Open Access
Add to BibSonomy
Abstract
We have developed a self-powered near-infrared photodetector (PD) with high detectivity using a tensile strained Ge layer capped with a thick Si layer. The Si layer acts as a stressor and maintains the strain of Ge with minimal dislocations by creating a rough surface. By using Raman spectroscopy, we confirmed that the Ge layer has a 1.83% in-plane tensile strain. The Ge PD exhibits a high responsivity of 0.45 A/W at -1 V bias voltage for 940 nm wavelength. The PD's dark current density is as low as ∼1.50 × 10−6 A/cm2 at -1 V. The high responsivity and low dark current result in a detectivity as high as 6.55 × 1011 cmHz1/2/W. This Ge PD has great potential for applications in light detection and ranging (LiDAR), Internet of Things (IoTs), and Optical Sensing Networks.
© 2024 Optica Publishing Group under the terms of the Optica Open Access Publishing Agreement
1. Introduction
The use of Light Detection and Ranging (LiDAR) technology is rapidly growing in various applications such as self-driving cars, security, and surveillance systems. To record signals from objects located at different distances, ranging from tens of centimeters to hundreds of meters, LiDAR systems require photodetectors (PD) with high density and high detectivity. The image resolution is crucial to identify and distinguish one object from another hence it is essential to have a high density of PDs. Furthermore, the light reflected from the objects generally has a low signal-to-noise ratio (SNR), making high-detectivity PDs even more necessary [1,2]. Furthermore, the diffused light caused by the rough surfaces can imply significant crosstalk between pixels, reducing the resolution. The data processing integrated circuit (IC) chips solve the crosstalk. The requirements of many ICs indicate that combining the PD and the processing circuits as photonic integrated circuits has the advantages of fewer data distortion, less crosstalk, fast decoding, and high resolution.
The near-infrared (NIR) wavelengths at 905 nm [3,4] and 940 nm [5,6] are commonly used in LiDAR systems. The choice of wavelength is determined by technological advancement and eye-safety considerations. The 1550 nm wavelength has the highest safety threshold per the maximum permissible exposure (MPE) value reported in the IEC regulations 60825. However, the 905 nm and 940 nm wavelengths are considered safe for the eyes and have mature semiconductor laser technology [4]. Therefore, the most commonly used LiDAR technology operates at a wavelength of either 905 or 940 nm.
However, the NIR wavelengths used are near the absorption edge of Si. At 940 nm, the absorption coefficient for Si is around 160 cm−1. The small absorption coefficient at this wavelength indicates that a thickness of approximately 60 µm is required to absorb around 60% of the incident light. Surface texturing, such as black-Si [7] or Si nanowires (SiNWs) [8], is typically utilized to improve the absorbance in Si solar cells. By this structure, the absorbance enhancement is approaching the 4n2 limit [9] in Si with a relatively thinner thickness [10]. Conducting polymer on the SiNW surface [11] or contacts to the backside of the substrate is commonly used to form electrodes to the surface nano-texturing. These contacting technologies challenge Si photodiodes because they cannot simultaneously facilitate high speed and high response. A gain structure was proposed to simultaneously achieve high absorption and speed to overcome this challenge. This structure uses the avalanche mechanism, which includes several thin layers responsible for light absorption and impact ionization amplification separately to achieve a high response. The thin layers ensure a short carrier transit time, which retains high speed. Examples of this structure include the avalanche PD (APD) or the Geiger-mode operated single-photon avalanche PD (SPAD) [8]. Germanium has an absorption coefficient of more than 150 times higher than Si at 940 nm. The absorption depth at 940 nm is ∼0.4 µm, far less than Si. In addition to high absorption, the process of Ge is compatible with Si, and the Ge PD can monolithically integrate the Si ICs on the Si wafer. However, the lattice constant of Si and Ge are different, having a ∼4% mismatch, making the epitaxial growth of Ge on Si with micrometer scale is challenging.
The growth of Ge on Si epitaxially exhibits stress because of the lattice constant difference. It causes high strain on the lattice in the thick Ge epitaxial layer, and the strain tends to be released by the lattice dislocation defects. In addition to the defects in the Ge by strain relaxation, the lower bandgap of Ge causes a higher reverse saturation current density (J0) in the diode. The dark current density in a PD contains the J0 and the generation current density (Js) in the Ge and GeSn PD [12,13]. The Js are generally larger than the J0, so reducing the Js enhances the SNR and, hence, the detectivity in a PD. According to the studies, the thin film's gap and surface states contribute to the generation current density, the Js. The dark current density reduction in the SiGe and Ge PD at NIR by silicon dioxide encapsulation and annealing at elevated temperatures indicates the surface states can be passivated effectively [14,15]. In addition to silicon dioxide, forming germanium oxide also passivated the Ge surface, reducing the dark current density [16]. Reducing the thickness of the active layer within the saturation thickness of epitaxial growth without strain relaxation could decrease the point defects and threading dislocation density, which would further suppress the dark current density of the PD [17,18]. However, the thin epitaxial layer indicates less absorbance unless the multiple quantum wells (MQWs) structure is used. In the MQWs, multiple thin Ge and Si layers are grown sequentially. The Si layer between two adjacent Ge layers relieves the high strain. Therefore, the total thickness of the multiple layers of Ge in the MQWs is higher than a single layer that exhibits higher absorbance [19].
