CHIN. PHYS. LETT. Vol. 30, No. 3 (2013) 036101 High Response in aTellurium-Supersaturated Silicon Photodiode * WANG Xi-Yuan(王熙元)1 , HUANG Yong-Guang(黄永光)1** , LIU De-Wei(刘德伟)1,2 , ZHU Xiao-Ning(朱小宁)1 , ZHU Hong-Liang(朱洪亮)1 1 Key Laboratory of Semiconductor Materials Science, Institute of Semiconductors, Chinese Academy of Sciences, P. O. Box 912, Beijing 100083 2 Department of Technology and Physics, Zhengzhou University of Light Industry, Zhengzhou 450002 (Received 28 November 2012) Single crystalline silicon supersaturated with tellurium are formed by ion implantation followed by excimer nanosecond pulsed laser melting (PLM). The lattice damaged by ion implantation is restored during the PLM process, and dopants are effectively activated. The hyperdoped layer exhibits high and broad optical absorption from 400 to 2500 nm. The n+ p photodiodes fabricated from these materials show high response (6.9 A/W at 1000 nm) with reverse bias 12 V at room temperature. The corresponding cut-off wavelength is 1258 nm. The amount of gain and extended cut-off wavelength both increase with increasing reverse bias voltage; above 100% external quantum efficiency is observed even at a reverse bias of 1 V. The cut-off wavelength with 0 V bias is shorter than the commercial silicon detector. This implies that the Burstein-Moss shift is due to hyperdoping. The amount of the extended cut-off wavelength increases with increasing reverse bias voltage, suggesting existence of the Franz–Keldysh effect. PACS: 61.72.uf, 61.72.U, 85.60.Gz DOI: 10.1088/0256-307X/30/3/036101 Because of silicon’s indirect band gap of 1.12 eV, the absorption and photoelectric response reduce sharply for wavelengths longer than 1100 nm.[1] The absorption coefficient of crystalline silicon is less than 103 cm−1 for wavelengths longer than 850 nm, therefore commercial silicon photodetectors only have high responsivity at wavelengths shorter than 850 nm.[2] Many efforts have been focused on extending the operating wavelength to improve responsivity. In recent years, silicon highly supersaturated with chalcogens beyond solid solubility limits has revealed strong and broad optical absorption from 400 nm to 2500 nm. These materials have been prepared by optical hyperdoping,[3−6] laser incorporation,[7] and ion implantation followed by pulsed laser melting (PLM).[8−10] However, experimental results[11,12] and first-principles calculations[13] show that the optical absorption above 1100 nm for silicon hyperdoped with sulfur, selenium, and tellurium all drops markedly after thermal annealing, and the decrease amplitude for tellurium is the lowest. The initial sub-bandgap optical absorption for silicon hyperdoped with tellurium is larger than sulfur and selenium by theoretical calculation,[13,14] too. Silicon photodiodes supersaturated with sulfur prepared by femtosecond laser irradiation in an SF6 atmosphere show high and broad response,[15] the reason for the high response is considered as a generation-recombination gain mechanism.[16] Said et al.[17] reported that n+ p photodiodes fabricated by sulfur or selenium ion implantation followed by pulsed laser melting exhibit high gain and a broad response wavelength range. Moreover, insulator-to-metal transition has been ob- served separately in silicon supersaturated with sulfur and selenium prepared by ion implantation followed by PLM.[18,19] In this Letter, we report an enhancement of both the wavelength range and photoelectric response exhibited by n+ p photodiodes fabricated by tellurium ion implantation followed by excimer nanosecond pulsed laser melting. Pan et al.[10] and Bob et al.