The deposition mechanism of boron doping in CVD silicon epitaxy has been investigated by exposing silicon substrates t o B,H,-H, doping gas mixtures a t epitaxy temperatures and examining the effect by dopant profile measuring in an afterwards intrinsically in-situ deposited epitaxial silicon layer. It has been shown that boron is deposited increasing its concentration on the surface linearly with prolonged exposition time and desorbed by purging the surface in pure hydrogen. I n the latter case its content decreases linearly proportional to the predeposited concentration. The desorbed boron builds up a secondary doping source which maintains a parasitic boron flow for reincorporation during following layer growth.Es wurde der Mechanismus des Boreinbaus wiihrend der Abscheidung von epitaktischen CVD Siliciumschichten untersucht, indein Siliciumsubstrate der Einwirkung von B,H,-H, Dotiergasrnischungen bei Epitaxietemperaturen ausgesetzt wurden. Der erreichte Effekt wurde durch die Auswertung der unmittelbar anschlieaend in einer dunnen und intrinsisch abgeschiedenen Siliciumschicht erhaltenen Dotierungsprofile gemessen. Danach nimnit die Oberfliichenkonzentration von Bor in der Vorabscheidungsphase linear mit der Expositionszeit zu und wiihrend einer Spultemperung in reinem Wasserstoff linear proportional zur Vorabscheidungskonzentration ab. Das desorbierte Bor gibt AnlaB zur Bildung einer Sekundarquelle die wiihrend einer nachfolgenden Sehichtabscheidung einen parasitiiren Boreinbau verursacht. 106.

From discussing the influorices of layer growth duration and dcpositiori temperaturo on the slope of the steep autodoping profile branch adjacent to the substrate, i t is concludcd that there exists a special autodoping part, termed redistribution autodoping, which is independent o n solid state diffusion effects, but should be rostricted to the bogiririing of the layer growth.Aus den Untersuchungsergebnissen ubor den EinfluB von Abscheidctcmpcratur und -dauer auf das AusmaB des steilen, substratseitigen Autodopingverlaufs iri Siliziumopitaxioschichten wird aiif das Vorhandenseiri cines von der Festkorperdiffusion unabharigigeri, aber auf die Anfangsphase der Schichtabscheidnng beschrlirikten Autodopinganteils geschlosson. Uieser Autodopiriganteil wird hervorgerufen durch eine l>otandenumvt.rteilung von der Substratoberfliiche zur Schichtoberfliiche wiihrend des Aufwachsens der erstcri Schichtatornlagori und Umverteilungsautodoping genannt. IritruductioiiIn epitaxial silicon layers, deposited on substrates, which contain at least locally limited, highly-doped regions of sanie conductivity as the layer itself, there exists a transition region of dopant density. This phenomenon, known as autodoping, is commonly attributed to two different niechanisnis, the solid state diffusion of dopants (GROVE) and the reincorporation of dopants, having been evaporated from the substrate during the prebake processes (BASSECHES et al.).As early as 1965 GROVE interpreted autodoping to be the only result of solid state diffusion, neglecting other factors in action. Indeed, an abrupt change of concentration within a seiiiiconductor volume is more and more flattened with prolonged thermal treatments at sufficiently high temperatures. The outdiffusion of arsenic from a buried layer into epitaxial silicon should, for example, be described by the well-known equation (1).This paper is concerned with the question, whether the steep branch of the autodopirig profile can theoretically be deduced from solid state diffusion only. A niethod, already previously published, was used in interpreting the really nieasured autodoping profile by two exponential branches, one of them approxiniating the gas phase autodoping part and the other constituting the steep autodoping part under consideration (KUIINE). I n this way, the extent of the autodoping effect can be obtained by nieasuring the slope of the straight lines found in the sernilogarithniated plots (Fig. 1). The values measured are independent of where the metallurgical transition from the substrate to the layer is actually located and of how high the intentional layer doping

The multi‐wafer problem of lateral autodoping has been investigated in a rectangular, horizontally arranged epitaxy reactor by arranging on the top‐side As‐diffused source wafers and smaller high‐resistivity, p‐type sensor wafers which were for indicating lateral autodoping dose after a short‐lasting intrinsic layer deposition. Wafers were evaluated by sheet resistance measurements, IR‐determination of layer thickness and spreading resistance profiling. A linear increase of lateral autodoping dose along the susceptor has been proved and a linearly increasing parasitic dopant partial pressure deduced. These findings give evidence for the validity of the multi‐wafer model of lateral autodoping previously developed by the authors.

A particular test structure is presented the design of which has been based on several buried layers formed as circular rings with a suitably selected distance from each other. I n an environment (concerning both substrate wafer and epitaxy layer) having an opposit,e type of (but low enough) conductivity with respect to buried layer, lateral autodoping will reveal a conducting channel between adjacent buried layers the sheet resistance of which can simply be measured. -It is shown that all those known effects influencing lateral autodoping as for instance buried layer doping level and substrate prebake can be revealed by sheet resistance measurement of the ring-shaped test structure, which additionally can be applied t o wafer mapping of lateral autodoping.Es wird eine speeielle Teststruktur zur Erfassung des lateralen Autodoping beschrieben. Ihr Aufbau beruht auf der Anwendung zentrisch angeordneter, kreisringformiger begrabener Gebiete, die einen geeigneten Abstand zueinander aufweisen. Der laterale Autodopingeffekt ruft einen elektrisch leitenden Kana1 zwischen zwei solchen Kreisringgebieten hervor, wenn das umgebende Silicummaterial einen zu den begrabenen Gebieten entgegengesetzten LeitfBhigkeitstyp aufweist und hinreichend hochohmig ist. Der Schichtwiderstand eines solchen Kanals kann in einfacher Weise gemessen werden. Es wird gezeigt, da13 die an und fur sich bekannten Effekte, durch die das laterale Autodoping herabgesetzt wird (z. B. Dotierungsniveau der begrabenen Gebiete, Substratvortemperung), durch die Messung des Schichtwiderstandes der Teststruktur quantitativ erfaat werden und auch Wafer-Mapping des lateralen Autodoping moglich ist.

No abstract