The noise equivalent power (NEP) and the corresponding detectivity are a PD's primary metrics for the lowest detection level. The detectivity is calculated by the relationship of the responsivity over the dark current density in a PD, which will be discussed in Section 3. A Si-based PD is a commonly used device for the 940 nm photodetection. A commercially available Si PD (Thorlabs FDS015) exhibits responsivity of 0.16 A/W at the 940 nm. The reported NEP is 5 × 10−11 W/Hz1/2 at 850 nm, the corresponding detectivity is calculated as 4.7 × 108 cmHz1/2/W. Improving the detectivity of a Si PD can use the light trapping structure by increasing the responsivity. A regularly arranged hole array with a diameter in a sub-micrometer enhances the light-harvesting. The responsivity at 940 nm is 0.25 A/W on a two µm thick Si-on-insulator (SOI) layer, and the dark current density is 50.09 mA/cm2 [20]. The detectivity is calculated to be 6.24 × 1010 cmHz1/2/W. Further improvement of the responsivity is implemented by the application of black-Si [21]. The black-Si layer was placed at the bottom of the Si wafer, which enhanced the absorbance at NIR. The responsivity and the dark current density are 0.55 A/W and 2.08 nA/cm2, respectively. The detectivity is 2.13 × 1013 cmHz1/2/W. However, the response time is as high as 200 ns, which will limit the bandwidth of the PD. Incorporating Ge will increase the responsivity, and that might increase the detectivity. The Ge is grown on a SOI layer for this application [22]. The thickness of the Ge is 400 nm. The responsivity at a wavelength of 900 nm is 0.33 A/W, and the dark current is 0.02 µA at -1 V bias, corresponding to 20 mA/cm2. The detectivity is calculated to be 4.12 × 109 cmHz1/2/W. Although the responsivity is enhanced, the detectivity is reduced because of the high dark current density. Utilizing the hole arrays and improving the quality of epitaxial Ge is capable to improve the detectivity [23]. The responsivity is 0.188 A/W at a wavelength of 1550 nm, and the dark current density is 6.6 mA/cm2, achieved at -1 V. The detectivity is 6.74 × 109 cm·Hz1/2/W. A Ge PD with low dark current density is reported using selected area growth through oxide windows on Si [24]. The thickness of the Ge is 0.6 µm, and the GaB is used as the junction material to the Ge. The responsivity is 0.15 A/W at 850 nm, and the dark current density is 35 mA/cm2. The detectivity is 4.48 × 1010 cmHz1/2/W. Although the detectivity of the Ge PD by a thinner absorption layer is improved to close the detectivity of the Si PD, there is still room for improvement. The reported dark current density for Ge PD is still high. A better Ge layer quality and appropriate passivation may help to reduce the dark current density.
In this work, we propose a growth method for a Ge epitaxial layer to retain strain with negligible defects for a high detectivity Ge PD in p-Si/i-Ge/n-Si (PIN) structure for NIR detection. The surface of epitaxial Ge was capped with a thick Si layer as a stressor. The high-temperature growth of the Si over Ge enhances Ge's tensile stress, as depicted schematically in Fig. 1. The Ge film may increase its area by creating a rough surface for tensile strain. A Ge PD is made on the tensile strain Ge layer with a Si stressor. A 940 nm laser connected with a fiber probe measures the photocurrent at different applied voltages. The responsivity and detectivity of the strained-Ge PIN PD are calculated by the photocurrent and dark current. The strained-Ge PIN PD has a high responsivity of 0.45 A/W at -1 V bias voltage at 940 nm in wavelength of. The external quantum efficiency at 940 nm is 59%. The dark current density of the PD is as low as ∼1.50 µA/cm2 at -1 V. The detectivity as high as 6.55 × 1011 cmHz1/2/W is obtained by the low dark current density and the high responsivity.
Fig. 1. The schematic diagram of how the surface of Ge extended during Si capping layer growth to ensure a high strain.
2. Experiment
A rapid thermal chemical vapor deposition (RP-CVD) system (ASM, Model Epsilon 2000+) is used to grow the Ge and Si layers. The epitaxial growth uses two Si wafers: one for PD fabrication and another for material characterization. After cleaning, an 8-inch Si wafer (n-type, 1-10 Ω-cm, 725 µm) was loaded into the RP-CVD chamber for Ge growth. After heating the substrate temperature to 400°C, a mixture of H2 and GeH4 (10% in H2) gases was conducted into the chamber, and the pressure was kept at 40 Torr for Ge growth. The Ge film was grown to approximately 1 µm at a growth rate of ∼3.6 Å/s. The Si capping layer was then grown by adding H2, dichlorosilane (DCS), and HCl gases at 825°C. The pressure was kept at 60 Torr with a growth rate of 5 Å/s for a thickness of 110 nm. After growth, one of the Si (110 nm)/Ge (1µm) wafers was annealed at 800°C for 60s to explore the effect of the thermal process for dopant activation after ion implantation in PD fabrication to explore the material structure of the Ge film. Material characterizations were carried out by scanning electron microscopy (SEM), atomic force microscopy (AFM), x-ray diffraction (XRD), Raman spectroscopy, transmission electron microscopy (TEM), and the UV-Vis-NIR spectrometer to understand the morphology, lattice orientation, strain status, micro-defects, and the optical reflectance.
An 8-inch wafer was used as the substrate to fabricate the Ge PIN PD. The process is illustrated in Fig. 2(a). After alignment mark fabrication, the boron ions were implanted into Si through screen oxide at a dosage of 5 × 1014 cm−2 at an energy of 20 KeV to form the p + layer. The screen oxide, deposited by a PECVD system (Oxford 100 PECVD), is 30 nm thick. The boron ions were then activated by a rapid thermal annealing (RTA) at 800°C for 60 s. This thermal process cannot be applied after n + layer ion implantation if the epitaxial grown temperature is not high enough. The p + ion implantation damages the lattice, and the epitaxial layer should grow on the crystalline surface. The activation process of the implanted ion can be done after N + ion implantation. After oxide removal, the Ge absorption layer in 1 µm thickness and the Si capping layer in 110 nm were grown sequentially. After epitaxial growth, the phosphorous ions were implanted into the Si capping layer with a 1 × 1015 cm−2 dosage at an energy of 20 KeV to form the n + layer. After implantation, the phosphorous ions were activated by RTA at 800°C for the 60 s. The Si/Ge/Si substrate layers were then etched by reactive ion etching (RIE) to define the detector area. An oxide in 200 nm was grown by CVD to form a passivation layer. The locations for p and n contacts were defined by photolithography and etched to exposure Si regions through the oxide layer. The Al layer is deposited 750 nm thick by the thermal evaporation process and is patterned by the lift-off process. Since the top Si is doping into n-type by phosphorus, and Al is a p-type dopant, we decided not to do the forming gas annealing for Al contact to prevent the possibility of Al diffusion into the Si, which may cause electron concentration reduction. The schematic layer structure of the Ge PIN PD is shown in Fig. 2(b). The 940 nm IR light is incident from the top of the diode. The optical image of the fabricated Ge PIN PD is shown in Fig. 2(c). The n and p contacts are not directly on the Device. They are extended to the outer contact pad to leave the device area for the fiber probe to guide the IR light illuminate on top of the Device.
