Optimization of Strain and Doping in Ge/GeSi Nanoscale Multilayers for GOI Short-Wave Infrared Imaging Applications

Abstract

In this study, in situ P-doping of Ge-based layers has been studied and compared with implanted layer profiles acting as absorbent top layer in PIN photodetectors. Several structures containing multilayers of n⁺-Ge/i-Ge, n⁺-GeSi/i-Ge, and n⁺-Ge/i-GeSi, were designed to regulate dopant out-diffusion and interface quality. The purpose of this study is to make an optimized n-type doping layer for PIN photodetectors with low dark current, high responsivity, and high quantum efficiency operating in short wavelength infrared (SWIR) region. The Ge-based structure on Si substrate was transferred to oxidized Si substrate and was finally back-etched from Si to form Ge-on-insulator (GOI) substrate. Comprehensive characterization using high-resolution X-ray diffraction (HR-XRD), secondary ion mass spectrometry (SIMS), scanning electron microscopy (SEM), transmission electron microscopy (TEM), atomic force microscopy (AFM), and photoluminescence (PL) have been applied at the first stage of our work. The initial Ge layer contains tensile strain of 0.15–0.17%. PL measurements further indicate a redshift of the Γ-LH transition and carrier-concentration-induced quenching at high doping levels, highlighting the competing effects of band filling and non-radiative recombination in heavily n-doped Ge structures. To circumvent this fundamental trade-off, we devised a decoupled device strategy in which the active absorption region employs an intrinsic Ge/GeSi nanoscale multilayer structure to preserve crystal and interface quality. Although, the epitaxial growth parameters were on the optimized conditions, still out-diffusion (in form of segregation and auto-doping) of P could not be impeded. Our final n-type layer in PIN structure was formed by implantation. This approach yields high-performance photodetectors with a peak responsivity of 0.99 A/W at 1550 nm, a corresponding external quantum efficiency of 79%, and low specific contact resistivities of 2.66 × 10⁻⁶ Ω·cm² (n-type) and 1.38 × 10⁻⁸ Ω·cm² (p-type). This work demonstrates that the strategic combination of multilayer/interface engineering and ion-implantation-based doping is a highly effective strategy for tailoring the optoelectronic properties of Ge-based nanomaterials for high-performance SWIR photodetection.

1. Introduction

High-performance, monolithic short-wave infrared (SWIR) photodetectors (PDs) are essential for the development of next-generation silicon optoelectronics, owing to their critical roles in emerging applications such as biomedical imaging [1,2], autonomous sensing [3,4], telecommunications [5,6], and quantum imaging [7,8]. Current technological paradigms are defined by fundamental material compromises. III–V compound semiconductors (e.g., InGaAs/InP) offer superior sensitivity but incur prohibitive costs and complexity in hybrid integration [9,10,11,12]. Colloidal quantum dots provide spectral agility through solution processing, yet face intrinsic limitations in long-term stability and lithographic compatibility [13,14,15,16]. While epitaxial Ge-on-Si affords a fully group-IV pathway, its performance is frequently constrained by defect-mediated recombination, dopant non-uniformity, and limited optical confinement [17,18]. Within this landscape, germanium-on-insulator (GOI) has emerged as a strategically optimized platform, combining enhanced optical isolation with full CMOS-process compatibility [19,20,21,22,23,24,25,26]. However, to fully exploit the potential of GOI for scalable SWIR systems, precise nanoscale control over dopants, strain, and band structure within the epitaxial Ge layer is essential—a challenge that motivates advanced multilayer structure engineering [27,28,29,30,31,32].

To transcend the performance limits of conventional Ge-based designs, band-structure engineering via epitaxial multilayer structures has been widely explored [33,34,35,36,37,38,39,40,41,42,43,44]. A prevalent approach employs Ge/GeSi nanoscale multilayers, which utilize pseudomorphic strain to introduce conduction-band offsets and improve carrier confinement [45,46,47,48,49,50]. However, strain-only designs encounter an intrinsic thermodynamic constraint: the Si fraction in GeSi barriers is typically limited to ≤30% to preserve epitaxial coherence, imposing a fundamental ceiling on band discontinuity and leading to performance plateaus in quantum efficiency [51,52,53,54,55,56,57]. This highlights a critical deficiency in the group-IV toolkit: the lack of a second, independently tunable parameter to cooperatively optimize band alignment and carrier population [58,59,60,61,62,63]. Heavy n-type doping has been proposed as such a complementary degree of freedom [64,65,66,67,68,69,70,71,72,73]. At high phosphorus concentrations (>1 × 10¹⁹ cm⁻³), pronounced band-gap renormalization and Fermi-level engineering can be achieved [74,75,76,77,78]. When combined with tensile strain, a synergistic mechanism emerges: strain reduces the L-Γ valley separation, while high electron concentrations raise the Fermi level to preferentially populate the Γ-valley, thereby enhancing the effective direct-gap character [79,80,81,82,83,84,85,86,87]. While theoretically promising, the practical implementation of this dual-parameter (strain and doping) strategy—particularly within the complex geometry of nanoscale multilayer structures—remains largely unexplored. Key challenges include managing dopant interdiffusion, mitigating doping-induced non-radiative recombination, and preserving interface integrity, all of which are critical for device performance [88,89,90,91,92,93].

