Excess Noise and Avalanche Multiplication in InAlAs

摘要

In recent years, InAlAs avalanche photodiodes (APDs) have reported high sensitivities in high bit rate 10 Gb/s [1] and even 40 Gb/s [2] optical communication applications. InAlAs is able to supersede InP as the multiplication layer in SAM-structure APDs due to its superior ionization characteristics that include larger bandgap and very dissimilar electron and hole ionization coefficients (α and β respectively), while remaining lattice matched to InP substrates. Preliminary studies also suggest relatively small temperature dependence of breakdown voltages compared to InP, which reduces the need for temperature stabilization. The significance of a large electron and hole ionization coefficient ratio is lower excess noise and increased sensitivity, without sacrificing the gain-bandwidth product. Furthermore, InAlAs/InGaAs APDs are electron initiated, which further improves the excess noise performance [3] should there be low-field ionization in the InGaAs absorption layer [4]. Accurate characterization of the InAlAs ionization properties such as excess noise and ionization coefficients are essential when designing and optimizing the InAlAs APD performance using simulators and analytical models. It is therefore surprising that the only data on submicron avalanche structures is by Saleh et al. [5] and even this only reports on electron-initiated multiplication. We present a systematic study of avalanche multiplication and excess noise characteristics of InAlAs on a series of p{sup}+-i-n{sup}+ and n{sup}+-i-p{sup}+ diodes with nominal intrinsic region widths from 0.1 μm to 2.5 μm. The carrier threshold energies and the ionization coefficient for enabled carriers between electric fields of 220 kV/cm to 980 kV/cm are extracted by fitting to the measured electron- and hole-initiated multiplication and excess noise characteristics by using the coupled integral equations technique described by Hayat et al. [6]. The enabled electron to hole ionization coefficient ratio, as plotted in Figure 1, varies from 32.6 to 1.2 with increasing field. The large ionization coefficient ratio at low fields accounts for the asymmetry in the multiplication characteristics, and confers low excess noise in thick structures, following local theory. With decreasing thickness, the excess noise initially increases, but the competing effects of dead space become significant at high fields, and in turn cause the excess noise to then decrease rapidly. For electron-initiated multiplication, the worst excess noise occurs for avalanche widths of 0.25 - 0.5 μm, and is lowest at the extremely thin and thick structures as shown in Figure 2. The advantage of reducing the structure widths is however limited by the tunneling mechanism that dominates the leakage currents for avalanche widths less than 0.2 μm, as shown in Figure 3. Thick structures on the other hand, will increase the minimum transit time through the avalanche region, which in turn decreases the bandwidth. Hence, care is needed to optimize the thickness of the multiplication region to optimize the noise performance of the APD without sacrificing the speed and maximum achievable gain.