Rapid determination of nanowire electrical properties using a dielectrophoresis-well based system

Abstract

The use of high quality semiconducting nanomaterials for advanced device applications has been hampered by the unavoidable variability in the growth of one-dimensional (1D) nanomaterials such as nanowires (NWs) and nanotubes, resulting in highly variable electrical properties across the population. Therefore, assessment of the quality of nanomaterials is vital for the fabrication of high-performance and reliable electronic devices. The controllable selection of high quality NWs has been recently demonstrated using a dielectrophoretic (DEP) NW assembly method; however, no convenient, rapid method has been adopted for the characterization of nanomaterial semiconducting properties. In this study, we solve this challenge with a low-cost, industrially scalable method for the rapid analysis of the electrical properties of inorganic single crystalline NWs, by identifying key features in the DEP frequency response spectrum (1 kHz – 20 MHz). NWs dispersed in anisole were characterized using a three-dimensional DEP chip (3DEP) in 60 seconds, and the resultant spectrum demonstrated a sharp change in NW response to DEP signal in 1 MHz – 20 MHz frequency rage. The 3DEP analysis, confirmed by field-effect transistor (FET) data, indicates that NWs with higher quality are collected at high DEP signal frequency range such as above 10 MHz. These results show that platforms such as the 3DEP, can be used for the characterization of rod-shaped nanoscale particles where the dipole moment is sufficiently large. It also shows that the 3DEP can be used to assess heterogeneous nanoparticle mixtures and identify nanomaterials with superior conductivity properties. The proposed methodology can be applied to any type of 1D nanomaterials. The 3DEP analysis coupled with dielectrophoretic assembly for the deposition of NWs at selected signal frequencies, leads to a reproducible fabrication of high quality NW FET devices.

Semiconducting single-crystalline nanowires (NWs) offer many advantages for the fabrication of solution-processed, low-cost printed electronic devices. Their unique characteristics make them potential key building blocks for many applications, such as chemical and biological sensors,^1–3 high performance FETs,^4–6 optical devices,^{7, 8} memory elements^9–11 and energy harvesting nanogenerators.^{12, 13}

However, NW bottom-up synthesis methods, e.g., vapor-liquid-solid (VLS) growth, often produce highly heterogeneous mixtures of NWs with a range of different electrical properties, and as a consequence, the purification of one-dimensional (1D) nanomaterials based on their electrical properties is a significant hurdle for the fabrication of reliable and high-performance devices. Whilst the deposition of NWs into high ordered arrays has been demonstrated with various techniques, such as Langmuir-Blodgett (LB),¹⁴ Blown-Bubble films,¹⁵ flow-directed assembly¹⁶ and electrostatic interactions,¹⁷ these techniques do not include a selection step to separate the most desirable NW based on their electrical performance.

We have previously demonstrated the direct selection of Supercritical Fluid–Liquid–Solid grown silicon (Si) NWs based on conduction properties from poly-disperse as-synthesized NWs using dielectrophoresis (DEP) coupled with impedance spectroscopy, highlighting the selective collection of NWs with different electrical properties at various applied frequencies.¹⁸ DEP is an electrostatic phenomenon of induced motion of a polarizable particle when subjected to an inhomogeneous electric field. Depending on the electrical properties of the particle and suspending medium, the induced DEP force causes the particle to either be attracted towards the region of high electric field gradient (positive DEP) or repelled (negative DEP) along the direction of the field gradient¹⁹; the magnitude and direction of the force is dependent on the electrical properties of particle and medium, and on the frequency of the applied field. Consequently, analyzing the DEP signal across a range of frequencies can be used to determine the properties of suspensions, whilst particles with different properties can be separated by selecting a frequency where the different populations experience positive and negative DEP, respectively¹⁸. The use of DEP characterization has been widely explored in the biosciences, but until recently was restricted to laboratories with expertise in the field. However, a new commercial platform that utilizes a disposable, three-dimensional (3D) well electrode chip (3DEP by DEPtech, Uckfield, UK) coupled with a prototype 3DEP system reader (Labtech, Uckfield, UK) makes DEP characterization suitable for a wider range of nanomaterials.²⁰ This 3DEP platform has been used in the measurement of the electrical properties of cells used in cancer studies^21–24 (incl. keratinocyte cells, OOPC cells, OSCC cells) and drug interaction.²⁵ The system is engineered principally for the rapid and near-real-time analysis of the electrical properties of cells, taking typically 10 seconds to analyze a population of ca. 20000 cells. As a rule of thumb, DEP electrodes are typically constructed with dimensions that reflect the size of the particles, and the 3DEP chip is of a size (typically 1 mm across, with 70 μm wide electrodes separated by 150 μm gaps) that would be considered far too large for the manipulation of 1D nanoparticles.

