NANO LETTERS
Relative Influence of Surface States and Bulk Impurities on the Electrical Properties of Ge Nanowires
2009 Vol. 9, No. 9 3268-3274
Shixiong Zhang, Eric R. Hemesath, Daniel E. Perea, Edy Wijaya, Jessica L. Lensch-Falk, and Lincoln J. Lauhon* Department of Materials Science and Engineering, Northwestern UniVersity, EVanston, Illinois 60208 Received May 14, 2009; Revised Manuscript Received July 3, 2009
ABSTRACT We quantitatively examine the relative influence of bulk impurities and surface states on the electrical properties of Ge nanowires with and without phosphorus (P) doping. The unintentional impurity concentration in nominally undoped Ge nanowires is less than 2 × 1017 cm-3 as determined by atom probe tomography. Surprisingly, P doping of ∼1018 cm-3 reduces the nanowire conductivity by 2 orders of magnitude. By modeling the contributions of dopants, impurities, and surface states, we confirm that the conductivity of nominally undoped Ge nanowires is mainly due to surface state induced hole accumulation rather than impurities introduced by catalyst. In P-doped nanowires, the surface states accept the electrons generated by the P dopants, reducing the conductivity and leading to ambipolar behavior. In contrast, intentional surface-doping results in a high conductivity and recovery of n-type characteristics.
One dimensional semiconductor nanowires are currently an intense subject of investigation in nanoscience because of their enabling potential in applications including nanoelectronics,1-6 optoelectronics,1,7 and chemical sensing.1 Catalytic chemical vapor deposition employing the vapor-liquidsolid (VLS) growth mechanism is routinely used to provide high-quality single crystal materials with controlled size and composition.8 While the incorporation of intentional impurities, or dopants, is exploited in numerous device applications,3,6 metal catalysts may introduce unintentional impurities into the nanowire, creating donor or acceptor levels that can adversely affect the electrical properties.9 A number of studies have explored the diffusion of Au catalyst atoms on or into Si nanowires,10-14 for example, but the influence of Au on Ge nanowires has not been determined. Because Ge has higher electron and hole mobilities than silicon, nanostructured Ge could exhibit superior performance in nanoelectronic applications and reveal more prominent quantum size effects.15 Indeed, high-performance field effect transistors (FET) have been demonstrated on p-type Ge nanowires (NWs)16,17 as well as Ge/i-Si core/shell NWs,18-20 and fully tunable double quantum dots have been demonstrated in Ge/Si heterostructure-based NWs.21,22 It has been shown, however, that surface states can strongly influence the electrical characteristics of nanowires.23-25 In the work of Hanrath et al.,25 nominally intrinsic Ge NWs grown by the Au nanocrystal-seeded supercritical * To whom correspondence should be addressed,
[email protected]. 10.1021/nl901548u CCC: $40.75 Published on Web 08/06/2009
2009 American Chemical Society
fluid-liquid-solid method exhibited p-type gate behavior that was attributed to surface hole accumulation due to trapped negative surface charge. The contribution of an acceptor impurity, such as Au, was not excluded. Wang et al.23 have demonstrated that p-type and n-type Ge nanowires grown by the VLS method have different oxidation routes leading to opposite band bending and an altered electrical response. The authors also claimed that the large hysteresis in the gate response is due to water molecules strongly bound to slow surface states in GeO2, complicating control of the Fermi level. Tutuc et al.26 concluded that Ge nanowires grown in the presence of significant phosphine partial pressures were insignificantly doped based on the ambipolar gate characteristics, though the influence of surface states was not a focus of the discussion. It is clear that a more quantitative understanding of the relative influence of surface states and bulk impurities on the electrical properties of Ge NWs is important from both a material and device design point of view, but prior studies have not established quantitative dopant concentrations that might be correlated with electrical properties. Recently, we determined the distribution of intentional and unintentional impurities in arbitrary regions of single Ge nanowires by atom probe tomography.27 This capability provides the opportunity to establish a more quantitative understanding of nanowire electrical properties. Here we report correlated electrical transport and atom probe tomography (APT)28 studies of undoped and P-doped
discussed below. The i-Ge nanowire conductivity is much greater than that of bulk intrinsic Ge (0.02 S cm-1), suggesting that the carrier concentration and/or mobility is enhanced in the nanowire geometry. This finding is consistent with the surface-induced hole accumulation reported previously.25 Surprisingly, the conductivity of the P:Ge nanowires is 2 orders of magnitude lower than that of the i-Ge nanowires, indicating that P doping reduces the carrier concentration and/or mobility. The conductivity of the P:Ge@Ge, however, is higher than that of i-Ge, indicating that surface doping with donors can increase the carrier concentration.
Figure 1. (a) Room temperature conductivities of five devices for each of three nanowire types: i-Ge, red circles; P:Ge, blue upward triangles; and P:Ge@Ge, green downward triangles. The solid lines show the average values: i-Ge, 2.5 S/cm; P:Ge, 0.04 S/cm; and P:Ge@Ge, 30 S/cm. (b) Conductivity versus 1/T for typical devices of each type. The solid curves are a fit to the Arrhenius relationship. The inset shows the activation energies obtained from the fits. The solid lines show the averaged values from five devices of each type.
