Home
Search
Collections
Journals
About
Contact us
My IOPscience
Conductivity control of as-grown branched indium tin oxide nanowire networks
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Nanotechnology 25 035701 (http://iopscience.iop.org/0957-4484/25/3/035701) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 130.108.121.217 This content was downloaded on 07/09/2014 at 11:45
Please note that terms and conditions apply.
Nanotechnology Nanotechnology 25 (2014) 035701 (9pp)
doi:10.1088/0957-4484/25/3/035701
Conductivity control of as-grown branched indium tin oxide nanowire networks J M LaForge1 , T L Cocker2,4 , A L Beaudry1 , K Cui3 , R T Tucker1 , M T Taschuk1 , F A Hegmann2 and M J Brett1,3 1 2 3
Department of Electrical and Computer Engineering, University of Alberta, Edmonton, AB, Canada Department of Physics, University of Alberta, Edmonton, AB, Canada NRC National Institute for Nanotechnology, Edmonton, AB, T6G 2M9, Canada
E-mail: [email protected] Received 3 September 2013, in final form 4 November 2013 Published 17 December 2013 Abstract
Branched indium tin oxide (ITO) nanowire networks are promising candidates for transparent conductive oxide applications, such as optoelectronic electrodes, due to their high porosity. However, these branched networks also present new challenges in assessing conductivity. Conventional four-point probe techniques cannot separate the effect of porosity on the long-range conductivity from the intrinsic material conductivity. Here we compare the average nanoscale conductivity within the film measured by terahertz time-domain spectroscopy (THz-TDS) to the film conductivity measured by four-point probe in a branched ITO nanowire network. Both techniques report conductivity increases with deposition flux rate from 0.5 to 3.0 nm s−1 , achieving a maximum of ∼10 ( cm)−1 . Modeling the THz-TDS conductivity data using the Drude–Smith model allows us to distinguish between conductivity increases resulting from morphological changes and those resulting from the intrinsic properties of the ITO. In particular, the intrinsic material conductivity within the nanowires can be extracted, and is found to reach a maximum of ∼3000 ( cm)−1 , comparable to bulk ITO. To determine the mechanism responsible for increasing conductivity with flux rate, we characterize dopant concentration and morphological changes (i.e., to branching behavior, nanowire diameter and nucleation layers). We propose that changes in the electron density, primarily due to changes in O-vacancy concentration at different flux rates, are responsible for the observed conductivity increase. This understanding will assist balancing structural and conductivity requirements in applications of transparent conductive oxide networks. Keywords: branched nanowires, nanowires, indium tin oxide, THz time-domain spectroscopy, conductivity (Some figures may appear in colour only in the online journal)
1. Introduction
candidate for initial studies of TCO networks. Several groups have fabricated different ITO architectures, such as nanotrees [12], nanopillars [13] and nanowires [14–16]. However, simply ensuring the correct morphology of TCO networks is insufficient for successful application, which also requires that TCO materials have suitable conductivity and transmissivity. In bulk-like films, transmission spectra and in-plane conductivity measured via four-point probe can readily assess the optical and electrical performance. However, TCO branched nanowire networks have reduced or zero
Nanostructured transparent conductive oxide (TCO) networks have applications as flexible conductive substrates and as charge collectors in organic photovoltaics [1–8]. Indium tin oxide (ITO) is a well-understood TCO used in many thin film applications due to its high optical transmittance and electrical conductivity [9–11], and thus ITO is a strong 4 Present address: Department of Physics, University of Regensburg,
Regensburg, Germany. 0957-4484/14/035701+09$33.00
1
c 2014 IOP Publishing Ltd Printed in the UK
Nanotechnology 25 (2014) 035701
J M LaForge et al
2. Experiment
in-plane conductivity, dependent on the morphology of the interconnections. Therefore, four-point probe cannot reliably measure the material’s intrinsic conductivity. Instead, these techniques measure long-range conductivity, or film conductivity, dependent on both the intrinsic material conductivity and the morphology of the material within the film (i.e. electrical connectivity). To address this, nanowire electrical properties have been characterized by placing individual wires across lithographically patterned contacts [17–19]. Alternatively, in situ direct contact techniques, such as scanning electron microscope (SEM) nanoprobe [20], allow for non-destructive measurements of nanowires with suitable spacing and mechanical robustness. While these techniques provide direct measurements of individual nanowires, scaling up to characterize a statistically significant ensemble of nanowires is difficult. If the aggregate electrical properties of larger areas are important, optical