M.-A. Nicolet

Diffusion barriers in thin films

Abstract Some of the conditions which thin film diffusion barriers should satisfy are enumerated. Various ways to try to meet these conditions are illustrated by means of examples. It is shown that metal films chosen for their mutual immiscibility with the adjoining metals (passive barriers) usually fail as barriers ( i.e. are non-barriers) because extended structural defects in the metal film constitute fast diffusion paths. Single-crystal barriers of such metals are effective but not practical. Barriers which are thermodynamically partially stable (partially stable barriers) and those which are fully stable (stable barriers) are discussed. Metal compounds of particular interest for such barriers are suggested. The concept of the sacrificial barrier, which is based on an irreversible loss of barrier material by interfacial reactions with the adjoining metals, is introduced and successful applications are presented. The stabilizing effect of impurities on an otherwise unstable barrier (stuffed barriers), the importance of mechanical stress and the critical influence of the fabrication process of a thin film barrier on its actual performance are described by practical examples. Values of the electrical resistivity for borides, carbides, nitrides and silicides of the early transition metals and values of the coefficient of linear thermal expansion of silicides are compiled for reference purposes.

Tantalum‐based diffusion barriers in Si/Cu VLSI metallizations

We have studied sputter-deposited Ta, Ta36Si14, and Ta36Si14N50 thin films as diffusion barriers between Cu overlayers and Si substrates. Electrical measurements on Si n + p shallow junction diodes demonstrate that a 180-nm-thick Ta film is not an effective diffusion barrier. For the standard test of 30-min annealing in vacuum applied in the present study, the Ta barrier fails after annealing at 500 °C. An amorphous Ta74Si26 thin film improves the performance by raising the failure temperature of a /Ta74Si26(100 nm)/Cu(500 nm) metallization to 650 °C. Unparalled results are obtained with an amorphous ternary Ta36Si14N50 thin film in the Si/Ta36Si14N50 (120 nm)/Cu(500 nm) and in the Si/TiSi2(30 nm)/Ta36SiN50 (80 nm)/Cu(500 nm) metallization that break down only after annealing at 900 °C. The failure is induced by a premature crystallization of the Ta36Si14N50 alloy (whose crystallization temperature exceeds 1000 °C) when in contact with copper.

Influence of the nature of the Si substrate on nickel silicide formed from thin Ni films

Abstract We investigate the rate of silicide formation and the composition of the resulting compound when thin (1000–5000 A) Ni films vacuum deposited onto various substrates are annealed from 200 °C to 325 °C and from 1 2 to 24 h . Only the phase Ni2Si is observed on substrates of (111), (100) and polycrystalline Si. On amorphous (vacuum-deposited) Si the two phases Ni2Si and NiSi form simultaneously in two distinct sublayers. In all cases, the compound layers grow at ( time ) 1 2 , and the activation energies all lie between 1.3 and 1.6 eV. The actual value of the growth rate depends on the substrate. On (100), polycrystalline and amorphous Si substrates the rates are very similar and correspond to about 1000 A at 1 h and 275 °C. On (111) Si substrates the rate is about a quarter of that. Transmission electron micrographs establish that silicide layers grown on different substrates have different microcrystalline structure.

Evaluation of amorphous (Mo, Ta, W)SiN diffusion barriers for 〈Si〉|Cu metallizations

Abstract Amorphous binary M(= Mo, Ta or W)-Si and ternary MSiN, r.f.-sputtered from M 5 Si 3 and WSi 2 targets, are assessed as diffusion barriers between silicon substrates and copper overlayers. By I ( V ) tests of the metallizations on n + p shallow junction diodes, the ternary MSiN barriers prevent copper from reaching the silicon at 800 °C or higher during a 30 min heat treatment in vacuum. Failure of the metallizations correlates with the crystallization temperature of the barrier, which is presumably a prelude to fast grain-boundary diffusion. Metal-rich MoSiN and WSiN barriers liberate nitrogen during annealing, which poses a limitation to their crystallization temperatures. No reaction products of copper with metal-rich MSi or MSiN barriers are observed, which is in agreement with our recent thermodynamic modelling of the MSiCu ternary systems.