In this letter, we demonstrate the efficacy of the 3DEP system as an efficient tool for the rapid evaluation of electrical properties of as-synthesized Si NWs ensemble using a frequency spectrum analysis. 3DEP chip frequency response of dispersed Si NWs shows a ‘threshold’ DEP signal frequency of ≈1 MHz, above which, the DEP starts to collects only high-quality NWs with the best conducting properties. The whole response spectrum (1 kHz – 20 MHz) characterization of NWs takes only several minutes. The direct correlation between DEP signal frequency and the Si NWs conductivity is confirmed by comparing field-effect transistor (FET) data for devices fabricated with NWs collected at various DEP frequencies (30Hz, 300 kHz, 10 MHz and 20 MHz).

VLS growth of Si NWs was catalyzed using commercially-available 60 nm Au nanoparticles randomly dispersed on poly-L-lysine-functionalized Si (111) substrates. Growth was performed at 900 °C and 80 kPa using 30 sccm (standard cm³/min) of SiCl₄ and 200 sccm of H₂ diluted with N₂ to a total flow rate of 1000 sccm. Small pieces of Si wafer with as-grown nominally undoped NWs were placed in small vials containing 2 mL of anisole. NWs were dispersed by using ultrasonic agitation at low power (≈200 W) for 10 s to 15 s. As-grown and dispersed Si NWs were analyzed by Scanning Electron Microscopy (SEM) and Transmission Electron Microscopy (TEM) and exhibited an average 22 μm length (FIG. 1 (a)) and ≈75 nm diameter (FIG. 1 (b)). TEM studies confirmed <111> growth direction of NWs. The NWs’ diameter was also evaluated and confirmed from the Atomic Force Microscopy (AFM) results (not shown).

FIG. 1.

(a) SEM image (30° tilt) of CVD-grown Si NW. (b) TEM image of a typical Si NW with ∼70 nm Si core and ∼3 nm thick shell of native oxide. Inset shows the corresponding Fast Fourier Transform (FFT) pattern.

A DEPtech 3DEP microwell electrode system was used for the determination of the DEP-frequency response of Si NWs dispersions. The gold-plated copper 3DEP electrode chip contained 20 wells of 1 mm diameter, each comprising eight gold-plated copper electrodes in a ring configuration separated by polyimide layers, with thicknesses of 70 μm and 150 μm, respectively, with a schematic shown in FIG. 2 (a). For each experiment, approximately 75 μL of Si NWs dispersed in anisole were injected into the 3D well electrode chip. The principle of operation of the 3DEP is based on the attraction of nanoparticles towards the imbedded electrodes during positive DEP, resulting in a reduction of light scattering, and repulsion from the electrodes during negative DEP, leading to an increased light scattering, when an alternating electric signal is applied to the electrodes. The relative strength of the dielectrophoretic force can be evaluated by measuring the changes of the transmitted light intensity through the well, for each applied DEP signal frequency. During the experiment, each of the 20 well received a different frequency energising signal in the range 1 kHz – 20 MHz, at peak-to peak bias (V_pk-pk) of 10 V. The applied frequency remained constant for each well during the experiment. Wells were excited with AC signal for 60 s, during which NWs were expected to experience positive DEP, and to move away from the well center. Software was then used to assess the change in NW distribution by measuring the change in light intensity within each well, as the NWs, suspended in anisole, reoriented and moved in response to the applied field. The change in light intensity was normalised to intensity of the image of the well captured before applying the electric field, and the relative dielectrophoretic force acting on NWs was then related to the changes of light intensity. When dispersed NWs experience a DEP force in a liquid, they reach their terminal velocity in milliseconds, and, so for any observed movement we can assume that velocity is proportional to the DEP force (F_DEP) exerted on the particle. F_DEP for NWs with cylindrical shape is commonly approximated to that of a prolate ellipsoid with major and minor axes equal to the length and diameter using the relationship given by Equation 1(a–c) ^26–31:

