LETTER pubs.acs.org/NanoLett
Direct Measurement of Individual Deep Traps in Single Silicon Nanowires E. Koren,† G. Elias,† A. Boag,† E. R. Hemesath,‡ L. J. Lauhon,‡ and Y. Rosenwaks*,† † ‡
Department of Physical Electronics, School of Electrical Engineering, Tel-Aviv University, Tel-Aviv 69978, Israel Department of Materials Science and Engineering, Northwestern University, Evanston, Illinois 60208, United States ABSTRACT: The potential of the metal nanocatalyst to contaminate vaporliquid solid (VLS) grown semiconductor nanowires has been a long-standing concern, since the most common catalyst material, Au, is known to induce deep gap states in several semiconductors. Here we use Kelvin probe force microscopy to image individual deep acceptor type trapping centers in single undoped Si nanowires grown with an Au catalyst. The switching between occupied and empty trap states is reversibly controlled by the backgate potential in a nanowire transistor. The trap energy level, i.e., EC ET = 0.65 ( 0.1 eV was extracted and the concentration was estimated to be ∼2 1016 cm3. The energy and concentration are consistent with traps resulting from the unintentional incorporation of Au atoms during the VLS growth. KEYWORDS: Silicon, nanowire, deep traps, Au atoms, KPFM
Si nanowires are commonly grown by the vaporliquidsolid (VLS) mechanism in which a metal catalyst forms eutectic droplets at the growing tips of the nanowires.16 Au has been favored as a catalytic material because of the deep eutectic reaction with Si (Au82Si18 at 360 °C), which enables low temperature growth of Si nanowires (460650 °C in most cases).6,7 In addition, both the low equilibrium solubility (∼2 1015 Au atoms/cm3 in Si at 650 °C)8 and the absence of stable silicides in the AuSi phase diagram should permit the growth of Si nanowires with relatively low impurity concentrations. However, high-angle annular dark-field scanning transmission electron microscopy (HAADF)9,10 measurements recently revealed higher numbers of Au atoms than expected from a simple extrapolation of the bulk solubility to the low growth temperature. Since Au is known to introduce both donor and acceptor deep traps in Si (with energies of 0.35 and 0.55 eV above the valence-band, respectively)8 which affect the electrical and optical properties,11,12 many groups have been motivated to work on alternative catalysts and growth techniques.1319 Consequently, it is very important to establish the influence of Au impurities on nanowire electronic properties.9 We use here Kelvin probe force microscopy (KPFM)2023 to image individual traps in undoped Si nanowires grown by the VLS method. The measured trapping centers show acceptor like behavior, i.e., they are negatively charged below the Fermi level and neutral above it, where the switching between occupied and empty trap states is reversibly controlled by changing the backgate potential (VBG) of intrinsic nanowire transistors. The energy level of these traps was directly extracted from quantitative measurements of fluctuations in the surface potential. Undoped Si nanowires were synthesized via low-pressure chemical vapor deposition at 460 °C with solution-deposited 50 nm Au catalyst at a total pressure of 40 Torr and flows of 30, r 2011 American Chemical Society
20, and 2 sccm of N2, He, and SiH4, respectively. For device fabrication, nanowires were suspended in isopropyl alcohol by sonication and drop-cast onto lightly p-doped Si substrates with a 10 nm of Si3N4 overlayer. Contact regions were defined by electron beam lithography using a PMMA resist. Thirty second, 50 W oxygen plasma cleaning of the electrode regions was followed by a 3 s etch in buffered hydrofluoric acid, after which the substrates were immediately loaded into an electron beam evaporator for Ni contact evaporation. The KPFM measurements were conducted using a Dimension 3100 (Veeco, Inc.) atomic force microscope system in a controlled nitrogen environment glovebox (less than ∼5 ppm H2O). Numerical simulations where performed using a Sentaurus TCAD device simulator (Synopsys, Inc.), to obtain the nanowire surface potential. This three-dimensional (3D) finite element simulator solves the Poisson equation coupled with the continuity equation for electrons and holes for the actual device geometry. Figure 1 shows the surface potential profiles measured along a single Si nanowire under different applied back gate potentials. Both the source and the drain contacts were grounded throughout the measurement to allow the injection of either holes or electrons into the wire. The potential distribution over the nanowire surface remains relatively uniform under negative back gate potentials while positive applied potentials lead to surface potential fluctuations, particularly negative “dips”. The appearance of negative dips as opposed to positive peaks is supported both by the surface potential images (Figure 1) and by the potential simulation of acceptor type traps, as will be addressed below. The images presented in Figure 1 are the 2D surface potential Received: March 27, 2011 Revised: May 2, 2011 Published: May 17, 2011 2499
dx.doi.org/10.1021/nl201019b | Nano Lett. 2011, 11, 2499–2502
Nano Letters
LETTER
Figure 1. Surface potential measurements of an intrinsic VLS grown Si nanowire under different back gate potentials. The upper and lower images are 2D potential distribution maps of the device under back gate potentials of 1 V (occupied state) and 1 V (unoccupied state), respectively. The scale is 200 mV.