To address these challenges, this work adopts a hierarchical experimental strategy that progresses from fundamental material assessment to structurally engineered nanoscale multilayer structures and culminates in a functional device demonstration. We first investigate uniform n-type Ge epitaxial films grown on Si with five distinct phosphorus doping concentrations: 2 × 10¹⁸, 3 × 10¹⁸, 4 × 10¹⁸, 9 × 10¹⁸, and 2 × 10¹⁹ cm⁻³. The doping-dependent strain, crystal quality, and bandgap characteristics of the films were analyzed by high-resolution X-ray diffraction (HR-XRD) and photoluminescence (PL) spectroscopy. Building on this understanding, we then probe the coupled effects in a more device-relevant geometry by designing and synthesizing doping-modulated nanoscale multilayer structures (n⁺-Ge/i-Ge, n⁺-GeSi/i-Ge, and n⁺-Ge/i-GeSi) that are designed and grown to probe the coupled effects of doping, composition, and strain; subsequently, a systematic multiscale analysis of their structure and composition was conducted using scanning electron microscopy (SEM), atomic force microscopy (AFM), cross-sectional transmission electron microscopy (TEM), and secondary ion mass spectrometry (SIMS). These studies uncover a critical constraint: while high doping can be spatially confined, it inevitably introduces significant non-radiative recombination and degrades interface sharpness, thereby compromising the optical efficiency of the active region. To transcend this material-level compromise, we introduce and demonstrate a decoupled design principle for the final device. The optical absorption region employs an intrinsic Ge/GeSi nanoscale multilayer structure, optimized for strain and carrier confinement while remaining undoped to maximize material quality. Conversely, the electrical p-n junction is precisely defined using ion implantation and rapid thermal annealing. This approach strategically isolates the optical optimization from the doping process. Consequently, the core innovation of this work evolves from merely exploring the strain–doping synergy to establishing a manufacturable integration strategy that leverages nanoscale material engineering for optical performance while utilizing conventional processing for electrical function. This work establishes a clear pathway from fundamental nanoscale material insights to high-performance, CMOS-compatible SWIR detectors, providing a practical framework for next-generation Si-based optoelectronic systems.

2. Materials and Methods

Phosphorus-doped Ge epitaxial films were grown on (100)-oriented p-type Si substrates using RPCVD (ASM Epsilon 2000, Almere, The Netherlands). To accommodate the large lattice mismatch between Ge and Si, a conventional low-temperature/high-temperature (LT/HT) two-step epitaxial strategy was employed. Specifically, a thin Ge buffer layer was first deposited at 400 °C to suppress island formation and promote continuous nucleation, followed by the growth of a thick, high-quality Ge layer at 650 °C. In situ phosphorus doping was introduced during the high-temperature growth step. To systematically investigate the influence of in situ doping concentration on structural and optical properties, five samples with monotonically increasing phosphorus concentrations were prepared, spanning from 2 × 10¹⁸ to 2 × 10¹⁹ cm⁻³. This well-controlled doping series enabled a quantitative assessment of dopant-induced strain modulation, defect evolution, and optical response. The total thickness of Ge was ~1 μm. After epitaxy, a 20 nm SiO₂ capping layer was deposited by plasma-enhanced chemical vapor deposition (PECVD) to suppress phosphorus out-diffusion during post-growth thermal treatment. Rapid thermal annealing (RTA) was subsequently performed at 550 °C for 60 s in a nitrogen ambient to activate dopant. While the strain and layer quality were preserved and no dopant out-diffusion was detectable in our structures after RTA treatment. Owing to the mismatch in thermal expansion coefficients between Ge and Si, a residual tensile strain was coherently introduced into the Ge epilayer during cool-down, providing an additional degree of nanoscale strain modulation.

