PROMPT aimed to study:
-the influence of H2O traces in an O2 ambient on oxidation mechanisms and kinetics in the thin-film (4-40 nm) regime
-the influence of silicon surface conditions on the oxidation of silicon (crystallography, damage, impurities, etc)
-the incorporation of impurities (eg H) during Al or poly-Si deposition to form MOS contacts
-the electrical properties of MOS test-sets, made under the conditions used above, correlated with growth conditions.
An improved understanding of the influence of material processing variables on the electrical characteristics of metal oxide silicon (MOS) structures was sought. An ultrahigh vacuum (UHV) processing facility was used to provide the ability to grow dry oxides on silicon and to deposit clean metal and polysilicon contacts. Surface and interface characterization techniques were applied. Understanding the processing variables will lead to improved fabrication techniques for ultra large scale semiconductor devices.
An ultraviolet (UV)/ozone dry cleaning station has been added to the cluster tool and a design study for the addition of a hydrogen atom source dry cleaning stage has been completed.
The effects of wet chemical cleaning procedures on MOS breakdown strength have been investigated. The cleaning procedures leave differing amounts of carbon on the native oxide produced during cleaning with levels below the detection limit. Little difference in MOS breakdown strength was observed providing the oxide growth step of fabrication was carried out with the native oxide intact. If the native oxide was desorbed by heating in UHV the MOS capacitors failed unless the desorption temperature was high and even then they perform poorly. Silicon carbide formation during oxide desorption was a serious problem for some cleaning procedures, while even when it was absent the silicon surface roughness was high and degraded the MOS devices.
A parallel study in conventional processing showed a similar correlation between silicon surface roughness and MOS breakdown performance.
The fabrication of metal contacts to silicon dioxide in UHV has shown that with appropriate attention to in situ removal of surface absorbates, improvements to contact adhesion strength of an order of magnitude can be achieved.
A recent study of gate oxide breakdown strength using a conduction atomic force microscope indicated that breakdown occurs even in the highest quality material, at small loca lized defects.
APPROACH AND METHODS
MOS test samples containing capacitor sets from 100 to 1000 micron diameter were fabricated on 15 mm square silicon samples in the UHV processing facility. Various processing stages, such as in situ cleaning, oxidation, and contact deposition, were carrie d out. The samples are then removed to a clean-room environment for photolithography and etch patterning. MOS test-sets made by conventional clean-room atmospheric furnace oxidation and contact deposition were examined for comparison with UHV processing. The examination of surfaces and interfaces by low and medium energy ion scattering, nuclear reaction analysis (for 18O and 16O and H) and scanning tunnelling microscopy (STM) have been carried out. These measurements are an essential part of the study to characterise the various process stages used in MOS fabrication.
In addition to these MOS processing studies, investigations of fundamental aspects of oxidation mechanisms and oxidation kinetics have been undertaken. These studies rely on the use of 18O and 16O isotopes as the oxidising ambient to give a means of high-accuracy nuclear reaction oxide thickness determination and the determination of the depth distribution.
PROGRESS AND RESULTS
-The UHV Cluster Tool we have built to fabricate MOS test sets on 15mm by 15mm Si samples has functioned effectively. The cluster tool in situ medium energy ion scattering and scanning tunnelling microscope diagnostic tools have been upgraded. A UV/ozone "dry-cleaning" station has been added to the cluster tool and a design study for the addition of a H atom source "dry-cleaning" stage has been completed. The in-situ diagnostics and "dry-cleaning" process stage have proved invaluable in our MOS process s tudies.
-The effects of wet chemical cleaning procedures (HF, RCA, Shiraki, etc.) before loading Si into the UHV Cluster Tool and UV/Ozone cleaning in the Cluster Tool on MOS breakdown strength have been investigated. The cleaning procedures leave differing amou nts of carbon on the "native-oxide" produced during cleaning with levels below the detection limit for our Cluster Tool ion scattering diagnostics in the UV/Ozone case. However, little difference in MOS breakdown strength (ca 13MV/cm) was observed providi ng the oxide growth step of fabrication was carried out with the "native-oxide" intact. If the "native-oxide" is desorbed by heating in UHV the MOS capacitors fail unless the desorption temperature is high (1150C) and even so they perform poorly (ca 10MV /cm). The STM Cluster Tool diagnostic showed that SiC formation during oxide desorption is a serious problem for some cleaning procedures, while even when it is absent the Si surface roughness is high and degrades the MOS devices.
-A parallel study in conventional processing also showed a similar correlation between Si surface roughness and MOS breakdown performance. Evidence was also found in this case that metallic contamination was correlated with reduced breakdown, either beca use it leads to surface roughening or because rough surfaces trap metals; more work is required to understand this.
-The fabrication of metal contacts to SiO2 in UHV has shown that, with appropriate attention to in-situ removal of surface absorbates, improvements to contact adhesion strength of about an order of magnitude can be achieved.
-A recent study of gate oxide breakdown strength using a conducting atomic force microscope indicates that breakdown occurs, even in the highest quality material, at small localised defects. This aspect of our research is at an early stage but indicatesthat gate oxide reliability is still limited by our process defects rather than by the "intrinsic" dielectric strength of the oxide. One thus deduces that improvements in gate oxide reliability are possible, perhaps by improving surface roughness or metalimpurity levels.
The results of the ultra clean processing investigations will enable the development of improved VLSI fabrication techniques to be made. The PROMPT Action has already led to an improved understanding of the relation between process variables and the consequent electrical performance of MOS structures. The prospect for further improvements, particularly in understanding the role of surface roughness at both the Si and SiO2 interfaces is encouraging.
The experience gained on process research with the UHV Cluster Tool is expected to provide useful input for the design and use of the next generation of Ultra Clean Cluster Tool equipment for VLSI fabrication. The surface diagnostic equipment developed inthis programme has lead to the spin-off of a company to provide custom built SPM instruments (East Coast Scientific Ltd, UK) and will be followed by a commercial launch of medium energy ion scattering instrumentation.
CB2 1PZ Cambridge