In this paper, an ultra-thin buffer technology for the epitaxial growth of Si<sub>x</sub>Ge<sub>1-x-y</sub>Sn<sub>y</sub> structures on Si or Si-on-Insulator substrates by using molecular beam epitaxy is presented. This technology builds the basis for integrated photonic devices as detectors, modulators and light sources. The paper discusses different device families with different material compositions, which all use a relaxed Ge virtual substrate with high quality. These are pseudomorphic Ge/Ge<sub>1-y</sub>Sn<sub>y</sub> structures, Si<sub>x</sub>Ge<sub>1-x-y</sub>Sn<sub>y</sub> structures lattice matched to Ge and (partially) relaxed Ge<sub>1-y</sub>Sn<sub>y</sub> virtual substrates. The photonic devices consist of heterojunction diodes with vertical pin doping structures. As an example, Ge/Ge<sub>1-y</sub>Sn<sub>y </sub>multi quantum well photodetectors which active regions made from Nx(Ge<sub>0.93</sub>Sn<sub>0.07</sub>/Ge) multi-quantum well structures are presented. Optical measurements at high frequencies are successfully performed on these photodetectors. A 3-dB bandwidth above 40 GHz is measured at the optical telecommunication wavelength of 1550 nm.
To enlarge the tensile strain in Ge light emission diodes (s-Ge LED) we applied a GeSn virtual substrate (VS) on Si (001) with a Sn content of 4.5 %, to produce s-Ge LEDs. The LED stack was grown by molecular beam epitaxy. Electroluminescence investigations of the s-Ge LED show a major direct Ge peak and a minor peak at lower energy, which is formed by the GeSn-VS and the s-Ge indirect transition. The main peak of a 100 nm thick s-Ge LED is red-shifted as compared to the Ge peak of an unstrained reference Ge LED grown on Ge-VS. At a temperature of T = 80 K the increased tensile strain, produced by the GeSn-VS, causes a redshift of the direct Ge peak from 0.809 eV to 0.745 and 0.769 eV, namely for the s-Ge LED with a 100 and 200 nm thick active layer. At T = 300 K the direct Ge peak is shifted from 0.777 eV of the reference Ge LED to 0.725 eV (for 100 nm) and 0.743 eV (for 200 nm). The peak positions do not differ much between the 50 and 100 nm thick s-Ge LEDs. The intensities of the direct Ge peak increase with the s-Ge layer thickness. Moreover, the intensity of the 50 nm thick s-Ge sample is found to be larger than that of the 100 nm thick reference Ge LED.
We analyzed Ge- and GeSn/Ge multiple quantum well (MQW) light emitting diodes (LEDs). The structures were grown
by molecular beam epitaxy (MBE) on Si. In the Ge LEDs the active layer was 300 nm thick. Sb doping was ranging
from 1×10<sup>18</sup> to 1×10<sup>20</sup> cm<sup>-3</sup>. An unintentionally doped Ge-layer served as reference. The LEDs with the MQWs consist
of ten alternating GeSn/Ge-layers. The Ge-layers were 10 nm thick and the GeSn-layers were grown with 6 % Sn and
thicknesses between 6 and 12 nm. The top contact of all LEDs was identical. Accordingly, the light extraction is
The electroluminescence (EL) analysis was performed under forward bias at different currents. Sample temperatures
between <300 K and 80 K were studied. For the reference LED the direct transition at 0.8 eV dominates. With increasing
current the peak is slightly redshifted due to Joule heating. Sb doping of the active Ge-layer affects the intensity and at
3×10<sup>19</sup> cm<sup>-3 </sup>the strongest emission appears. It is ~4 times higher as compared to the reference. Moreover a redshift of the
peak position is caused by bandgap narrowing.
