Chalcogen-hyperdoped germanium for short-wavelength infrared photodetection Light-emitting diodes with AlN polarization-induced buried tunnel junctions: A second look

Obtaining short-wavelength-infrared (SWIR; 1.4 μ m–3.0 μ m) room-temperature photodetection in a low-cost, group IV semiconductor is desirable for numerous applications. We demonstrate a non-equilibrium method for hyperdoping germanium with selenium or tellurium for dopant-mediated SWIR photodetection. By ion-implanting Se or Te into Ge wafers and restoring crystallinity with pulsed laser melting induced rapid solidification, we obtain single crystalline materials with peak Se and Te concentrations of 10 20 cm − 3 (10 4 times the solubility limits). These hyperdoped materials exhibit sub-bandgap absorption of light up to wavelengths of at least 3.0 μ m, with their sub-bandgap optical absorption coefficients comparable to those of commercial SWIR photodetection materials. Although previous studies of Ge-based photodetectors have reported a sub-bandgap optoelectronic response only at low temperature, we report room-temperature sub-bandgap SWIR photodetection at wavelengths as long as 3.0 μ m from rudimentary hyperdoped Ge:Se and Ge:Te photodetectors. we report dopant-mediated sub-bandgap photodetection in Ge doped with supersaturated concentrations of Se or Te. We fabricate these materials using a scalable, non-equilibrium hyperdoping method, consisting of ion implantation of Se or Te into Ge followed by nanosecond pulsed laser melting (PLM) induced rapid solidifi-cation. This hyperdoping process produces single-crystal materials with peak Se and Te dopant concentrations of 10 20 cm − 3 , which are four orders of magnitude above the Ge:Se and Ge:Te solubility limits. The hyperdoping increases the dopant-mediated sub-bandgap absorption coefficients α of these materials to values comparable to those of commercial SWIR III-V and II-VI materials. We test rudimentary photodetectors made from hyperdoped Ge:Se and Ge:Te for sub-bandgap optoelectronic response and demonstrate that these materials can be used for room-temperature sub-bandgap photodetection.Whilerecent sub-bandgap photodetection studies have focused on Si, we work with Ge because of its higher carrier mobility. Previous studies of dopant-mediated Ge photodetectors incorporating different dopants (S, Te, Zn, B, Cu, Cd, Zn, and Au) through various doping methods have reported sub-bandgap response only at low temperature, which is impractical for many SWIR applications.


I. INTRODUCTION
Short-wavelength-infrared (SWIR) (1 μm-3 μm) photodetection is used for numerous commercial, military, and scientific applications. [1][2][3][4] Conventional semiconductors such as silicon and germanium, however, do not absorb and detect SWIR photons, which have an energy below that of the intrinsic bandgap of these materials. Current state-of-the-art SWIR photodetectors are made from narrow bandgap III-V or II-VI semiconductors (e.g., InGaAs, InAs, Pb 1−x Sex, and Hg 1−x CdxTe) that are heterogeneously integrated with Si-CMOS electronics. 1,4,5 Three main issues constrain the applicability of photodetectors made in this manner. First, III-V and II-VI semiconductors are chemically incompatible with Si-CMOS processing, limiting the size of final photodetector focal plane arrays, and are often expensive or toxic. 1,5,6 Second, their heterogeneous integration is a complex, low-throughput process that further increases final device cost. Third, photodetectors made from the majority of these materials must be cooled to low temperature to attain satisfactory signal-to-noise ratios. The lack of a low-cost, nontoxic material capable of room-temperature SWIR photodetection limits applications involving SWIR photodetection.