[9] reported silicon supersaturated with tellurium. The silicon wafers were implanted with high dose tellurium ions (130 Te+ of 1 × 1016 cm−2 ) and followed by XeCl 308 nm excimer nanosecond pulsed laser melting (50 ns duration, laser fluence of 0.6–0.7 J/cm2 and 2.3 J/cm2 , respectively), the material shows strong sub-bandgap optical absorption and high free carrier density, but no high response photodetector has been reported. Our experimental parameters are different: tellurium ion implantation dose of 126 Te+ 2 × 1015 cm−2 and KrF 248 nm excimer nanosecond pulsed laser (30 ns duration, laser fluence of 0.3 J/cm2 ). In addition, we use double side passivation with a back surface point contact structure to reduce the surface recombination. Our photodiode results and those of Said et al. demonstrate that silicon photodiodes supersaturated with sulfur, selenium or tellurium can exhibit a high and broad photoelectric response. This class of materials has potential applications for photodetectors. In our experiment, 370-µm-thick polished Czochralski p-type monocrystalline Si (100) wafers with resistivity of 30–50 Ω·cm were ion implanted at room temperature with 245 keV 126 Te+ to a dose of 2 × 1015 ions/cm2 . All wafers were tilted by ∼7∘ off the incident beam axis to minimize the channelling * Supported by the Beijing Natural Science Foundation under Grant No 4122080, the National Basic Research Program of China under Grant No 2012CB934202, and the Chinese Academy of Sciences (No Y072051002). ** Corresponding author. Email: yghuang@red.semi.ac.cn © 2013 Chinese Physical Society and IOP Publishing Ltd 036101-1 CHIN. PHYS. LETT. Vol. 30, No. 3 (2013) 036101 ) -1 n+ layer p-Si SiO2 ) ) -3/2 SK 3/2 100 -3 Metal ln( S W cm 103 102 101 100 10-1 10-2 10-3 (cm SiO2 sition. The inset is an Arrhenius plot of carrier concentration versus temperature. The linear fit slope demonstrates a deep level trap state of 0.183 eV (very close to reported value[23] ). The measured sheet carrier concentration is 1.436 × 1014 cm−2 at 300 K, the corresponding carrier-to-donor ratio (the ratio of free carrier density to total Te atom concentration) is 7.18%. The average carrier bulk concentration is approximately 4.1 × 1018 /cm3 , at least 2 orders of magnitude larger than Te solubility in c-Si.[24] Conductivity ( effect. The maximum implantation depth is approximately 350 nm according to SRIM-2011 software.[20] For effectively activating dopants, all ion-implanted wafers were irradiated with a 1 pulse spatially homogenized, KrF pulsed excimer laser (248 nm, 30 ns pulse duration). The wafers were translated by a horizontal displacement set to make sure that the laser spot could irradiate the whole area. The single pulse laser fluence was 0.3 J/cm2 . Laser melting and the subsequent rapid solidification form a single crystal region nearly free of defects, which retains most of the implanted dopants.[21,22] Raman spectral, optical property, electrical property, and Rutherford backscattering spectroscopy (RBS) were performed in sequence. Before each experiment and measurement, wafers were cleaned ultrasonically. 200 35.0 34.8 34.6 34.4 34.2 34.06 8 1000/ 300 400 10 (1/K) 12 500 (K) Metal (arb. units) Figure 1 shows the cross-sectional view of the photodetector. We use these materials to fabricate photodiodes by using a standard complementary metal oxide semiconductor (COMS) process. The 200 nm SiO2 films were deposited on both sides by plasma enhanced chemical vapour deposition (PECVD) for passivation. The electrode windows were prepared by lithography and HF solution etching. The 150 nm top electrode was deposited by thermal evaporation (background pressure 1 × 10−4 Pa), and the 500 nm bottom electrode was deposited by electron beam evaporation (background pressure 1 × 10−6 Pa). The electrodes of two sides were rapid thermal annealed with pure N2 ambient protection to form the ohm contact. Lastly, the wafers were diced to small units. We measured the 𝐼–𝑉 characteristics and photoresponse at room temperature. We measured the Hall effect of the Te-doped silicon