Ultrascaled Germanium Nanowires for Highly Sensitive

Jul 11, 2018 - related hole-trapping states at the NW surface and can be described by a ... highly resolved imaging, light effect transistors and sing...
0 downloads 0 Views 2MB Size
Letter Cite This: Nano Lett. 2018, 18, 5030−5035

pubs.acs.org/NanoLett

Ultrascaled Germanium Nanowires for Highly Sensitive Photodetection at the Quantum Ballistic Limit Philipp Staudinger,† Masiar Sistani, Johannes Greil, Emmerich Bertagnolli, and Alois Lugstein*

Nano Lett. 2018.18:5030-5035. Downloaded from pubs.acs.org by EASTERN KENTUCKY UNIV on 08/19/18. For personal use only.

Institute for Solid State Electronics, Technische Universität Wien, Floragasse 7, 1040 Vienna, Austria ABSTRACT: We report an experimental study on quasi-one-dimensional Al− Ge−Al nanowire (NW) heterostructures featuring unmatched photoconductive gains exceeding 107 and responsivities as high as 10 A/μW in the visible wavelength regime. Our observations are attributed to the presence of GeOx related hole-trapping states at the NW surface and can be described by a photogating effect in accordance with previous studies on low-dimensional nanostructures. Utilizing an ultrascaled photodetector device operating in the quantum ballistic transport regime at room temperature we demonstrate for the first time that individual current channels can be addressed directly by laser irradiation. The resulting quantization of the photocurrent represents the ultimate limit of photodetectors, allowing for advanced concepts including highly resolved imaging, light effect transistors and single photon detectors with practically zero off-state current. KEYWORDS: Photoconductive gain, quantum ballistic transport, photogating, germanium, nanowire

Q

channel conductivity. Thus, trapping of photogenerated carriers enhances the sensitivity of field-effect-based photodetectors as already reported in previous studies for other nanostructures such as MoS2,13 black phosphorus,14 ZnO NWs,15 or PbS quantum dots.16 Figure 1c schematically depicts this trapping mechanism and the electrostatic response with respect to an externally applied gate voltage (VG). Without an external bias, the partially filled acceptor-like surface states in the GeOx interface layer cause an upward bending of the bands inducing the p-type behavior commonly observed in nominally intrinsic Ge NWs17−19 (Figure 1c, left). Applying a negative gate voltage leads to further hole accumulation and an initial increase in conductivity as expected for p-type semiconductors. However, due to the imbalanced trap population with respect to the Fermi level (Figure 1c, center), filled surface states begin to discharge over time (Δt) to establish a new equilibrium resulting in intermediate band bending (Figure 1c, right). Experimental evidence for this phenomenon and the impact on the electrical transport are shown in Figure 1d. After applying a negative gate voltage (VG = −15 V) at room temperature, the drain current (ID) through the back-gated Ge NW FET device decreases exponentially by more than 3 orders of magnitude. Relaxation time constants are in the range of several minutes in excellent accordance with trapping times of slow surface states in Ge reported in previous studies.20,21 Since redistribution processes are kinetically limited,18 no such decay occurs when conducting the experiment at T = 80 K. To

uasi 1D semiconductor nanowires (NWs) have attracted considerable attention because of their mechanical, electrical, and optical properties that are distinctly different from those of their bulk material counterparts. Some of these qualities can be independently controlled or even enhanced by leveraging the singular aspects of the 1D electronic states. Particularly Ge nanostructures are of renewed interest not only for next generation field effect transistors1 but also, due to the moderate direct bandgap, for various electrooptical applications such as near-infrared photodetectors2 or even lasers.3 Although Ge NWs have been considered as photoconductive elements in the past,4−6 an ultrahigh gain such as it is commonly observed in 2D materials7 has not been reported to date. In this work, we demonstrate a quantum ballistic photodetector based on individual Ge NWs integrated into backgated field effect transistors (FET) with extraordinary high photosensitivity accomplished by controlling the surface trap population. Channel lengths are aggressively scaled by utilizing a thermally induced exchange reaction8 to achieve axial metal− semiconductor−metal (Al−Ge−Al) heterostructures allowing for quantum confinement and ballistic transport even at room temperature.9 Thus, by exploiting scaling and photogating effects we demonstrate ultrahigh gain quantum ballistic Ge photodetectors with a spatial footprint of less than 500 nm2. A schematic illustration of the back-gated FET device under test and an exemplarily TEM image of the actual photodetector are depicted in Figure 1a and 1b, respectively. To achieve reliable and reproducible electrical performance, the NWs are coated by a protective ALD-deposited Al2O3 layer. However, a thin native GeOx layer remains at the interface10 providing a large number of trapping states,11 which further act as a highly efficient local gate,12 directly modulating the © 2018 American Chemical Society

