Low-Dark-Current and Wide-Dynamic-Range InGaAs/InAlAs Avalanche Photodiodes with a Dual-Charge Layer

Abstract

This study explores the impact of a dual-charge layer structure on the performance of InGaAs/InAlAs avalanche photodiodes (APDs) with a separate absorption, charge, multiplication, charge, and transit (SACMCT) structure. The dual-charge layer, consisting of p-doped and n-doped charge layers on either side of the avalanche layer, is designed to precisely control the internal electric field, effectively reduce the dark current, and extend the dynamic range. Simulation results guided the fabrication of a backside-illuminated APD, which achieved a linear operating range of 10–30 V and a dark current as low as 80 nA. The optimized design significantly reduced the dark current and increased the breakdown voltage compared to previously reported APDs. These improvements demonstrate the potential of dual-charge-layer APDs for high-speed optical communications and precision photodetection applications.

1. Introduction

Compared to the traditional PD and amplifier configuration, avalanche photodiodes (APDs) provide internal gain, which enhances signal detection without requiring additional amplification. This feature improves the signal-to-noise ratio, reduces power consumption, and simplifies system design. Moreover, APDs demonstrate superior performance in long-distance optical communication by maintaining signal integrity and stability, especially in weak signal conditions, making them a choice for high-speed optical communication systems [1,2,3]. A low, dark current helps suppress noise, enhancing detection sensitivity in low-light conditions. Meanwhile, a wide dynamic range allows APDs to operate effectively across varying optical power levels, preventing saturation under high-strength illumination and ensuring stable performance in fluctuating signal environments. These characteristics collectively improve the reliability and efficiency of APD-based optical communication and sensing systems [4].

Early separate absorption and multiplication structures have limitations in controlling the electric field between the absorption and multiplication layer [5]. This lack of control results in excessive dark currents and noise, limiting the dynamic range. To address these limitations, a charge layer was introduced between the absorption and multiplication layers, leading to the development of separate absorption, charge, and multiplication (SACM) or separate absorption, grading, charge, and multiplication structures (SAGCMs). This design enables a more balanced electric field distribution, thereby preventing an excessive electric field in the absorption layer and reducing the dark current. Additionally, a uniform electric field distribution prevents premature breakdown and extends the linear operating range [6]. These advancements have improved the dynamic range, noise performance, and stability of APDs, making SACM APDs more efficient for high-speed and high-sensitivity applications. To further enhance control over the dynamic range and electric field distribution of APDs, a dual-charge layer structure has been introduced [7,8]. Most studies focus on enhancing the dynamic range of APDs by optimizing either the absorption or avalanche layers [9,10,11], particularly through the use of cascaded avalanche layers [12,13,14]. However, a single-charge layer limits precise control across electric fields in both the absorption and avalanche layers [15,16,17,18]. High-performance APDs require more precise internal electric field control to ensure their proper operation.

In 2014, Nada et al. proposed an APD with a 1 μm absorption layer and a 20 μm diameter, achieving a GBP of 235 GHz [19]. However, the thin avalanche layer led to a dark current exceeding 1 μA with a near 90% breakdown voltage of 26.5 V, thereby limiting its dynamic range and signal-to-noise ratio. In 2018, Nada et al. reported an APD with a 600 nm absorption layer and a 14 μm diameter, achieving a responsivity of 0.7 A/W at 1300 nm [20], yet exhibiting a high dark current of 2 μA. In 2019, they further reported an APD with a thinner 300 nm absorption layer, reaching a bandwidth of 42 GHz [11], which is the highest reported bandwidth among InGaAs/InAlAs APDs. However, it exhibited a dark current of 1 μA and a breakdown voltage of approximately 26 V.

To address the issues of a high dark current and low breakdown voltage, we simulated InGaAs/InAlAs separate absorption, charge, multiplication, charge, and transit (SACMCT) APDs with a dual-charge layer configuration. In this design, one charge layer controls the electric field in the absorption layer, while the other regulates the electric field in the avalanche layer. The dual-charge layer, consisting of p-doped and n-doped charge layers on either side of the avalanche layer, effectively reduces the dark current and expands the dynamic range by precisely controlling the internal electric field.