To extend the investigation beyond uniformly doped Ge films and explore dopant behavior in more device-relevant architectures, three types of nanoscale multilayer structures-n⁺-Ge/i-Ge, n⁺-GeSi/i-Ge, and n⁺-Ge/i-GeSi-were fabricated on Si substrates. These structures were designed to emulate key building blocks of SWIR PDs. Sharp doping profiles and abrupt heterointerfaces are critical for device performance. To this end, each nanoscale multilayer featured a deliberately graded phosphorus profile, increasing from the interface toward the surface. This design allowed systematic evaluation of dopant diffusion kinetics, segregation tendencies, and interface abruptness under thermal processing. In particular, the comparison among Ge-only and GeSi-based structures allowed isolation of compositional effects on phosphorus redistribution and strain relaxation behavior. High-resolution X-ray diffraction (HR-XRD) ω-2θ scans around the (004) reflection was performed to extract the out-of-plane lattice parameter and evaluate strain and crystalline quality (Section 3.1). Prior to the characterization, the SiO₂ capping layer was removed by wet etching in dilute hydrofluoric acid. Room-temperature photoluminescence (PL) spectra of the uniformly doped n-Ge films were acquired. Excitation was provided by a 785 nm diode laser with an incident power of 1 mW. Emission was collected in a back-scattering configuration through a 50× objective and spectrally resolved using a liquid-nitrogen-cooled InGaAs detector over the wavelength range of 1200–2300 nm (Section 3.2). Surface and cross-sectional morphologies of the nanoscale multilayer structures were examined using field emission scanning electron microscopy (FE-SEM) and atomic force microscope (AFM) in tapping mode. For atomic-scale analysis, cross-sectional specimens were prepared by focused ion beam (FIB) milling and characterized using aberration-corrected transmission electron microscope (TEM) to assess interface sharpness, defect distribution, and strain contrast. Phosphorus concentration profiles and interlayer diffusion lengths were quantified via secondary ion mass spectroscopy (SIMS), providing insight into dopant segregation and redistribution across the nanoscale multilayer structures. The comprehensive characterization of multilayer structures established fundamental design rules governing dopant distribution, strain, and interface quality (Section 3.3). The insights gained—particularly the identified trade-off between high doping concentration and optical material quality—directly informed the subsequent device design strategy, which strategically employs an intrinsic absorption region and ion-implanted contacts to achieve high-performance PDs. Based on these material-level insights, Ge/GeSi nanoscale multilayer structure SWIR PDs were fabricated to validate the proposed decoupled design approach. The detailed fabrication process and characterization methods for these devices are described in Section 3.4.

3. Results and Discussions

3.1. X-Ray Characterization

As the crystal quality and strain state directly affect the electronic and optical performance of Ge epilayers, HR-XRD characterization was performed on n-Ge films with five different phosphorus doping concentrations to quantify the corresponding strain levels. The doping concentrations of these samples were independently measured using SIMS. The strain was calculated based on the lattice mismatch relative to unstrained bulk Ge. Figure 1 presents the HR-XRD results. The Ge diffraction peaks of all samples are nearly symmetric, with no obvious shoulder on the high-angle side. This indicates that Ge/Si interdiffusion during thermal treatment is negligible, allowing the originally abrupt interface to be preserved without forming an intermediate Si_1−xGe_x transition layer. All Ge peaks exhibit a shift toward higher angles relative to bulk Ge, consistent with the presence of tensile strain. This tensile strain arises from the mismatch in thermal expansion coefficients between the Ge epilayer and the Si substrate. The perpendicular lattice constant (${\mathit{\alpha }}^{\perp }$) of the epilayer can be extracted from the XRD diffraction peak position using Bragg’s law [94].

$${\mathit{\alpha }}^{\perp }={\displaystyle \frac{2\mathit{\lambda }}{\sin \left(\frac{{\mathit{\omega }}_{Ge}}{2}\right)}}$$

(1)

where $\mathit{\lambda }$ is the incident wavelength of the radiation (Cu K_a1 line, $\mathit{\lambda }$ = 1.5406 Å) and ${\mathit{\omega }}_{Ge}$ is the angular position of the Ge (004) peak measured by HR-XRD on the Si (004) substrate. The in-plane lattice constant (${\mathit{\alpha }}^{\parallel }$) can be defined as:

$${\mathit{\alpha }}^{\parallel } = {\displaystyle \frac{1+\nu }{2\nu }}\left[{\mathit{\alpha }}_{Ge}-{\mathit{\alpha }}^{\perp }\left({\displaystyle \frac{1-v}{1+v}}\right)\right]$$

(2)

where $\nu$ is the elastic modulus of Ge ($\nu$ = 0.271) [57,73], and ${\mathit{\alpha }}_{Ge}$ is the unstrained Ge lattice constant (${\mathit{\alpha }}_{Ge}$ = 5.6576 Å). The residual strain of Ge epilayer can be calculated by the following equation:

$$\mathit{\epsilon } = {\displaystyle \frac{{\mathit{\alpha }}_{Ge}^{\parallel }-{\mathit{\alpha }}_{Ge}}{{\mathit{\alpha }}_{Ge}}}$$

