U. Starke; J. Bernhard, M. Franke, J. Schardt, A. Seubert, and K. Heinz (Universit¨at Erlangen) High power, high frequency and high
tempera-ture device applications have made silicon car-bide an interesting wide-bandgap semiconduc-tor. However, the quality of SiC material avail-able is still hampering an industrial application.
This holds for the growth of substrate material as well as for the deposition of epitaxial metal films for Schottky contacts or oxides of suffi-ciently low interface trap density for the gen-eration of MOS devices. These problems are ultimately related to the ability to control and understand the atomic structure of surfaces and interfaces of SiC.
On SiC(0001) which is the most widely used SiC surface in electronics, several stable surface phases exist, distinguished by the stoichiometry of the surface region. These phases include Si enriched, carbonized and oxide covered struc-tures. On the Si rich side of the phase diagram on SiC(0001) a (33) phase represents the first ordered phase. It is prepared in ultra-high vac-uum (UHV) by annealing at 800–850ÆC
un-der simultaneous Si deposition. In this phase the surface is covered by excess Si as indi-cated by Auger electron spectroscopy (AES) and X-ray photoelectron spectroscopy (XPS) whereby clear evidence for Si–Si bonds are ob-served.
From scanning tunneling microscopy (STM) and low-energy electron diffraction (LEED) holography the existence of a single adatom per unit cell in a three-fold bonding coordi-nation can be deduced [Reuter et al., Physical Review Letters 79, 4818 (1997)]. Parallel ef-forts of a quantitative LEED structure deter-mination and ab initio calculation using den-sity functional theory (DFT) corroborated these experimental findings and resolved the remain-der of the reconstruction geometry[Starke et al., Physical Review Letters 80, 758 (1998)]. As shown in Fig. 100, in addition to the adatom and its three next neighbors the surface is covered by a Si adlayer that in turn is covering the topmost SiC substrate bilayer.
Figure 100: (33)-SiC(0001): (a) Top view: Lat-eral relaxations within the silicon adlayer are indi-cated with respect to the ideal positions in the top-most SiC bilayer. (b) Side view along [11¯20] di-rection. Red and blue atoms as revealed by LEED holography.
A strong rotation (‘twist’) of 9Æ2Æwithin the adlayer (Fig. 100(a)) is observed which results in displacements of up to 0.74 ˚A. As a con-sequence the epitaxial mismatch of the Si ad-layer to the SiC substrate is relieved and the in-teratomic distances within the adlayer are be-tween 2.31 ˚A and 2.35 ˚A close to the value of 2.35 ˚A for an ideal Si–Si bond length. In addi-tion as shown in Fig. 100(b) all atoms are sit-uated in a single layer being three-fold coordi-nated to their Si neighbors with 120Æ bond an-gles and one-fold coordinated to the Si atoms of the substrate bilayer. Thus, these Si atoms are effectively sp2-hybridized and their four bonds fully saturated. The only and single remaining dangling bond per unit cell is located at the Si
The SiC crystal can be constructed by stacking hexagonal bilayers on top of each other. The crystal structure is determined by the mutual in-plane orientation of adjacent bilayers. Dif-ferent stacking sequences lead to difDif-ferent so-called polytypes, of which a large number ex-ists (Fig. 101(c) displays the 4H-SiC with a ro-tation after every two bilayers). This polytyp-ism imposes a severe problem to single crys-tal growth. However, the high degree of bond saturation present in the (33) reconstruction provides a very effective passivation of the sur-face which explains the good homoepitaxial growth possible under Si rich growth condi-tions in chemical vapor deposition (CVD) and molecular beam epitaxy (MBE): A successful growth process in this manner requires the sub-strate to be cut slightly tilted with respect to the basal plane, i.e. in a so-called off-axis ori-entation. On such a substrate with the (33) phase present (as observed during MBE growth [Tanaka et al., Applied Physics Letters 65, 2851 (1994)]) the surface passivation leads to a high mobility of incoming particles such that they can diffuse along the terraces and attach them-selves to a step. The new material continues the periodic structure of the bilayer and thus re-produces the stacking sequence of the substrate which is exposed and interrupted at the steps.
