1. Introduction

Avalanche photodiodes (APDs) have been widely used in telecommunication systems, light detection and ranging (LIDAR), single photon counting, and data centers owing to their potential for higher receiver sensitivity compared to p-i-n photodiodes [1]. The signal-to-noise ratio (SNR) of APDs can be expressed as:

The dark current of photodetectors is frequently described in terms of the bulk and surface components, which scale with area and perimeter, respectively. The bulk portion is primarily related to the material quality. The dark current that originates at the surface depends on fabrication processing, a critical aspect of which passivation of surface defects from dangling bonds, crystalline defects, and impurities is essential. This is especially true for mesa-structure detectors that tend to have relatively large exposed surfaces. APDs have the additional issue of high electric fields at the surface. Many passivation methods have been explored to reduce surface leakage, for example, SU-8 [4], benzocyclobutene (BCB) [5,6], SiO₂, and Si_xN_y [7]. However, these passivation methods have proved to be only partially successful in reducing surface-related dark current [8]. Recently, Nada et al. [9,10] reported a triple-mesa-structure InGaAs/InAlAs separate absorption, charge, and multiplication (SACM) APDs that achieved high bandwidth with small active areas. The low dark current exhibited by these APDs was attributed to suppression of surface leakage by the triple mesa [11–13]. In this work, we report a reach-through APD with modified charge layer doping that is different from the structure in [9–12]. The reach-through structure can achieve better restriction of the electric field in the center of the multiplication region, which results in a relatively low electric field at the surface. This approach is approximately equivalent to passivation with the same semiconductor material, which effectively eliminates surface dangling bonds. Moreover, it can be easily extended to different semiconductors.

2. Device design and fabrication

A schematic diagram of the triple-mesa structure is shown in Fig. 1(a). It consists of three mesas with increasing area from top to bottom. Figure 1(b) is an SEM picture of a fabricated triple mesa APD. The epitaxial layers of the triple-mesa reach-through APD are shown in Fig. 1(c). From the top to bottom, the structure consists a 100 nm InAlAs P-type contact layer, an 80 nm InAlAs unintentionally-doped buffer layer, an 80 nm InAlAs P-type charger layer, an 800 nm InAlAs unintentionally-doped multiplication layer, a 160 nm InAlAs N-type contact layer, and an InP N-type substrate. The mesas of the triple-mesa APDs were defined with standard photolithography and formed by wet etching. The lowest mesa was etched with a solution of H₂SO₄: H₂O₂: H₂O, which has a fast etch rate and good anisotropy. The second and third mesas were etched with a solution of C₆H₈O₇: H₃PO₄: H₂O₂: H₂O, which has a stable slow etch rate. Top and bottom contacts were deposited by evaporation and lift-off of Ti/Au.

 

Fig. 1 (a) Schematic diagram of triple-mesa APDs. (b) SEM picture of one fabricated triple-mesa APD. (c) Cross sections of triple-mesa reach-through InAlAs APDs.

Download Full Size | PDF

The triple-mesa structure can restrict the surface electric field more than a double-mesa [12]. The purpose of the charge layer is to tailor the vertical electric field profile, and the triple-mesa structure performs a similar function laterally. Therefore, the electric field can be precisely controlled in the center of the multiplication region. The simulated electric field at – 40 V bias of an InAlAs double-mesa APD and a triple-mesa (reach-through) APD are shown in Fig. 2. In the double-mesa structure, the electric field in the multiplication layer is influenced by the first mesa region, however, electric-field crowding is observed at the foot of the first mesa sidewall, as shown in Fig. 2(a). This can cause premature breakdown at the edge [14]. The electric field profile of the triple-mesa APD is shown in Fig. 2(b).

 

Fig. 2 Simulated electric field profiles of (a) double-mesa and (b) triple-mesa APDs.

Download Full Size | PDF

3. Triple-mesa results

The gain versus bias voltage of the InAlAs reach-through APD is shown in Fig. 3(a). The APDs were illuminated with a 543 nm laser. For InAlAs at least 99% of the incident light is absorbed before the multiplication region, which ensures pure electron injection. The measured excess noise versus gain is shown in Fig. 3(b). The InAlAs reach-through APDs have similar avalanche characteristics as that of InAlAs p-i-n APDs.

 

Fig. 3 (a) Gain versus bias voltage under 543 nm laser, and (b) excess noise of InAlAs reach-through APD.