(3)

The sign of $\mathit{\epsilon }$ reflects the strain nature: ε > 0 indicates tensile strain, and ε < 0 indicates compressive strain in the Ge epilayer. For the five n-doped Ge samples, the extracted strain values range from 0.15% to 0.17%, confirming the presence of tensile strain. Although samples with higher phosphorus concentrations exhibit slightly larger strain values, the variation is minimal. This consistent tensile strain originates predominantly from the thermal expansion coefficient mismatch between the Ge epilayer and the Si substrate during final cool-down of the substrate. The minor variations observed in the measured strain values are within experimental uncertainty, confirming that phosphorus doping has negligible influence on the strain state for the doping concentrations investigated. Four-probe electrical measurements were subsequently performed to determine the active carrier concentrations of the samples. As shown in

Table 1. Four-probe electrical measurements and extracted active carrier concentrations for n-doped Ge samples with varying phosphorus concentrations.

Sample	Dopant Concentration (cm−3)	Sheet Resistance (Ω)	ρ (Ω * cm)	Active Concentration (cm−3)
A	2 × 1018	165	0.006	1 × 1018
B	3 × 1018	61	0.0057	2 × 1018
C	4 × 1018	25	0.0025	4 × 1018
D	9 × 1018	17	0.0017	8 × 1018
E	2 × 1019	14	0.0014	1 × 1019

, the active carrier concentrations for some low-doped samples are lower than the nominal dopant concentrations, suggesting partial dopant activation. The other results indicate that the measured active concentrations closely match the phosphorus concentrations obtained from SIMS, confirming that the majority of incorporated dopants are electrically activated. Increasing the concentration of activated n-type doping in epitaxial Ge primarily reduces contact and series resistance, thereby improving the external quantum efficiencies (EQEs) and responsivity of PDs. Heavily activated n-type doping also enables effective band engineering by filling the L valleys, which is essential for quasi-direct bandgap operation. However, excessive n-type activation can lead to increased Auger recombination and free-carrier absorption, which can increase dark current and decrease optical efficiency.

3.2. Photoluminescence Measurements

The influence of n-type doping concentration on PL was investigated using five Ge-on-Si samples with phosphorus concentrations ranging from 2 × 10¹⁸ cm⁻³ to 2 × 10¹⁹ cm⁻³. As shown in Figure 2a, the dominant emission peak observed between 1600 and 1800 nm corresponds to the direct optical transition at the Γ-valley of tensile-strained Ge. The shoulder feature at longer wavelengths (1800–2000 nm) can be attributed to indirect bandgap emission. For Sample A, the direct band-to-band transition occurs at 0.765 eV (1620 nm), which is 35 meV lower than the direct bandgap of bulk Ge (0.80 eV, 1550 nm). This redshift in the bandgap relative to bulk Ge arises from the combined effects of tensile strain and heavy n-type doping in the Ge epilayer. The light-hole (LH) and heavy-hole (HH) valence bands in Ge are split at the zone center of the Brillouin zone (BZ) due to epitaxial strain-induced tetragonal deformation. Theoretically, this splitting should give rise to separate PL peaks corresponding to the LH and HH transitions. In practice, these two transitions partially overlap, resulting in a broadening of the overall PL peak. For Samples A (2 × 10¹⁸ cm⁻³) and B (3 × 10¹⁸ cm⁻³), the PL intensity increases with higher n-type doping. However, when the carrier concentration exceeds 3 × 10¹⁸ cm⁻³, PL intensity begins to decrease due to carrier-concentration quenching, balancing the initial enhancement from band-filling effects. The strain-induced valence band splitting leads to two closely spaced optical transitions. Figure 2b shows the PL spectrum of Sample E, in which the peak splitting is clearly visible. The spectrum comprises transitions from the Γ-valley to both the HH and LH bands, as well as overlapping contributions from the L-valley to the HH and LH bands, which collectively broaden the high-wavelength side of the PL peak.