distributions for negative (1 V, bottom) and positive (1 V, top) back gate potentials. When the back gate potential is positive (negative), the Fermi level energy increases (decreases), due to injection of electrons (holes) into the nanowire. Consequently, both acceptor and donor type traps can be detected by the KPFM in their charged state. An occupied acceptor (donor) state with a localized negative (positive) charge will lead to a dip (peak) in the surface potential with a height and width that depend on the electrostatic screening in the Si nanowire. Therefore, the energy level of a trap can be extracted directly from the voltage dependence of the onset of such potential fluctuations. The fact that potential fluctuations are only observed when applying positive back gate potentials suggests that we are observing active acceptortype states as opposed to donor-type traps. However, the presence of a high density of surface states close to the valence band could obscure bulk donor-like traps. We have calculated the effect of a single acceptor trap inside the nanowire for different energy levels and for several distances from the nanowire surface. A single trapping center was simulated in the nanowire by defining a 1 nm3 volume unit containing 1 1021 traps/cm3, i.e., an occupied trap of one elementary charge. Panels a and b of Figure 2 show the calculated potential distribution for a single occupied acceptor trap (VBG = 0.6 V) located 1 nm below the nanowire surface at an energy of EC ET = 0.65 eV. Figure 2c shows the evolution of the calculated surface potential dip as a function of the applied back gate bias. The potential dip becomes visible at VBG ∼ 0.1 V (when the Fermilevel approaches the trap energy level) and saturates at VBG ∼ 0.6 V. The actual dip height for a given VBG depends on the electrostatic screening inside the nanowire, which is affected by the dopant concentration, charge carriers induced by the back-gate voltage, surface charge, and the trap distance from the nanowire surface. When the trap is located deeper inside the nanowire, the
Figure 2. 3D (a) and 2D (b) views of potential simulation of a single occupied trap (VBG = 0.6 V) located 1 nm from the nanowire surface with energy of EC ET = 0.65 eV (scale is 130 mV). (c) Simulated potential profiles along the nanowire showing the evolution of the potential dip over the nanowire surface as the back gate potential is increased. (d) Simulated potential profiles for different depths of traps below the upper surface of the nanowire at VBG= 0.6 V. The dip magnitude decreases exponentially as a function of the trap depth (inset graph). The shaded area of the inset indicates the region of the nanowire in which traps can be detected by the KPFM.
magnitude of the potential dip decreases exponentially due to this screening, eventually eliminating the possibility of detecting it with KPFM (Figure 2d). Therefore, we attribute the measured dips to states that are located within ∼5 nm of the nanowire surface (shaded region of Figure 2d inset). The concentration was estimated to be ∼2 1016 cm3, by simply dividing the number of dips along the nanowire by the top shell volume (assuming that each dip is related to a single trap and traps which are located more than 5 nm from the upper surface of the nanowire are not imaged by the KPFM). In order to extract the energy level of the traps, it is necessary to consider the difference between the measured and the simulated flatband potentials of the nanowire device; we expect the device to have a nonzero trapped charge density in the oxide that is not included in the simulation, leading to an offset. The experimental flatband potential was determined by inspection of 2500
dx.doi.org/10.1021/nl201019b |Nano Lett. 2011, 11, 2499–2502
Nano Letters
LETTER
Figure 4. (a) Potential dip magnitude vs VBG both for the measured device (squares) and for simulations (lines) of traps located 0.55, 0.65, and 0.75 eV below the conduction band minimum. The inset picture is a schematic illustration of the energy level of the acceptor type trap relative to the Si band gap. (b) Scaling of a trap potential dip height due to longrange electrostatic forces acting on the probe during the measurements. Scaling is calculated for different tipsample distances, under different back-gate potentials, for a cantilever width of 30 μm, a cantilever length of 130 μm, a bottom sphere radius of 30 nm, and a cone half opening angle of 17.5°. The probe had a tilting angle of 10° relative to the surface.
