Frequency response .1 Bit stream input

As an input, the bit stream below in Fig. 5‐9is pseudo random and has the most important transitions because it includes all the combinations of three bits.

Fig. 5‐9 The bit stream input contains combinations three bits. 

5.4.2 Eye diagram

To evaluate system speed, eye diagrams are often used. An open eye interprets the nice transient respond of an electronic device. In Fig. 5‐10, simulations of transients responds to the bit stream inputs with frequencies 1GHz, 10GHz, 20GHz, 40GHz, and 100GHz are performed. A regular trend can be seen that as the signal frequency is higher than the speed that GePD can handle with, signals distortion can be read as the “eye” getting closed. In realistic cases, eye diagrams needs to be extended with parasitic components, noises, and loads of amplifiers, depending on various interconnect systems.

Fig. 5‐10 Eye diagram of the GePD with speed 30GHz. 

5.5 Conclusion

In this chapter, it is seen that the speed of GePD increases with the total width (wGe), while for each wGe there is an optimum doping separation (Wi) value. A maximum 32GHz is found at wGe equals to 2 µm, and Wi equals to 0.2 µm. For optimization in length direction, we should note that length of GePD needs to be compromised with the absorption efficiency of light signals and the junction capacitance.

Value of junction capacitance of GePD with different length is obtained by C‐V simulation. The simulation of C‐V might seem redundant, since capacitance is of course increasing linearly with effective capacitor area (hGe × LGe), but it helps us to demonstrate the model of equivalent circuit of GePD, which would be shown in chapter 6.

Finally, frequency respond is shown in eye diagrams while a bit stream of optical signal with various frequencies.

Chapter 6 Conclusion

As one of the crucial components in optical interconnect system, germanium photodiodes are promising due to its corresponding bandgap to absorb infra‐red light and the compatibility to silicon‐based photonics and electronics. However, many researchers have shown that GePD is suffering from high dark current, and efforts were done to suppress the dark current and to increase gain‐bandwidth product by improving GePD efficiency and speed.

Several guidelines toward the GePD structure are provided in this thesis with the aid of investigation of measured dark current and TCAD simulation, to help fabricating high‐efficiency and low‐power consumption germanium photodetector with the structure resembles following figure. A table of guidelines is put at the end of this chapter after statements of conclusions chapter by chapter.

In chapter 2, we show the investigation of physical mechanisms behind the dark current in germanium photodiode via experimental data. It can be verified that dark current is in the combination behavior of diffusion current and Shockley‐Read‐Hall generation current. At room temperature, and at the bias ‐0.5V, SRH generation is the main component in dark current. There is a good qualitative agreement between formula trend and measured data.

In chapter 3, TCAD simulation is involved with inclusion of SRH model. We constructed base structures similar to measured photodiodes in TCAD sprocess tool with defect densities are estimated as 5e18/cm³ and 1e20/cm³ for simulations of constant defect profile and exponential defect profile. Alterations were made in the base structures to find the structure that has minimum dark current. Generally speaking, to make a germanium photodiode with low leakage, a thin germanium layer is required to reduce the SRH volume, i.e. number of traps, as much as possible.

A narrow doping separation is crucial to keep a small SRH volume within the large electric field region. Finally, deep junctions (junction depth ≈ hGe) are needed to circumvent the dark current arouse by the spreading electric field of shallow junctions.

In chapter 4, we start to simulate the photocurrent of GePD with optical signals. To have high quantum efficiency, we have an optimum when the width of germanium equals to 3.4 µm, the thickness of germanium equals to 0.5 µm, and the doping separation equals to half of the width, which is 1.7 µm. To maintain high quantum efficiency while keeping low level of dark current, deep‐and‐shallow junction is proposed. Deep junction is with slightly doping 3e13/cm³ to suppress dark current, and shallow junction is degenerately doped to prompt photo currents. The final GePD structure with optimized geometries and junction engineering is shown below, increases as total width (wGe) decreases, while for each wGe there is an optimum doping separation (Wi) value. A maximum 32GHz is found at wGe equals to 2 µm, and Wi equals to 0.2 µm. For optimization in length direction, we should note that length of GePD needs to be compromised with the absorption efficiency of light signals and the junction capacitance. The value of junction capacitance of GePD with different length is obtained by C‐V simulation to demonstrate the model of equivalent circuit of GePD.

To summarize the whole thesis, contributions from above chapters are generalized in two aspects, as described in the following two paragraphs:

In device aspect, we provide guidelines toward the GePD structure, which are summarized above, and arranged in Table 6‐1on next page. Each parameter is discussed while others are fixed.

In system aspect, we provide a simulation framework such that values of electronic components in the equivalent circuit of GePD become available or computable. In this way, evaluation of speed at the system level could be achieved by applying equivalent circuit of germanium photodiode connects to arbitrary known interconnection systems. For a germanium photodiode with length L, components can be obtained as shown in Table 6‐2.

Parameters Guidelines Reasons To lower Dark current

hGe Smaller, better

Idark increases linearly with hGe, due to number of traps increases as germanium has a larger volume.

wGe ≥ 3 µm

(Wi=1.8µm)

For same Wi, many traps at the tapered facets are more close to the middle in a narrow GePD, where electric field is stronger.

Wi Smaller, better Number of traps increases at the intrinsic region where electric field is large.

Junction Deeper, better Electric field spreading out from shallow junction would increase Idark.

To increase Quantum efficiency

hGe 0.5 µm The optimum reached in the structure which

lights mostly contribute in the center of GePD while not being squeezed out of germanium.

wGe 3.4 µm

Wi

½

Junction Shallower, better Strong electric fields from shallow junction improve QE

Low Dark Current and High QE at same time Optimized geometry

and

Deep‐and‐shallow two‐steps implant

Strong electric field of shallow junctions is away from defects at bottom of GePD, and it provides a strong driven force to carriers.

Deep junction with slight doping can screen the electric fields spreading from top to defects at the bottom.

To boost speed

LGe Depends on application L influences absorption of light and parasitic capacitance

wGe Smaller, better Narrow GePD means shorter distance for carriers to travel to electrodes.

Wi The optimum exists,

depending on wGe

Maximum happened at different Wi values in GePD with different wGe, because there are two mechanisms competing each other:

Junction capacitance and drift velocity Table 6‐1 Several Guidelines for designing GePD. 

Components Resource Expression

Iph (A) Ch 4 of this thesis Photonsabsorbed(L) × QE(ratio) RP (ohm) Ch 2, Ch3 of this thesis Vbias(V) / Idark (A/µm) × L(µm) CPD (Farads) Ch 5 of this thesis As simulated in Sec. 5.3.1 CP, LP, RS Estimated from circuit layout

RL Depends on next circuit stage (amplifier...etc.)

Table 6‐2GePD equivalent circuit [20][21]and acquired value of components. 

Appendices

Appendix A: Command codes in