wafer annealed by PLM over a temperature range of 80–500 K using van der Pauw technique. The sign of the Hall coefficient is of n-type, implying the formation of a p–n junction. Figure 2 shows the temperaturedependent conductivity of Te-supersaturated silicon. From 500 K to 80 K, the conductivity monotonically reduces to 1/10 implying semiconductor characteristics rather than metal characteristics. It demonstrates that the Te-supersaturated silicon in this doping concentration does not establish insulator-to-metal tran- Fig. 2. (Color online) Temperature-dependent conductivity of Te-supersaturated silicon. The conductivity reduces monotonically with decreasing temperature, it suggests semiconductor characteristics rather than metal characteristics. Inset: an Arrhenius plot of carrier concentration versus temperature, revealing a deep level trap state with the activation energy of 0.183 eV. Raman spectral intensity Fig. 1. (Color online) Cross-sectional view of the Tesupersaturated silicon photodiode. The n+ layer is Tedoped silicon, and 200-nm-thick SiO2 films were deposited by PECVD on both sides for passivation. The 150 nm metal film was deposited by thermal evaporation as the top electrode, another 500-nm-thick metal film was deposited by electron beam evaporation as the bottom electrode. 2000 1500 1000 500 0 1500 1000 500 Silicon wafer Silicon wafer implanted with Te 0 2000 Te-supersaturated silicon after PLM 1500 1000 500 0 450 460 470 480 490 500 510 520 530 540 550 Wavenumber(cm -1 ) Fig. 3. (Color online) Raman spectra of the silicon wafer, the silicon wafer implanted with Te, and the Tesupersaturated silicon after PLM. Figure 3 shows the Raman spectra of the silicon wafer, the silicon wafer implanted with Te, and the Te supersaturated silicon after PLM (silicon wafer and Te-supersaturated silicon after PLM were excited by the same intensity laser). The Raman peak of the silicon wafer is 520.894 cm−1 (FWHM of 1.638 cm−1 ), the Raman hump from 450 cm−1 to 530 cm−1 of the silicon wafer implanted with Te is derived from the amorphous silicon phase[25] and defects induced by ion implantation, the Raman peak of the Te supersaturated silicon after PLM is 521.003 cm−1 (FWHM of 1.420 cm−1 ). After PLM, the hump disappears and 036101-2 CHIN. PHYS. LETT. Vol. 30, No. 3 (2013) 036101 the Raman signal peak is almost completely from the crystalline silicon phase, and the FWHM is even narrower than that of silicon wafer without any processing. It is suggested that PLM is efficient in restoring the crystalline lattice. Moreover, the Raman peak shifts to 521.003 cm−1 after the PLM process. This suggests existence of compressive stress in the lattice, which may be due to lattice distortion induced by plenty of large-radius Te atom incorporation and not completely recovering the volume reduction (the volume of liquid silicon is smaller than that of solid silicon) after the rapid melting-solidification process during PLM. Random Total Te content 100 Substitutional Te 10 1 500 750 1000 1250 1500 Current (mA) RBS yield (Counts) Channelling 1000 ter PLM ranges from 400 nm to 2500 nm. The total absorptance A was determined from the directly measured reflectance 𝑅 and transmittance 𝑇 , according to 𝐴 = 1 − 𝑅 − 𝑇 . The absorptance of silicon wafer without any processing is less than 0.03 from 1200 nm to 2500 nm. After ion implantation, the absorptance changes slightly from 400 nm to 1000 nm, the gradient absorptance line from 1000 nm to 2500 nm implies the existence of amorphous silicon, which is demonstrated by Raman spectral measurement above. After PLM, the absorptance from 400 nm to 1000 nm restores like silicon wafer without any processing, and the absorptance from 1150 nm to 2500 nm (corresponding to subbandgap of silicon) is approximately 0.25–0.30, which may arise from direct defect-to-conduction-band transitions induced by Te ion implantation followed by PLM.