2. Methods

This section introduces a backside-illuminated SACMCT APD model aimed at analyzing the influence of the dual-charge layer on the photocurrent, dark current, electric field strength, and linear operating range, as shown in Figure 1. A dual-charge layer configuration, consisting of both p-doped and n-doped layers, was employed to precisely control the electric field distribution. The APD structure comprises three mesas: the first mesa consists of a contact layer, followed by the second, which includes a transit layer and an n-doped charge layer. The final mesa comprises an avalanche layer, a p-doped charge layer, an undoped absorption layer, and a p-doped absorption layer. To reduce the leakage current caused by the electric field at the sidewall, the smaller first mesa was designed to confine the electric field beneath it. When the APD is fully depleted, the sidewall of the undoped InGaAs layer becomes the most vulnerable region owing to its minimized bandgap. Theoretically, the electric field in this region should be zero to minimize the leakage current. To enhance carrier transport speed, a hybrid absorption layer consisting of both p-doped and undoped layers was used to replace the conventional intrinsic absorption layer [21]. Given that the electron drift velocity significantly exceeds the holes, the transit time is determined by the drift velocity of the holes. When the electric field in the undoped absorption layer exceeds 50 kV/cm, holes are able to move through the depletion region at saturation velocity [22].

The simulation was conducted using Silvaco TCAD, integrating various physical models, including the Shockley–Read–Hall model, Fermi–Dirac carrier distribution, Auger model, analytic mobility model, BTB tunneling, and Selberherr’s impact ionization. The incident light had a wavelength of 1310 nm and a power of 4 μW, while the material’s physical properties are detailed in

Table 1. Physical properties employed in the simulation [23].

Parameter	Unit	In0.52Ga0.48As	In0.53Al0.47As	In0.52Al0.15Ga0.33As
Bandgap	eV	0.75	1.44	1.1
Electron Affinity	eV	4.5	4.25	4.38
Permittivity	F/m	13.9	12.2	12.5
SRH recombination	μs	0.1	0.01	1
Selb An	cm−1		8.6 × 106
Selb Ap	cm−1		2.3 × 107
Selb Bn	Vcm−1		3.5 × 106
Selb Bp	Vcm−1		4.5 × 106
Selb β	1		1
BBT.A	V−1 cm−1 s−1	1 × 1011	2.1 × 1013
BBT.B	Vcm−1	3 × 106	2.2 × 106
BBT.γ	1	2	2

.

3. Simulation and Optimization

The dynamic range of an APD can be expressed as follows:

$$\mathrm{D}\mathrm{y}\mathrm{n}\mathrm{a}\mathrm{m}\mathrm{i}\mathrm{c}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{g}\mathrm{e}=10{\log }_{10}\left(P_{max}/P_{min}\right)$$

(1)

where P_max represents the maximum detectable optical power determined by the linear operating range and saturation current; P_min represents the minimum detectable optical power, typically limited by the dark current and noise currents. The dynamic range can be effectively improved by optimizing the dual-charge layer to enhance the linear operating range and reduce the dark current.

In this section, a SACMCT APD model is developed to explore how variations in the doping concentration and thickness of the dual-charge layer affect the electric field distribution, linear operating range, dynamic range, and dark current.

As shown in Figure 2a, the I-V characteristics show the behavior of the linear operating range and dark current under varying doping concentrations in the p-doped charge layer. The electric field strength in the depletion region under varying p-doped charge layer doping concentrations is shown in Figure 2b. When the doping concentration increased from 6 × 10¹⁷ to 9 × 10¹⁷ cm⁻³, the linear operating range contracted from 7.5 to 33 V to 19–25 V, while the electric field strength in the avalanche layer rose from 677 kV/cm to 767 kV/cm; this indicates that higher doping increases the electric field in the InAlA avalanche layer, thereby reducing the breakdown voltage. Simultaneously, the electric field in the undoped absorption layer is reduced, limiting the depletion and leading to a high punch-through voltage.

Figure 2c,d show the I-V characteristics and cross-sectional electric field profile for varying thicknesses of the p-doped charge layer. As the thickness increased from 40 to 70 nm, the linear operating region contracted from 23 V to 0, the breakdown voltage decreased from 31 V to 20 V, and the electric field within the avalanche layer rose from 694 to 815 kV/cm, leading to an increase in the dark current and a reduction in the breakdown voltage. When the thickness surpassed 60 nm, the punch through the voltage neared the breakdown voltage, greatly limiting the linear operating region. If the thickness continues to increase, the device may undergo breakdown before reaching depletion, potentially resulting in device failure. Thus, it is crucial to ensure that the breakdown voltage exceeds the punch-through voltage.