Fig. 2. (a) The process flow for making a Ge PD on an epitaxial grown Ge with a Si capping layer, (b) the schematic structure of the Ge PD, and (c) the top view optical images of the Ge PD.
3. Results and discussion
3.1 Material characterization of epitaxial Si/Ge layer
The surface and cross-section morphology of the grown Ge were observed using SEM, and the results are presented in Fig. 3. The cross-section image of the as-deposited Si (110 nm)/Ge (1 µm) sample is shown in Fig. 3(a). The epitaxial film exhibited a rough surface. The Si-Ge interface appeared interlaced, indicating possible interdiffusion. The interdiffusion region was observed to be up to 200 nm thick. The surface morphology of this sample is shown in Fig. 3(b), where many distinct structures in a specific shape were found. The structures were not tightly connected, and some vacancies were observed between them. The separated structures with specific shapes suggest that the strain may have induced them, and the strain will be measured by Raman spectroscopy. The cross-sectional image of the annealed sample is shown in Fig. 3(c), where an even more apparent interlacing of Si and Ge regions with a non-uniform surface was observed. The surface morphology shown in Fig. 3(d) indicates that the pieces of the specific structures were slightly melted on their edge and connected. However, the structure structures are still clearly distinguishable, with some vacancies. The cross-section image of the Ge PIN PD is shown in Fig. 3(e), with a thin layer of silicon oxide deposited on the surface. An EDX line scan was performed for element identification and along the growth direction. The intensity represents the relative concentration. The intermixing between Si and Ge was observed, with Si atoms diffusing deeply into the Ge film and Ge diffusing into the Si region.
Fig. 3. (a) The cross-section, and (b) the surface morphology of the as-deposited Ge epitaxial layer with Si capping layer by SEM. (c) The cross-section, and (d) the surface morphology of the thermally annealed Ge epitaxial layer with Si capping layer by SEM. (e) The Si and Ge elements distribution along the growth direction of the Ge and Si capping layer was taken by EDX. The Si and Ge are inter-diffused at its interface.
The AFM surface morphology on the as-deposited sample is shown in Fig. 4(a). The scan range is 5 × 5 µm2. Many rock-like structures were distributed on the surface. They are well separated, with the highest structure having a height of 200.2 nm. The features of some rocks are rectangular at the base, a terrace on the top, and some in the shape of pyramids. After annealing, those specific shapes disappear, forming irregular or merged rocks on the surface in the AFM image shown in Fig. 4(b). The height is also slightly reduced to 190.2 nm. The enlarged part of the surface morphology by SEM of the as-grown Si/Ge epitaxial layer was shown in Fig.4c. Many pyramid-like structures with small terraces on the top were found over the images. Two of them around the center were marked. The base is square with a side length of approximately 350 nm for the larger one and 280 nm for the smaller one. Aside from the pyramid is a rectangle with a long-side size of 430 and 280 nm for the short-side length. Other structures are approximately in this range. Some structures are connected to form irregular shapes. According to the facet of the diamond structure, the side walls in the pyramid and rectangle are most probably associated with (111) faces, as sketched in Fig.4d.
Fig. 4. The AFM surface morphology of the (a) as-deposited and (b) thermally annealed Ge layer with Si capping. (c) The enlarged part of SEM surface morphology on the as-deposited Ge layer with Si capping. The red rectangles are used to help identify the pyramid structure. (d) The schematic diagram of the facet structure of the surface morphology.
The Si/Ge layers were considered to be relaxed with a rough surface. By the significant lattice constant mismatch between two layers, the highly strained film tends to relax by creating large amounts of dislocations when the thickness of the strained layer exceeds the critical thickness or having another highly stress film grown on it. The epitaxial growth of the Ge on Si is examined by the TEM. The TEM image containing the Si substrate, Ge layer, and Si capping is shown in Fig. 5(a). No cluster of dislocations was found in the images, indicating the strain remained in the epitaxial layer. Some stacking defects are observed near the surface around the Si and Ge interfaces, as shown in Fig. 5(b).
Fig. 5. (a)The TEM images of the Ge film with Si capping layer. (b) The amplified images show the only stacking fault in the Ge film.
The XRD diffraction patterns are used to understand the preferred orientation and the out-of-plane lattice constant. The XRD was taken by the Cu Kα1 characteristic x-ray emission with a wavelength of 1.5405 Å. The preferred orientations for the Ge and Si are (400), as shown in Fig. 6(a). The diffraction peaks for the as-deposited Ge and Si are at the 2θ equal to 66.218° and 69.211°, respectively. The Ge and Si bulk positions in XRD have diffraction angles of 65.99° and 69.13°, respectively. The lattice constants are, therefore, to be 5.6412 Å for the Ge film and 5.4248 Å for the Si. The shift of Ge (400) to a higher diffraction angle indicates an out-of-plane compressive strain. These lattice constants are less than the bulk values for Ge and Si, which are 5.6578 Å and 5.4306 Å [25]. The out-of-plane strain ${\varepsilon _ \bot }$ can be expressed by the difference in the lattice constant [26],
The dfilm is the lattice of the measured epitaxial film, and the dbulk is the film's lattice in its bulk form. Therefore, according to (1), the Ge and Si films are calculated to exhibit 0.29% and 0.11% compressive strain in the out-of-plane direction, respectively. The XRD peaks have very similar 2θ for the Ge and Si in the epitaxial layers after thermal annealing, which are 66.174° and 69.160°, corresponding to the lattice constant of 5.6440 Å and 5.4287 Å. The compressive strain reduced to 0.24% and 0.04% for Ge and Si in the epitaxial layers. The strain in the Si layer almost diminishes after thermal annealing. The microstructures of the as-deposited and thermal annealed epitaxial layers were explored by the Raman scattering with the excitation wavelength at 632.8 nm. The Raman spectra are shown in Fig. 6(b). Three vibrations were found in the Raman spectra, which correspond to the Ge-Ge, Si-Ge, and Si-Si vibrations [27]. The Ge-Ge longitudinal phonon vibration (ωGe-Ge) was found at 291.93 cm−1 for the as-deposited and the thermal annealed Ge in the epitaxial films. The identical Ge-Ge vibration scattering for both samples indicates that the thermal annealing in the fabrication process for the Ge PIN PD has a negligible change in the strain situation and microstructures. This result is reasonable because the Ge layer already has two high-temperature processes during and after epitaxial growth: the growth of the Si (825°C) and the post-implantation activation annealing (800°C).
