Silicon-based Intermediate-band Infrared Photodetector realized by Te Hyperdoping

Si-based photodetectors satisfy the criteria of low-cost and environmental-friendly, and can enable the development of on-chip complementary metal-oxide-semiconductor (CMOS)-compatible photonic systems. However, extending their room-temperature photoresponse into the mid-wavelength infrared (MWIR) regime remains challenging due to the intrinsic bandgap of Si. Here, we report on a comprehensive study of a room-temperature MWIR photodetector based on Si hyperdoped with Te. The demonstrated MWIR p-n photodiode exhibits a spectral photoresponse up to 5 {\mu}m and a slightly lower detector performance than the commercial devices in the wavelength range of 1.0-1.9 {\mu}m. We also investigate the correlation between the background noise and the sensitivity of the Te-hyperdoped Si photodiode, where the maximum room-temperature specific detectivity is found to be 3.2 x 10^12 cmHz^{1/2}W^{-1} and 9.2 x 10^8 cmHz^{1/2}W^{-1} at 1 {\mu}m and 1.55 {\mu}m, respectively. This work contributes to pave the way towards establishing a Si-based broadband infrared photonic system operating at room temperature.


Introduction
Broadband infrared photodetectors have been attracting the interest of many researchers due to their wide variety of applications, such as telecommunication, security equipment, environmental sensing, and biomedicine. [1][2][3][4][5] Nowadays, commercially available photodetectors are mostly fabricated with mercury cadmium telluride (MCT, HgCdTe), PbS and III-V quantum-well/dot (QWIP/QWID). [6][7][8][9][10][11][12][13] These photodetectors exhibit high device performance in infrared light detection, but suffer from some crucial drawbacks, such as high cost, problematic environmental impact, operation at cryogenic temperatures and especially incompatibility with the complementary metal-oxide-semiconductor (CMOS) fabrication routes. Alternatively, Si-based photodetectors overcome these disadvantages, [8,14] but their infrared photoresponse is fundamentally restricted to the near infrared (NIR) spectral range due to the 1.12-eV indirect bandgap of Si (λ = 1.1 μm). Therefore, the development of a roomtemperature broadband infrared Si-based photodetector is of great interest in the realm of all-Si photonic systems.
One of the most promising approaches to further extend the room-temperature optical response of Si to the short-and mid-wavelength infrared (SWIR, MWIR) range consists of introducing deep-level dopants (e.g. transition metals and chalcogen dopants) into Si at concentrations in excess of the solid solubility limit. [15][16][17][18][19][20][21][22][23][24][25] This process leads to the broadening of the deep-level states into an intermediate band (IB) with finite width that allows for the strong optical absorption of photons with an energy lower than that of the Si bandgap. [26] Moreover, utilizing deep-level impurities provides a path to obtain extrinsic Si-based photodetectors for room-temperature operation, which is not possible with shallow-level impurities. [15,18,19,[21][22][23] However, deep-level impurities such as transition metals have very high diffusion velocities in Si. [27] Therefore, they will diffuse to the surface and form the so-called cellular breakdown, preventing the dopant incorporation and eventually the fabrication of CMOS-compatible devices. [28] Si hyperdoped with chalcogen dopants does not exhibit celluar breakdown and has also shown potential applications for infrared photodetectors. [15,22] Unfortunately, due to the lack of a meticulous design of the device architecture and of the doping homogeneity, the achieved S-or Se-hyperdoped Si photodiodes show quite low sub-bandgap external quantum efficiency (EQE) and only operate in the SWIR wavelength range. Moreover, S-and Sehyperdoped Si lose their ability to absorb infrared light even after short-duration thermal treatments at low temperatures. [29,30] This will compromise the application of hyperdoped Si for integration with the existing CMOS-compatible processes involving temperature-dependent steps. Unlike S and Se, Te impurities have much lower diffusivity. [12] Te-hyperdoped Si shows stable IR-absorption upon thermal processing up to 400 °C with a duration of 10 min. [29,32] Therefore, Si hyperdoped with Te holds promise for fabricating photodetectors with broad spectral response and enhanced detectivity as well as their integration into manufacturing processes.
In this work, we report on a room-temperature MWIR Si p-n photodiode working in photovoltaic mode based on Si hyperdoped with Te. The hyperdoped Si layers are homogeneous, free of cellular breakdown and surface flat. The materials and device fabrication are fully CMOS-technology compatible. The fabricated photodiode exhibits an enhanced performance figure of merit, e.g. spectral photoresponse, specific detectivity, bandwidth and response speed.
These results point out the potential of Te-hyperdoped intermediate-band Si photodetectors for room-temperature high-performance MWIR detection as the new generation of Si-based photonic systems.