Received: May 6, 2018 Revised: June 30, 2018 Published: July 11, 2018 5030

DOI: 10.1021/acs.nanolett.8b01845 Nano Lett. 2018, 18, 5030−5035

Letter

Nano Letters

Figure 1. Ge NW photodetector. (a) Schematic illustration of the device architecture, with an Al2O3 coated Ge NW monolithically contacted by Al leads. (b) TEM image of the Al−Ge−Al NW heterostructure achieved by a thermally induced exchange reaction (see Methods) with a diameter of d = 30 nm and a segment length of LGe = 230 nm. The thin GeOx interface layer is not visible in this image. Scale bar is 100 nm. (c) Schematic band diagram visualizing the charge trapping mechanism causing a strong surface gating effect. The filling of acceptor-like surface traps below the Fermi level is the origin of cross-sectional band bending and the observed p-type behavior of nominally intrinsic Ge NWs. Applying a negative gate voltage leads to the accumulation of additional holes, which further increases the conductivity in the NW. Discharging of filled traps above the Fermi energy over time counteracts this effect. (d) Transient behavior of the drain current (ID) for a device with a long Ge segment of LGe = 2 μm at a bias voltage VDS = 0.1 V at room temperature and 80 K directly after applying a gate voltage of VG = −15 V. (e) Same measurement as in Figure 1d at room temperature but with incident chopped laser light with λ = 532 nm and f mod = 0.1 Hz modulation frequency. (f) Transfer characteristics measured at VDS = 0.1 V with and without applied laser light (logarithmic plot depicted in inset). The gating effect of trapped photogenerated carriers causes the threshold voltage to increase under laser irradiation.

compared to Figure 1d is due to statistical variations in the discharging process of the surface trapping states. As shown in Figure 1f laser illumination causes the threshold voltage to increase by approximately 1 V for the actual Ge NW FET device demonstrating the tremendous potential of this socalled photogating effect4,7,12,14 for highly sensitive photodetection. To investigate scaling effects, we characterize Ge NW photodetectors featuring channel lengths of 650, 85, and 45 nm. The plot in Figure 2a shows the photocurrent for these devices exposed to chopped green laser light (λ = 532 nm, modulation frequency f mod = 127 Hz) as a function of irradiance (EL). For low illumination densities (EL < 2 W/m2) the photocurrent appears to be directly proportional to the irradiance and levels with a sublinear increase at higher incident light power with exponents as low as 0.2 due to saturation of surface traps. The data were fitted according to the Hornbeck−Haynes model13,14 taking into account trapassisted photocurrent enhancement:

demonstrate the influence of the surface traps on the photoresponse, chopped laser light is applied at frequencies much higher than the observed detrapping times. Figure 1e shows the instantaneous and exceptional high photoresponse as a function of time, i.e. filling level of surface traps, indicating the importance of the trapping process on the mechanisms involved in the photocurrent generation. Electron−hole pairs are generated when photons with energies higher than the bandgap are absorbed in a semiconductor. If one carrier species is efficiently trapped into localized surface states, they act as a local gate, effectively modulating the transistor threshold voltage (Vth). Thus, the photocurrent is directly related to the transconductance (gm) of the transistor through the relation:

Iph = gmΔVth

(1)

where ΔVth represents the shift in threshold voltage due to laser light exposure. Since gm itself is dependent on the threshold voltage of the transistor, Iph decreases as the trapping states are depleted in Figure 1e. The lower dark current as 5031