Download Full Size | PDF

In order to verify that the triple-mesa structure can suppress surface leakage, single-mesa and triple-mesa APDs were fabrication from the same InAlAs reach-through wafer. The designed diameters of single-mesa and triple mesa APDs are shown in Table 1. Figure 4(a) shows the dark currents of both structures for different single mesa diameters. The actual diameters of the single mesas, after undercutting in the wet etch, were 55 µm, 75 µm, and 135. For the triple-mesa, we used the effective diameters, which were estimated to be 50 µm, 70 µm, and 124 µm. The dark current of the 50 µm-diameter triple-mesa device is < 1 pA at −45 V. The dark current densities of the single-mesa and triple-mesa are plotted in Fig. 4(b). For bias > - 15 V, the dark current density of the single-mesa APDs is ~1.5 $\mu \text{A}/{\text{cm}}^2$, while that of the triple-mesa is ~30 $\text{nA}/{\text{cm}}^2$, i.e., ~50 times lower than the single-mesa.

Table 1. Designed Diameters of Single-mesa and Triple-mesa APDs*

View Table

 

Fig. 4 Comparison of single-mesa and triple-mesa InAlAs reach-through APDs: (a) dark current, and (b) dark current density.

Download Full Size | PDF

By plotting dark current versus diameter, it is possible to determine the relative magnitudes of the bulk and surface components of the dark current. As Figs. 5(a) and 5(b) show, the dark current of the single-mesa APDs varies linearly with mesa diameter, which indicates that surface leakage dominates. However, the quadratic relationship of the triple-mesa devices indicates that bulk leakage is the most significant dark current component. It follows that the triple-mesa design can effectively suppress surface leakage.

 

Fig. 5 Dark current versus device diameter: (a) single-mesa, and (b) triple-mesa APDs.

Download Full Size | PDF

To measure the spatial uniformity of the photoresponse of the triple-mesa APDs, two-dimensional scans were carried out. Figure 6 shows the unity gain response of a triple-mesa APD under 543 nm CW laser illumination. Nearly all the response current is confined in the third mesa (smallest one). There is almost no photoresponse in the second and first mesa. The circular top contact blocks light and creates a valley between the edge and the center.

 

Fig. 6 Two-dimensional photo response of the InAlAs triple-mesa APDs.

Download Full Size | PDF

4. Design and fabrication tolerances

While the triple-mesa structure can suppress the surface leakage better than the single-mesa, this comes at the cost of more complex design and fabrication. One issue is radii of the first and second mesas. Figure 7 illustrates the electric field distributions of different radii. As shown in Fig. 7(a), when the radial difference between the first and second mesas are both 1 µm, at the foot of each mesa there is electric-field crowding similar to that in the double-mesa. Increasing the radial difference of the second mesa to 3 µm while keeping that of the first mesa at 1 µm, eliminates the electric-field crowding. This represents the smallest triple-mesa radius tolerance; the radius of the first mesa should be at least 1 µm bigger than that of the second mesa, and the second mesa should be more than 3 µm larger than the third mesa.

 

Fig. 7 Comparison of electric field distribution of triple-mesa InAlAs reach-through APDs with (a) 1 µm and 1 µm surplus radiuses, and (b) 3 µm and 1 µm surplus radiuses.

Download Full Size | PDF

Another important point is that the triple-mesa requires that etching terminate at the interfaces of layers. If this is not achieved, the electric field distribution will be changed, and may adversely affect the dark current performance, especially at high reverse bias voltage. Figure 8(a) illustrates the electric field distribution when the third mesa is over-etched. The electric field on the top mesa sidewall is much higher than that when the etch stops at the interface (i.e. Figure 2(b)). This is confirmed by the measurements of the dark current shown in Fig. 8(b). At low bias, < −15 V, the over-etched triple-mesa APDs exhibits similar low dark current density of ~30 $\text{nA}/{\text{cm}}^2$. However, the dark current increases abruptly at high bias, owing to the high surface electric field around the top mesa.

 

Fig. 8 Over-etched triple-mesa APDs: (a) electric field distribution, and (b) comparison of dark current density.

Download Full Size | PDF

5. Conclusion

We report triple-mesa reach-through APDs in which the surface-related dark current is effectively suppressed by reducing the surface electric field. The InAlAs triple-mesa APDs exhibit ~50 times lower dark current density than single-mesa APDs fabricated from the same wafer. The bulk dark current dominates for the triple mesa devices while that of the single mesa is surface leakage. Two-dimensional scans show almost no photoresponse from the wider mesas. Tolerances of triple-mesa design and fabrication have also been discussed.