its precursor was seen: as mentioned above, inter-integral-order beam streaking was often visible, and such streaking often only occurred in one domain, i.e., in one direction. These streaks would then sharpen with additional heat and coalesce into (1/2)-order beams. One cannot distinguish a domain-averaged 2x1 from a 2x2 LEED pattern by the mere presence of fractional beams, but one can do so by examining the relative intensity of (1/2)-order beams. Referring to Figure 6.1, patterns a, b, and c are schematic LEED patterns for three single domains of a Si{111}2x1 surface. Pattern d is a combination of the previous three patterns. Pattern e is a schematic diffraction pattern for a Si{111}2x2 surface. The (1/2,0 beam of pattern a, the (-1/2,1/2) beam of pattern b, and the (0,-1/2) beam of pattern c are degenerate under rotation of the surface unit-mesh (domains) with respect to the bulk-like surface unit-mesh. These three beams, and others, would combine to form pattern d, if all three the domains were present on the surface, but they would only have equal intensity if the domains were present in equal concentration. The same three beams in pattern e are degenerate under the symmetry operations of the 2x2 surface unit-mesh and should always have equal intensities at normal incidence. In the present experiment, these beams did not always have equal intensities at normal incidence. Therefore, the surface unit-mesh was in fact 2x1.

What surface structure of Ce and Si would form a 2x1 LEED pattern? First, experimental observations (thermal stability, permeance of Ce, intensities of beams) indicate that the surface is not a simple adsorbate-superstructure, as in the case of Al, Ga, In, etc.... on silicon. A silicide is most likely. Second, the stable silicides of cerium are CeSi₂, CeSi, Ce_5Si_4, Ce_3Si₂, and Ce_5Si_3 [118]. The CeSi₂ is the most prevalent and occurs in bulk form in two modifications, a tetragonal ThSi₂ and an orthorhombic GdSi₂ structure. The latter is a distortion of the former's unit-cell, from (4.3x4.3x13.8 Ang.) to (4.2x4.1x13.9 Ang.). Figures 6.2.a and b are schematics of the unit-cell of the ThSi₂ structure and a resultant LEED pattern. The pattern was calculated using a kinematic structure factor, the same symbols represent degenerate intensities, the pattern is domain-averaged, the reciprocal unit-mesh of the ThSi₂ structure is indicated with dotted lines, and the reciprocal unit-mesh of the Si{111}1x1 surface is indicated with solid lines. This calculated pattern accurately represents the observed 2x1 diffraction patterns discussed in the previous section. The calculated diffraction pattern indicates several things: a) the ThSi₂ pattern is not commensurate with the Si{111}1x1 pattern, b) the (1/4)-order beams are weak, c) the (1/2)-order beams are most prominent, and d) there are small beam-clusters where the Si{111}1x1 beams would be located. The most important fact is, however, the presence of a ``2x1-like'' diffraction pattern created by this ThSi₂ or a GdSi₂ unit-mesh. Hence, Ce on Si{111} may form a silicide.

2x1 LEED patterns on Si{111} have also been reported for Gd, Yb, Eu, Ca, and Ba silicides [114, 119]. The 2x1 patterns for Gd and Yb silicide can be associated with the previous ThSi₂ discussion. The other three silicides probably have 2x2 patterns created from a bulk-terminated CaSi₂ structure. Interestingly, CaSi₂ can also be formed in the GdSi₂ structure, under high-pressure and temperature 120].

6.3.2 Si{111}√3x√3-30-Ce

There are five main disilicide-structures that rare-earth metals may form: tetragonal ThSi₂, orthorhombic GdSi₂, hexagonal AlB₂ (where the Al and B atoms are replaced by rare-earth atoms and Si atoms, respectively), hexagonal AlB₂-x (where x varies between 0 and 0.3), and trigonal/rhombohedral CaSi₂ (where the Ca atoms are replaced by rare-earth atoms). ThSi₂, see Figure 6.2.a, has a quasi-hexagonal mesh of Si atoms, with the rare-earth atoms in the interstices. The building block of the unit cell is indicated with solid lines. This subunit contains four Si atoms (filled squares) at the corners of a rectangle (axb) and one rare-earth atom (empty circle) in the center, below the plane of Si atoms. Alternate stacking (in and out of the plane) and translation (by 1/2 a x^hat) of these subunits create the ThSi₂ unit-cell, with glide-plane symmetry. The AlB₂ unit-cell, see Figure 6.2.c, is created from the same subunit when the ratio of the rectangular sides (a/b) is √3 and the translation vector is 1/2 a x^hat + 1/2 b y^hat. The AlB₂-x structure has one-in-six B (Si) atoms removed, depicted in the figure by the empty triangular symbol. This structure has a √3x√3-30 surface unit-mesh when compared to Si{111}1x1. The CaSi₂ structure is similar to AlB₂, but the stacking is of alternate Si, Si, and Ca 1x1 hexagonal meshes. This structure has a 2x2 surface unit-mesh when compared to Si{111}1x1.

Everyone of the rare-earth elements is reported to form in of the above structures with Si, and most form at least two polymorphs. The silicides of Sc, Y, Nd, Gd, Tb, Dy, Ho, Er, Tm, Yb, and Lu have all been reported to crystallize in the AlB₂ or the AlB₂-x structure. The silicides of (Ca, Y, La, Ce, Pr, Nd, Sm, Gd, Tb, Dy, and Ho) and (Y, La, Ce, Pr, Nd, Sm, Eu, Gd, and Dy) have been reported [121] to form the GdSi₂ and the ThSi₂ structures, respectively. Additionally, the silicides of Y, Gd, and Er have √3x√3-30 LEED patterns (mixed with 2x1) arising from the AlB₂-x surface unit-mesh [19, 114, 115, 116, 117]. Since there exist so many interconnected polymorphic structures, we believe that the Si{111}√3x√3-30-Ce LEED pattern arises from a surface-stabilized AlB₂-x-polymorph of CeSi₂. We cannot rule out, however, the presence of a √3x√3-30 adsorbate-superstructure on a bulk-terminated CeSi₂ surface.

In conclusion, the (2x1) and the √3x√3-30 LEED patterns of Ce on Si{111} probably arise from silicide structures. We find no evidence to support the presence of Si{111}-Ce adsorbate superstructures, as opposed to the cases of most group-III, -IV, and -V metal-atoms on Si{111}.