DOI: 10.1021/acs.nanolett.8b01845 Nano Lett. 2018, 18, 5030−5035

Letter

Nano Letters

Figure 2. Photodetector characteristics. (a) Photocurrent as a function of irradiance of incident light for devices with Ge segment lengths of LGe = 650, 85, and 45 nm and diameters of d ≈ 30 nm measured at constant longitudinal electrical field EDS = 104 V/cm (VG = 0 V, T = room temperature, λ = 532 nm, f mod = 127 Hz). Solid lines represent a fit according to the Hornbeck−Haynes model13,14 (eq 2). (b) Photoconductive gain and responsivity as a function of the absorption rate as extracted from Figure 2a. (c) Wavelength dependency of the photocurrent measured at constant irradiance EL = 18 kW/m2 (VG = 0 V, T = room temperature, f mod = 127 Hz). (d) Schematic band diagram visualizing the proposed phototrapping mechanism. Photogenerated electrons, trapped in acceptor-like surface states, act as a local gate.

Iph = qGph,0

Fph n

1 + (Fph /Fph,0)

geometrical and bulk material properties (see Methods). For low illumination densities we obtain responsivities as high as 10 A/μW and photoconductive gains exceeding 107 at room temperature, which to the best of our knowledge is the highest ever reported for Ge NWs and comparable to values reported for both highest performing 1D (ZnO NWs,15 Gph ∼ 108) and 2D (Graphene-MOS2 heterostructures,23 Gph ∼ 108) photoconductors. A further increase of irradiance and thus the number of absorbed photons and photogenerated carriers causes a steady decrease of the photoconductive gain due to the saturation of surface traps. The spectral response plotted in Figure 2c demonstrates the high sensitivity obtained throughout the whole visible spectral range. The decrease at higher wavelength is expected when the excitation energy becomes less than the bandgap of the respective absorbing material. For Ge, the generation of electron−hole pairs is inefficient for energies below the bandgap corresponding to the direct transition at the Γ-point (EΓ1 = 0.8 eV), resulting in a steep decrease of the photocurrent. The preceding decrease at about 1 eV is due to the kinetic barriers for electrons to be trapped efficiently into the oxide, as previously described by Hanrath et al.18 Based on these findings we expand the model, introduced earlier to explain the transient behavior of the FET device (Figure 1c), with respect to the observed electro-optical phenomena. Incident photons of sufficient high energy are absorbed in the Ge channel of the FET device generating electron−hole pairs. Due to the high number of acceptor-like surface states18 and the small diameter of the NWs, the

(2)

where q is the elementary charge, Gph,0 the photoconductive gain at low illumination density, Fph the number of absorbed photons per unit time (see Methods), Fph,0 the photo absorption rate at which trap saturation occurs, and n a phenomenological fitting parameter giving the sublinear increase in the trap saturated regime. From the experimental data, n = 0.8 and Fph,0 = 7.5 × 104 s−1 (corresponding to EL = 2 W/m2) have been deduced for the device with a channel length of 650 nm. We find a slight increase of n and decrease of Fph,0 for devices with shorter channel lengths. However, the overall device performance remains similar, which is quite remarkable considering the reduction of feature sizes to values far below the wavelength of the incident light. Photoconductive detectors are often benchmarked by their photoconductive gain (Gph), which is the ratio between the number of extracted electrons and the number of absorbed photons per unit time.7 To calculate the number of incident photons we use the physical cross section of the respective NW device. Resonance effects which possibly could increase the absorption cross section are reported to play only a minor role for these NW geometries. In fact, calculations show that the absorption efficiency of Ge NWs having a diameter of around 30 nm is well below 1 for all wavelengths investigated in our study.22 In Figure 2b, (Gph) is plotted together with responsivity (R) as a function of absorbed photons per second (Fph) estimating an absorption (η) of approximately 70−80% based on pure 5032

DOI: 10.1021/acs.nanolett.8b01845 Nano Lett. 2018, 18, 5030−5035

Letter

Nano Letters

Figure 3. Quantum ballistic germanium nanowire photodetector. (a) Conductance (G) as a function of gate voltage (VG) for a device with LGe = 18 nm with (λ = 532 nm, EL = 27 kW/m2) and without laser illumination at room temperature. The conductance was directly obtained from the measured current according to G = ID/VDS. The inset shows schematically how conductance quantization originates in these structures by either applying a gate voltage or laser light. (b) Time resolved measurement at VG = −20.6 V and pulsed laser light (f mod = 0.25 Hz, λ = 532 nm, EL = 27 kW/m2).