Material characterization
The process flow of fabricating the Te-hyperdoped Si photodetector is shown in Figure 1(a)-(b). Detailed microstructure investigations of the Te-hyperdoped Si layer were carried out by Rutherford backscattering spectrometry (RBS) and transmission electron microscopy (TEM). [32][33][34] Single-crystalline regrowth of the Te-hyperdoped Si with a flat sample surface is achieved by the pulsed laser melting (PLM) treatment (as shown in Figure 1(c)-(f)). Moreover, Te is found to be uniformly distributed within the top 120 nm of the Si wafer without the formation of extended defects, secondary phases, or cellular breakdown. Figure 1(g) shows the absorption spectrum for the PLM-treated Te-hyperdoped Si sample. A virgin Si sample is also measured as a reference. More detailed information about the calculation of the absorption coefficient (α) can be found in the supporting information (SI). As described in our previous work, [34] the Te-hyperdoped Si layer exhibits strong broadband sub-bandgap optical absorption in the MIR region, as compared with a bare Si sample. This is consistent with the previously reported results about S-and Se-hyperdoped Si. [22,35] In particular, a well-defined broad absorptance band peaking at around 0.36 eV is observed, which is attributed to the presence of an intermediate band.
To get insight into the optical capture cross-section σ and the probability distribution of the binding energy ETe of Te-induced deep-level states, the absorption spectrum is fitted by where NTe is the concentration of Te states and αe is the absorption coefficient for the subbandgap absorption by assuming that electrons are promoted from Te deep-level states to the conduction band (CB). [36] This assumption is reasonable since the optical transition from the IB to the CB is more intense than that of the valence band (VB) to the IB, as shown by the density functional theory calculations. [37,38] σe is the electron-related optical cross-section, which can be obtained as follows in case of deep-level impurities in Si [39] describes the single discrete deep-level state excitation (ℏ is the photon energy). [39] This this term convolved with a Gaussian distribution of the deep-level states energies, with a mean binding energy ETe and a standard deviation Eσ, is used to describe the broadening of the deep-level states into an IB with finite width in Te-hyperdoped Si. [36,37] The spectral fit to αe is shown in Figure 1 [29,38,40] The energy probability distribution of deep-level states determined by the spectral fit is plotted in the inset of Figure 1(g) with the x-axis being the energetic distance to the CB. The energy probability distribution of the Te deep-levels suggests that the IB is not merged with the CB.