$${\mathit{F}}_{\mathit{D}\mathit{E}\mathit{P}}=\frac{\mathbf{2}\mathit{L}\mathit{\pi }{\mathit{r}}^{\mathbf{2}}}{\mathbf{3}}{\mathit{\epsilon }}_{\mathit{m}}\mathit{R}\mathit{e}\left\{{\mathit{K}}_{\mathit{f}}\right\}\mathbf{\nabla }{\mathit{E}}^{\mathbf{2}}$$ (1a)

$$K_f=\frac{{\tilde{\epsilon }}_p-{\tilde{\epsilon }}_m}{{\tilde{\epsilon }}_m}$$ (1b)

$${\tilde{\epsilon }}_p=\epsilon -j\frac{{\sigma }_p}{\omega },{\tilde{\epsilon }}_m=\epsilon -j\frac{{\sigma }_m}{\omega }$$ (1c)

where, r is NW radius, L is NW length, ε_m is the permittivity of the fluid medium and ∇E² is the gradient of the electric field strength squared, K_f is the Clausius-Mossotti factor and Re denotes “the real part of”. ${\tilde{\epsilon }}_p$ is the permittivity of NW, ${\tilde{\epsilon }}_m$ is the permittivity of the medium, ε_o is the vacuum permittivity, σ_p is the conductivity of NW, σ_m is the conductivity of the medium and ω is the angular frequency of the electric field.

FIG. 2.

(a) A schematic of the 3D well electrode chip used in this work. The nanomaterials can either be attracted towards the electrodes (the walls of the well) by applying a positive DEP or repelled from the electrodes by applying a negative DEP. The movement of the nanomaterials is tracked by image analysis and detected by measuring the light intensity of the beam passing through the well for a period of 60 seconds. (b) Relative DEP force (arbitrary units) as a function of DEP signal frequency for Si NWs suspended in anisole. The points represent the mean DEP response of three independent experiments. The solid red line represents the best fit. Error bars were evaluated from the fluctuations of light intensity due to NWs random movement in the liquid.

Examination of Equation 1a shows that the DEP force scales with particle volume. This means that traditionally, nanoparticles have required the reduction in volume to be balanced by an increase in ∇E², so that electrodes designed for manipulating nanoparticles typically have inter-electrode gaps of a few μm and are fabricated using thin-film photolithographic methods that produce very sharp cross-sections; such electrodes are usually only able to trap for a few tens of μm from the inter-electrode gaps. However, the electrodes used here do not resemble these electrodes at all; instead the inter-electrode gap is 150 μm. Nevertheless, the nature of NWs is that the dipole along the long axis is sufficiently substantial to induce enough DEP force to move the nanoparticles in the suspension medium; consequently, DEP alignment is observed even at a distance from these large electrodes.

The dielectrophoretic force spectrum of Si NWs obtained by this method is shown in FIG. 2 (b). The points represent the mean (n=3) DEP system response at different frequencies; the solid line represents the best fit of the data. The graph shows a plateau response between 1 kHz and 100 kHz, followed by a decrease of relative DEP force at around 1 MHz signal frequency. At frequencies 10 MHz – 20 MHz, the DEP force reaches lower, however non-zero, values, indicating that a smaller population of NWs is still able to respond to the DEP force.