An alternative to using III-V or II-VI materials is to induce a low-cost, Si-compatible material such as germanium to detect SWIR-light through dopant-mediated photoconductivity. In this ARTICLE scitation.org/journal/adv paper, we report dopant-mediated sub-bandgap photodetection in Ge doped with supersaturated concentrations of Se or Te. We fabricate these materials using a scalable, non-equilibrium hyperdoping method, consisting of ion implantation of Se or Te into Ge followed by nanosecond pulsed laser melting (PLM) induced rapid solidification. This hyperdoping process produces single-crystal materials with peak Se and Te dopant concentrations of 10 20 cm −3 , which are four orders of magnitude above the Ge:Se and Ge:Te solubility limits. The hyperdoping increases the dopant-mediated subbandgap absorption coefficients α of these materials to values comparable to those of commercial SWIR III-V and II-VI materials. We test rudimentary photodetectors made from hyperdoped Ge:Se and Ge:Te for sub-bandgap optoelectronic response and demonstrate that these materials can be used for room-temperature sub-bandgap photodetection.
While recent sub-bandgap photodetection studies have focused on Si, we work with Ge because of its higher carrier mobility. [7][8][9][10][11][12] Previous studies of dopant-mediated Ge photodetectors incorporating different dopants (S, Te, Zn, B, Cu, Cd, Zn, and Au) through various doping methods have reported sub-bandgap response only at low temperature, which is impractical for many SWIR applications. [13][14][15][16][17] In this paper, we chose Se and Te as dopants because they are deeplevel-dopants 18 (supplementary material Fig. 1), and among deeplevel Ge-dopants, they have relatively high equilibrium solubility limits. 19,20 Compared to Ge with shallow-level dopants, Ge doped with deep-level-dopants demonstrates lower thermal ionization at room-temperature, reducing background free-carrier concentrations and improving device signal-to-noise ratios. 13,14 A higher equilibrium solubility limit permits attaining even higher concentrations through hyperdoping, 21 leading to a higher sub-bandgap absorption coefficient. 13,14

II. EXPERIMENTAL FABRICATION AND CHARACTERIZATION
To fabricate chalcogen-hyperdoped Ge, we first implant ptype (100) Ge wafers of 5 Ω cm-10 Ω cm resistivity with either 60-keV 80 Se + ions or 80-keV 130 Te + ions to a dose of either 10 14 cm −2 or 10 15 cm −2 . The projected range for both doses is 27 nm. We performed all implantations at liquid-nitrogen temperature to suppress ion-beam-induced porosity and dynamic annealing. During implantation, all substrates were oriented 7 ○ with respect to the [100] crystal axis to minimize ion channeling.
To characterize the resulting amorphous layer, we performed channeling Rutherford backscattering spectrometry (RBS) on the samples with a 2-MeV He + ion beam aligned along the [100] axis. We used the SIMNRA program to simulate the RBS spectra (supplementary material Figs. 3 and 4). After implantation, the 10 14 -cm −2 and 10 15 -cm −2 dose samples have an amorphous-layer thickness of 60 nm and 75 nm for Se samples and 60 nm and 76 nm for the Te samples, respectively, as measured by RBS. To restore crystallinity after implantation damage, we laser-melted each sample with a 355nm, 0.49-J/cm 2 , single 4-ns FWHM pulse from a Nd:YAG laser [supplementary material Fig. 2(a)].
We chose this fluence to reach a melt depth of 210 nm (predicted by numerical solutions to the heat equation), which is expected to be safely beyond the deep tail of the implant distribution and beyond the implant-amorphized and implant-damaged regions of each sample [supplementary material Fig. 2(d)]. 22 Under this scenario, the melt front reaches the underlying perfect crystalline substrate, permitting defect-free single-crystal epitaxial regrowth. During PLM, time-resolved reflectivity measurements indicate the following melt durations: 33 ns-54 ns for the 10 14 -cm −2 samples and 39 ns-66 ns for the 10 15 -cm −2 samples. These measured meltdurations are consistent with numerical simulations [supplementary material Fig. 2 To measure Se and Te concentration-depth profiles, we carried out secondary ion mass spectrometry using a 5.5-keV Cs + ion beam and collected 133 Cs 80 Se + and 133 Cs 130 Te + secondary ions, respectively. To examine the impact of PLM on sub-surface implantation damage, we carried out bright-field and high-resolution crosssectional transmission electron microscopy (XTEM) at 200 keV. We prepared the samples for XTEM using a focused ion beam and in situ lift-out. To quantify absorptance, we measured transmittance (T) and reflectance (R) using a UV-VIS-NIR spectrophotometer and a Fourier-transform infrared spectrometer (FTIR). To minimize gasabsorption lines in absorptance data, we purged the FTIR chamber with liquid nitrogen before all measurements.