techniques, such as terahertz time-domain spectroscopy (THz-TDS) [21–23] can non-destructively measure the average nanoscale conductivity of >106 structures simultaneously, making these techniques attractive for non-contact characterization of TCO networks. Terahertz spectroscopy has been used to study a wide range of nanostructures in recent years [21–23], including nanowires [24–31]. Recently, we have used glancing angle deposition (GLAD) to provide a directional vapor flux during the self-catalyzed vapor–liquid–solid (VLS) growth of branched ITO nanowires. This technique, named VLS-GLAD, allows for improved control over several morphological properties such as nanowire diameter, branch orientation and height placement [12, 32, 33]. We have fabricated nanowire networks with VLS-GLAD that are well suited for three-dimensional electrodes (e.g. for an organic photovoltaic device). The utility of THz-TDS for characterizing the conductive properties of both bulk ITO [34, 35] and VLS-grown ITO nanocolumns [30, 31] has been demonstrated. In one of the latter reports [30], the films were 20 nm morphological features observed in figure 1(a). We took four-point-probe measurements of the film conductivity (across a distance of ∼1 mm) in both linear probe and van der Pauw configurations [44]. These data are compared to the film conductivity observed with THz-TDS by taking the dc extrapolation of the Drude–Smith conductivity fits (σ˜ film (ω = 0)), as presented in figure 2(a), which shows agreement between the trends of the four-point probe and
c-parameter −0.985 ± 0.007 −0.95 ± 0.1 −0.970 ± 0.006 −0.972 ± 0.005 −0.94 ± 0.01 −0.94 ± 0.01
THz film conductivities, especially at low and high flux rates. The magnitudes of the dc conductivities measured by the two techniques are generally close, though the consistently smaller film conductivities measured by four-point probe suggest that more complicated conduction pathways through the nanowire networks are sampled by the four-point-probe measurements. Regardless, the agreement between the conductivity trends in figure 2(a) implies that morphological changes in the films have limited impact on the conduction pathways measured by four-point probe. A notable deviation occurs for samples at 1.0 and 1.5 nm s−1 , which may point to morphological changes playing a role in the suppression of conductivity over long ranges (∼1 mm). This point will be discussed in more detail below. The four-point-probe film conductivities were calculated from the sheet resistances (figure 2(b)) and total film thicknesses measured by SEM. Several morphological changes occur with the increase in flux rate. Nanowire diameter measured at the film base increases with flux rate, as shown in figure 2(c). However, the diameter at the film’s top does not appear to change significantly (figure 2(c)). Thus, the volume averaged nanowire diameter increases with flux rate somewhere between these two trends. Additionally, the network’s effective mass density (ρ/ρbulk ) increases slightly with flux rate, as seen in figure 2(d). In the Drude–Smith fits to the THz film conductivity, the electron density averaged across the THz interaction volume nfilm is used as a fit parameter (figure 1(e)). This average electron density for the film can be used to estimate the electron density within the nanowires nnw via the filling fraction (mass density) of the nanowire film network, nnw = nfilm (ρbulk /ρ) [38, 39]. The nanowire electron density is shown as a function of deposition flux rate in figure 1(e) 4
Nanotechnology 25 (2014) 035701
J M LaForge et al
σ˜ nw (ω) = σ˜ film (ω)(ρbulk /ρ). Equivalently, nnw e2 τDS 1 c 1+ . (2) σ˜ nw (ω) = m∗ 1 − iωτDS 1 − iωτDS The dc extrapolation of σ˜ nw (ω) is the average dc conductivity of an individual nanowire. Therefore, changes to nanowire conductivity with deposition flux rate can be separated from changes to the mass density of the nanowire network. Comparison of the dc film conductivities found by THz-TDS (σ˜ film (0)) and the dc nanowire conductivities (σ˜ nw (0)) in figure 2(a) reveals nearly identical trends with deposition flux rate, indicating that the strong conductivity increase with flux rate is not due to an increase in mass density. Within the Drude–Smith model, localization effects are primarily fit using the c-parameter (though the scattering time also contains some nanoscale information through Matthiessen’s rule [37–40]). The c-parameters in our Drude–Smith fits describe confinement of charge carriers inside nanowires on a scale that is smaller than the nanowire diameter. Therefore, this confinement is still present in the nanowire conductivities σ˜ nw (ω). Conversely, a new conductivity, σITO , can be defined that corresponds to the intrinsic dc conductivity of the Drude gas of carriers that experiences the localization inside the nanowires. Within the framework of the Drude–Smith model, the intrinsic dc conductivity of the carriers in the ITO nanostructures in the absence of structural confinement is σITO =
σ˜ nw (0) , 1+c
(3)
