晶片表面取向对极薄氧化物质量的影响

We studied effects of silicon wafer surface orientation on very thin oxide quality, testing Si(100)and (111) wafers. It has been found that the very thin oxide structure is dominating by thesilicon wafer surface orientation, and that when Si(111) is oxidized, SiO,/Si(111) interfacemicroroughness increases as oxide gets thicker than 10 nm, resulting in a degradation of oxidefilms quality on Si(111). When oxide thickness is decreased to less than 10 nm, SiO,/Si interfacesmoothness is maintained similar to Si(100) and (111) but the SiO/Si interface exhibits largerinterface charges and larger flatband voltage shift for Si(111) than for Si( 100).

Perfect control of the silicon wafer surface is essentialin order to realize future ultralarge scale integrated(ULSI) devices. As a gate oxide gets thinner together withhigher integration of large scale integrated devices, electrical characteristics of the oxide are strongly affected by sil-icon wafer quality. Silicon wafer surface orientation is regarded as one of the factors that governs performance ofmetal-oxide-semiconductor (MOS) devices. It is wellknown that the interface charges of Si(100) arelower than that of Si(111) and that the electron channelmobility of metal-oxide-semiconductor field-effect transis-tor (MOSFET) on Si(100) is higher than that onSi(111).! Therefore, a Si(100) wafer is mainly employedin the current manufacturing line, but the effects of siliconwafer surface orientation on a very thin oxide quality havenot yet been sufficiently revealed.

Smoothness of the SiO,/Si interface is a very impor-tant factor to realize high quality very thin gate oxide films.It has been reported that the increase of surface microroughness degrades the electrical properties of the thin ox-ide films.2-6 It has also been reported that the increase ofsurface microroughness caused by NHOH/H,O/HO(APM) cleaning greatly depends on the types of Siwafer, such as Czochralski, floating zone, and epitaxial.2-6However, the relationship between surface microroughnessand the silicon wafer surface orientation has not yet beenrevealed.At present, surface microroughness of the siliconwafer is evaluated with a scanning tunneling microscope(STM) and atomic force microscope (AFM).We are ableto obtain more detailed data on microroughness.

This letter describes the effects of silicon wafer surfaceorientation on a very thin oxide quality. It also touchesupon the behavior of microroughness of the SiO/Si inter-face.

In this experiment, the oxidation was carried out in theultraclean environment characterized by extremely lowmetal and airborne impurity concentrations to reveal theeffect of silicon wafer orientation only.7-9 The very thinoxide was formed by dry oxidation at 900 C, where theseexist, 0.4 nm preoxide formed at 300 CMOS devices were fabricated after field oxide formation by wet oxidationat 1000°C in conventional furnace, where oxide was removed for the device area. Just (100) and (111) orientedsilicon wafers were used. The surface microroughness wasmeasured by AFM, where the height accuracy was cali-brated at the order of 0.1 nm by using standard samples.10Si/SiO interface microroughness was measured withAFM after removing the oxide with the advanced bufferedHF(BHF).

Figure 1 shows the time dependence of oxide thicknesson Si(100) and (111) formed by ultraclean dry oxidationat 900 °C, where the oxide thicknesses were determined byx-ray photoelectron spectroscopy (XPS) calibrated withellipsometry for oxides with thicknesses thinner than 14nm or by ellipsometry for the others.12 It is seen in Fig. 1that the oxidation rate of Si(111) is higher than that ofSi(100). The oxide growth on Si(100) obeys a simple par-abolic law, exhibiting a slope of 0.5. It can be explained bythe Deal--Grove model even when oxide thickness is lessthan 30 nm.On the other hand, the oxide growth onSi(111) features slope of 0.65 when oxide thickness is below 20 nm.

Figure 2 shows the Si XPS spectra of 1.7 and 2.7 nmoxides on Si(100)and Si(111), where Sizp spectra forSi(100) and (111) substrates have been confirmed to coincide with each other at 10 meV. The binding energy ofthe oxide peak for Si(100) is higher than that for Si(111)by a factor of 0.20 eV. These results suggest that the verythin oxide structure is dominated by the silicon wafer sur-face orientation.