Device characterization: Electrical Properties
For the temperature-dependent electrical measurements, the Te-hyperdoped Si p-n intermediate-band Si p-n photodiode is placed inside a helium closed-cycle Janis cryostat with a ZnSe window. by fitting the dark I-V curve using the single-diode equation as follows: [41] = 0 [ where 0 is the saturation current, is the electron charge, is the ideality factor, kB is the Boltzmann constant and is the temperature; whereas and ℎ are the series and parallel resistances, respectively. A room-temperature ideality factor of 2.1 was found by fitting the experimental data. This rectifying behavior is directly related with the n-type character of the Te-hyperdoped Si layer, [28] forming a p-n junction between the p-type Si substrate and the ntype Te-hyperdoped layer. At -50 mV reverse bias voltage, the room-temperature dark current density of 0.2 mA/cm 2 is lower than that of the reported Au-hyperdoped Si and Se-hyperdoped Si photodetectors. [18,22] Under white-light illumination, an abrupt increase of the reverse current by more than one order of magnitude is observed, showing the operational principle of a photodetector. An open circuit voltage of about 0.2 V was deduced. . For the ideality factor, we observe two different behaviors depending on the temperature range of operation. For the high-temperature range (T>160 K), the ideality factor shows an almost temperature-independent value between 2.0 and 2.2, suggesting that the generation/recombination processes in the depletion region could be the dominant transport mechanism. [22] On the other hand, for the low-temperature range (T<140 K), the ideality factor is very sensitive to temperature, increasing from 2.2 to 6.7 as the temperature decreases. This temperature dependence of the ideality factor and its value higher than 2 at temperatures below 140 K are indications of a change in the conduction mechanism, i.e. a temperature-independent tunnelling conduction mechanism [42] . Figure 2(d) represents the temperature dependence of the saturation current. We observe that the saturation current also shows two different behaviors. Both follow the Arrhenius law of the form: where is a constant and is the activation energy of the transport mechanism. It is worth noting that both exponential behaviors have been well-fitted in a wide range of current. From fitting to Equation 4 in the high-temperature range (T>160 K), an activation energy of 0.51 eV was deduced. This value is close to about half the silicon band gap (0.56 eV). This supports the previous interpretation of the transport mechanism that is driven by a recombination process in the depletion region in this temperature range. [43] At low temperature (T<160 K), a smaller activation energy of 43 meV was found. Such a small activation energy together with the temperature behavior of the ideality factor support the existence of a tunnelling current [44] . We believe that the origin of this tunnelling current could be related with a multi-tunnelling captureemission process through the Te-localized states in the junction region, although further work is in progress to elucidate it exactly. Figure 3 shows the temperature-dependent spectral responsivity and the band diagram of the Te-hyperdoped Si p-n photodiode. The infrared responsivity is estimated at zero bias (i.e. the photovoltaic mode) to prove the pure photovoltaic effect of the photodetector. The Tehyperdoped Si p-n photodiode shows a strong sub-bandgap responsivity up to 5 µm, whereas the responsivity of a commercial Si photodetector reaches the noise floor as expected at wavelengths longer than 1.2 µm. As shown in Figure 3(a), the Te-hyperdoped Si p-n photodiode exhibits a room-temperature below-bandgap responsivity of 79 mAW -1 at 1.12 µm. Moreover, at the 1.55 µm-telecommunication wavelength, a room-temperature responsivity of around 0.3 mAW -1 was obtained, which is comparable to that reported for hyperdoped Si-based photodiodes and solar cells at this wavelength. [45,46] The room-temperature EQE at 1.55 µm was found to be 6×10 -4 , which is comparable to other deep level impurity-hyperdoped Si p-n photodiodes. [18,22] The responsivity of the Te-hyperdoped Si photodetector at different temperatures is displayed in Figure 3(a)-(g). Different behaviors in terms of line shape and responsivity can be identified for three temperature ranges. In the region of 300-200 K, the photoresponse extends up to the MWIR range with a kink in the spectrum, where the responsivity reaches the noise floor at around 4 µm. Interestingly, as the temperature decreases, an additional broad photoresponse band spanning from 1.9 to 3.7 µm is clearly observed (≤ 160 K). Particularly, the responsivity extends well up to 5 µm in the temperature range of 60-20 K. As shown in Figure 3(h) and discussed in the previous work, [34] the sub-bandgap photoresponse observed here corresponds to the excitation of charge carriers from the VB to the CB through the IB (VB →IB (process II) and IB→CB (process III)). Moreover, this observation of sub-bandgap optical transitions indicates that the IB is separated from the CB.

Detectivity
The mostly quoted performance merit for a photodetector is its sensitivity, i.e. the input-output signal efficiency compared to the output noise signal. This merit is exemplified as the noise equivalent power (NEP), which is defined as the signal power at which the photocurrent can no longer be differentiated from the noise floor, expressed as Where Rph is the spectral responsivity and 2 1/2 is the root mean square of the total noise current. [8] Here where Rph is the responsivity with the units of A/W and A is the photosensitive area (0.082 cm 2 ) of the detector. The calculated specific detectivity (D*) of the device is shown in Figure 4. The maximum room-temperature D* of around 3.2 × 10 12 cmHz 1/2 W -1 is achieved at 1 μm (see Figure 4(a)), which is above two times larger than that of Ag-hyperdoped Si photodetectors. [21] Importantly, the achieved room-temperature D* at 1 μm is comparable to that of a commercial Si Photodiode (i.e. FDS1010).
Figure 4(b) shows the D* of the Te-hyperdoped Si photodetector at different temperatures, from 20 to 300 K. The D* related to the room-temperature sub-bandgap photoresponse is in the range of 10 11 -10 6 cmHz 1/2 W -1 , which is 4 orders of magnitude larger than that of Tisupersaturated Si photodetector. [19] With increasing temperature, the detector exhibits a decreasing D*, which is mainly due to the increase of the noise as the temperature increases. However, this performance can be improved by optimizing the manufacturing process. In the future, efforts must be focused towards an advanced-device design to boost the device efficiency.
The device architecture can be optimized by proper surface passivation, better optical active area design for improving absorption, different contact geometries and metal electrodes for improving carrier collection, as well as antireflection-coating for improving the responsivity.
Frequency-dependent responsivity is a key factor related to the response time of the devices. to a time constant τ of 37 µs) at room temperature, which is more than two times larger than that of Ti-supersaturated Si photodetectors. [19] This indicates a high bandwidth of the Tehyperdoped Si photodetector, which also excludes the thermal (bolometric) contribution to the observed responsivity. [8]  was measured to be 39 µs, the time required by the photodetector to generate a photocurrent from 10 to 90 of the peak output value after the light illumination is turned on. The decay time was found to be 42 µs.