To test whether chalcogen-hyperdoped Ge exhibits roomtemperature sub-bandgap optoelectronic response, we fabricated rudimentary photodetectors using 10 15 -cm −2 dose hyperdoped samples. We created a bottom contact to the p-type wafer substrate by thermally evaporating a 200-nm thick Al-layer. On the wafer side opposite the Al back-contact, we formed an n++ Ge:Se or Ge:Te layer via the hyperdoping process detailed above. Using photolithography and SF 6 reactive ion etching, we then formed a mesa structure out of the n++ layer to electrically isolate the region from the rest of the substrate. The mesas were 4 μm deep and 1 × 1 mm 2 in area. We then used photolithography, e-beam evaporation, and lift-off to form 200-nm thick Ni-bar contacts on opposite sides of the mesa structure (separated 1000 μm apart) to create the top photodetector device contacts. All contacts were Ohmic. We attached the bottom contact to a printed circuit board (PCB) with silver paste and wirebonded the top contacts to the PCB. All photodetectors rectified when current was passed between the top and bottom contacts.
We tested the photodetectors using a 2.0-μm laser photoconductivity setup. We illuminated the photodetectors with chopped light from a continuous-wave laser diode (Brolis semiconductor) at a wavelength of 2.0 μm. The laser light was mechanically chopped at 23 Hz and focused on a 20-μm spot size in the middle of the mesastructure surface. Using a lock-in amplifier, we measured the AC current generated between the photodetector's two top-bar-contacts and the bottom-contact.
To determine the spectral photoresponse of the photodetectors, we substituted each detector in an FTIR spectrometer with a photodetector. In this FTIR photoconductivity setup, we connected a photodetector to a trans-impedance amplifier (TIA) for amplification of the photocurrent. No bias voltage was applied to the photodetector during measurement. The FTIR resolution was set to 100 cm −1 . To reduce noise, we set the band pass filter of the TIA to the 0.1 kHz to 10 kHz range. We obtained the photoconductive spectrum in arbitrary units since the incident power from the Globar was unknown. The measured photoresponse of each detector was normalized to the emission spectrum of the Globar light source. When measuring dark noise, we blocked the illumination from the FTIR ARTICLE scitation.org/journal/adv spectrometer just before the objective lens to ensure that no light reached the photodetector.