Fig. 6. (a) The XRD spectrum and (b) Raman scattering of the as-deposited and thermally annealed Ge layer with Si capping.
The amount of biaxial strain in the Ge film was calculated using the formula Δω = bɛ|| [28], where Δω represents the difference in vibration wavenumber between the Ge-Ge vibration (ωGe-Ge) present in the epitaxial film and the Ge-Ge vibration present in bulk (ω0Ge-Ge). The ω0Ge-Ge is 300.17 cm−1 [29]. The b is a phonon strain shift coefficient reported as -390 cm−1 [30] or -450 cm−1 [31]. And ɛ|| is the in-plane strain. The Δω is -8.24 cm−1, corresponding to a tensile strain from 1.83 to 2.11%. The tensile strain in the epitaxial-grown Ge film is attributed to the difference in the thermal expansion coefficient [32,33]. The thermal expansion coefficient (α) for silicon (Si) is 2.6 × 10−6 °C−1, and for germanium (Ge) is 5.9 × 10−6 °C−1. The correlation between the α and the difference in length (ΔL) of a substance with length (L) at a particular temperature and a temperature difference (ΔT) can be expressed as α = ΔL/LΔT.
Ge is grown at high temperatures with the exact dimensions in all directions (in-plane) as the substrate Si. As a result, after cooling down from high temperature to room temperature after Ge growth, Ge tends to shrink more in dimension than Si to the natural lattice constant. However, this shrinkage is limited by the Si substrate, which stretches the Ge lattice to form a larger lattice constant, resulting in a tensile strain. The difference in the thermal expansion coefficient of Ge and Si causes ɛ||, which the following equation can express: [34],
T1 is the growth temperature, and T0 is the room temperature. The ${\alpha _{Ge}}(T ),\; and\; {\alpha _{Si}}(T )$ are the thermal expansion coefficients for Ge [35] and Si [36] as a function of temperature. According to (2), the ɛ|| is calculated to be approximately 0.28% at the temperature of 825°C. The theoretical value is far less than the strain obtained from Raman scattering. The significant difference indicates the simple thermal expansion difference is not the dominant mechanism. The stress established during the epitaxial growth by the lattice mismatch between Ge and Si may cause the highly tensile strain. The thick capping Si causes heavy compressive stress to the Ge epilayer. The out-of-plane compressive stress will enhance the in-plane tensile stress by the Poisson ratio, γ, expressed in the following relationship for a flattened layer.
The Poisson ratio for the Ge thin film on the Si substrate was determined experimentally to be 0.25 [37]. According to (3), the in-plane strain is calculated to be 1.16%. While this value is still lower than the strain obtained by Raman spectroscopy, it can be considered comparable. The Ge epitaxial layer is more complex as the rough surface has many rock-like structures. This rough surface may affect the expression of Eq. (3) or lead to a smaller Poisson ratio in Eq. (3).
The peaks at around 400 cm−1 correspond to the SiGe molecule vibrations [38,39]. This vibration can be deconvoluted into two Gaussian peaks. The peak positions for the epitaxial Si/Ge layer are 386.66 and 398.83 cm−1. The composition of the Ge and the strain in the SiGe layer can be expressed as an empirical algebraic equation by considering the peak position in the Raman spectroscopy [31,40,41].
The ${\omega _{Si - Ge}}({x,\; \varepsilon } )$ is the measured peak position of the Si-Ge vibrations, the $\omega _{Si - Ge}^0$ is the initial value, which is 400.1 cm−1. The x is the percentage of Ge in the Si1-xGx alloy. The ${b_{SiGe}}$ is the phonon strain-shift coefficient, which is -570 ± 50 cm−1 for SiGe. Assume the strain in SiGe is the average of the top Si and the underlying Ge layer. It is 1.16% after calculation, according to (3). Therefore, the vibration peaks at 398.83 cm−1 correspond to a Ge concentration of approximately 23%, and the 386.66 cm−1 vibration corresponds to a Ge concentration of 90%. Since each peak is broad in the Raman spectroscopy, the concentration is assumed to have a wide distribution. We presume the 398.83 cm−1 peak is attributed to the top Si capping layer in which the Ge diffuses into Si to form a SiGe layer with minimal Ge. The 386.66 cm−1 peaks are the Si diffuses into Ge to create a high Ge content SiGe layer. This Si and Ge interdiffusion was observed in the SEM elemental scanning analysis in Fig. 3(e). The peaks for the Si-Si vibration in the annealed and the as-deposited samples are at 516.17 and 515.63 cm−1, respectively. The phonon strain-shift coefficient for Si-Si is reported to be -725 ± 15 cm−1 [42]. The shift of the Si-Si in vibration peaks in the Si/Ge epitaxial layers to the bulk Si is -3.83 and -4.37 cm−1, corresponding to the tensile strain of 0.52 and 0.60%, for the annealed and the as-deposited sample, respectively. This indicates the tensile strain is predominant in the Ge and SiGe films.
