Enhancing Ge0.86 Sn0.14 Photodiode Detectivity through Layer Transfer

Introduction

Germanium-tin (GeSn) alloys have gained significant attention in recent years due to their potential for fabricating low-cost mid-wave infrared (MWIR) photodetectors on silicon (Si) substrates. By incorporating tin (Sn) into the germanium (Ge) lattice, the conduction and valence bands of the material shift towards lower energies, resulting in a direct bandgap for Sn contents above 6.4% in strain-relaxed layers. This direct bandgap property enables stronger absorption of infrared photons compared to the indirect bandgap of pure Ge, making GeSn an attractive material for photodetector applications.

However, the detectivity of a photodetector, which is a measure of its sensitivity, depends on both its responsivity (the ability to generate photocurrent) and noise characteristics. GeSn layers are typically grown on Ge virtual substrates, such as thick Ge strain-relaxed buffers (SRBs) on Si substrates. The large lattice parameter difference between GeSn and Ge can lead to the formation of dislocations that propagate through the active layer, degrading the noise characteristics of the photodiode.

Furthermore, the sub-micron thickness of GeSn layers used in devices may not be sufficient to absorb all incoming infrared photons, even with the enhanced absorption due to the direct bandgap.

This article presents a novel approach to improve the optical absorption in the active layers of GeSn photodiodes without degrading their noise characteristics.

The Layer Transfer Approach

The key idea is to transfer the grown GeSn photodiode stack upside down onto a gold layer, remove the initial Si substrate and the underlying Ge SRB used for epitaxy, and then process the top contacts of the photodiode. This approach aims to enhance the optical absorption by creating a Fabry-Perot cavity between the gold layer and the GeSn/air interface, while also eliminating the dislocated layers that could contribute to noise.

Material Growth and Device Fabrication

The process begins with the growth of a p-i-n type stack consisting of a 620 nm thick non-intentionally-doped Ge0.86 Sn0.14 active layer sandwiched between an n-type GeSn layer and a p-type Ge carrier-collecting layer. This stack is grown on a 200 mm Si(001) substrate using Reduced Pressure Chemical Vapor Deposition (RPCVD).

Figure 1: (a) Schematic cross-section of the heterostructure grown on a 200 mm Si wafer, the transferred photodiode (sample A), and the reference photodiodes (sample B). (b) Scanning Electron Microscopy image of Ge0.86Sn0.14 photodiodes with layer transfer (sample A).

After the growth, the GeSn-carrying coupons are transferred onto a gold-covered Si wafer by thermocompression bonding. The initial Si substrate and the bottom part of the Ge SRB are then removed by grinding and dry etching, exposing the surface of the GeSn active layer. The transferred layers are processed into 200 μm mesa photodiodes ("Sample A"), while reference photodiodes ("Sample B") are fabricated from the as-grown, non-transferred stack.

Photodiode Characterizations

The relative spectral responses of the photodiodes are obtained using a Fourier Transform Infrared (FTIR) spectrometer, and the absolute responsivities are measured under a 1.55 μm laser beam illumination. Noise measurements are conducted at 0 V bias using a transimpedance amplifier and an electrical spectrum analyzer.

Results and Discussion

The transferred photodiodes (Sample A) showed a significant responsivity improvement compared to the reference photodiodes (Sample B). As seen in Figure 2, the underlying gold layer, together with the air-GeSn interface, created an efficient Fabry-Perot cavity, resulting in strong modulation in the responsivity spectrum.

Figure 2: Responsivity of transferred photodiodes A (solid line) and reference photodiodes B (dashed line) at 77 K and 300 K.

Although the absorbing layer thickness was reduced by 35%, transferring the stack onto a gold layer multiplied the responsivity at 2.5 μm by 2.9 at 300 K and 4.8 at 77 K.

Electrical characterizations revealed that the two types of photodiodes had similar dark current values and noise levels, as shown in Figure 3. The removal of the underlying dislocated layers in the transferred sample (Sample A) did not reduce the noise level, as only defects located in the depletion region contribute to noise.

Figure 3: The noise spectrum of transferred photodiodes (solid curves) and reference photodiodes (dashed curves) at 77 K and 300 K are almost identical.

With the unchanged noise level and significantly improved responsivity, the detectivity of the transferred photodiodes (Sample A) was increased by a factor of 3 at 300 K and 5.4 at 77 K at 2.5 μm, as summarized in Table 1.

Table 1: Characteristics of the two samples at 77 K and 300 K.

Sample	T [K]	Average noise [A.Hz⁻¹/²]	Responsivity at 2.5 µm [A.W⁻¹]	Specific detectivity at 2.5 µm [cm.Hz¹/².W⁻¹]
Transferred Photodiode A	77	4.91 x 10⁻¹²	0.53	2.56x10⁹
300	2.15 x 10⁻¹¹	0.57	6.20x10⁸
Reference Photodiode B	77	5.52 x 10⁻¹²	0.11	4.71x10⁸
300	2.30 x 10⁻¹¹	0.2	2.07x10⁸

Conclusion and Perspectives

The layer transfer approach demonstrated a significant increase in the detectivity of GeSn photodiodes by improving the responsivity through the creation of a Fabry-Perot cavity, while maintaining the noise level unaffected by the transfer process. The detectivity at 2.5 μm was improved by a factor of 3 at room temperature and 5.4 at 77K.

This technique not only enhances the performance of GeSn photodiodes but also opens up new avenues for exploring the advantages of layer transfer for other optoelectronic devices, such as GeSn light-emitting diodes (LEDs).

Reference

[1] C. Cardoux, E. Kroemer, L. Casiez, M. Frauenrath, N. Pauc, V. Calvo, S. Assali, J. M. Hartmann, N. Coudurier, P. Rodriguez, J. Chrétien, J. Widiez, O. Gravrand, A. Chelnokov, V. Reboud, "Enhancing the Detectivity of Direct Bandgap Ge0.86Sn0.14 Photodiodes by Layer Transfer," Univ. Grenoble Alpes, CEA, LETI, and CEA, Grenoble INP, IRIG, PHELIQS, Grenoble, France, 2024.