or equivalently, nnw e2 τDS , (4) m∗ where we have assumed that the increase in scattering time due to the removal of boundary scattering constitutes a small correction to the intrinsic dc conductivity. Accounting for this would lead to an increase of σITO by a few per cent. As can be seen in figure 2(a), the intrinsic conductivity of the ITO nanowire networks reaches a maximum of 3300 ± 400 ( cm)−1 , comparable to values of ∼104 ( cm)−1 measured in a different study via four-point probe along a single ITO nanowire at room temperature with similar characteristics to ours (d ∼ 100 nm, n ∼ 1020 cm−3 and Fermi wavenumber kF ∼ 109 m−1 ) [17]. Based on the large nanowire conductivity measured in [17], the type of structural disorder that causes carrier localization in our nanowires appears to be absent in their samples. The nanowires in [17] were grown at 850–900 ◦ C, which should lead to fewer growth defects. Note that post-growth thermal annealing is known to improve the conductivity of our ITO nanotrees by a factor of 10–100 [12]. The large difference we observe between σ˜ nw (0) and σITO therefore suggests a possible explanation for the annealing-induced conductivity improvements observed previously. Annealing ITO nanowires may remove disorder within the structures that acts as an impediment to carrier conduction on the scale of a few nanometers. Further study of this process should help determine the structural σITO =
Figure 2. Comparison of the changes in (a) conductivity as
measured with four-point probe and THz-TDS. Dc conductivities of the ITO films extracted from THz-TDS via the extrapolation to ω = 0 of the Drude–Smith fits correspond to σ˜ film (0). Average dc conductivities of the nanowires in the films obtained by correcting σ˜ film (0) using the deposited mass densities are given by σ˜ nw (0). Finally, σITO denotes the intrinsic conductivity of the ITO that makes up the nanowires determined from the Drude–Smith model fit parameters. (b) Sheet resistance in response to changes in deposition flux rate. (c) Nanowire trunk diameter measured at the top and bottom of the film. (d) The effective film density relative to the expected bulk value of ITO (ρ0 = 7120 µg mm−3 ).
(brown, left), and is an order of magnitude larger than the average electron density for the film (blue, right). Table 2 summarizes the Drude–Smith fit parameters including carrier densities. In analogy to the nanowire electron density, we can define the conductivity within the nanowires as 5
Nanotechnology 25 (2014) 035701
J M LaForge et al
Figure 3. Comparison of the base of the nanowire films from table 1 ((b)–(g)) and ∼1 µm control films ((a), (h)) deposited at 0.5 nm s−1
(a) and 3.0 nm s−1 (h). Images ((b)–(g)) show the samples measured by THz-TDS in order of increasing flux rate.
characteristics that limit the conductivity in the as-grown nanotrees presented here. Finally, we note that the intrinsic conductivity of the nanowires, σITO , increases with deposition flux rate (figure 2(a)) as a result of the increase in electron density nnw within the nanowires (figure 1(e)). However, the relative rate of increase across deposition flux rates is substantially reduced in σITO (4×) compared to other measures, σ˜ nw (0) (21×) and σ˜ film (0) (28×), which suggests that the nanoscale disorder of the as-deposited material is significantly affected by the deposition rate. We note that electronic surface depletion layers have led to stronger-than-expected confinement in other semiconductor systems examined by THz-TDS [42]. Strong doping in ITO should limit surface depletion layers to a few Angstroms [45], though deeper chemical depletions layers caused by filling of oxygen-vacancies near the surface may also be present [45–47]. These surface effects are expected in both annealed and as-deposited samples and are therefore unable to account for either the 10–100 times increase in conductivity we have observed after annealing similar nanowires previously [12] or the carrier confinement observed here. Conversely, short-range disorder that confines carriers in as-deposited nanowires and is subsequently removed by annealing appears to be the more likely. Images of the nucleation layer shown in figure 3 demonstrate a thickness increase with flux rate. However, assuming a bulk-like material in the nucleation layer, the increased nucleation layer is responsible for only a small component (1%–8%) of the mass density increase seen in figure 2(d). In comparison to a previous study [30], where the bulk nucleation layer comprised 15–25% of the total film thickness, the nucleation layer accounts for
Search
Collections
Journals
About
Contact us
My IOPscience
Conductivity control of as-grown branched indium tin oxide nanowire networks
This content has been downloaded from IOPscience. Please scroll down to see the full text. 2014 Nanotechnology 25 035701 (http://iopscience.iop.org/0957-4484/25/3/035701) View the table of contents for this issue, or go to the journal homepage for more
Download details: IP Address: 130.108.121.217 This content was downloaded on 07/09/2014 at 11:45
Please note that terms and conditions apply.