As noted by Sun et al. [79], the PL intensity of Ge layers generally increases with carrier density. In contrast, as illustrated in Figure 2, the PL signal corresponding to transitions from the Γ-valley to the valence band decreases when the carrier concentration exceeds 3 × 10¹⁸ cm⁻³. For compound semiconductors such as GaAs and GaP, the phenomenon where light emission intensity decreases is carrier concentration quenching. It is typically attributed to non-radiative recombination pathways, including vacancy-complex formation, precipitate generation, and band-to-band or impurity-assisted Auger recombination. For Ge layers grown on Si in this work, carrier-concentration quenching counteracts the anticipated benefits of heavy n-type doping, particularly in the context of Ge lasing. As the dopant concentration increases, the loss in emission intensity due to quenching can outweigh the gain from band-filling effects. This behavior may be attributed to degraded crystallinity caused by point defects, which is consistent with prior observations reported by Oehme [77] and Kasai et al. [81], revealed the similar trend. Furthermore, with increasing n-type doping, the Γ-HH peak position remains essentially unchanged, whereas the Γ-LH peak exhibits a pronounced redshift. This observation highlights that heavy-hole states are less sensitive to doping-induced bandgap renormalization, whereas light-hole states shift more significantly, emphasizing the nuanced impact of heavy n-type doping on valence subbands. Due to its larger effective mass and higher density of states, the HH band is relatively insensitive to bandgap renormalization and local strain variations, resulting in an essentially fixed Γ-HH transition energy. In contrast, the LH band, with a smaller effective mass, is more susceptible to bandgap renormalization, enhanced valence band splitting, and doping-induced strain redistribution, leading to a reduction in the Γ-LH transition energy and a corresponding redshift of the PL component. These observations imply that, beyond a certain threshold, increasing n-type doping in the Ge active layer does not enhance, and may even reduce, the radiative efficiency of optoelectronic devices.

3.3. n⁺-Ge/i-Ge, n⁺-GeSi/i-Ge, and n⁺-Ge/i-GeSi Nanoscale Multilayer Structures

In most cases to form the n-type doped layer in a PIN detector structure is applying implantation followed by an RTA treatment for dopant activation. Therefore, our initial idea to form a PIN photodetector is the possibility to grow the whole structure since in such a case, the processing of detector could be shortened. Therefore, to explore the growth behavior of P-doped Ge/GeSi multilayers, we investigated three representative structures: n⁺-Ge/i-Ge, n⁺-GeSi/i-Ge, and n⁺-Ge/i-GeSi. A series of nanoscale multilayer films were fabricated with alternating n-type and intrinsic layers. For the n⁺-Ge/i-Ge multilayer, a 50 nm n⁺-Ge layer and a 100 nm intrinsic Ge layer were alternated four times, followed by a 400 nm intrinsic Ge capping layer. The n⁺-GeSi/i-Ge multilayer consisted of four repetitions of a 100 nm n⁺-GeSi/200 nm i-Ge bilayer. In the n⁺-Ge/i-GeSi multilayer, four periods of a 100 nm i-GeSi/200 nm n⁺-Ge bilayer were grown, capped with a 200 nm intrinsic Ge layer for surface protection. Figure 3 illustrates the design schematics of these nanoscale multilayer structures. The phosphorus concentration in the n-type layers was deliberately graded, increasing from the bottom to the top n-type layer to investigate dopant diffusion and segregation effects. The GeSi layers in this study contain approximately 30% Si.

The SEM result in Figure 4a shows that the surface of the n⁺-Ge/i-Ge multilayer was very rough and the AFM result in Figure 4d shows the roughness of the wafer surface is as high as 47.9 nm. The pronounced surface roughness (a transition from layer-by-layer of three-dimensional growth) and large-scale undulation observed in the n⁺-Ge/i-Ge nanoscale multilayer grown on Si indicate significant degradation of the epitaxial quality. Such morphology typically reflects substantial strain accumulation and partial relaxation within the Ge/Si structure, given the large lattice mismatch between Ge and Si. The relaxation process likely generates misfit and threading dislocations, whose propagation toward the surface produces cross-hatch features and height modulations. In addition, the high n-type doping level in the Ge layers can suppress adatom mobility and promote three-dimensional island-like growth, further enhancing surface roughening and undulation. The repeated n/i multilayer also amplifies interface roughness through vertical propagation across the nanoscale multilayer periods. The observed severe surface corrugation reflects a combination of strain-induced relaxation, defect-mediated roughening, and dopant-modulated growth kinetics, resulting in a substantially nonideal epitaxial morphology on Si substrates.

As shown in Figure 4b,c,e,f, the n⁺-GeSi/i-Ge and n⁺-Ge/i-GeSi multilayers grown on Si, in contrast to the n⁺-Ge/i-Ge multilayer, exhibit considerably smoother surfaces with reduced roughness and more clearly defined layer interfaces. Compared with n⁺-Ge/i-Ge multilayer sample, the surface of n⁺-GeSi/i-Ge and n⁺-Ge/i-GeSi multilayer samples are remarkably better, with the roughness of 18.5 nm and 14.6 nm. This improvement suggests that the P incorporation in GeSi is higher than Ge. In general, the dopant incorporation in Ge (or Si) is ruled by thermodynamical theory which indicates the influence of strain effect and chemical effect. The latter effect is ruled by the tendency toward making Ge–Ge bond compared to P–P or P–Ge bond. This effect is expressed in heat sublimation energy and it is in favor of P–P bonding compared to P–Ge. Meanwhile, there is also strain effect which is directed by the difference in atomic radius between Ge and P. Both these effects are dominant for incorporation of P in Ge and it results in low P incorporation. Meanwhile the above two effects in GeSi are very different. The chemical effect is better in Si compared to Ge and the difference in atomic radius of Si and P is almost minor. Therefore, the P incorporation in GeSi is higher than Ge alone. In addition, GeSi layers generally provide higher adatom mobility and more favorable surface diffusion kinetics compared to highly doped Ge, which helps maintain two-dimensional layer-by-layer growth and reduces interface roughening during the repeated multilayer process. Consequently, the relatively smooth morphology and sharp interfaces indicate that partial Si alloying promotes more coherent epitaxy.