3.2 Photoluminescence and Reflectance spectrum of the Si-capped Ge epilayer
The photoluminescence spectra are taken on the as-deposited Si/Ge/Si substrate, and the annealed Si/Ge/Si substrate are shown in Fig. 7(a) and Fig. 7(b), respectively. The excitation wavelength is 405 nm. The photoluminescence spectrum of a bulk Ge consists of three phonon-assisted radiative emission peaks, which are longitudinal acoustic (LA), transverse optic (TO), and transverse acoustic (TA) phonons [43,44]. They are well separated at low temperatures (12 K). The TA, LA, and TO phonon-assist photoluminescence spectrum has peak positions at 0.73, 0.71, and 0.702 eV [45]. These peaks are broadening and convoluted and do not shift explicitly at elevated temperatures. In contrast to the phonon-assist emissions, the bandgap emission from the Ge moves to lower energy at high temperatures. Theoretical studies indicate the peaks for each phonon-assist photoluminescence do red-shift to lower energy with the temperature increase. However, the dominant emissions change at different temperatures, causing the overall peaks seems to remain at the same energy [44]. For example, at 100 K, the LA phonon-assisted emission is dominant. Increase the temperature to 160 K. The dominant emissions are the LA phonon-assisted and the TA phonon-assisted emission. At 295 K, the dominant emission became the TA phonon-assisted emission, which red-shifted to the position of LA phonon-assisted emission at 100 K.
Fig. 7. The room-temperature photoluminescence spectra of the (a) as-deposited and (b) the thermally annealed Ge layer with Si capping. (c) the surface reflectance of the as-deposited and the thermally annealed Ge layer with Si capping.
The emissions in the photoluminescence spectrum for Ge in this experiment exhibit three peaks that can be deconvoluted to 0.767, 0.786, and 0.792 eV for the as-deposited epitaxial layers. The peaks have positions at 0.780 and 0.794 eV for the annealed layers. The 0.767 eV peak disappears. The possible reason for the 0.767 eV peak is the strain-enhanced direct transition [46]. The direct transition has a peak at 0.78 eV for the 0.12% strained-Ge. And the peak energy reduced as the strain increased. Therefore, the 0.78 eV peak may reduce to around 0.767 eV as the strain in this work is ∼1.83%. In addition, the intensity ratio of the direct emission to the phonon-assist emission also enhanced as the strain increased. However, this peak has a very low intensity compared to the phonon-assisted emission, which might not be detectable after annealing.
The reflectance of the Si/Ge layers with rough surfaces was measured by a UV-Vis-NIR spectrometer (Hitachi, U-4100) with an integrated sphere, de-axis by 7°. The reflectance spectrum is shown in Fig. 7(c). The reflectance for the flattened Ge surface is 40 to 35% for the wavelength increased from 800 to 1550 nm. The reflection (R) of a light incident to a flat semiconductor surface is R = (n-1)2/(n + 1)2, in which n is the refractive index of the semiconductor. And the n ∼4.4 for Ge at around the measured wavelength. The R is calculated as 39.6%, consistent with the measurement. With the rough surface established by high strain, the reflectance reduced from 39 to 24% at 940 nm for the annealed Si/Ge layer. The reflectance exhibits a relative 38% reduction. After 940 nm, the thickness of the Si/Ge film causes light interference. The lowest reflectance is 22.5% at a wavelength of 1480 nm, reduced from R = 35% for the flat Ge, corresponding to a relatively 35.7% reduction. The reduced reflectance indicates that the light coupling into the Si/Ge film is enhanced, contributing to the responsivity.
3.3 Optical properties of the Ge PIN diode at 940 nm
A laser with a light emission wavelength of 940 nm is used to measure the PD properties in the dark room. The light is incident on the device surface via a fiber probe, as shown in Fig. 8(a). The laser intensity was measured by a reference PD through a fiber probe. The highly strained Ge PD exhibits a minimal dark current, as low as 1.497 × 10−6 A/cm2 (0.47 nA) at -1 V with a diameter of 200 µm, as shown in Fig. 8(b). The minimum dark current is not at zero bias, indicating the dark room may not be in completely dark conditions. This dark current density is relatively low, approaching a Ge diode's ideal saturation current density, around 10−7 A/cm2. The reverse saturation current density is written as following equation [47],
Fig. 8. (a) the set-up of Ge PD measurement. (b) the J-V characteristic of the Ge PD under dark and laser illumination. The calculated (c) responsivity and (d) detectivity of the Ge PD.