III. RESULTS AND DISCUSSION
A. Chalcogen dopant-profile Figures 1(a) and 1(b) show the Ge:Se and Ge:Te concentration depth profiles, respectively, of the low (10 14 cm −2 ) and high (10 15 cm −2 ) dose samples before and after laser melting, obtained by secondary ion mass spectrometry. The as-implanted concentration depth profiles each have an expected Gaussian-like shape. In the laser-melted hyperdoped samples, the Se and Te concentrationdepth profiles are still Gaussian-like but have undergone some impurity redistribution compared to the as-implanted profiles. This broadening in concentration-depth profiles after laser melting is characteristic of liquid-phase impurity diffusion in the melt. We interpret the surface spikes in the first 10 nm of all the profiles as SIMS surface-transient measurement artifacts. 23 The noisy concentration signals below 5 × 10 17 cm −3 in the Se profile and below 1 × 10 17 cm −3 for Te profile reflect the SIMS sensitivity limit. All four laser-melted samples shown in Figs. 1(a) and 1(b) have supersaturated dopant concentrations. The peak Se concentration in the high-dose laser-melted hyperdoped sample, 1 × 10 20 cm −3 , is four orders of magnitude larger than the Ge:Se solubility limit 19 of 5 × 10 15 cm −3 . Likewise, the peak Te concentration in the high-dose laser-melted hyperdoped sample, 1 × 10 20 cm −3 , is four orders of magnitude larger than the Ge:Te solubility limit 19 of 2 × 10 15 cm −3 . We confirmed that Te was substitutional in the Ge lattice following PLM using RBS and channeling. We can estimate the retained impurity-doses after PLM by comparing integrations of the SIMS profiles before and after PLM. For these estimations, we integrate the dopant profile of each sample from a depth of 10 nm to the depth at which each profile reaches the respective SIMS sensitivity limit. We begin integrating from 10 nm because of artifacts in the SIMS measurement near the sample surface. These integrations indicate that in the Se samples, after PLM, 67 ± 5% of the as-implanted low-dose and 84 ± 5% of the as-implanted high-dose are retained. For the Te samples, 71 ± 5% of the as-implanted low-dose and 90 ± 5% of the as-implanted high-dose are retained after PLM.

Figures 2(a) and 2(b) show XTEM and HRXTEM (inset)
micrographs of the high-dose hyperdoped Ge:Se and Ge:Te samples, respectively. The Se and Te micrographs appear very similar and exhibit the same essential features. The micrographs show that the bulk of each sample is single-crystalline. No extended defects, secondary phases, or cellular breakdown features are visible in either micrograph. The hyperdoped layer thicknesses, evaluated by SIMS, of the laser melted Se and Te samples are 105 nm and 115 nm, respectively. For each sample, the hyperdoped layer appears identical to the substrate beneath. The micrographs demonstrate that pulsed laser melting fully restores the crystallinity of the implantation-damaged region.
Three other features of the micrographs are worth noting. First, both samples exhibit an amorphous looking surface layer that is characteristic of a surface oxide layer obtained after PLM. Second, the interface between the amorphous surface layer and the bulk regions contains small, diffuse dark regions. We suspect these regions are bend contrast that arises due to localized strain produced by localized dopant-concentration peaks at the surface. We note that the high-resolution imaging and selected area diffraction yield no evidence of crystallites which might otherwise be responsible for such features. Third, spots are visible throughout each micrograph that are characteristic of focused-ion-beam-induced damage in Ge. 24,25 Because these spots are found throughout the specimen (well beyond the hyperdoped layers), we conclude that they do not arise from the hyperdoping process.  Fig. 3(b), we show the αd product of the samples with sub-bandgap absorptance. By modeling each sample as a two-layer thin-film absorbing stack, 7 we estimate the product of the absorbing layer thickness, d, and the absorption coefficient, α, from the data in Fig. 3(a). In this estimate, we assume hyperdoped samples have a reflection coefficient at 2.4 μm equivalent to that of a virgin-Ge-air interface (0.366 at 2.4 μm). Further assuming that the hyperdoped layer in each sample uniformly absorbs sub-bandgap light, we obtain the following α values at a wavelength of 2.4 μm: α = 1300 cm −1 for a 10 19 cm −3 effective concentration (for low-dose Te, taking the layer thickness to be 100 nm), α = 5000 cm −1 for a 10 20 cm −3 effective concentration (for high-dose and a Se thickness of 100 nm), and α = 3700 cm −1 for a 8.3 × 10 19 cm −3 effective concentration (for high-dose Te and a thickness of 120 nm). These sub-bandgap α values are comparable to those of commercially available direct-bandgap semiconductors used for SWIR photodetection at the same wavelength. 26 Estimated wavelength-dependent αd products of each sample, calculated from estimates of α and d described above, are presented in Fig. 3(b).