The LEDs with undoped GeSn/Ge-MQWs as active layer show a very broad luminescence band with a peak around
0.65 eV, pointing to a dominance of the GeSn-layers. The light emission intensity is at least 17 times stronger as
compared to the reference Ge-LED. Due to incorporation of Sn in the MQWs the active layer should approach to a direct
semiconductor. In indirect Si and Ge we observed an increase of intensity with increasing temperature, whereas the
intensity of GeSn/Ge-MQWs was much less affected. But a deconvolution of the spectra revealed that the energy of
indirect transition in the wells is still below the one of the direct transition.
This work presents the limiting factors of fast Germanium p-i-n photodetectors for optical on-chip communication. The photodetectors are grown by molecular beam epitaxy on Silicon and Silicon on insulator substrates. On-wafer RF and optical RF measurements up to 40 GHz are performed at a wavelength of 1.55 μm. Different de-embedding procedures are used to obtain the amplitude and phase of the device impedance and the equivalent circuit description. An analysis of the reflection coefficient compared to the equivalent circuit explains the frequency characteristic and it is used to determine background doping of the intrinsic layer and the expansion of the space charge width. The optical bandwidth is measured for different bias voltages and background doping. The RC limitation of the detectors is shown and analyzed leading to adjusted parameters for high speed detectors at zero-bas.
This work concentrates on the device characteristics and performance of Ge-on-Si p-i-n diodes for the use as absorption modulators. At first, the impact of temperature on electrical and on optical characteristics of these p-i-n diodes is investigated. Secondly, the feasibility of optical modulation using the Franz-Keldysh effect is demonstrated for temperatures up to 359 K. The Ge-on-Si p-i-n diodes are grown using a molecular beam epitaxy system. The layer structure includes a double Si/Ge-heterojunction and an intrinsic zone with a thickness of 500 nm. During the growth process several annealing steps are performed to reduce the dislocation density and incorporate tensile strain in the intrinsic zone. The dark current is proportional to the diode area and amounts to 40 mA/cm<sup>2</sup> at a reverse voltage of 1 V. An analysis of the temperature dependence of the dark current shows that it is dominated by generation/recombination of carriers probably at threading dislocations. The optical absorption spectra recorded show a shrinkage of the infrared cut off wavelength of about 0.6 nm/K. In addition the change of absorption at the direct bandedge with different applied biases, i.e. the Franz- Keldysh effect, is demonstrated for temperatures from 300 K to 359 K. With regard to modulation of an optical signal the on/off ratio is evaluated as function of the voltage swing. With a moderate voltage swing of 2 V the maximal absorption change is 300 cm<sup>-1</sup> and the optimal working regime shifts from 1625 nm at 300 K to 1665 nm at 337 K.
GeSn LED’s with Sn contents up to 4% exhibit light emission from the direct band transition although GeSn of low Sn contents is an indirect semiconductor.. The emission wavelength is red shifted compared to Ge. The redshift of the direct band transition is confirmed by different optical characterization techniques as photoluminescence, electroluminescence, photodetection and reflectivity. The photon emission energy decreases from 0.81 eV to 0.65 eV for compressively strained GeSn of 0% to 4% Sn content. Growth of GeSn up to 12% Sn is performed for which preliminary characterization results are given.
In this study, Strained silicon Quantum Wells (QW) were characterised using a variety of micro-scopical techniques. Among the techniques used were Transmission Electron Microscopy (TEM), Elemental Electron Loss Spectroscopy (EELS), and micro-Raman spectroscopy. A combination of these methods facilitates investigation of the structure, the strain, and the dislocations present in such materials. Both conventional and High Resolution Transmission Electron Microscopy (HRTEM) are used to analyse strained silicon quantum wells (QW). These techniques allow for structure analysis at the atomic level. Elemental Electron Loss Spectroscopy (EELS) is used in tandem with other analytical techniques in order to give a quantitative analysis of the structures. The presence of various layers is independently verified using EELS, while layer depth and concentration profiles are also established. Relaxation levels in the virtual substrate as well as the strain in Si quantum wells are calculated using Raman spectroscopy.