Whilst using monodisperse solution of Au nanoparticles to catalyze growth of Si NWs could be expected to provide a mono-dispersed collection of Si NWs, the frequency range over which the DEP dispersion takes place (beginning at ca. 1 kHz and continuing to over 20 MHz) suggests a range of NW heterogeneity; the dispersion of a homogeneous NWs would typically take one decade to move from <10% of the transition to >90% ³². As the frequency increases, the dipole across the nanomaterials weakens, causing the relative DEP force to drop, so that NWs with the lowest conductivity will exhibit a dielectric dispersion at the lowest frequencies. The variations in conductivity are expected to originate from the crystal quality of the NWs, such as the concentration of surface and crystalline defects and associated traps. It is also expected that the Si NWs will not experience negative DEP, given their high conductivity value compared to the medium (anisole). However, at higher frequencies we predict that F_DEP will reach a plateau and then decline close to zero DEP force. We also anticipate that DEP collection frequency will vary according to NWs’ different doping level and corresponding charge carrier mobility.

The electrical properties of the NWs collected at different frequencies were investigated by FET analysis to establish a direct correlation with the effective DEP force spectrum. The FET devices were prepared on Si/SiO₂ substrates, with pre-patterned source-drain electrodes with 10 μm gap (FIG. 3), where highly doped Si substrate served as the bottom-gate electrode. The dielectrophoretic bottom contact structures consisting of 2 nm Ti acting as an adhesion metal followed by 50 nm Au or 50 nm Pd layers were patterned with photolithographic lift-off techniques. The choice of the contact metals is dictated by p-type conductivity that is usually observed in nominally undoped Si NWs.³³ Both gold and palladium metals are closely matched to the valence band edge of the Si to enable near-Ohmic contact for p-type transport, thus we did not expect any substantial variations in the FET device performance.³⁴ The alignment of the Si NWs was performed using DEP by applying an AC field across the two parallel electrodes (at V_pk-pk = = 12 V) forming the source-drain contacts of the FET device. The substrate was placed on an inclined surface (≈30° versus the horizontal plane) to allow NW dispersion to flow along the substrate and to cross the dielectrophoretic electrodes, as described previously.¹⁸ A 10 μL - 20 μL drop was placed on top of the dielectrophoretic structures after the AC field was applied. For the DEP alignment an AIM-TTI TG120–20 MHz function generator was used. After alignment, the substrates were gently rinsed with IPA to remove weakly aligned NWs across the source-drain electrodes, and then were gently dried with a N₂ flow. The fabrication of the FET devices was completed by performing a second lift-off process to deposit 80 nm thick metal contacts (of the same metal used for the bottom contacts) on top of the aligned NW for the improvement of the charge carrier injection. Prior to the metal deposition, the edges of the NWs at the source-drain contacts regions underwent a diluted hydrofluoric acid (HF) treatment for 12 s. It is expected that the HF treatment removes the native oxide layer of the NWs (see Refs. ^35–37) thereby forming a “clean” semiconductor-electrode contact region. The final step included the post-anneal of the FET device at 250 °C – 300 °C for 45 min to improve the metal-semiconductor contact for optimum charge transport characteristics.⁴ A schematic of the bottom-gate FET device with gold contacts is shown in FIG. 4(a). The transistors were characterised using a Keithley 4200 SCS analyser system in a N₂-filled glove box in order to minimise effects of atmospheric contamination.

FIG. 3.

(a) Polarised microscope image (POM) and (b) SEM image of Si NWs aligned across two parallel electrodes (10 μm gap) via DEP.

FIG. 4.