Conclusion
In conclusion, we have fabricated a MWIR infrared photodetector based on Te-hyperdoped Si.
We have demonstrated that the Te implanted Si layer is fully recrystallized after PLM and exhibits a Te-mediated broad infrared absorptance up to the MWIR region. In addition, we have Te-hyperdoped Si is a promising material for room-temperature Si-based infrared photodetectors. Moreover, the full process, including ion implantation and short-time annealing, is CMOS-compatible. To be competitive with commercial infrared photodetectors, further efforts must be made toward an advanced device design to boost the device efficiency of this prototype of infrared photodetector.

Material preparation and characterization:
A p-type double-side polished Si (100) wafer with a thickness of 380 ± 5 µm and a resistivity of 1-10 Ωcm was implanted with Te ions at room temperature. To obtain a homogeneous distribution of Te ions in the implanted layer, the sample was implanted by two sequential implantations with a dose of 1.6×10 15 cm -2 and 6.2×10 14 cm -2 and implantation energies of 150 keV and 50 keV, respectively. The implanted layer is 120 nm thick with a peak Te concentration of 2.5×10 20 cm -3 (0.5 %), which has been calculated by the SRIM code and then verified byRBS. [47] After ion implantation, the sample was annealed by a spatially homogenized XeCl excimer laser with 308 nm wavelength and 28 ns duration to achieve the re-crystallization of the implanted layer. [32][33][34]  and measurements can be found elsewhere. [32][33][34] Device Fabrication: The structure of the Te-hyperdoped Si p-n photodiode comprises a thin Te-hyperdoped Si layer on top of the p-Si substrate (as shown in Figure 1 and Figure S1). In detail, the Te-hyperdoped Si sample was immersed into 10% hydrogen fluoride (HF) solution to remove the native-SiO2 layer. Next, a photolithography process was performed to prepare the top electrode. A 0.084 cm 2 illuminated area was obtained by defining fingers with a separation of 5 μm resulting in frame-like Au top electrodes on top of the n-type Te-hyperdoped Si layer . [33,34] The bottom electrode was made by an In/Ga eutectic layer to form an ohmic contact at the bottom surface with a certain distance from the sample edges in order to reduce possible parasitic electrical conduction through the edges of the sample. signal of the lock-in amplifier. Therefore, the frequency of the TTL signal is adjustable to generate pulsed light and thus perform a frequency scan of the responsivity. Moreover, the output power of the LED is also adjustable with a maximum incident power density of 1.83 mW/cm 2 . The cut-off frequency of the LED is 0.1 GHz, which is sufficiently higher than that of the Te-hyperdoped Si p-n photodetector under investigation. In the time-resolved experiments, a low-noise current amplifier was employed to amplify the photocurrent signal from the device under 1.55 µm-LED excitation. A digital oscilloscope was used to record the time-resolved photocurrent.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author. Si is also included for comparison. The spectra have been vertically offset for clarity. (g) Material properties. Optical sub-bandgap absorption spectra from FTIR measurements for virgin Si (black), and the PLM-treated Te-hyperdoped Si with a Te concentration of 0.50 (blue). The inset shows the energy probability distribution of Te deep-level states which is determined by fitting the absorption spectrum with Equation (3) (red). The raw data for the absorptance of sample Te-0.5% can also be found in ref. [43].