[19] 1750 Scattered energy (keV) Fig. 4. (Color online) Rutherford backscattering spectrometry of Te-supersaturated silicon after PLM. Voltage (V) 0.8 Fig. 6. (Color online) The current-voltage characteristics of the Te-supersaturated silicon photodiode. 0.6 Absorptance 50 40 30 20 10 0 -10-2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 Silicon wafer implanted with Te 0.4 Te-supersaturated silicon after PLM 0.2 Silicon wafer 0.0 500 1000 1500 2000 2500 Wavelength(nm) Fig. 5. (Color online) The absorptance of silicon wafer, silicon wafer implanted with Te, and Te supersaturated silicon after PLM ranges from 400 nm to 2500 nm. The reflectance 𝑅 and the transmittance 𝑇 were measured directly, the absorptance was calculated by the formula 𝐴 = 1 − 𝑅 − 𝑇. Figure 4 shows the Rutherford backscattering spectroscopy (RBS) of Te supersaturated silicon after PLM. The peak of Te in the C-RBS spectrum originates from substitutional tellurium atoms in the crystalline silicon lattice, the peak of Te in the R-RBS spectrum originates from the total tellurium atoms (substitutional and interstitial). The Te atom substitutional rate is approximately 37.9%. Figure 5 shows the optical absorptance of silicon wafer, silicon wafer implanted with Te, and Te supersaturated silicon af- Figure 6 shows the current-voltage characteristics of the Te-supersaturated silicon photodiode. The forward threshold voltage is 0.55 V. We measured the responsivity of these photodiodes at a series of reverse bias voltages, using light from a 300 W halogen tungsten lamp which was filtered through a monochromator with a spectral resolution of 0.1 nm. We scanned the monochromator in steps of 20 nm from 400 to 1600 nm, keeping the optical power to be on the order of 1 µW at each wavelength. The light was focused, and the illuminated area was much smaller than the photodiode top active region. The light was positioned to give the maximum photoresponse for each photodiode. The monochromatic light was chopped so that the photodiode response could be measured using a lock-in amplifier. A commercial Si photodiode and a commercial InGaAs photodiode were used to set the output power at each wavelength. The commercial Si photodiode was used for measurement from 400 to 1000 nm, and the commercial InGaAs photodiode from 1000 to 1600 nm. Contributions from second-order wavelengths for measurement above 800 nm were eliminated by passing the light through a high pass filter. In all measurements, light was incident on the top surface of photodiode at room 036101-3 CHIN. PHYS. LETT. Vol. 30, No. 3 (2013) 036101 temperature.[26] Figure 7 depicts the result of measurement. For wavelengths above 1270 nm, the responsivity is almost undetectable due to the background noise signal in the measurement system. At 0 V, the responsivity is low, and the cut-off wavelength (0.03 A/W) is 1136 nm, shorter than that of a commercial silicon photodiode (1160 nm). The blue shift suggests that the Burstein–Moss effect[27] may be induced by hyperdoping. The responsivity increases with increasing reverse bias voltage, EQE (external quantum efficiency) beyond 100% is observed even at −1 V. A remarkable gain (6.9 A/W at 1000 nm) at −12 V is exhibited, the corresponding EQE is more than 850% due to EQE=[1.24 × 𝑅(𝐴/𝑊 )]/𝜆 (µm). In addition, by increasing reverse bias voltage, the optoelectronic response of the Te-supersaturated silicon photodiode extends into the infrared, the amount of the red shift also increases with the increasing reverse bias voltage, the cut-off wavelength is 1256 nm at −12 V. The red shift may originate from the Franz–Keldysh effect[28] and further study is in progress. 