controlling the sub-band population by light exposure. A schematic of this mechanism is given in the inset of Figure 3a, where (photo)gating induces a shift of the obtained 1D subbands of the channel. Thus, for a particular back-gate voltage, light exposure enhances the conductance by entirely opening an additional transfer channel in the device as demonstrated in Figure 3b. Transmission via the second sub-band is switched on and off as the laser is modulated causing an effective quantization of the obtained photocurrent. The sensitivity of the photodetector operating in this regime is significantly enhanced by the high transconductance originating from the steep current increase in-between individual conductance plateaus (compare eq 1) and corresponds to a photoconductive gain of approximately 2 × 104, which is remarkable considering the spatial footprint of the sensing element of less than 500 nm2. The manipulation of individual transmission single current channels represents the ultimate limit of photodetectors, allowing for advanced concepts such as light effect transistors27 and light sensing elements with practically zero off-state current. In summary, we present Ge NW photodetectors featuring unmatched photoconductive gains and responsivities in the whole visible spectral range. A range of devices are studied in detail, and our findings are discussed in the framework of a charge trapping model under the consideration of the photogating effect in accordance with previous studies on low-dimensional photoconductors. The extraordinary high photosensitivity allows for reducing channel lengths down to sizes at which ballistic transport occurs, enabling the first demonstration of an ultrahigh gain quantum ballistic Ge photodetector at room temperature. These findings have the potential to pave the way toward advanced photoelectric devices fully compatible with today’s CMOS technology.

photogenerated electrons are efficiently trapped in the interfacial oxide layer, where they further act as a local gate, effectively modulating the channel conductivity. This mechanism can result in an extraordinary high gain up to 108,15,23 while photodetectors such as photomultipliers, phototransistors, and avalanche photodiodes typically show multiplication factors of less than 106.24 It is known that two types of surface trapping states exist in Ge, with time constants in the range of minutes and microseconds, respectively.11 According to the modulation frequency of 127 Hz of the incident light, the faster type is responsible for the photoconductive gain in our electrooptical measurements. This photogating effect is schematically depicted in Figure 2d. Higher photon energies promote the generation of electrons which can efficiently fill high energetic traps, causing the photoconductive gain to increase with decreasing wavelengths. Above a certain illumination density, negative surface charges prevent electrons from further occupying traps in the GeOx, which causes a pronounced saturation mechanism. For the proposed quantum ballistic photodetector, the size of the Ge NW segment must scale with the electron coherence length. Very recently we have shown that, for ultrascaled Ge NWs integrated in back-gated FET devices, well-resolved conductance plateaus develop in the G(Vg) trace, attributed to the electrostatically modulated population of single spindegenerated 1D sub-bands.9 The transfer characteristic of such a device with a physical channel length of LGe = 18 nm is depicted in Figure 3a (black curve). Clear conductance quantization steps are visible with each step corresponding to a quantum unit of conductance G0 = 2e2/h.25 In addition, distinct conductance anomalies are obtained at sub-G0 values in agreement with previous studies on ballistic transport in semiconductor NWs.26 Under laser illumination (red curve) the photogating effect causes a shift of the transistor threshold voltage toward higher values (compare Figure 1c), without affecting the fundamental step height of the plateaus obtained due to conductance quantization. It is emphasized that such a voltage shift is a unique feature of the photogating effect, in contrast to the photoconductive effect in which a constant current increase is expected.13 For the quantum ballistic device this implies that ID is increased under illumination by opening an additional transfer channel, which effectively allows for



METHODS Device Fabrication. Vapor−liquid−solid (VLS) grown Ge NWs were uniformly coated with 20 nm Al2O3 in an ALD process and deposited on a SiO2 (100 nm) coated highly doped Si substrate, which serves as the common back-gate. Contact paths to individual NWs were fabricated by electron beam lithography, 23 s buffered HF (7:1) dip, 5 s HI (14%) dip, and 100 nm Al sputtering and following lift-off process. 5033