The photocurrent density achieves 1.49 × 10−2 A/cm2 after laser illumination with a power intensity of 10. 51 µW corresponds to a power density of 33.45 mW/cm2. Increase the laser intensity to 154.26 mW/cm2, the photocurrent density increases to 2.1 × 10−2 A/cm2. It is worth noting that the photocurrent is extensive for the applied voltage larger than 0 V and falls at 0.33 V. This behavior represents a photovoltaic effect, in which the diode generates electricity upon light illumination. It is possible that certain Internet of Things (IoT) devices and optical sensor networks may not have high-capacity energy storage units such as batteries or capacitors. They may also be located in hard-to-reach areas where battery replacement is not feasible. In such situations, the generated Ge PDs in photovoltaic mode operation can potentially be utilized as an alternative energy source to recharge the battery by integrating into the IoTs and optical sensor network system [48]. The responsivity is 0.45 A/W at the voltage of -1 V for the 10.51 µW laser intensity. The responsivity reduced to 0.13 A/W with the increased incident laser intensity to 48.47 µW, as shown in Fig. 8(c). However, the responsivity at zero bias is relatively low, ∼2 orders of magnitude lower than that at -1 V bias, and the responsivity shows voltage-dependence behavior. This voltage-dependent responsivity has been found in some published results on the Ge-on-Si PDs [49,50]. One of them attributed it to photoconductive gain without further discussion. Photoconductive gain is a phenomenon that occurs due to the unequal mobility of electrons and holes in a material. In the case of Ge, the electron mobility is 3900 (cm2V−1s−1), and the hole mobility is 1900 (cm2V−1s−1). Due to the higher mobility of electrons over holes, photo-generated holes remain in the absorption layer while the electrode collects electrons. The excess holes then attract electron injection, resulting in a signal amplification known as gain. An increase in the bias voltage applied to the material causes an increase in the electrical field, increasing the photoconductive gain. In contrast, some literature reported Ge-on-Si PD without voltage-dependent responsivity [51–53]. On different types of Si substrates, one study reported the photocurrent, dark current density, and responsivity of the Ge-on-Si PD [53]. The report concludes that Ge PD made on p, p+, and n + Si substrates exhibit flat and voltage-independent responsivity. However, only Ge PD made on the n-type Si substrate exhibits voltage-dependent responsivity. The report further explains that the p-Ge/i-Ge/n-Si PD has a band diagram that shows a barrier at the i-Ge/n-Si heterojunctions for electrons. This barrier slows down the electrons generated by photons in the i region, making it harder for them to reach the n-Si. However, increasing the reverse voltage in PD operation can reduce this barrier, leading to an increase in photocurrent with the applied voltage. The report suggests that the interface barrier between i-Ge/n-Si could be the reason for the voltage-dependent responsivity in our Ge PD. However, to confirm this assumption, we need to conduct further experiments. For instance, we can replace n-Si with n-Ge or increase the doping concentration in the n-Si in our future Ge PD. Therefore, the Ge-on-Di PD with p-Ge/i-Ge/n-Ge homojunction structure does not have voltage-dependent responsivity [51].
In addition to the responsivity, the minimum incident optical power that the PD can distinguish from the noise is also a figure of merit, noise equivalent power (NEP) [54].
R is the responsivity, q is the value of electron charge (1.6 × 10−19 Coulombs), and Jd is the dark current density. According to (6) and (7), due to the low dark current, the specific detectivity is 6.55 × 1011 cmHz1/2/W, as shown in Fig. 8(d). This value is higher than the reported Ge PD and is comparable to the Si PD [55]. The high detectivity might be due to the high-quality strain Ge layer after annealing and the better SiO2 device passivation.
The comparison of the performance properties of responsivity, dark current density, and detectivity to similar devices fabricated on either Si or Ge were listed in Table 1. The responsivity of this work at 940 nm is higher than others at comparable wavelengths. This work's dark current density also exhibits a relatively low value of 1.6 × 10−6 A/cm2, which is more than a thousand times lower. The low dark current may be due to the lower defects by the better epitaxial growth and oxide passivation. As a result, a high detectivity of 6.55 × 1011 cmHz1/2/W is achieved in this work.
Table 1. The comparison of the performance properties of Si or Ge PDs
4. Conclusions
This work presents a highly tensile strained Ge layer with a thick Si capping exhibiting a rough surface morphology. The surface morphology is explored by the SEM and AFM. Pyramid-like Ge rocks were found all over the surface with a top (100) and four sides (111) surface. The XRD and Raman scattering spectrum explored the epitaxial film's lattice structure and strain. The (400) only diffraction peaks indicate an epitaxial growth layer with 0.28% out-of-plane compressive strain. And the shift of Ge-Ge vibration in the Raman scattering spectrum suggests a 1.83% of in-plane tensile strain. A mechanism regarding the Poisson ratio and thermal expansion is considered responsible for the highly tensile strain. The positions of the Si-Ge vibrations reveal the interdiffusion of the Si and Ge. The photoluminescence spectrum indicates the appearance of Ge direct band transition, which is the effect of high strain. The rough surface also implies a light coupling into the Ge PD. The Ge PD has a high responsivity of 0.45 A/W at a wavelength of 940 nm. The high detectivity of 6.55 × 1011 cmHz1/2/W is obtained by the low dark current and the high responsivity. This Ge PD is suitable for applications in the field of light detection and ranging (LiDAR).
Funding
Ministry of Education (Higher Education Sprout Project, Innovation Development Center, of Sustainable Agriculture); National Science and Technology Council (NSTC-112-2221-E-005 -066 -).
Acknowledgments
Z. Pei thanks the Taiwan Semiconductor Research Institute (TSRI) for the help of Ge growth and the semiconductor fabrication processes in this work.
Disclosures
The authors declare no conflicts of interest.
Data availability
Data underlying the results presented in this paper are not publicly available at this time but may be obtained from the authors upon reasonable request.
References
1. S. Royo and M. Ballesta-Garcia, “An Overview of Lidar Imaging Systems for Autonomous Vehicles,” Appl. Sci. 9(19), 4093 (2019). [CrossRef]
2. F. M. Paul, B. Paul Stuart, D. B. Jeffrey, et al., “Comparison of flash lidar detector options,” Opt. Eng. 56(3), 031223 (2017). [CrossRef]
3. S. Monfray, S. Guerber, A. Montagné, et al., “Optical Phased Array for 905-nm LIDAR applications integrated on 300 mm Si-Photonic Platform,” in 2022 Optical Fiber Communications Conference and Exhibition (OFC)(2022), pp. 1–3.
4. S. Donati, G. Martini, Z. Pei, et al., “Analysis of Timing Errors in Time-of-Flight LiDAR Using APDs and SPADs Receivers,” IEEE J. Quantum Electron. 57(1), 1–8 (2021). [CrossRef]
5. A. T. Hsiao, C. H. Liu, P. H. Chen, et al., “Real-time LiDAR module with 64 × 128-pixel CMOS SPAD array and 940-nm PCSEL,” in 2022 IEEE Sensors Applications Symposium (SAS)(2022), pp. 1–6.
