Doping Effects in Direct Bandgap GeSn Light Emitter

Abstract

N-type doping is necessary for many applications in Ge and GeSn optoelectronics. In recent years the need for energy efficient electronic integrated circuits has led to the exploration of different approaches in order to bring light emitters on chips. Group-IV lasers are being considered, along with heterogeneous integration and flip-chip bonding. To overcome the lack of direct bandgap in Ge, options like injecting high amounts of tensile strain and/or n-type doping were investigated, aiming to achieve a sufficiently large electron population in the Γ-valley to support laser action. Investigations of highly n-doped Ge resulted in strong luminescence enhancement and in an optically pumped laser action. Alternatively, it has been shown that alloying Sn and Ge changes the electronic band structure so that the conduction band at Γ decreases below the L-valley, resulting in a transition from an indirect to a fundamental direct bandgap semiconductor. Optically pumped GeSn lasers were demonstrated, but their operation is still limited to temperatures of about 130 K. The naturally appearing question is whether n-doping can improve the laser performance of these direct bandgap materials, which still have a low directness (the separation between Γ and L valleys is typically 0-50 meV). Calculations indicate an increase of net gain up to a certain doping concentration, beyond which it saturates and then decreases. Here, we concentrate on the experimental study of n-type doping in direct bandgap Ge0.875Sn0.125 alloys. A series of samples with different doping concentrations, in the order of 1018 cm-3, were grown via reactive gas source epitaxy on Si(100) wafers buffered by a Ge virtual substrate. Digermane (Ge2H6), Tin tetrachloride (SnCl4) and phosphine (PH3) were used as precursors. The material is analyzed, regarding crystalline quality, strain and composition, by transmission electron microscopy (TEM), X-ray diffraction reciprocal space mapping (XRD-RSM) and Rutherford Backscattering Spectrometry (RBS). Photoluminescence under a chopped 532 nm laser excitation shows the influence of increased doping concentration on the material quality, regarding carrier recombination due to defects. The optical analysis of the bulk material also shows a very slight change in the bandgap, which can be attributed to bandgap narrowing caused by increasing doping concentration. The impact of the doping concentration is further studied in microdisk resonator lasers. The microdisks are defined by electron beam lithography, and mesas are etched using reactive ion etching (RIE) by Cl2/Ar plasma. To relax the strain of the GeSn layer, and to increase the optical mode confinement, the disks are under-etched by an isotropic CF4 plasma. Finally the structures are passivated with a 10 nm thick Al2O3layer deposited via atomic layer deposition (ALD). The advantage of using microdisk structures is that the deep under-etching of the GeSn disks results in the full strain relaxation of the active layer, eliminating the adverse effect of compressive strain on the band structure. Photoluminescence/laser studies were performed using an excitation with a pulsed Nd:YAG laser at 1064 nm wavelength, and a pulse width of 5 ns. Power and temperature dependence of the laser emission were studied. Figure 1 shows normalized photoluminescence from bulk GeSn and from an under-etched microdisk with a diameter of 8 µm made of n-doped GeSn. The lasing spectrum shows the contribution of multiple modes inside the disk. Undercutting the disks relaxes the strain, resulting in a decreased bandgap leading to a redshift of the laser peak. The photoluminescence peak energies depend on various effects: the strain in the material, the band filling due to high excitation densities and doping, and finally the doping-induced bandgap narrowing, while the laser emission peak position is also influenced by the free-carrier contribution to the refractive index. All these effects will be discussed, based on the systematic analysis of the above described material and laser structures. Additionally, calculations are used to investigate the influence of doping on optical gain. With L-valleys sitting only slightly above, and having a much larger density of states than the Γ valley, the majority of the photo-generated electrons will populate the L valleys if no n-type doping is employed. Only a fraction of electrons will therefore contribute to gain. This can be improved by n-doping, (e.g. by filling the L valleys), as already used in Ge lasers (mandatory due to the indirect bandgap, but also useful in cases of moderate directness). However, additional electrons also increase the free-carrier absorption, and the optimal doping will be a tradeoff between the two effects. Figure 1