DOI: 10.1021/acs.nanolett.8b01845 Nano Lett. 2018, 18, 5030−5035

Nano Letters



Annealing at 350 °C in forming gas atmosphere for several minutes induces a material exchange process between Al and Ge allowing for control of the actual Ge channel length down to only several nanometers. More details of this process, including TEM and high-resolution energy-dispersive X-ray characterization, are given elsewhere.28 Electrical/Optical Measurements. The laser beam of an NKT SuperK Extreme laser source was guided through polarization maintaining optical fibers and an optical lens to individual NW devices, such that the electrical field was oriented parallel to the NW axis at a fwhm of ∼70 μm. The overall laser power was measured beforehand with a calibrated photodiode and was subsequently reduced with optical filters. Incident light was chopped at a certain frequency (mostly 127 Hz) by the use of a SuperK COMMAND device, and the current through the device under test was amplified using a Keithley 428 current amplifier. The measured current difference was evaluated by an EG&G 7265 Lock-In Amplifier using the modulation frequency as a reference signal. Calculation of Irradiation, Absorption, and Photoconductive Gain. Irradiance (EL) was calculated by numerically integrating a 2D Gauss function (fwhm = 70 μm) over the physical Ge NW channel area as seen from the top (LGe · d), dividing by this area and multiplying by the overall measured laser power. The absorption (η) was estimated by using the Beer−Lambert law for a circular cross section with an absorption coefficient of 5.6 × 105 cm−1, which roughly gives 70% to 80% for the investigated devices depending on diameter. The photoconductive gain calculates to Gph = Fel/ Fph, where Fel = Iph/q is the number of extracted electrons per unit time and Fph = EL · LGe · d · η/(h · ν) is the number of absorbed photons per unit time.