The I-V characteristics for different doping concentrations of the n-doped charge layer are shown in Figure 3a. The p-doped charge layer, situated between the absorption and avalanche layers, influences both breakdown and punch-through voltage. A higher doping concentration leads to an increase in the breakdown voltage while simultaneously lowering the punch-through voltage. In contrast to the p-doped layer, the n-doped charge layer affects both voltages in the same direction, causing a simultaneous rise in the breakdown and punch-through voltage. This is achieved by designing a transit layer, which allows part of the electric field in the depletion region to be distributed. As shown in Figure 3b, reducing the transit layer in the electric field results in an increase in both the absorption and avalanche layers of the electric field, enhancing both the punch-through and breakdown voltages.

The I-V characteristics of the APD with varying n-doped charge layer thicknesses are shown in Figure 3c. The punch-through and breakdown voltages increased with thickness. The dark current increased when the thickness of the charge layer increased. This is due to the high tunneling current, which can be represented as follows:

$$J_{bbt}=A·E^2\mathrm{e}\mathrm{x}\mathrm{p}\left(-E/B\right)$$

(2)

where E represents the effective electric field strength, and A and B represent material-dependent constants. This formula indicates that the tunneling current increases exponentially with electric field strength. As shown in Figure 2d and Figure 3d, to achieve a trade-off between suppressing the tunneling current in the avalanche layer and ensuring that the electric field in the absorption layer allows holes to reach saturation drift velocity, the thicknesses of the p-doped and n-doped charge layers were set to 50 nm and 65 nm, respectively.

Figure 4 shows that reducing the doping concentration of the dual-charge layer increases the breakdown voltage and extends the dynamic range. However, this also increases the electric field in the absorption layer, potentially leading to elevated tunneling currents. To mitigate this risk, the electric field should remain below 200 kV/cm [2]. Therefore, the p-doped charge layer was set to 50 nm with a doping concentration of 8 × 10¹⁷ cm⁻³, while the n-doped charge layer was adjusted to 65 nm with a doping concentration of 6 × 10¹⁷ cm⁻³.

4. Fabrication and Discussion

According to the epitaxial structure in

Table 2. Epitaxial layers of the SACMCT APD.

Layer	Material	Doping (cm−3)	Thickness (nm)
N-doped contact	InP	1 × 1019	150
Transit	InP	1 × 1016	300
N-doped charge	InP	6 × 1017	65
Avalanche	In0.53Al0.47As	1 × 1015	100
P-doped charge	In0.53Al0.47As	8 × 1017	50
Undoped absorption	In0.52Ga0.48As	1 × 1015	300
P-doped absorption	In0.52Ga0.48As	1 × 1016~1 × 1018	300
P-doped contact	In0.52Al0.15Ga0.33As	1 × 1019	500

, the InGaAs/InAlAs APD was epitaxially fabricated on a semi-insulating InP substrate using molecular beam epitaxy (MBE).

To achieve a low, dark current in APDs, wet etching is typically employed due to its minimal damage to the sidewalls. Figure 5a depicts the structure following a triple-mesa etching process. The first and third mesas were etched using an HBr:HCl:H₂O₂ = 10:5:1 solution diluted with deionized water. This solution enables the etching of InP, InGaAs, InAlAs, and InAlGaAs, with the etching rate adjustable by varying the deionized water ratio. The second mesa was etched using a CH₃COOH:H₃PO₄:HCl = 1:1:1 solution, which was also diluted with deionized water. This solution selectively etches InP while preserving InAlAs, allowing for precise depth control. However, it causes excessive lateral etching, making it unsuitable for prolonged etching durations.

Following the etching process, the DektakXT (Bruker, Billerica, MA, USA) stylus profiler was used to measure the mesa depths, as illustrated in Figure 5b. The measured depths of the first, second, and third mesas were 157.3 nm, 375.3 nm, and 1210.7 nm, respectively, closely aligning with the designed values of 150 nm, 365 nm, and 1250 nm. These results confirm the accuracy and reliability of the etching process.

After etching, a 300 nm SiO₂ passivation layer was deposited via Plasmalab100 (Oxford Instruments, Abingdon, Oxfordshire, UK) plasma-enhanced chemical vapor deposition (PECVD) to neutralize the dangling bonds on the sidewalls. To reduce capacitance, a 3 μm benzocyclobutene (BCB) layer was applied to electrically isolate the metal electrodes from the device. The BCB layer effectively reduces the device capacitance to the tens of fF range, which is critical for high-speed applications. To ensure precise alignment, coplanar metal electrodes were selected and deposited via WINVAC 400 L (Winter, Beijing, China) electron-beam evaporation. The N-electrode comprised Ni-AuGe-Ni-Au (20 nm—30 nm—20 nm—300 nm), while the P-electrode consisted of Ti-Pt-Au (40 nm—20 nm—300 nm), as illustrated in Figure 6. Following the lift-off of metal electrodes, alloying was performed at 380 °C to form ohmic contacts. Finally, the wafer was thinned to 200 μm and polished, followed by the deposition of a 267 nm SiO₂ anti-reflection layer on the back surface.

