The electrical effects of dislocations has been studied by modeling zero-bias resistance-area product (R^sub 0^A) of long wavelength infrared diodes fabricated in molecular beam epitaxy (MBE)-grown HgCdTe-Si epitaxial films. Results show that dislocations influence both 40 K and 78 K R^sub 0^A products in high dislocation density (HgCdTe/Si) material. In low dislocation density samples (HgCdTe/CdZnTe), the variations in 78 K R^sub 0^A are limited by the composition (x) variations in Hg^sub 1-x^Cd^sub x^Te material, whereas dislocation contribution dominates the variations at 40 K. The origin of relatively large spread in 40 K R^sub 0^A in both types of samples is traced to the statistical variations in the core charges of dislocations. It is concluded that additional alternatives besides the reduction of dislocation density (such as control of core charges), may also need attention in order to make Si a viable substrate material for the growth of HgCdTe epitaxial layers suitable for devices operating at 40 K.

Key words: HgCdTe-on-Si, infrared detectors, long wavelength infrared (LWIR), zero-bias resistance-area productIn pursuit of developing low-cost technology to fabricate high density, large size, long wavelength infrared (LWIR) detector arrays for thermal imaging applications, recent years have witnessed considerable activity on the growth of HgCdTe epitaxial layers on Si substrates by molecular beam epitaxy (MBE) technique.1-8 Short wavelength infrared (SWIR) and mid-wavelength infrared (MWIR) arrays have been previously reported in the literature, using MBE-grown HgCdTe epitaxial layers on 4-in. (211) Si substrates.2-6 These arrays are generally operated at 77 K, where contribution of dislocations is within tolerable limits for some tactical applications. High dislocation density in the range of mid-10^sup 6^ cm^sup -2^ in these epitaxial layers has, however, prevented the realization of uniform LWIR arrays suitable for operation at 40 K for some strategic applications. Thus, present thinking appears to revolve around efforts to reduce the dislocation densities in HgCdTe/Si below 10^sup 6^ cm^sup -2^, as is the case with the HgCdTe/CdZnTe system. Given the large (19%) mismatch between the lattice-parameters of Si and HgCdTe (a^sub Si^ = 5.43 A^sup 0^, a^sub HgTe^ = 6.453 A^sup 0^, and a^sub CdTe^ = 6.48 A^sup 0^), as compared to lattice-matched CdZnTe substrates, it may not be entirely possible to match the dislocation density levels in the two systems. With such a possibility in mind, this paper proposes to explore the basic origin of the large spread of 40 K R^sub 0^A product by modeling the reported results of LWIR diodes fabricated in HgCdTe/Si epitaxial layers.

MODEL

The Rockwell group used (211) Si/ZnTe/CdTe substrates to grow HgCdTe epilayers by MBE. The structure of the active layers consisted of an n-HgCdTe absorber layer (LWIR) followed by a wider bandgap P-layer. Diodes were fabricated in double-layer planar heterostructure (DLPH) architecture by selective area ion implantation. Arsenic was implanted through photoresist windows and followed by annealing for cap layer diffusion and activation of arsenic. The diodes were finally passivated with a CdTe layer. Similar diodes fabricated on the epitaxial layers grown on bulk CdZnTe substrates were also studied for comparison. It was reported that, at 78 K, diode performances of MBE LWIR HgCdTe on Si are comparable to that of LWIR HgCdTe on bulk CdZnTe substrates, both in terms of R^sub 0^A values and their variations. However, at 40 K, comparable diode performance in terms of R^sub 0^A value is occasionally obtained due to very large variations in R^sub 0^A. In the following paragraphs, we discuss the comparison of the experimental results with the theory in both types of diodes, with the objective of tracing the origin of large variations in R^sub 0^A at 40 K.

Figure 1 shows a plot of R^sub 0^A product of small area (~10^sup -6^-10^sup -5^ cm^sup 2^) LWIR HgCdTe diodes fabricated in epitaxial layers grown on bulk CdZnTe substrates. The dislocation density in the epitaxial layers has been stated as 8 × 10^sup 4^ cm^sup -2^. The experimental data (discrete points) shown here has been taken from Figs. 4a and Ua of Ref. 8. It is clearly observed that R^sub 0^A product of the diodes exhibit a much larger spread at 40 K than at 78 K. Let us first discuss R^sub 0^A variations at 40 K, which can be explained in more than one way as discussed below.

Discussions given in the following paragraphs are based on the understanding that, at 40 K, thermal diffusion current contribution is negligibly small and that the impedance of dislocations that intersect the junction dominates the diode impedance.

Tobin et al.14 explained variations of 40 K R^sub 0^A product in a variable area diode array by assuming Poisson distribution of defects/dislocations within the base material, implying that the dislocations were nonuniformly distributed. It can be clearly seen from Eq. 1, that R^sub 0^A will be independent of diode area for a uniform distribution of dislocations. Poisson statistics were then used to calculate the probability that a diode of a given size had a particular number of dislocations. In this model, small area diodes in an array exhibited a larger spread than the larger area diodes. Though the model is able to qualitatively explain R^sub 0^A variations in a variable area array of HgCdTe diodes at 40 K, it does not appear to conform to the real-time situation encountered in practice. For example, consider the 40 K R^sub 0^A spread shown in Fig. 1. Experimental values vary by an order of magnitude from 10^sup 5^ to 10^sup 6^ ohm-cm^sup 2^. At the given dislocation density of 8 × 10^sup 4^ cm^sup -2^ in the base material, a small area diode (10^sup -6^-10^sup -5^ cm^sup 2^) can either have zero or one dislocation within the diode area, which is insufficient to explain the large spread shown in Fig. 1. In fact, an order of magnitude variation of R^sub 0^A in diodes of similar areas here demands variations in the number of dislocations (within the diode area) from 0 to 9. This is only possible if some local regions in the base material have variable dislocation densities varying up to 9 × 10^sup 6^ cm^sup -2^. In practice, this appears to be a highly unlikely event, especially when efforts are focused to fabricate uniform arrays with very high operability.