(a) Schematic of a bottom-gate Si NW FET device with gold source-drain contacts. (b) Transfer characteristic of FETs of NWs aligned via DEP at 30 Hz, 300 kHz, 10 MHz and 20 MHz. The plot is normalised to the number of NWs across the FET channel, thus showing an average current per NW. Inset: Output characteristic of the devices shown on (b). At 30 and 300 kHz the output currents overlap. (c-d) Typical performance data, including sub-threshold swing (s-s), trap density (N_trap) and mobility (μ) for a number of FET devices with nanowires aligned at 30 Hz, 300 kHz, 10 MHz and 20 MHz.

Typical transfer characteristics of FETs made of NWs aligned at various DEP frequencies are shown in FIG. 4(b). The transfer characteristics were normalized to the number of NWs that bridge the source – drain electrodes. The inset exhibits the output characteristic of the same FETs at V_G = −40 V. DEP aligned NWs at higher frequencies (10 MHz and 20 MHz) demonstrated higher output currents highlighting their superior conductivity properties.

In order to evaluate the FET performance, the sub-threshold slope (s-s), the trap density (N_trap) and mobility were calculated as a function of DEP frequency. The sub-threshold behaviour (V/decade) indicates how much gate voltage is needed to turn-on the transistor, and, in addition to the FET geometrical parameters, it also depends on the density of defect/trap states on the NW surface or at the NW/dielectric interface. Thus, higher trap density results in less steep current increase and high sub-threshold values. The sub-threshold slope is given by Equations 2 (Refs. ^{38, 39}) and 3 (Ref. ³⁴), and is related to the trap density (N_trap) according to Equation 4 (Refs. ^{40, 41}):

$$\mathit{s}-\mathit{s}=\mathbf{ln}(\mathbf{10})\frac{\left(\mathbf{1}+\frac{{\mathit{c}}_{\mathit{i}\mathit{t}}}{{\mathit{c}}_{\mathit{b}\mathit{o}\mathit{x}}}+\frac{{\mathit{c}}_{\mathit{s}\mathit{i}}}{{\mathit{c}}_{\mathit{b}\mathit{o}\mathit{x}}}\right)}{\left(\frac{\mathit{q}}{\mathit{k}\mathit{T}}-\frac{\mathbf{1}}{{\mathit{E}}_{\mathit{b}\mathit{n}\mathit{w}}\mathbf{2}\mathit{r}}\right)}$$ (2)

$$\mathit{s}-\mathit{s}=\frac{\mathit{\Delta }{\mathit{V}}_{\mathit{G}}}{\mathit{\Delta }\mathbf{log}{\mathit{I}}_{\mathit{D}}}$$ (3)

$${\mathit{N}}_{\mathit{t}\mathit{r}\mathit{a}\mathit{p}}=\left[\frac{\mathit{q}(\mathit{s}-\mathit{s})\mathbf{log}(\mathit{e})}{\mathit{k}\mathit{T}}-\mathbf{1}\right]\frac{{\mathit{C}}_{\mathit{N}\mathit{W}}}{\mathbf{2}\mathit{\pi }\mathit{N}\mathit{r}\mathit{L}\mathit{q}}$$ (4)

where, k is the Boltzmann’s constant, T is absolute temperature, q is the electron charge, E_bnw is the field at the bottom of the Si NW, r is the NW radius, C_it is the capacitance of interface states at the Si oxide-shell/Si region of the NW, C_si is the capacitance of the Si NW, and C_box is the back oxide capacitance (SiO₂), C_NW is the total gate capacitance, N is the total number of NWs across the channel, L is the channel length.

The carrier mobility (μ) was calculated using the cylinder-on-plate model which takes into consideration the electrostatic fringing effect acting on the finite number of NWs in the channel as shown in Equations 5 and 6 (Refs, ^{18, 42–45}):

$$C_{NW}=N\frac{2\pi {\epsilon }_o{\epsilon }_{r}L}{cos{h}^{-1}(\frac{r+d}{r})}$$ (5)

$$\mu =\frac{L^2}{C_{NW}}\frac{1}{V_{SD}}{g}_m$$ (6)

where, C_NW is the total gate capacitance, ε_o is the absolute permittivity, ε_r is the SiO₂ dielectric constant (3.9), d is the thickness of the gate dielectric (230 nm), μ is the mobility, V_SD is the applied source-drain voltage and g_m is the transconductance $({\text{g}}_{\text{m}}=\frac{\partial {\text{I}}_{\text{D}}}{\partial {\text{V}}_{\text{G}}})$.

