Photonic-plasmonic mode coupling in nanopillar Ge-on-Si PIN photodiodes

Incorporating group IV photonic nanostructures within active top-illuminated photonic devices often requires light-transmissive contact schemes. In this context, plasmonic nanoapertures in metallic films can not only be realized using CMOS compatible metals and processes, they can also serve to influence the wavelength-dependent device responsivities. Here, we investigate crescent-shaped nanoapertures in close proximity to Ge-on-Si PIN nanopillar photodetectors both in simulation and experiment. In our geometries, the absorption within the devices is mainly shaped by the absorption characteristics of the vertical semiconductor nanopillar structures (leaky waveguide modes). The plasmonic resonances can be used to influence how incident light couples into the leaky modes within the nanopillars. Our results can serve as a starting point to selectively tune our device geometries for applications in spectroscopy or refractive index sensing.


Electric field profiles for parameter variations
In the following, we show simulation results for the electric field components ( , , ) ( = , , ) in our devices under illumination with 0 ‖ .

Variation of superstrate refractive index sup
By varying the refractive index on top of the device the plasmonic modes can be shifted. Due to their high field strength, the LSPR excited under illumination with 0 ‖ tend to react much more strongly to changes within their surroundings.
SiO 2 Ge Supplementary Figure S4. Plots of ( , , ) with = , , for two different values for ℎ. Cross-sectional device image in direction of under illumination with 0 ‖ at λ = 1483 nm. NP = 300 nm, = 100 nm. Horizontal plane approximately 250 nm below the top Si-Ge interface.

Electric device characteristic
From the current-voltage characteristic ( ) of a device the following figures of merit can be calculated through where s is the saturation current, D the voltage across the diode, T the thermal voltage, η the ideality factor and S the series resistance of the diode.

Electro-optical device characteristic: Comparison with reference
Here, we provide the responsivity spectrum of a reference detector as baseline and compare it to the NP-PD responsivity spectra.
The reference detector is a (non-structured) bulk photodetector with an identical semiconductor layer stack. When measuring the reference detector responsivity spectrum under vertical incidence, however, the topmost Si layer can only be assumed to be transparent for incident wavelengths above ~1150 nm as a consequence of the indirect bandgap of Si. The extraction of the responsivity of Ge-on-Si photodetectors grown in the same MBE chamber and with a similar semiconductor layer structure except for individual layer thicknesses is discussed in detail in Ref. [S1]. The wavelength dependence of the responsivity of a photodetector is influenced by the wavelength-dependence of the external quantum efficiency ( ) [S2]: where is the elementary charge, is the speed of light, ℎ is the Planck constant and is the wavelength of the incident light. Here, we combine the measured spectrum of the reference detector in the wavelength range of 1210 -1700 nm with a linear extrapolation of the responsivity to lower wavelengths, this approach is motivated by results obtained for similar devices (see Fig. 3.5 in Ref. [S1]). The scaled reference spectrum is also used as a baseline for peak fitting (Fig. 5) in order to consider contributions to the responsivity originating from a remaining planar Ge layer underneath the nanopillar structures (Fig. 1). The baseline is rescaled for each NP-PD device to take process variations into account.
Several competing effects can be expected to play a role when comparing the NP-PD responsivity to that of a reference device under identical illumination conditions, e.g.
a reduction in the absorbing volume of Ge in the NP-PD photodetectors compared to bulk devices, -an increase in absorption for NP-PD photodetectors compared to bulk devices due to the antireflective properties of the patterned surface as well as an increase in the effective penetration depth of the incident light as a result of structuring, -a reduction in absorption for NP-PD photodetectors compared to bulk devices as a result of the metallic top layer that leaves only small apertures for light to impinge onto the Ge nanopillar structures, -an increase in absorption for NP-PD photodetectors compared to bulk devices as a result of plasmonic enhancement sustained by the nanocrescent holes.
Instead of a detailed discussion of these effects, which exceeds the scope of this work, we provide two different comparisons of the NP-PD responsivity spectra with that of the reference device. The responsivity spectra as measured are shown in Fig. S6. Here, it can be seen that the responsivities of the NP-PD devices are drastically reduced compared to that of a reference detector.
In a second comparison we take into account the effects of a reduction in the absorbing volume of Ge in the NP-PD devices compared to a reference detector by normalizing to the actual available absorbing device volume as follows opt,norm = opt ⋅ Λ 2 π i 2 .
Here, Λ indicates the distance between two nanopillars in the square lattice, i.e. the lattice pitch, and is the nanopillar diameter. The comparison of devices with different detector diameter indicates an enhancement in normalized responsivity for specific wavelength ranges (s. Fig. S7) which can exceed the responsivity spectrum of a non-structured device.