Thus, by allowing for a step flow growth mode the (33) phase facilitates a homo-polytype epi-taxy.
Figure 101: (
3
3)R30Æ phase on SiC(0001) with the adatom marked in light blue: (a) Side view projection along the [11¯20] direction. (b) Top view.
(c) Bulk structure of 4H-SiC. The stacking sequence
Annealing the (33) phase in UHV leads to a (
3
3)R30Æ phase. This phase represents a simple reconstruction with a Si adatom on top of the SiC substrate in a so-called T4 po-sition, i.e. three-fold coordinated to the top-most Si atoms and on top of a carbon atom in the next layer (see Fig. 101(a) and (b) for a side and top view of this structure). A surprising concomitant of this reconstruction is the deli-cate dependence of the substrate bilayer stack-ing on the chemical environment durstack-ing prepa-ration which we explain for the example 4H-SiC (Fig. 101(c)): While heating a chemically prepared sample leads to a bulk truncated sub-strate layer sequence with the two topmost bi-layers in the same orientation, the Si rich prepa-ration conditions represented by annealing the (33) phase result in a surface slab of three bi-layers in the same orientation which breaks the 4H-SiC bulk symmetry.
From morphology changes observed in STM we deduce that strong material transport leads to this addition of an additional bilayer in the same orientation on top of the regular zig-zig-zag-zag stacking in the bulk [Starke et al., Physical Review Letters 82, 2107 (1999)]. Oxy-gen adsorption/desorption cycles lead to the re-versed situation with again two identically ori-ented bilayers which indicates that the bilayer rearrangement is strongly sensitive to the chem-ical environment.
A well ordered oxide termination can be pre-pared on SiC by a thermal hydrogen treatment under atmospheric pressure. Hydrogen serves as an etching agent to generate a flat SiC face. The outermost layer of this native SiC sur-face apparently gets oxidized, presumably by residual oxygen present during or after the re-action. Brought into UHV without further treat-ment, the surface displays a (
3
3)R30Æ LEED pattern with bright and sharp spots and practically no background on surfaces of both polarities, i.e. SiC(0001) and SiC(000¯1). (Note, that the spot intensities are markedly different from the above (
3
3)R30Æ phase, indicat-ing a completely different surface structure.)
About one monolayer of oxygen is present in these phases as determined by AES which also indicates a SiO2-type bonding in this layer. The atomic arrangement was determined by LEED structure analyses as a silicon oxide monolayer above an otherwise bulk-truncated crystal.
Figure 102: (
3
3)R30Æ-silicate adlayer on SiC: (a) Top view showing the honeycomb-type ar-rangement with Si–O–Si bonds. (dark shaded area indicates the (11), light shaded the (
3
3)R30Æ unit cell). (b) Side view projection along the [01¯10]
direction for SiC(000¯1). (c) Equivalent side view of the oxide structure on SiC(0001), where linear Si–O–Si bonds connect the silicate layer and the SiC substrate.
As shown in Fig. 102(a) the layer consists of a honeycomb-like arranged sublayer of two Si atoms per (
3
3)R30Æ unit cell con-nected by an overlayer of two-fold coordinated oxygen atoms. This layer which resembles a silicate-like structure, is directly connected to the topmost SiC bilayer by a Si–C bond (panel (b)) on SiC(000¯1) while on SiC(0001) a linear Si–O–Si bridge mediates the contact (panel (c)) [Bernhardt et al., Applied Physics
Letters 74, 1084 (1999)]. The lateral unit vector of the (
3
3)R30Æ periodic lattice matches that of bulk SiO2 within 95%. The only dif-ference between the silicate monolayers on SiC and the bulk structure of a high temperature SiO2 phase known as β-tridymite is the posi-tion of the Si atoms. In the bulk structure a sili-cate layer consists of three sublayers with the Si atoms alternatingly positioned below and above
the oxygen atoms. Hypothetically, the sili-cate adlayer found on SiC can be transformed into this structure simply by shifting one of the two Si atoms in the unit cell upwards into this upper Si sublayer position. In that way the (
3
3)R30Æ silicate layer may be a useful seed for further growth of oxide films and help to improve the performance of SiC-MOS struc-tures.