D. Sub-bandgap optoelectronic response
Figure 4(a) shows the setup for measuring the response of the photodetectors made from chalcogen-hyperdoped Ge under 2.0-μm laser light. In the planar-device structure shown in the figure, the chalcogen-hyperdoped layer sits atop a raised mesa structure. We chose this planar-device structure for its simplicity and potential suitability for SWIR-imaging-array applications. The hyperdopedlayer atop the mesa is isolated from the substrate to reduce leakage currents and to ensure that any applied voltage passes through the device's p-n++ junction. In each photodetector, the hyperdoped layer forms a rectifying junction with the p-type substrate, as shown in the dark I-V curve in Fig. 4(b). Figure 4(c) shows the difference between the I-V curves obtained under 2.0-μm laser illumination and no illumination. This photocurrent difference demonstrates sub-bandgap photocurrent. Increasing reverse bias across the junction increases (and then saturates) sub-bandgap photocurrent due to enhanced collection of charge carriers excited by the subbandgap light. Under zero bias voltage and 2.0-μm illumination at room temperature, the Ge:Se and Ge:Te photodetectors exhibit an external quantum efficiency (EQE) of 1.0 × 10 −5 and 6.0 × 10 −5 , respectively. The low EQE of these initial rudimentary devices can be increased straightforwardly by optimizing hyperdoping fabrication to enhance sub-bandgap absorptance (i.e., the αd product) and by increasing photon and carrier collection through improved device design and fabrication. Figure 4(d) shows the broadband response of the chalcogenhyperdoped photodetectors, obtained by substituting each detector in the FTIR spectrometer with a hyperdoped photodetector. The response of each detector is normalized to the intensity spectrum of the FTIR Globar. Because we do not know the incident power per unit wavelength of the FTIR Globar illuminating the photodetectors and because the response is obtained from a Fourier transform of the phase-corrected time-domain interferogram, we report the normalized photoconductive response in arbitrary units. Between 2.2 μm and 3.0 μm, the response of the hyperdoped Ge:Se is consistently larger than that of the Ge:Te photodetector. We note that the 10 15 cm −2 -dose hyperdoped Ge:Se sample also shows higher sub-bandgap absorptance between 2.2 μm and 3.0 μm than the Ge:Te sample (Fig. 2). The dark noise for the Ge:Se and Ge:Te photodetectors is 1.0 × 10 −5 and 2.0 × 10 −5 , respectively. The Ge:Se and Ge:Te photodetectors both exhibit a room-temperature, subbandgap response out to wavelengths beyond the detection edge of extended InGaAs, which has a bandgap of 2.6 μm.

IV. CONCLUSION
Chalcogen-hyperdoped Ge exhibits room-temperature SWIR photodetection. Ion implantation, followed by pulsed laser melting induced rapid solidification, produces single crystal materials with peak Se and Te concentrations of 10 20 cm −3 (10 4 times the Ge:Se and Ge:Te solubility limits). We estimate that 10 15 -cm −2 dose hyperdoped Ge:Se and Ge:Te samples have an average subbandgap absorption coefficient at λ = 2.4 μm of 5000 cm −1 and 3700 cm −1 , respectively, which is comparable to those of commercial materials at the same wavelength. Rudimentary photodetectors made of hyperdoped Ge:Se exhibit a room-temperature optoelectronic response between 2.0 μm and 3.0 μm, which is consistently higher than that of hyperdoped Ge:Te photodetectors between 2.2 μm and 3.0 μm. These results suggest that chalcogen-hyperdoped Ge has the potential to be used for low-cost, room-temperature, silicon-compatible SWIR-photodetection.

SUPPLEMENTARY MATERIAL
Additional figures are presented in the supplementary material: Ge:Se and Ge:Te energy-defect levels, laser-melting setup and simulations, Ge:Se RBS spectra, and Ge:Te RBS spectra.