The details in the TEM images in Figure 5 also show that the n⁺-Ge/i-GeSi nanoscale multilayer exhibits noticeably sharper interfaces, and its highly doped top n-Ge layer is able to maintain the intended thickness with minimal degradation. In contrast, the top n-GeSi layer in the n⁺-GeSi/i-Ge multilayer becomes thinner and increasingly interfacial roughness when the P concentration has increased to 2 × 10¹⁹ cm⁻³, deviating from the designed thickness. This suggests that the presence of high P atoms during the growth may strongly affect the growth kinetics. These results indicate that n⁺-Ge/i-GeSi nanoscale multilayer provide a more compliant and kinetically favorable growth, whereas heavily doped GeSi in n⁺-GeSi/i-Ge is more susceptible to strain- and kinetics-driven degradation.

To date, SIMS could probe dopant profile and obtain the depth resolution inside such nanoscale multilayer structure. Figure 6a shows the SIMS result and it is observed that it is difficult to form a clear interface between n-Ge and i-Ge, which is consistent with the TEM characterization surface results. Although cross-sectional TEM indicates that the n⁺-Ge/i-GeSi multilayer exhibits sharper morphological interfaces and that both the top heavily doped n-Ge and the intrinsic GeSi layers retain their designed thicknesses, SIMS depth profiles reveal the chemical picture regarding material doping concentration. The n⁺-GeSi/i-Ge multilayer in Figure 6b shows much steeper concentration gradients, preserves the intrinsic doping level of the i-Ge layer, and supports >6 × 10¹⁹ cm⁻³ doping in the top n-GeSi, while in the n⁺-Ge/i-GeSi multilayer structure in Figure 6c, significant interdiffusion occurs in the nominal intrinsic GeSi layer, with the doping concentration of the top n-Ge layer reaching only about 5 × 10¹⁸ cm⁻³. The presence of Si in the doped GeSi layer alters the thermodynamics and kinetics of dopant solubility and migration which can suppress dopant out-diffusion and produce sharper SIMS profiles in the n⁺-GeSi/i-Ge case. High n-type doping in pure Ge may decrease surface adatom mobility and promote segregation or clustering during growth or subsequent thermal steps, resulting in net loss of dopants from the intended top layer, thereby lowering the measured peak concentration. The n⁺-GeSi/i-Ge multilayer appears to create a chemical and defect environment that is more retentive of dopants, whereas heavily doped Ge adjacent to GeSi is more susceptible to kinetically enhanced intermixing and dopant depletion, yielding broadened profiles and reduced top-layer peaks. The comprehensive characterization of the three in situ doped nanoscale multilayer structures (n⁺-Ge/i-Ge, n⁺-GeSi/i-Ge, and n⁺-Ge/i-GeSi) provides critical insights into the fundamental trade-offs between doping, strain, and interfacial integrity. A pivotal conclusion from this study is that despite enabling nanoscale dopant and strain engineering, in situ doped Ge/GeSi multilayers inherently suffer from significant dopant interdiffusion and interface degradation, as unequivocally revealed by SIMS and TEM. These material-level shortcomings would directly manifest as junction broadening and elevated leakage currents in photodetectors, fundamentally limiting device performance. These SIMS features, combined with interface roughening and dopant diffusion observed in in situ doped multilayers, indicate that devices based on these structures would be performance-limited. This motivates the use of ion-implanted multilayers, which decouple optical absorption and electrical conduction, preserving interface quality. Therefore, this study concludes that integrating such in situ doped multilayers directly as the active region is impractical for high-performance detectors. This material-level understanding directly motivated the decoupled device strategy employed in this work: using an intrinsic Ge/GeSi multiple quantum well (MQW) for optimal optical absorption, while relegating the doping function to ion-implanted contacts, thereby circumventing the core limitations identified here. These growth-related effects, coupled with interface roughening in in situ doped structures, would inevitably limit device performance. Consequently, ion-implanted multilayers are used for device fabrication, ensuring high crystalline quality and well-confined doping.