Figure 3. Measured (a) and simulated (b) surface potential profiles across the nanowire (the profiles are offset (Y axis) to allow easier comparison (lower scale)). (c) The charge carrier concentration as a function of the back gate potential was used to extract the flatband of the simulated device as ∼0.3 V (from the point at which the concentration for holes is equal to the concentration of electrons).
the potential line scans at the nanowire sides, under different applied back-gate potentials (Figure 3a,b). The potential lines are offset (Y axis) to allow easier comparison (lower scale). The surface potential at the nanowire sides is flat (flatband point) at VBG = 0.2 ( 0.1 V for the measurement (Figure 3a) and at VBG = 0.3 V for the simulation (Figure 3b). We note that for the simulation, the flatband potential can be determined with higher accuracy from Figure 3c, from the point where the concentration of electrons and holes is equal, i.e., VBG = 0.3 V. The 0.1 V
difference between the measured and the simulated flatband potentials was accounted for in the extraction of the trap energy level by shifting the threshold voltage at which the dips appear by 0.1 V relative to the simulation data (i.e., shifting the measured profile in Figure 4a 0.1 V to the right). Figure 4a presents the magnitude of the potential dips versus VBG both for the measured device (symbols) and for the simulations (solid lines) of traps located 0.55, 0.65, and 0.75 eV below the conduction band minimum. The experimental data points are each averages of all the potential dips for a given back-gate potential (Figure 1); such results were consistent for more than 10 nanowire devices. The calculated profile for a trap located at EC ET = 0.65 eV is in good agreement with the measured profile. This energy is similar to what is expected for Au atoms incorporated during VLS growth.8 The measured potential in Figure 4a is scaled by a factor of 22 to obtain the observed fit with the simulations. This scaling is due to the convolution24 effect of the measuring KPFM probe (tip þ cantilever), but it does not affect the onset of the curves in Figure 4a, which is the parameter sensitive to the energy level of the trap. The convolution is a result of the long-range 2501
dx.doi.org/10.1021/nl201019b |Nano Lett. 2011, 11, 2499–2502
Nano Letters electrostatic forces acting on the probe that cause the measurement to be a weighted average of the surface potential. As a result, the magnitude of a local potential perturbation surrounded by a larger region of homogeneous potential must be scaled.25 This is the case encountered in the measurement of a nanowire and a trap; due to their different size, their potential will be scaled by a different factor (compared to their real value). The convolution was implemented using our previously described model,26 which calculates the effect of the long-range electrostatic forces on the measured KPFM image. Unlike earlier approaches,27,28 the simulated probe geometry included the entire cantilever and the probe tilting angle relative to the surface corresponding to the experimental conditions. The convolution was performed on the calculated surface potential profile of a single trap. Figure 4b shows the scaling of the potential dip height following the convolution relative to the calculated height for different tip sample distances under different back-gate potentials. It is observed that for probesample distances between 15 and 25 nm, which corresponds to the experimental distance, the scaling varies between 20 and 28 for all back-gate potentials. The ratio of 22 between the calculated and the measured dips in the fit of Figure 4a is therefore within the expected range. In summary, we have used KPFM to image individual deep acceptor type trapping centers in undoped Si nanowires grown with an Au catalyst. The energy level of the traps (EC ET = 0.65 ( 0.1 eV), which was extracted from the onset point of the measured surface potential dips, is in good agreement with the predicted value for Au in Si, and the estimated concentration is consistent with recent experimental observations.9,10 We note, however, that if Au is indeed responsible for the observed traps, it could result from direct incorporation from the catalyst or the inward diffusion of surface Au atoms.