6. K.-B. Hong, W.-T. Huang, H.-C. Chung, et al., “High-Speed and High-Power 940 nm Flip-Chip VCSEL Array for LiDAR Application,” in Crystals(2021). [CrossRef]
7. M. Garín, T. P. Pasanen, G. López, et al., “Black Ultra-Thin Crystalline Silicon Wafers Reach the 4n2 Absorption Limit–Application to IBC Solar Cells,” Small 19(39), 2302250 (2023). [CrossRef]
8. S. Thiyagu, B. P. Devi, and Z. Pei, “Fabrication of large area high density, ultra-low reflection silicon nanowire arrays for efficient solar cell applications,” Nano Res. 4(11), 1136–1143 (2011). [CrossRef]
9. E. Yablonovitch, “Statistical ray optics,” J. Opt. Soc. Am. 72(7), 899–907 (1982). [CrossRef]
10. S. Thiyagu, B. P. Devi, Z. Pei, et al., “Ultra-low reflectance, high absorption microcrystalline silicon nanostalagmite,” Nanoscale Res. Lett. 7(1), 171 (2012). [CrossRef]
11. Y.-T. Lee, F.-R. Lin, C.-H. Chen, et al., “A 14.7% Organic/Silicon Nanoholes Hybrid Solar Cell via Interfacial Engineering by Solution-Processed Inorganic Conformal Layer,” ACS Appl. Mater. Interfaces 8(50), 34537–34545 (2016). [CrossRef]
12. S. Ghosh, G. Sun, T. A. Morgan, et al., “Dark Current Analysis on GeSn p-i-n Photodetectors,” in Sensors(2023). [CrossRef]
13. M. R. M. Atalla, S. Assali, S. Koelling, et al., “Dark current in monolithic extended-SWIR GeSn PIN photodetectors,” Appl. Phys. Lett. 122(3), 031103 (2023). [CrossRef]
14. Z. Pei, L. S. Lai, H. P. Hwang, et al., “Si1−xGex/Si multi-quantum well phototransistor for near-infrared operation,” Phys. E 16(3-4), 554–557 (2003). [CrossRef]
15. W. H. Chang, A. T. Chou, W. Y. Chen, et al., “Room-temperature electroluminescence at 1.3 and 1.5 µm from Ge/Si self-assembled quantum dots,” Appl. Phys. Lett. 83(14), 2958–2960 (2003). [CrossRef]
16. B. Son, Y. Lin, K. H. Lee, et al., “High speed and ultra-low dark current Ge vertical p-i-n photodetectors on an oxygen-annealed Ge-on-insulator platform with GeOx surface passivation,” Opt. Express 28(16), 23978–23990 (2020). [CrossRef]
17. Q. Chen, H. Zhou, S. Xu, et al., “A Route toward High-Detectivity and Low-Cost Short-Wave Infrared Photodetection: GeSn/Ge Multiple-Quantum-Well Photodetectors with a Dielectric Nanohole Array Metasurface,” ACS Nano 17(13), 12151–12159 (2023). [CrossRef]
18. B. Son, Y. Lin, K. H. Lee, et al., “Dark current analysis of germanium-on-insulator vertical p-i-n photodetectors with varying threading dislocation density,” J. Appl. Phys. 127(20), 203105 (2020). [CrossRef]
19. P. Zingway, C. S. Liang, L. S. Lai, et al., “A high-performance SiGe-Si multiple-quantum-well heterojunction phototransistor,” IEEE Electron Device Lett. 24(10), 643–645 (2003). [CrossRef]
20. Y. Gao, H. Cansizoglu, S. Ghandiparsi, et al., “High Speed Surface Illuminated Si PDUsing Microstructured Holes for Absorption Enhancements at 900–1000 nm Wavelength,” ACS Photonics 4(8), 2053–2060 (2017). [CrossRef]
21. Z. Zheng, Q. Luo, K. Xu, et al., “All-silicon PIN photodetector based on black silicon microstructure,” Opto-Electron. Eng. 48, 200364-200361 (2021). [CrossRef]
22. G. Dehlinger, S. J. Koester, J. D. Schaub, et al., “High-speed Germanium-on-SOI lateral PIN photodiodes,” IEEE Photonics Technol. Lett. 16(11), 2547–2549 (2004). [CrossRef]
23. H. Zhou, Q. Chen, S. Wu, et al., “Grating and hole-array enhanced germanium lateral p-i-n photodetectors on an insulator platform,” Opt. Express 30(4), 4706–4717 (2022). [CrossRef]
24. A. Sammak, M. Aminian, L. K. Nanver, et al., “CMOS-Compatible PureGaB Ge-on-Si APD Pixel Arrays,” IEEE Trans. Electron Devices 63(1), 92–99 (2016). [CrossRef]
25. Y. Y. Fang, J. Tolle, R. Roucka, et al., “Perfectly tetragonal, tensile-strained Ge on Ge1−ySny buffered Si(100),” Appl. Phys. Lett. 90(6), 061915 (2007). [CrossRef]
26. S. Dolabella, A. Borzì, A. Dommann, et al., “Lattice Strain and Defects Analysis in Nanostructured Semiconductor Materials and Devices by High-Resolution X-Ray Diffraction: Theoretical and Practical Aspects,” Small Methods 6(2), 2100932 (2022). [CrossRef]
27. T. S. Perova, J. Wasyluk, K. Lyutovich, et al., “Composition and strain in thin Si1−xGex virtual substrates measured by micro-Raman spectroscopy and x-ray diffraction,” J. Appl. Phys. 109(3), 033502 (2011). [CrossRef]
28. G. Capellini, C. Reich, S. Guha, et al., “Tensile Ge microstructures for lasing fabricated by means of a silicon complementary metal-oxide-semiconductor process,” Opt. Express 22(1), 399–410 (2014). [CrossRef]
29. G. Capellini, G. Kozlowski, Y. Yamamoto, et al., “Strain analysis in SiN/Ge microstructures obtained via Si-complementary metal oxide semiconductor compatible approach,” J. Appl. Phys. 113(1), 013513 (2013). [CrossRef]