Nanotechnology Nanotechnology 25 (2014) 035701 (9pp)
doi:10.1088/0957-4484/25/3/035701
Conductivity control of as-grown branched indium tin oxide nanowire networks J M LaForge1 , T L Cocker2,4 , A L Beaudry1 , K Cui3 , R T Tucker1 , M T Taschuk1 , F A Hegmann2 and M J Brett1,3 1 2 3
Department of Electrical and Computer Engineering, University of Alberta, Edmonton, AB, Canada Department of Physics, University of Alberta, Edmonton, AB, Canada NRC National Institute for Nanotechnology, Edmonton, AB, T6G 2M9, Canada
E-mail: [email protected] Received 3 September 2013, in final form 4 November 2013 Published 17 December 2013 Abstract
Branched indium tin oxide (ITO) nanowire networks are promising candidates for transparent conductive oxide applications, such as optoelectronic electrodes, due to their high porosity. However, these branched networks also present new challenges in assessing conductivity. Conventional four-point probe techniques cannot separate the effect of porosity on the long-range conductivity from the intrinsic material conductivity. Here we compare the average nanoscale conductivity within the film measured by terahertz time-domain spectroscopy (THz-TDS) to the film conductivity measured by four-point probe in a branched ITO nanowire network. Both techniques report conductivity increases with deposition flux rate from 0.5 to 3.0 nm s−1 , achieving a maximum of ∼10 ( cm)−1 . Modeling the THz-TDS conductivity data using the Drude–Smith model allows us to distinguish between conductivity increases resulting from morphological changes and those resulting from the intrinsic properties of the ITO. In particular, the intrinsic material conductivity within the nanowires can be extracted, and is found to reach a maximum of ∼3000 ( cm)−1 , comparable to bulk ITO. To determine the mechanism responsible for increasing conductivity with flux rate, we characterize dopant concentration and morphological changes (i.e., to branching behavior, nanowire diameter and nucleation layers). We propose that changes in the electron density, primarily due to changes in O-vacancy concentration at different flux rates, are responsible for the observed conductivity increase. This understanding will assist balancing structural and conductivity requirements in applications of transparent conductive oxide networks. Keywords: branched nanowires, nanowires, indium tin oxide, THz time-domain spectroscopy, conductivity (Some figures may appear in colour only in the online journal)
1. Introduction
candidate for initial studies of TCO networks. Several groups have fabricated different ITO architectures, such as nanotrees [12], nanopillars [13] and nanowires [14–16]. However, simply ensuring the correct morphology of TCO networks is insufficient for successful application, which also requires that TCO materials have suitable conductivity and transmissivity. In bulk-like films, transmission spectra and in-plane conductivity measured via four-point probe can readily assess the optical and electrical performance. However, TCO branched nanowire networks have reduced or zero
Nanostructured transparent conductive oxide (TCO) networks have applications as flexible conductive substrates and as charge collectors in organic photovoltaics [1–8]. Indium tin oxide (ITO) is a well-understood TCO used in many thin film applications due to its high optical transmittance and electrical conductivity [9–11], and thus ITO is a strong 4 Present address: Department of Physics, University of Regensburg,
Regensburg, Germany. 0957-4484/14/035701+09$33.00
1
c 2014 IOP Publishing Ltd Printed in the UK
Nanotechnology 25 (2014) 035701
J M LaForge et al
2. Experiment
in-plane conductivity, dependent on the morphology of the interconnections. Therefore, four-point probe cannot reliably measure the material’s intrinsic conductivity. Instead, these techniques measure long-range conductivity, or film conductivity, dependent on both the intrinsic material conductivity and the morphology of the material within the film (i.e. electrical connectivity). To address this, nanowire electrical properties have been characterized by placing individual wires across lithographically patterned contacts [17–19]. Alternatively, in situ direct contact techniques, such as scanning electron microscope (SEM) nanoprobe [20], allow for non-destructive measurements of nanowires with suitable spacing and mechanical robustness. While these techniques provide direct measurements of individual nanowires, scaling up to characterize a statistically significant ensemble of nanowires is difficult. If the aggregate electrical properties of larger areas are important, optical techniques, such as terahertz time-domain spectroscopy (THz-TDS) [21–23] can non-destructively measure the average