Ge nanowires to quantify the relative influence of surface states and impurities on electrical conductivity. We confirm that the conductivity of undoped Ge nanowires is dominated by holes that accumulate in response to occupied surface states, and we rule out significant contributions from impurities such as Au and O. Interestingly, we find that phosphorus doping of ∼1018 cm-3 reduces the conductivity of Ge nanowires by compensating the surface hole accumulation, explaining prior reports of insignificant doping by the VLS process.26 We quantify the concentration of occupied surface states that are needed to provide a consistent explanation for the differences in conductivity between doped and undoped nanowires. Phosphorus doped and nominally undoped Ge nanowires were grown using Au nanoparticles in a VLS process using GeH4 and PH3 precursors. Details of the growth and device fabrication procedures are provided in the Supporting Information. The conductivities of nanowires were measured by the four probe method in which the current is sourced through two outer electrodes while a voltage drop is measured between two inner electrodes. Figure 1a compares the room temperature conductivities of three types of nanowires, i-Ge, P:Ge, and P-doped Ge with enhanced surface doping (P:Ge@Ge); the impurity levels will be Nano Lett., Vol. 9, No. 9, 2009
Temperature-dependent conductivity measurements were carried out on back-gated devices to explore variations in carrier concentration and mobility with temperature and doping. The conductivity of P:Ge@Ge is nearly independent of temperature from 80 to 385 K (Figure 1b), which is consistent with degenerate doping and a mobility limited by ionized impurity carrier scattering and/or surface scattering. The conductivities of the i-Ge and P:Ge NWs increase with increasing temperature, and thermally activated behavior is observed above ∼260 K. This activated behavior is attributed to the generation of carriers by impurities and/or defects because: (1) the intrinsic nanowires are not intentionally doped and (2) the mobility is unlikely to increase in this temperature range, as further justified below. The temperature-dependent conductivities from 10 nanowires (five of each type) were fit to an Arrhenius relationship σ ∼ exp
( ) -EA kBT
between 260 and 385 K to extract activation energies associated with carrier generation. The average activation energies were found to be EA ) 0.09 ( 0.01 eV for i-Ge NWs and EA ) 0.19 ( 0.01 eV for P:Ge NWs (Figure 1b, inset). For i:Ge, the experimentally measured activation energy EA may be compared with the ionization energies Ei of impurity levels that one might expect, such as Au acceptors (EA ) Ei,Au/2 ∼ 0.08 eV)29 from the catalyst or oxygen donors (EA ) Ei,O/2 ∼ 0.10 eV)29 from the growth environment. As discussed further below, however, Au and O are not present in sufficient concentrations to explain the activated conductivity. For P:Ge, the primary impurity is phosphorus, but the measured activation energy does not correspond to that expected for P donors (EP/2 ∼ 0.006 eV). We therefore consider whether surface states may lead to the thermal generation of carriers. Temperature-dependent transconductance measurements were made to determine the majority carrier type and look for variations in the field effect mobility with temperature (Figure 2). Similar behaviors were observed for at least 10 devices of each type. Nominally undoped (i-Ge) nanowires exhibit a p-type gate response at zero gate bias as found by other workers25 (Figure 2a). In general, P:Ge nanowires exhibited ambipolar behavior similar to that reported by Tutuc et al.26 (Figure 2b). We note that the current at positive gate bias in Figure 2b, which corresponds to accumulation in an n-type channel, is also influenced by the contact barriers. Significant hysteresis was noted in both i-Ge and 3269
Figure 2. Typical transfer characteristics with the drain source voltage Vds ) 0.5 V: (a) i-Ge; (b) P:Ge; the dashed lines are the tangents to the transconductance in the linear region. (c) Threshold voltages extrapolated from the gate transfer curves of i-Ge and P:Ge devices. The downward arrows correspond to gate voltage sweeps from positive to negative (backward), and the upward arrows correspond to gate voltage sweeps from negative to positive (forward). The average threshold voltages for i-Ge are 6.6 and 0.27 V for the backward and forward directions, respectively. The average threshold voltages for P:Ge are -2.0 and -7.8 V for the backward and forward directions, respectively. (d) Typical transfer characteristics of P:Ge@Ge with Vds ) 0.5 V. (e) Field effect mobility of i-Ge and P:Ge as a function of temperature.
P:Ge devices, consistent with a large number of unpassivated surface states as discussed in detail below. The average threshold voltage of the P:Ge devices was significantly lower than that of the i-Ge devices (Figure 2c), indicating that P dopants are acting as donors. Clear n-type behavior was only observed for Ge nanowires that had a heavily doped shell (P:Ge@Ge, Figure 2d). Typical room temperature field effect mobilities for unpassivated i-Ge nanowires, without correcting for the series contact resistance, were ∼15 cm2 V-1 s-1, while those for P:Ge were ∼1 cm2 V-1 s-1. These values did not vary with temperature over the range used to extract the activation energy (Figure 2e). It should be noted that these values represent a lower bound for the mobilities because (1) the voltage drop at the contact may be limiting the transconductance, (2) the capacitance is estimated by using a cylinder-on-plane model (see Supporting Information), which assumes a uniform dielectric, and (3) the model neglects the depletion capacitance, which may be significant for moderate carrier concentrations such as those discussed here. The degenerate doping of the P:Ge@Ge nanowires precluded an estimation of the field effect mobility. 3270
Figure 3. 3-D reconstruction of a portion of a P-doped Ge nanowire viewed parallel (a) and perpendicular (b) to the growth direction: Ge (green), P (gray spheres); reconstruction dimensions, 32 × 32 × 100 nm3. (c) [111] atomic planes visible in the center of the reconstruction: dimensions, 3.3 × 2.6 nm2.
Table 1. Concentrations of Bulk Impurities in i-Ge and P:Ge NWs As Determined by Pulsed Laser Atom Probe Tomographya P:Ge nanowires
i-Ge
core
shell
Au concn (1017 cm-3)