30. A. Z. Al-Attili, S. Kako, M. K. Husain, et al., “Whispering Gallery Mode Resonances from Ge Micro-Disks on Suspended Beams,” 2, (2015). [CrossRef]
31. F. Pezzoli, E. Bonera, E. Grilli, et al., “Phonon strain shift coefficients in Si1−xGex alloys,” J. Appl. Phys. 103(9), 093521 (2008). [CrossRef]
32. T. Imajo, T. Suemasu, and K. Toko, “Strain effects on polycrystalline germanium thin films,” Sci. Rep. 11(1), 8333 (2021). [CrossRef]
33. Y. S. Li and J. Nguyen, “Tensilely Strained Ge Films on Si Substrates Created by Physical Vapor Deposition of Solid Sources,” Sci. Rep. 8(1), 16734 (2018). [CrossRef]
34. D. D. Cannon, J. Liu, Y. Ishikawa, et al., “Tensile strained epitaxial Ge films on Si(100) substrates with potential application in L-band telecommunications,” Appl. Phys. Lett. 84(6), 906–908 (2004). [CrossRef]
35. H. Singh, “Determination of thermal expansion of germanium, rhodium and iridium by X-rays,” Acta Cryst A 24(4), 469–471 (1968). [CrossRef]
36. Y. Okada and Y. Tokumaru, “Precise determination of lattice parameter and thermal expansion coefficient of silicon between 300 and 1500 K,” J. Appl. Phys. 56(2), 314–320 (1984). [CrossRef]
37. J. Bharathan, J. Narayan, G. Rozgonyi, et al., “Poisson Ratio of Epitaxial Germanium Films Grown on Silicon,” J. Electron. Mater. 42(1), 40–46 (2013). [CrossRef]
38. J. H. Zhao, C. H. Li, Q. D. Chen, et al., “Surface Detection of Strain-Relaxed Si1−xGex Alloys With High Ge-Content by Optical Second-Harmonic Generation,” IEEE J. Quantum Electron. 51(12), 1–6 (2015). [CrossRef]
39. P. M. Mooney, F. H. Dacol, J. C. Tsang, et al., “Raman scattering analysis of relaxed GexSi1−x alloy layers,” Appl. Phys. Lett. 62(17), 2069–2071 (1993). [CrossRef]
40. F. Pezzoli, E. Grilli, M. Guzzi, et al., “Strain-induced shift of phonon modes in Si1-xGex alloys,” Mater. Sci. Semicond. Process. 9(4-5), 541–545 (2006). [CrossRef]
41. G. Capellini, M. De Seta, Y. Busby, et al., “Strain relaxation in high Ge content SiGe layers deposited on Si,” J. Appl. Phys. 107(6), 063504 (2010). [CrossRef]
42. S. Nakashima, T. Mitani, M. Ninomiya, et al., “Raman investigation of strain in Si∕SiGe heterostructures: Precise determination of the strain-shift coefficient of Si bands,” J. Appl. Phys. 99(5), 053512 (2006). [CrossRef]
43. G. A. Thomas, E. I. Blount, and M. Capizzi, “Indirect recombination mechanisms in germanium,” Phys. Rev. B 19(2), 702–718 (1979). [CrossRef]
44. J. Menéndez, C. D. Poweleit, and S. E. Tilton, “Temperature-dependent photoluminescence in Ge: Experiment and theory,” Phys. Rev. B 101(19), 195204 (2020). [CrossRef]
45. R. R. Lieten, K. Bustillo, T. Smets, et al., “Photoluminescence of bulk germanium,” Phys. Rev. B 86(3), 035204 (2012). [CrossRef]
46. T. H. Cheng, K. L. Peng, C. Y. Ko, et al., “Strain-enhanced photoluminescence from Ge direct transition,” Appl. Phys. Lett. 96(21), 211108 (2010). [CrossRef]
47. D. Neamen, “The pn Junction Diode,” in Physics of Semiconductor Devices(McGraw Hill, 2012), pp. 276–323.
48. S. Schmickl, T. Faseth, and H. Pretl, “Microwatt power management: challenges of on-chip energy harvestingMicrowatt Power Management: Herausforderungen bei On-Chip Energy Harvesting,” Elektrotech. Inftech. 138(1), 31–36 (2021). [CrossRef]
49. Y. Lin, D. Ma, K. Hong Lee, et al., “PIC-integrable, uniformly tensile-strained Ge-on-insulator photodiodes enabled by recessed SiNx stressor,” Photonics Res. 9(7), 1255–1263 (2021). [CrossRef]
50. P. H. Huang, C. H. Chou, and C. L. Hsin, “Ge Photodetector Monolithically Integrated on Si by Rapid-Melting-Growth Technique,” IEEE Photonics Technol. Lett. 27(12), 1254–1256 (2015). [CrossRef]
51. J. Osmond, L. Vivien, J.-M. Fédéli, et al., “40 Gb/s surface-illuminated Ge-on-Si photodetectors,” Appl. Phys. Lett. 95(15), 151116 (2009). [CrossRef]
52. J. Michel, J. Liu, and L. C. Kimerling, “High-performance Ge-on-Si photodetectors,” Nat. Photonics 4(8), 527–534 (2010). [CrossRef]
53. C. Masini, L. Calace, G. Assanto, et al., “High-performance p-i-n Ge on Si photodetectors for the near infrared: from model to demonstration,” IEEE Trans. Electron Devices 48(6), 1092–1096 (2001). [CrossRef]
54. X. Gong, M. Tong, Y. Xia, et al., “High-Detectivity Polymer Photodetectors with Spectral Response from 300 nm to 1450 nm,” Science 325(5948), 1665–1667 (2009). [CrossRef]
55. J. Y. Li, L. Q. Shan, L. Y. Liang, et al., “Light Trapping in Conformal CuO/Si Pyramids Heterojunction for Self-Powered Broadband Photodetection,” IEEE Photonics Technol. Lett. 34(21), 1163–1166 (2022). [CrossRef]
56. T. H. Loh, H. S. Nguyen, R. Murthy, et al., “Selective epitaxial germanium on silicon-on-insulator high speed photodetectors using low-temperature ultrathin Si0.8Ge0.2 buffer,” Appl. Phys. Lett. 91(7), 073503 (2007). [CrossRef]
