Physical Model For Contact Degradation At High Current Densities

Abstract

A new physical model on electrical degradation of the metaYpolysilicodsilicon contact at high current densities is presented and verified by experiments. Results from submicron n+poly/n+emitter/p-base/n+collector bipolar transistors indicate that the observed contact instabilities are reaction rate limited by transient dehydrogenation and hydrogenation at the n+poly/n+emitter interface and steady-state dehydrogenation at the metal/n+poly interface. Introduction Scaling down transistors in present and future integrated circuits increases the current densities which pass through the interconnect/silicon contacts. Thus, the long-term stability of the contacts in BJTs and MOSTs can be a potential reliability problem. Instabilities of the emitter contact resistance RE and current gain pF were observed in submicron B JT’s with metal/n+poly-silicon/n+siliconemitter (metaVn+poly/n+emitter) structure when stressed at h i g h f o r w a r d e m i t t e r c u r r e n t d e n s i t y JE-s t ress > -1.0mA/pm2 [ 1-31. Detail experiments on contact instability were reported recently [3] showing increasing and decreasing pF with stress time and two rate constants both with A,J, + A,Ji dependency. These dependencies were attributed to the hydrogenation and dehydrogenation reactions at the metal/poly-Si and poly-Si/c-Si interfaces. However, a viable physics-based theory to explain and model the instability dependencies on current density was not given in [3] which is presented here. Model and Experiments Figure 1 (a) shows the hydrogenation-dehydrogenation pathways in the metal/n+poly/n+emitter structure of the BJT used in the experiment and the physical model development [4]. Electrons (filled circles) flow into the contact which is forward-biased to V, = 1V to 2V. Holes (open circles) are injected from the p-base into the n+poly and n+emitter layers. The limiting transient reaction is the hydrogenation-dehydrogenation at the n+poly/n+emitter interface which decrease-increase respectively the density of the interfacial silicon and oxygen dangling bonds, Si. and SiO., where electrons and holes recombine to give I, and control pF. The rate equation for the density of one interfacial trap species, nIT(t), its solution for a steady-state hydrogen concentration, and the rate constants are given below which account for all the experimental variations observed in [31. Metal n+polySi ncEmitter p-Base Fig.1 Creation and destruction of interface traps i n the metal/n+poly/n+emitter/p-base contact layers at high forward current densities. (a) Cross-section with kinetics pathways. (b) Transition energy band diagram. NITT in (3) i s t h e t o t a l i n t e r f a c i a l t r a p d e n s i t y , hydrogenated plus not-hydrogenated. The initial and final value, nIT(t=O)=NIT0 and nIT(t=m)ENITm determines whether pF decreases (NIT0 < NIT,) or increases (NIT? > NIT,) with s t r e s s t i m e , wh ich i s c o n s i s t e n t wi th 41 4-93081 3-75-1 197 1997 Symposium on VLSl Technology Digest of Technical Papers exper iments [3] a t l o w and h igh cur ren t dens i t ies respectively. The initial (pre-stress) value is determined by fabrication technology and the final state-steady value, NIT,, is determined by the steady-state mobile (detrapped) hydrogen concentration which is reached immediately after JE i s turned on, which i s much faster than the t ime dependence of nIT(t), because the source of the hydrogen is at the metal/n+poly interface where the hydrogen trapping site concentration is very high and the detrapped and mobile hydrogen concentration is relatively low, and because the hydrogen diffusion rate through the thin n-tpoly layer is large compared with the time dependence of nIT(t) a t the n+poly/n+emitter interface. The two reaction mechanisms at the metal/n+poly interface suggested in Fig. l ( a ) give a s teady-state hydrogen concentration proportional to dJ, and JE. The rate of nIT(t) given by ( 5 ) and (6) has the linear hydrogen capture part, and a hydrogen emission part determined by hydrogen bond-b reak ing r a t e wh ich is p ropor t iona l t o t h e concentration of the hot electrons generated by Auger recombination of two thermal electrons and one hole whose rate i s proportional to N2P. At low JE the electron concentration N at the n+poly/n+emitter region is given by t h e d o n o r c o n c e n t r a t i o n , NDD+, w h i l e t h e h o l e concentration P is proportional to JE if high injection level is reached in the p-base, thus, eH 0~ Ho to H' as indicated in ( 5 ) . A t h igh JE, N and P a r e both asymptot ica l ly proportional to JE so that eH = J i as indicated in (6). These current density dependencies were observed in the stress time dependence of increasing or decreasing OF shown in Fig.2 as -AIB/IBo. The kinetics fits well the exponential dependence given by (2) for each of the two interface trap species with time constants and T~ observed in [3], attributed to the dominant n+poly/n+Si Si. interface traps and the SiO. interface traps at the SiO,/Si interfaces from the residual oxide patches shown in Fig.l(a). The model also accounts for the experimental variations in the emitter res is tance RE observed in [3]. Figure 3 shows the experimental data and theoretical fit of the JE dependence of the PF(t) rate constants extrapolated to a constant device temperature T=25"C from high temperature stress at T-150°C [3]. The least-squares-fit of the these rate data to (6 ) gives