nanoscale conductivity of >106 structures simultaneously, making these techniques attractive for non-contact characterization of TCO networks. Terahertz spectroscopy has been used to study a wide range of nanostructures in recent years [21–23], including nanowires [24–31]. Recently, we have used glancing angle deposition (GLAD) to provide a directional vapor flux during the self-catalyzed vapor–liquid–solid (VLS) growth of branched ITO nanowires. This technique, named VLS-GLAD, allows for improved control over several morphological properties such as nanowire diameter, branch orientation and height placement [12, 32, 33]. We have fabricated nanowire networks with VLS-GLAD that are well suited for three-dimensional electrodes (e.g. for an organic photovoltaic device). The utility of THz-TDS for characterizing the conductive properties of both bulk ITO [34, 35] and VLS-grown ITO nanocolumns [30, 31] has been demonstrated. In one of the latter reports [30], the films were 20 nm morphological features observed in figure 1(a). We took four-point-probe measurements of the film conductivity (across a distance of ∼1 mm) in both linear probe and van der Pauw configurations [44]. These data are compared to the film conductivity observed with THz-TDS by taking the dc extrapolation of the Drude–Smith conductivity fits (σ˜ film (ω = 0)), as presented in figure 2(a), which shows agreement between the trends of the four-point probe and
c-parameter −0.985 ± 0.007 −0.95 ± 0.1 −0.970 ± 0.006 −0.972 ± 0.005 −0.94 ± 0.01 −0.94 ± 0.01
THz film conductivities, especially at low and high flux rates. The magnitudes of the dc conductivities measured by the two techniques are generally close, though the consistently smaller film conductivities measured by four-point probe suggest that more complicated conduction pathways through the nanowire networks are sampled by the four-point-probe measurements. Regardless, the agreement between the conductivity trends in figure 2(a) implies that morphological changes in the films have limited impact on the conduction pathways measured by four-point probe. A notable deviation occurs for samples at 1.0 and 1.5 nm s−1 , which may point to morphological changes playing a role in the suppression of conductivity over long ranges (∼1 mm). This point will be discussed in more detail below. The four-point-probe film conductivities were calculated from the sheet resistances (figure 2(b)) and total film thicknesses measured by SEM. Several morphological changes occur with the increase in flux rate. Nanowire diameter measured at the film base increases with flux rate, as shown in figure 2(c). However, the diameter at the film’s top does not appear to change significantly (figure 2(c)). Thus, the volume averaged nanowire diameter increases with flux rate somewhere between these two trends. Additionally, the network’s effective mass density (ρ/ρbulk ) increases slightly with flux rate, as seen in figure 2(d). In the Drude–Smith fits to the THz film conductivity, the electron density averaged across the THz interaction volume nfilm is used as a fit parameter (figure 1(e)). This average electron density for the film can be used to estimate the electron density within the nanowires nnw via the filling fraction (mass density) of the nanowire film network, nnw = nfilm (ρbulk /ρ) [38, 39]. The nanowire electron density is shown as a function of deposition flux rate in figure 1(e) 4
Nanotechnology 25 (2014) 035701
J M LaForge et al
σ˜ nw (ω) = σ˜ film (ω)(ρbulk /ρ). Equivalently, nnw e2 τDS 1 c 1+ . (2) σ˜ nw (ω) = m∗ 1 − iωτDS 1 − iωτDS The dc extrapolation of σ˜ nw (ω) is the average dc conductivity of an individual nanowire. Therefore, changes to nanowire conductivity with deposition flux rate can be separated from changes to the mass density of the nanowire network. Comparison of the dc film conductivities found by THz-TDS (σ˜ film (0)) and the dc nanowire conductivities (σ˜ nw (0)) in figure 2(a) reveals nearly identical trends with deposition flux rate, indicating that the strong conductivity increase with flux rate is not due to an increase in mass density. Within the Drude–Smith model, localization effects are primarily fit using the c-parameter (though the scattering time also contains some nanoscale information through Matthiessen’s rule [37–40]). The c-parameters in our Drude–Smith fits describe confinement of charge carriers inside nanowires on a scale that is smaller than the nanowire diameter. Therefore, this confinement is still present in the nanowire conductivities σ˜ nw (ω). Conversely, a new conductivity, σITO , can be defined that corresponds to the intrinsic dc conductivity of the Drude gas of carriers that experiences the localization inside the nanowires. Within the framework of the Drude–Smith model, the intrinsic dc conductivity of the carriers in the ITO nanostructures in the absence of structural confinement is σITO =