The sub-threshold swing, trap density and mobility as a function of DEP frequency of the devices presented in FIG. 4 (b) are shown in TABLE I. Furthermore, similar data from devices fabricated at a range of DEP signal frequencies (30 Hz, 300 kHz, 10 MHz, 20 MHz) are also shown on FIG. 4 (c–d). It is clear, that Si NWs aligned at the low DEP frequencies (30 Hz and 300 kHz) showed low maximum on-state current at the gate voltage (V_G) of −40 V. For NWs collected at the high frequencies (10 MHz and 20 MHz), the FETs showed a significantly improved performance, with two orders of magnitude higher current, 50 % lower sub-threshold swing and more than 100 times higher mobility values (see FIG. 4 (d)). The increase of the mobility values from low to high frequencies is attributed to the various densities of defects and traps states which all affect the dipole in fast oscillating external non-uniform DEP field.¹⁸

TABLE I.

Sub-threshold swing, trap density and mobility values of the devices shown in FIG. 4 (b).

DEP Frequency (Hz)	s-s (V/decade)	NTrap (cm−2)	μ (cm2 V−1 s−1)
30	4.09	1.73E13	0.1
300 k	4.97	2.10E13	0.06
10 M	2.13	8.88E12	7.05
20 M	2.61	1.09E13	8.71

Notably, devices fabricated at the high frequency range exhibited high on-current and high mobility values. An example of high current FET I-V curves is shown in FIG. 5, with the device exhibiting high on/off ratio of 10⁵, on-currents up to ≈0.65 mA when measured at high biased voltages, and the hole mobility of 16 ± 2.3 cm² V⁻¹ s⁻¹.

FIG. 5.

Example of a high current FET (comprising of 200 NWs) with NWs being aligned at 20 MHz. Left y-axis shows the linear scale and right y-axis shows the logarithmic scale at V_D = −5 to −25 V, in −5 V steps. Maximum device on-current is ∼0.65 mA.

These results demonstrate that NWs grown in the same controllable CVD process, still have significant variations in properties as evidenced by the FET data. By analyzing the frequency spectrum obtained using the 3DEP, we obtain a direct correlation between the properties of NWs collected at different frequencies and the corresponding FET characteristics (FIG. 4 (c–d)). Thus, by altering the dielectrophoretic frequency, high or low quality NWs can be discriminated. We propose that 3DEP analysis can be applied to other semiconducting nanomaterials.

In conclusion, we have experimentally demonstrated that 3DEP DEP-microwell electrode system is a powerful tool, which can provide a rapid discrimination of Si NWs with different electrical properties, providing a guidance for the selection and collection of high quality NWs with high electrical conductivity. The analysis provided by 3DEP was experimentally confirmed by FET data. DEP aligned NWs in the high frequency range (10 MHz and 20 MHz) showed 100 times higher on-current, low sub-threshold swing and trap densities, and more than 100 times higher mobility values, compared to the lower frequency range (30 Hz, 300 kHz). Also, the high quality Si NW FETs demonstrated near mA on-currents, high on/off ratios of 10⁵ and high effective hole mobility of 16 cm² V⁻¹ s⁻¹.

Beyond its immediate application for Si NWs, this work also demonstrates that appropriate electrode geometries on the tens-of-microns scale can be used to evaluate and manipulate 1D nanostructures such as NWs and potentially other 1D nanostructures such as carbon nanotubes. Due to large dipole moments induced in high-aspect-ratio particles, the DEP force is effective even at some distance from relatively large electrodes, providing large volume for nanostructures separation. This also suggests that large-scale DEP sorting process of 1D nano-objects may also be feasible and warrants further investigation.