3.4. Design for SWIR Photodetectors with Intrinsic Ge Nanoscale Multilayer Heterostructures

Observations of interface roughening and dopant diffusion in in situ doped multilayers highlighted the limitations of such structures for device performance. Accordingly, the ion-implanted design was chosen to preserve optical quality while controlling electrical properties. Building upon the material insights from Section 3.3, we fabricated a GOI-based SWIR PD with a decoupled architecture: an intrinsic Ge/GeSi nanoscale multilayer structure as the optical absorption region and ion-implanted regions for electrical junction formation. This approach directly addresses the limitations of in situ doped structures by eliminating doping-induced defects from the active zone and utilizing a mature, high-resolution technique for contact doping. First, an intrinsic low temperature Ge buffer layer and a high temperature Ge layer were sequentially grown on a p-type Si (001) donor wafer using a LT/HT two-step epitaxy method in a RPCVD reactor. Subsequently, a four-period GeSi/Ge multilayer was directly grown, with each period consisting of a 20 nm thick GeSi layer and a 90 nm thick Ge layer. The nanoscale multilayer structure design—specifically the use of a GeSi barrier—was optimized to replicate the favorable strain accommodation and interfacial stability observed in the best-performing n⁺-Ge/i-GeSi model structure from Section 3.3. It is important to note that the primary purpose of employing the Ge/GeSi nanoscale multilayer structure is to enhance the optical responsivity and quantum efficiency via strain engineering and carrier confinement. This design does not inherently improve, and may even influence, the dark current characteristics, which are predominantly governed by the material’s defect density and the junction quality. Following the growth of the intrinsic epitaxial multilayer, a 300 nm PECVD SiO₂ layer was deposited as a bonding interface. A separate Si handling wafer with a 500 nm thermal oxide SiO₂ layer was prepared. The donor wafer and the handling wafer were then directly bonded. The donor substrate and the low-temperature Ge buffer layer were subsequently removed by grinding and wet etching, transferring the intrinsic nanoscale multilayer structure onto the insulating substrate to form the GOI platform. The electrical junctions were then formed exclusively by ion implantation, completely avoiding n-type in situ doping in the device integration flow. The fabrication of the top n-type contact layer was achieved through a two-stage ion implantation process, with each step employing a dose of 1 × 10¹⁵ cm⁻² and an energy of 18 keV. A dedicated activation annealing cycle was performed immediately after each implantation step at 500 °C for 60 s in a hydrogen ambient. This process created a highly doped n⁺–Ge region with a concentration of approximately 2 × 10²⁰ cm⁻³. The main process flow is illustrated in Figure 7a. The final n-type dopant distribution throughout the PIN structure was confirmed by SIMS, as shown in Figure 7c. Direct structural or optical comparison between in situ doped and ion-implanted multilayers is not performed, because material imperfections in the in situ doped structures would dominate device response. Instead, ion-implanted multilayers are used to decouple optical absorption and electrical conduction, ensuring meaningful device characterization and high-performance evaluation. The two approaches are conceptually linked: limitations observed in in situ doped multilayers directly informed the adoption of ion-implanted contacts. This clear linkage between material characterization and device design ensures continuity from in situ doped studies to the implementation of ion-implanted SWIR photodetectors, providing a coherent and traceable narrative for the reader.

A critical determinant of PD performance is the quality of the ohmic contact, specifically the associated contact resistance [95,96,97,98]. In this investigation, we sought to minimize this resistance by fabricating a NiGe layer directly onto the N-doped and P-dope Ge region within the p-i-n structure. To decouple the contact resistance from the intrinsic device resistance and quantify it independently, we employed the circular transmission line model (CTLM). Where $L$ represents the inner circle radius, and $d_i$ represents the difference between the outer circle radius and the inner circle radius. The mask design incorporated $d_i$ values ranging from 5 to 30 μm in uniform increments of 5 μm. The derived metrics, specifically the total resistance ($R_T$) and the contact resistivity (${\rho }_c$), are determined using the following established formulas [99]:

$$R_T = {\displaystyle \frac{R_{sh}}{2\pi L}}(d_i+2L_T)$$

(4)

$${{\rho }_c=R}_{sh}·L_T^2$$

(5)

Specifically, the relationship for $R_T$ is defined as a function of the sheet resistance of the semiconductor ($R_{sh})$, the transfer length $\left(L_T\right)$, and the inner circle radius ($L$). The CTLM analysis confirmed excellent ohmic contact performance for both doping types. Specifically, the data fitting, as detailed in Figure 8, yielded an n-type contact resistivity (${\rho }_c$) of 2.66 × 10⁻⁶ Ω cm² and a p-type contact resistivity (${\rho }_c$) of 1.38 × 10⁻⁸ Ω cm². These values demonstrate the highly efficient contact formation achieved on both doped Ge surfaces.