σ˜ nw (0) , 1+c
(3)
or equivalently, nnw e2 τDS , (4) m∗ where we have assumed that the increase in scattering time due to the removal of boundary scattering constitutes a small correction to the intrinsic dc conductivity. Accounting for this would lead to an increase of σITO by a few per cent. As can be seen in figure 2(a), the intrinsic conductivity of the ITO nanowire networks reaches a maximum of 3300 ± 400 ( cm)−1 , comparable to values of ∼104 ( cm)−1 measured in a different study via four-point probe along a single ITO nanowire at room temperature with similar characteristics to ours (d ∼ 100 nm, n ∼ 1020 cm−3 and Fermi wavenumber kF ∼ 109 m−1 ) [17]. Based on the large nanowire conductivity measured in [17], the type of structural disorder that causes carrier localization in our nanowires appears to be absent in their samples. The nanowires in [17] were grown at 850–900 ◦ C, which should lead to fewer growth defects. Note that post-growth thermal annealing is known to improve the conductivity of our ITO nanotrees by a factor of 10–100 [12]. The large difference we observe between σ˜ nw (0) and σITO therefore suggests a possible explanation for the annealing-induced conductivity improvements observed previously. Annealing ITO nanowires may remove disorder within the structures that acts as an impediment to carrier conduction on the scale of a few nanometers. Further study of this process should help determine the structural σITO =
Figure 2. Comparison of the changes in (a) conductivity as
measured with four-point probe and THz-TDS. Dc conductivities of the ITO films extracted from THz-TDS via the extrapolation to ω = 0 of the Drude–Smith fits correspond to σ˜ film (0). Average dc conductivities of the nanowires in the films obtained by correcting σ˜ film (0) using the deposited mass densities are given by σ˜ nw (0). Finally, σITO denotes the intrinsic conductivity of the ITO that makes up the nanowires determined from the Drude–Smith model fit parameters. (b) Sheet resistance in response to changes in deposition flux rate. (c) Nanowire trunk diameter measured at the top and bottom of the film. (d) The effective film density relative to the expected bulk value of ITO (ρ0 = 7120 µg mm−3 ).
(brown, left), and is an order of magnitude larger than the average electron density for the film (blue, right). Table 2 summarizes the Drude–Smith fit parameters including carrier densities. In analogy to the nanowire electron density, we can define the conductivity within the nanowires as 5
Nanotechnology 25 (2014) 035701
J M LaForge et al
Figure 3. Comparison of the base of the nanowire films from table 1 ((b)–(g)) and ∼1 µm control films ((a), (h)) deposited at 0.5 nm s−1
(a) and 3.0 nm s−1 (h). Images ((b)–(g)) show the samples measured by THz-TDS in order of increasing flux rate.
characteristics that limit the conductivity in the as-grown nanotrees presented here. Finally, we note that the intrinsic conductivity of the nanowires, σITO , increases with deposition flux rate (figure 2(a)) as a result of the increase in electron density nnw within the nanowires (figure 1(e)). However, the relative rate of increase across deposition flux rates is substantially reduced in σITO (4×) compared to other measures, σ˜ nw (0) (21×) and σ˜ film (0) (28×), which suggests that the nanoscale disorder of the as-deposited material is significantly affected by the deposition rate. We note that electronic surface depletion layers have led to stronger-than-expected confinement in other semiconductor systems examined by THz-TDS [42]. Strong doping in ITO should limit surface depletion layers to a few Angstroms [45], though deeper chemical depletions layers caused by filling of oxygen-vacancies near the surface may also be present [45–47]. These surface effects are expected in both annealed and as-deposited samples and are therefore unable to account for either the 10–100 times increase in conductivity we have observed after annealing similar nanowires previously [12] or the carrier confinement observed here. Conversely, short-range disorder that confines carriers in as-deposited nanowires and is subsequently removed by annealing appears to be the more likely. Images of the nucleation layer shown in figure 3 demonstrate a thickness increase with flux rate. However, assuming a bulk-like material in the nucleation layer, the increased nucleation layer is responsible for only a small component (1%–8%) of the mass density increase seen in figure 2(d). In comparison to a previous study [30], where the bulk nucleation layer comprised 15–25% of the total film thickness, the nucleation layer accounts for
