Neutron Transmutation Doped Silicon for Power Semiconductor Devices

Abstract

Power semiconductor devices operate at high currents and voltages. The fabrication of these devices requires high-resistivity silicon wafers with good spatial uniformity over large areas. Neutron transmutation doping (NTD) has been demonstrated to be an excellent technique for achieving these requirements. The development of commercially available NTD silicon has consequently led to substantial improvements in the yield of high- voltage power rectifiers and thyristors. In addition, the greater spatial uniformity, as well as the precise control over the resistivity achievable by using the NTD process, has led to a substantial increase in the breakdown voltage capability of thyristors.

This paper reviews the impact of using neutron transmutation doped silicon for power-device fabrication. Experiments conducted to verify that defects produced during NTD processing can be subsequently annealed out will be described. These experiments were crucial to allaying initial concerns regarding the quality of neutron transmutation doped silicon and were instrumental in achieving the present widespread use of this process for fabrication of starting material for the power-device industry.

The year 1982 marks the twenty-fifth anniversary of the development of the power thyristor. It is therefore appropriate at this juncture to review the impact of neutron transmutation doping (NTD) of silicon upon the fabrication of high-voltage, large-area power thyristors. The basic concept of doping silicon by creating phosphorus atoms by the absorption of thermal neutrons was discussed in 1961 by Tanenbaum and Mills (1). However, it was not developed into a production process until recently. In 1975 Herman and Herzer wrote a paper (2) in which they stated that:

The results indicate that transmutation doping may be a rather good technique for production of homogeneously doped silicon material... Further work must be undertaken to establish whether transmutation doping will play a dominant role in addition to the standard methods for growing silicon crystals.

on which Burtscher (3) commented in 1976:

Herman and Herzer write about the chances of neutron-irradiated silicon perhaps somewhat too cautiously... The use of neutron- irradiated silicon for manufacturing large area devices (especially high power devices) is by no means a question of the future... The use of neutron-irradiated silicon for the production of high voltage, high power devices has become standard.

Since 1976 this statement has been reinforced to the degree that almost all high-voltage, large-area power thyristors are presently fabricated by using neutron transmutation doped silicon as the starting material. This trend is expected to continue into the future.

In order to examine the impact of the development of neutron-doped silicon upon power thyristors, it is interesting to examine the growth in the power ratings of these devices over the last 25 years. The progressive increase in the blocking-voltage capability (which is related to the device breakdown voltage) and the current-handling capability (which is related to the device area) are shown in Figure 1 and Figure 2. A careful look at these figures indicates that after rather gradual increases in the blocking-voltage and current-handling capability from 1958 to 1975, these power ratings of the thyristors have shown a much more rapid increase from 1975. This upsurge in the power ratings is directly attributable to the availability of the improved starting material for thyristor fabrication as a result of the development of NTD silicon.

When the concept of using neutron-doped silicon for power devices was initially proposed in the early 1970s, it was clear that significant improvements in device power ratings would accrue

from the projected improvement in doping homogeneity within the wafers. However, several important concerns with regard to this new material were expressed by the power-device industry. These concerns were related to the secondary reactions listed in Figure 3 which cause the production of sulfur due to thermal neutron absorption by the phosphorus (P³¹) and production of magnesium by the fast neutron absorption by silicon (Si³⁰). These elements even at low concentration levels were expected to reduce the minority-carrier lifetime because they create deep lying levels in the silicon energy gap which constitute recombination centers. In addition, it was known that after the neutron doping had been performed, the silicon lattice contained a high level of damage due to gamma recoil during the transmutation reaction and the irradiation of the crystal by the high energy P particles emitted during the decay of Si³¹ to P³¹. Further, the nuclear reactors in which the thermal neutron capture was performed for the doping were known to contain a background of fast neutrons. The collision of fast neutrons with silicon atoms was known to create localized damage clusters which could represent serious problems because they might influence the diffusion front during device fabrication. This would severely degrade device breakdown voltage negat-ing the sought-after advantages of the homogeneous doping.