The photo currents of the fabricated PDs with undoped Ge/GeSi multilayer were measured at room temperature to confirm the optoelectronic performance. The device fabrication and photocurrent measurements were not performed on in situ doped multilayers because material characterization revealed interface roughening and dopant diffusion, which would inevitably degrade device performance. Instead, an ion-implanted doping strategy was adopted to decouple optical absorption and electrical conduction, ensuring reliable device evaluation. Photocurrent measurements were conducted using an Agilent B1500A semiconductor parameter analyzer integrated with a manual probe station. To assess the spectral response, a fixed-wavelength laser (1310 nm) and a tunable laser source covering the 1500–1630 nm range were utilized. Optical excitation was achieved by coupling incident light through a single-mode fiber-oriented perpendicular to the device surface. The incident optical power was precisely regulated at 1 mW, with calibration verified via a high-precision commercial reference detector. The responsivity (R), a key figure of merit for detector efficiency, is defined by the following expression:

$$R ={\displaystyle \frac{I_{ph}}{P_o}}={\displaystyle \frac{\eta q\lambda }{hc}}$$

(6)

where $I_{ph}$ denotes the photocurrent, $P_o$ represents the incident optical power, $\eta$ is the external quantum efficiency (EQE), $q$ is the elementary charge, $\lambda$ is the incident wavelength, $h$ is Planck’s constant, and $c$ is the speed of light in a vacuum.

The responsivity characteristics for the PD configurations at wavelengths of 1550 nm and 1310 nm are illustrated in Figure 9a. The optical responsivity of the Ge/GeSi nanoscale multilayer structure PD at 1550 and 1310 nm was 0.99 A/W and 0.83 A/W at −1 V, corresponding to EQEs of 79.2% and 78.6%, respectively. The responsivity remained nearly constant across the reverse bias range, indicating high carrier collection efficiency throughout operation. These results demonstrate that the fabricated PDs maintain efficient photogenerated carrier collection at the nanoscale, validating both the nanoscale multilayer structure design and the ion implantation doping strategy, which collectively ensure optimal junction formation. The spectral response of the PD was characterized at room temperature using a tunable laser source, as shown in Figure 9b. Due to the optical confinement provided by the QW nanoscale structure on insulator, a resonant response peak was clearly observed in the C-band, with the highest responsivity successfully obtained at 1550 nm. The maintenance of high-performance optoelectronic characteristics is primarily attributed to the synergistic effect of low contact resistivity and high crystalline quality in the nanoscale absorption multilayer, which minimizes carrier recombination and maximizes quantum efficiency. Furthermore, these results validate the effectiveness of ion implantation in forming well-controlled n-type layers in nanoscale SWIR detectors, ensuring that device performance is not significantly compromised during junction engineering. These findings highlight the critical role of nanoscale material design, interface engineering, and precise dopant placement in achieving high-performance, low-bias PDs suitable for low-power SWIR imaging applications.

4. Conclusions

We systematically investigated the effects of n-type in situ doping on Ge epitaxial films and Ge/GeSi nanoscale multilayer structures on Si. PL measurements showed initial enhancement of Γ-valley emission with increasing doping, but high concentrations (>1 × 10¹⁹ cm⁻³) led to quenching, revealing the fundamental trade-off between doping and radiative efficiency. Among all the nanoscale multilayers, n⁺-GeSi/i-Ge exhibited the steepest dopant profile (>6 × 10¹⁹ cm⁻³) and suppressed interdiffusion, whereas n⁺-Ge/i-GeSi maintained sharp nanoscale interfaces but lower peak doping (~5 × 10¹⁸ cm⁻³). These findings unequivocally demonstrated that while in situ doping enables band engineering, it inherently introduces defects that limit optical performance. Driven by the recognition that in situ doping inherently compromises optical material quality, we devised a decoupled device strategy: the active region adopted an intrinsic Ge/GeSi nanoscale multilayer structure to preserve interface quality and minimize non-radiative loss, while the electrical junctions were formed by ion implantation to ensure precise doping control. Consequently, optimized Ge/GeSi SWIR PDs achieved responsivity of 0.99 A/W at 1550 nm and 0.83 A/W at 1310 nm, EQE of 79.2% and 78.6%, and low contact resistivity (2.66 × 10⁻⁶Ω ·cm² for n-type, 1.38 × 10⁻⁸ Ω·cm² for p-type). More broadly, the decoupled design strategy establishes a general paradigm for leveraging nanoscale heterostructures for optimal optical properties while employing conventional processing for electronic functions, which could be extended to other material systems for imaging applications.