’ AUTHOR INFORMATION
LETTER
(10) Oh, S. H.; van Benthem, K.; Molina, S. I.; Borisevich, A. Y.; Luo, W. D.; Werner, P.; Zakharov, N. D.; Kurnar, D.; Pantelides, S. T.; Pennycook, S. J. Nano Lett. 2008, 8 (4), 1016–1019. (11) Queisser, H. J. Solid-State Electron. 1978, 21 (111), 1495–1503. (12) Grimmeiss, H. G. Annu. Rev. Mater. Sci. 1977, 7, 341–376. (13) Wagner, R. S.; Ellis, W. C. Trans. Metall. Soc. AIME 1965, 233 (6), 1053. (14) Nebol’sin, V. A.; Shchetinin, A. A. Inorg. Mater. 2003, 39 (9), 899–903. (15) Miyamoto, Y.; Hirata, M. Jpn. J. Appl. Phys. 1976, 15 (6), 1159–1160. (16) Wagner, R. S.; Ellis, W. C.; Jackson, K. A.; Arnold, S. M. J. Appl. Phys. 1964, 35 (10), 2993–3000. (17) Wang, N.; Tang, Y. H.; Zhang, Y. F.; Yu, D. P.; Lee, C. S.; Bello, I.; Lee, S. T. Chem. Phys. Lett. 1998, 283 (56), 368–372. (18) Schmidt, V.; Senz, S.; Gosele, U. Z. Metallkd. 2005, 96 (5), 427–428. (19) Morales, A. M.; Lieber, C. M. Science 1998, 279 (5348), 208–211. (20) Martin, Y.; Abraham, D. W.; Wickramasinghe, H. K. Appl. Phys. Lett. 1988, 52 (13), 1103–1105. (21) Koren, E.; Rosenwaks, Y.; Allen, J. E.; Hemesath, E. R.; Lauhon, L. J. Appl. Phys. Lett. 2009, 95 (9), 092105. (22) Koren, E.; Berkovitch, N.; Rosenwaks, Y. Nano Lett. 2010, 10 (4), 1163–1167. (23) Koren, E.; Hyun, J. K.; Givan, U.; Hemesath, E. R.; Lauhon, L. J.; Rosenwaks, Y. Nano Lett. 2011, 11 (1), 183–187. (24) Jacobs, H. O.; Leuchtmann, P.; Homan, O. J.; Stemmer, A. J. Appl. Phys. 1998, 84 (3), 1168–1173. (25) Koley, G.; Spencer, M. G.; Bhangale, H. R. Appl. Phys. Lett. 2001, 79 (4), 545–547. (26) Strassburg, E.; Boag, A.; Rosenwaks, Y. Rev. Sci. Instrum. 2005, 76, (8). (27) Rosenwaks, Y.; Shikler, R.; Glatzel, T.; Sadewasser, S. Phys. Rev. B 2004, 70 (8), 085320. (28) Hudlet, S.; Saint Jean, M.; Guthmann, C.; Berger, J. Eur. Phys. J. B 1998, 2 (1), 5–10.
Corresponding Author
*E-mail:
[email protected].
’ ACKNOWLEDGMENT This research was generously supported by Grant No. 2008140 from the United StatesIsrael Binational Science Foundation [BSF] and the National Science Foundation Grant DMR1006069. ’ REFERENCES (1) Wagner, R. S.; Ellis, W. C. Appl. Phys. Lett. 1964, 4 (5), 89–90. (2) Givargizov, E. I. J. Cryst. Growth 1975, 31 (Dec), 20–30. (3) Ross, F. M.; Tersoff, J.; Reuter, M. C. Phys. Rev. Lett. 2005, 95, 146104. (4) Werner, P.; Zakharov, N. D.; Gerth, G.; Schubert, L.; Gosele, U. Int. J. Mater. Res. 2006, 97 (7), 1008–1015. (5) Kodambaka, S.; Hannon, J. B.; Tromp, R. M.; Ross, F. M. Nano Lett. 2006, 6 (6), 1292–1296. (6) Schmidt, V.; Wittemann, J. V.; Senz, S.; Gosele, U. Adv. Mater. 2009, 21 (2526), 2681–2702. (7) Cao, L. Y.; Garipcan, B.; Atchison, J. S.; Ni, C. Y.; Nabet, B.; Spanier, J. E. Nano Lett. 2006, 6 (9), 1852–1857. (8) Collins, C. B.; Carlson, R. O.; Gallagher, C. J. Phys. Rev. 1957, 105 (4), 1168. (9) Allen, J. E.; Hemesath, E. R.; Perea, D. E.; Lensch-Falk, J. L.; Li, Z. Y.; Yin, F.; Gass, M. H.; Wang, P.; Bleloch, A. L.; Palmer, R. E.; Lauhon, L. J. Nat. Nanotechnol. 2008, 3 (3), 168–173. 2502
dx.doi.org/10.1021/nl201019b |Nano Lett. 2011, 11, 2499–2502