Letter

REFERENCES

(1) Heyns, M.; et al. Advancing CMOS beyond the Si roadmap with Ge and III/V devices. Technol. Dig. - Int. Electron Devices Meet. IEDM 2011, 13.1.1. (2) Assefa, S.; Xia, F.; Vlasov, Y. A. Reinventing germanium avalanche photodetector for nanophotonic on-chip optical interconnects. Nature 2010, 464, 80−84. (3) Wirths, S.; et al. Lasing in direct-bandgap GeSn alloy grown on Si. Nat. Photonics 2015, 9, 88−92. (4) Kim, C. J.; Lee, H. S.; Cho, Y. J.; Kang, K.; Jo, M. H. DiameterDependent Internal Gain in Ohmic Ge Nanowire Photodetectors. Nano Lett. 2010, 10, 2043−2048. (5) Lee, H. S.; et al. Large Electroabsorption Susceptibility Mediated by Internal Photoconductive Gain in Ge Nanowires. Nano Lett. 2012, 12, 5913−5918. (6) Otuonye, U.; Kim, H. W.; Lu, W. D. Ge nanowire photodetector with high photoconductive gain epitaxially integrated on Si substrate. Appl. Phys. Lett. 2017, 110, 173104. (7) Koppens, F. H.; et al. Photodetectors based on graphene, other two-dimensional materials and hybrid systems. Nat. Nanotechnol. 2014, 9, 780−793. (8) Brunbauer, F. M.; Bertagnolli, E.; Lugstein, A. Gate-Tunable Electron Transport Phenomena in Al−Ge−Al Nanowire Heterostructures. Nano Lett. 2015, 15, 7514−7518. (9) Sistani, M.; et al. Roomerature Quantum Ballistic Transport in Monolithic Ultrascaled Al-Ge-Al Nanowire Heterostructures. Nano Lett. 2017, 17, 4556−4561. (10) Zhang, L.; et al. Selective Passivation of GeO2/Ge Interface Defects in Atomic Layer Deposited High- k MOS Structures. ACS Appl. Mater. Interfaces 2015, 7, 20499−20506. (11) Kingston, R. H. Review of germanium surface phenomena. J. Appl. Phys. 1956, 27, 101−114. (12) Thunich, S.; et al. Photocurrent and photoconductance properties of a GaAs nanowire. Appl. Phys. Lett. 2009, 95, 83111. (13) Furchi, M. M.; Polyushkin, D. K.; Pospischil, A.; Mueller, T. Mechanisms of Photoconductivity in Atomically Thin MoS2. Nano Lett. 2014, 14, 6165−6170. (14) Guo, Q.; et al. Black Phosphorus Mid-Infrared Photodetectors with High Gain. Nano Lett. 2016, 16, 4648−4655. (15) Soci, C.; et al. ZnO Nanowire UV Photodetectors with High Internal Gain. Nano Lett. 2007, 7, 1003−1009. (16) Konstantatos, G.; Sargent, E. H. PbS colloidal quantum dot photoconductive photodetectors: Transport, traps, and gain. Appl. Phys. Lett. 2007, 91, 173505. (17) Tsipas, P.; Dimoulas, A. Modeling of negatively charged states at the Ge surface and interfaces. Appl. Phys. Lett. 2009, 94, 12114. (18) Hanrath, T.; Korgel, B. A. Influence of Surface States on Electron Transport through Intrinsic Ge Nanowires. J. Phys. Chem. B 2005, 109, 5518−5524. (19) Zhang, S.; et al. Relative Influence of Surface States and Bulk Impurities on the Electrical Properties of Ge Nanowires. Nano Lett. 2009, 9, 3268−3274. (20) Kingston, R. H.; McWhorter, A. L. Relaxation Time of Surface States on Germanium. Phys. Rev. 1956, 103, 534−540. (21) Ahn, Y. H.; Park, J. Efficient visible light detection using individual germanium nanowire field effect transistors. Appl. Phys. Lett. 2007, 91, 162102. (22) Cao, L.; et al. Engineering light absorption in semiconductor nanowire devices. Nat. Mater. 2009, 8, 643−647. (23) Zhang, W.; et al. Ultrahigh-Gain Photodetectors Based on Atomically Thin Graphene-MoS2 Heterostructures. Sci. Rep. 2015, 4, 1−8. (24) Soci, C.; et al. Nanowire Photodetectors. J. Nanosci. Nanotechnol. 2010, 10, 1430−1449. (25) Chuang, S.; et al. Ballistic InAs Nanowire Transistors. Nano Lett. 2013, 13, 555−558. (26) Micolich, A. P. What lurks below the last plateau: experimental studies of the 0.7 × 2 e 2/h conductance anomaly in one-dimensional systems. J. Phys.: Condens. Matter 2011, 23, 443201.

AUTHOR INFORMATION

Corresponding Author

*E-mail: [email protected]. ORCID

Philipp Staudinger: 0000-0003-3377-8575 Masiar Sistani: 0000-0001-5730-234X Alois Lugstein: 0000-0001-5693-4775 Present Address †

IBM Research - Zürich, Säumerstrasse 4, 8803 Rüschlikon, Switzerland. Author Contributions

P.S., M.S., and A.L. conceived and designed the experiments. P.S. and M.S. fabricated the nanostructures, carried out the optical measurements, and prepared the manuscript. J.G. contributed to the design of the experimental setup. All authors analyzed and discussed the results extensively. Notes

The authors declare no competing financial interest.



ACKNOWLEDGMENTS The authors gratefully acknowledge financial support by the Austrian Science Fund (FWF) Project No. P28175-N27. The authors thank the Center for Micro- and Nanostructures (ZMNS) for providing the cleanroom facilities and Martien den Hertog and Loung Minh Anh from Univ. Grenoble Alpes, CEA, INAC, for TEM support. 5034

DOI: 10.1021/acs.nanolett.8b01845 Nano Lett. 2018, 18, 5030−5035

Letter

Nano Letters (27) Marmon, J. K.; Rai, S. C.; Wang, K.; Zhou, W.; Zhang, Y. LightEffect Transistor (LET) with Multiple Independent Gating Controls for Optical Logic Gates and Optical Amplification. Frontiers in Physics 2016, arXiv Prepr. arXiv1601.04748. (28) Kral, S.; et al. Abrupt Schottky Junctions in Al/Ge Nanowire Heterostructures. Nano Lett. 2015, 15, 4783−4787.

5035

DOI: 10.1021/acs.nanolett.8b01845 Nano Lett. 2018, 18, 5030−5035