The I-V characteristics were tested with an optical power of 4 μW at 1310 nm. The simulated and experimental I-V curves are displayed in Figure 7. The APD entered the linear operating region at 10 V and reached breakdown at approximately 30 V. We observed discrepancies between the experimental results and simulations, which we attribute to variations in the doping concentration of the undoped absorption layer. In the simulations, the doping concentration was set to 1 × 10¹⁵ cm⁻³; however, the actual undoped absorption layer exhibited non-uniform doping due to growth conditions and the influence of adjacent highly doped layers. This doping induces variations in the electric field in the absorption layer. The minimum electric field strength is lower than that in the simulations, resulting in a lower dark current. Conversely, the higher average electric field reduces the fraction of the electric field allocated to the avalanche layer, leading to a lower photocurrent level. Compared with the linear operating range of 14–26.5 V and the dark current of 1 μA reported by Nippon Telegraph and Telephone Corporation [19], our optimized dual-charge-layer APD extended the linear operating range to 10–30 V and obtained a dark current of 80 nA. Based on Equation (1), with an equivalent saturation photocurrent, an increased breakdown voltage raised P_max by approximately 13%, while reducing dark current lowered P_min by approximately 92%, resulting in a dynamic range improvement of around 23 dB.

The capacitance–voltage (C-V) characteristics with various mesa diameters are presented in Figure 8. The APD starts operating as the punch-through voltage attains 10 V. When the voltage surpasses this threshold, the depletion layer completely expands, stabilizing the capacitance. The tested capacitance comprises both junction capacitance (C_j) and parasitic capacitance (C_p). Since the electrode pattern, fabrication process, and testing conditions are consistent across varying diameters, their C_p values remain the same. However, C_j is directly related to the mesa area and can be represented as follows:

$$C_j={\displaystyle \frac{\epsilon S}{d}}$$

(3)

Here, ε denotes the dielectric constant; d denotes the depletion region thickness; and S denotes the junction area. C_j is directly proportional to S, leading to the ratio $\sqrt{60 fF-C_p}$:$\sqrt{80 fF-C_p}$:$\sqrt{110 fF-C_p}$ = 16:20:24. As a result, C_p is estimated to be around 20 fF.

The APD’s frequency response was measured at 20 V, demonstrating a 3 dB bandwidth of 25.5 GHz, as illustrated in Figure 9. The total bandwidth was constrained by the transit time, multiplication build-up time, and the RC-limited time. The RC-limited bandwidth is determined using the following equation:

$$f_{RC}={\displaystyle \frac{1}{2\pi RC}}$$

(4)

Here, the total resistance R consists of an ohmic contact resistance of approximately 30 Ω and a matched load of 50 Ω. Similarly, the capacitance C includes C_j and C_p. For a minimum mesa diameter of 16 μm, the RC-limited bandwidth reaches 33 GHz. The total 3 dB bandwidth is expressed as follows:

$${\displaystyle \frac{1}{{f_{3db}}^2}}={\displaystyle \frac{1}{{f_{tr+M}}^2}}+{\displaystyle \frac{1}{{f_{RC}}^2}}$$

(5)

The transit bandwidth and multiplication bandwidth are typically difficult to separate and are considered together. Consequently, the combined transit bandwidth and multiplication bandwidth is approximately 40 GHz, with the RC bandwidth being the primary factor constraining the total 3 dB bandwidth. Further bandwidth enhancement requires the optimization of the fabrication process to minimize resistance and capacitance.

5. Conclusions

The dual-charge layer significantly improved the linear operating range and reduced the dark current of InGaAs/InAlAs SACMCT APDs. By optimizing these layers, the dynamic range was extended, highlighting the impact of this configuration. Guided by the simulation results, an APD with a 600 nm absorption layer and a 100 nm avalanche layer was successfully fabricated. The optimized dual-charge layer design increased the breakdown voltage to 30 V and reduced the dark current to 80 nA. Optimizing the dual-charge layer to reduce the dark current and extend the dynamic range provides valuable insights into improving the signal-to-noise ratio and the stability of photodetectors. These advancements support the application of InGaAs/InAlAs APDs in high-speed and long-distance photodetection.