10 -12 V Responsivity (A/W) -4 V -2 V Q E = 10 0% and so does the amount of extended cut-off wavelength into infrared. The responsivity is 6.9 A/W at −12 V at 1000 nm corresponding to EQE 855%. According to our high response, Te-supersaturated silicon photodiodes and the result of Said et al., current chalcogen elements (sulfur, selenium, and tellurium) supersaturated silicon photodiodes all exhibit a high and broad response. It is illustrated that this class of materials indeed has novel optoelectronic properties. The strong gain mechanism may be the generation-recombination gain mechanism resulted from the high trap density like sulfur supersaturated with silicon,[16] and there may be no direct relationship between insulatorto-metal transition and high response (EQE>100%). The present result has potential applications in faint light signal detection. The authors would like to acknowledge Jianming Li and Jiadong Xu of the Semiconductor Integration Technology and Engineering Center, Institute of Semiconductors, Chinese Academy of Sciences, for ion implantation, Yong Zeng, Yan Zhao and Professor Tao Chen of the Institute of Laser Engineering, Beijing University of Technology for PLM, Shaoxu Hu of the State Key Lab on Integrated Optoelectronics, Institute of Semiconductors, Chinese Academy of Sciences, for the optoelectronic response testing. 1 - 1 V Commercial Si 0V References detector 0.1 400 600 800 1000 1200 1400 1600 Wavelength (nm) Fig. 7. (Color online) Responsivity of Te-supersaturated silicon photodiode at different reverse bias voltages. The responsivity of a commercial silicon photodetector at 0 V is shown for reference, and its responsivity does not vary with reverse bias voltages. The EQE>100% is exhibited even at −1 V, and Raman spectra of Te-supersaturated silicon after PLM shows only a crystalline silicon phase. This evidence suggests that the physical mechanism of high gain in this material system could not be avalanche gain. Such a small bias voltage is insufficient to drive appreciable amounts of impact ionization in silicon.[29] The strong gain mechanism may be the generationrecombination gain mechanism resulted from high trap density like sulfur supersaturated with silicon,[16] and it has potential applications for photodetectors. In summary, we have fabricated n+ p silicon photodiodes with Te ion implantation and followed by nanosecond excimer pulsed laser melting. The responsivity increases with increasing reverse bias voltage, [1] Sze S M 1981 Physics of Semiconductor Devices (New York: Wiley-Interscience) [2] Csutak S M et al 2002 IEEE Photon. Technol. Lett. 14 516 [3] Wu C et al 2001 Appl. Phys. Lett. 78 1850 [4] Younkin R et al 2003 J. Appl. Phys. 93 2626 [5] Sheehy M A et al 2006 Mater. Sci. Eng. B 137 289 [6] Sher M J et al 2011 Mater. Res. Bull. 36 439 [7] Tabbal M et al 2010 Appl. Phys. A 98 589 [8] Kim T G et al 2006 Appl. Phys. Lett. 88 241902 [9] Bob B P et al 2010 J. Appl. Phys. 107 123506 [10] Pan S H et al 2011 Appl. Phys. Lett. 98 121913 [11] Crouch C H et al 2004 Appl. Phys. A 79 [12] Tull B R et al 2009 Appl. Phys. A 96 327 [13] Shao H et al 2012 Europhys. Lett. 99 46005 [14] S ánchez K et al 2010 Phys. Rev. B 82 165201 [15] Carey J E et al 2005 Opt. Lett. 30 1773 [16] Huang Z H et al 2006 Appl. Phys. Lett. 89 033506 [17] Said A J et al 2011 Appl. Phys. Lett. 99 073503 [18] Winkler M T et al 2011 Phys. Rev. Lett. 106 178701 [19] Ertekin E et al 2012 Phys. Rev. Lett. 108 026401 [20] Ziegler J F et al 2010 Nucl. Instrum. Methods Phys. Res. Sect. B 268 1818 [21] Boyd I W and Wilson J I B 1983 Nature 303 481 [22] Recht D et al 2012 Appl. Phys. Lett. 100 112112 [23] Grimmeiss H G et al 1981 Phys. Rev. B 24 4571 [24] Sheehy M A 2004 Ph. D. Thesis (Harvard University) [25] Voutsas A T et al 1995 J. Appl. Phys. 78 6999 [26] Hu S X et al 2012 Semicond. Sci. Technol. 27 102002 [27] Burstein E 1954 Phys. Rev. 93 632 [28] Franz W 1958 Z. Naturforsch. A: Phys. Sci. 13 484 [29] Chang C et al 1985 J. Appl. Phys. 57 302 036101-4