REACTIVELY SPUTTERED TI-SI-N FILMS. I. PHYSICAL PROPERTIES

Films of Ti-Si-N were synthesized by reactively sputtering TiSi2, Ti5Si3, or Ti3Si targets in an Ar/N2 gas mixture. They were characterized in terms of their composition by MeV 4He backscattering spectrometry, their atomic density by thickness measurements combined with backscattering data, their microstructure by x-ray diffraction and high-resolution transmission electron microscopy, and their electrical resistivity by four-point-probe measurements. All films have a metal–to–silicon ratio close to that of their respective targets. The as-deposited films are either entirely amorphous or contain inclusions of TiN-like nanometer-sized grains when the overall atomic composition of the films approaches the TiN phase in the ternary Ti-Si-N diagram. A correlation between the resistivity of the as-deposited films and their position in the ternary phase diagram is evident, indicating that at the atomic scale, the spatial arrangement of atoms in the amorphous phase and their bonding character can approximate those...

Reactively sputtered RuO2 diffusion barriers

The thermal stability of reactively sputtered RuO2 films is investigated from the point of view of their application as diffusion barriers in silicon contact metallizations with an Al overlayer. Backscattering spectra of Si/RuO2/Al samples and electrical measurements on shallow junction diodes with Si/TiSi2.3/RuO2/Al contacts both show that RuO2 films are effective diffusion barriers between Al and Si for 30-min annealing at temperatures as high as 600°C.

Iron silicide thin film formation at low temperatures

Abstract The rate kinetics of the formation of compound phases from thin layers of 1000–1500 A α-Fe deposited onto single-crystal Si have been studied by MeV 4He+ backscattering spectrometry. Si is observed to dissolve into the thin α-Fe layer before compound formation. A compound layer of FeSi is produced at about 450°C and FeSi2 formation begins at about 550°C. The (100) surface of Si is slightly more reactive than the (111) surface. An inert diffusion marker of implanted Xe was used to investigate the relative movement of the two species. X-ray diffractometry identifies the structure of the compound species as identical to bulk FeSi and FeSi2. The compounds formed on both (111) n-Si and (100) n-Si are apparently polycrystalline and untextured.

Effects of electrically active impurities on the epitaxial regrowth rate of amorphized silicon and germanium

Abstract The influence of electrically active n-type ( 75 As) and p-type ( 11 B) impurities on the solid phase epitaxial regrowth of ion-implanted amorphized Si〈100〉 and Ge〈100〉 has been studied for low temperature furnace annealing. Both types of impurity increase the rate of regrowth of both silicon and germanium at a concentration level of 10 20 cm -3 . Above this level, 75 As retards regrowth in germanium. In compensated surface layers, the regrowth rate slows down to the values observed in self-implanted or intrinsic crystals for both silicon and germanium. The results can be qualitatively explained in terms of electrically induced generation of point defects at the amorphous-crystalline interface.

An improved forward I-V method for nonideal Schottky diodes with high series resistance

Two methods are described to obtain the value of the series resistance<tex>(R)</tex>of a Schottky diode from its forward I-V characteristic. The value of<tex>R</tex>is then used to plot the curve ln (<tex>I</tex>) versus<tex>V_{D} (= V - IR)</tex>which becomes a straight line even if ln<tex>(I)</tex>versus<tex>V</tex>does not. The ideality factor<tex>n</tex>and the Schottky-barrier height<tex>\Phi_{B0}</tex>of the diode then follow from the standard procedure. The main advantages of the methods are: 1) a linear regression can be used to calculate the value of<tex>R</tex>, 2) many data points are used over the whole data range which raises the accuracy of the results, and 3) the validity of constant<tex>R</tex>assumption can be checked by the linearity of the ln<tex>(I)</tex>versus V<inf>D</inf>curve. The methods are illustrated on the experimental data of a real diode.

The first phase to nucleate in planar transition metal-germanium interfaces

Abstract The first compound to nucleate in binary reaction couples at subeutectic temperatures can be predicted by a rule proposed by Walser and Bene. This paper reports on the formation of germanides of Co, Hf, Mn, Ni, Pd and Rh at temperatures ranging between 100 °C and 600 °C. With the exception of the PdGe system, the first compounds formed agree with the rule of Walser and Bene. Possible nucleations of a second and third compound are studied also.