Thermally tunable VO₂-SiO₂ nanocomposite thin-film capacitors

Tunable dielectrics are of broad interest for a range of devices, including capacitive sensors, thermal metamaterials, signal processing, and energy storage systems. Temperature tunable materials are being widely explored in the context of thermal concentrators¹ and thermal cloaks.^2,3 Previous studies have demonstrated that nanocomposite dielectrics can be used for thermally tunable metamaterials and switchable capacitors.^4–14 The tunable dielectric properties (e.g., permittivity and loss tangent) of a nanocomposite strongly depend on the composition, particle sizes, morphology, distribution, ratio, and interaction among constituents.^15,16

The phase-transition material vanadium dioxide (VO₂) displays a strong and sharp temperature-driven insulator-metal transition (MIT) that features large changes in carrier density from its low-temperature monoclinic structure to the high-temperature tetragonal rutile structure.^17–24 The dielectric constant of VO₂ varies by orders of magnitude across the MIT temperature (T_MIT).^25,26 The transition behavior of thin-film VO₂ can be further engineered via doping,²⁷ defect engineering,²⁸ or altering the deposition conditions.^29–31 Consequently, VO₂-containing systems are promising for the realization of thermally tunable devices. However, the highly conducting nature of the metal phase of VO₂ can lead to current leakage, which is detrimental for capacitor applications. Thereby, it is important to keep the metal-phase VO₂ spatially constrained to minimize leakage losses in a capacitor, especially when the VO₂ is in the metallic state. The controlled dispersion of VO₂ in an insulating dielectric matrix (such as SiO₂) is one potential approach.^32–35 Pioneering studies on VO₂-based composites fabricated by ion implantation of vanadium into silica, followed by annealing in an oxygen environment, illustrated a phase-transition evolution that was strongly dependent on the size of VO₂ inclusions.³⁶ However, the use of such composites in tunable capacitors has not yet been explored.

In this study, VO₂-SiO₂ nanocomposites were prepared by a co-sputtering approach, and sandwiched between thin SiO₂ layers. For comparison, a VO₂ control layer and a SiO₂/VO₂ multilayer were also deposited under identical conditions. We found that dispersing VO₂ inclusions in a silica matrix results in gradual tuning of the dielectric properties as opposed to an abrupt phase transition seen in pure VO₂.

The schematics of our metal-insulator-metal (MIM) capacitor structures are shown in Fig. 1. Radio-frequency (RF) sputtering was used to fabricate each dielectric layer. First, a 10 nm SiO₂ layer was deposited onto a Pt (111) (150 nm) /  Ti (10 nm) /SiO₂ (300 nm) / n-Si substrate at 400 °C using an SiO₂ target. Then, three types of structures were fabricated using V₂O₅ and SiO₂ targets: a VO₂ control layer (100 nm), a superlattice VO₂/SiO₂ multilayer (5 nm for each, 20 layers in total), and a composite VO₂-SiO₂ layer (100 nm). The composite film is the primary focus of this study. The growth temperature and pressure were 600 °C and 5 mTorr, respectively. The RF power supplied to the SiO₂ was 150 W, and that to the V₂O₅ was varied to achieve different VO₂/SiO₂ ratios in the composite layers. (V₂O₅ power of 100 W, 46 W, and 12 W, for 50%, 30%, and 10% VO₂ molar content respectively). Finally, another SiO₂ layer of 10 nm was deposited on top to form a sandwich structure. Gas flows of 49.75 SCCM of argon (Ar) and 0.25 SCCM of oxygen (O₂) were used during deposition. Ti (5 nm in thickness) / Au (150 nm in thickness) were deposited on the SiO₂ layer as top contacts using an electron-beam evaporator. Round metal contacts with a diameter of 200 μm were defined using a shadow mask.

The effective permittivity of dielectric media can be calculated from the capacitance as

where C is the capacitance of device, A is the Ti/Au contact area, d is the dielectric media thickness, ξ is the effective permittivity of the dielectric material, and ξ₀ is the vacuum permittivity of 8.86 × 10⁻¹²F m⁻¹.

The detailed parameters of each sample are shown in Table I.

The current-voltage (I-V), capacitance-frequency (C-F), and capacitance-voltage (C-V) data were collected by using a Keithley 236 Source Measure Unit and an Agilent 4284A Precision LCR Meter with a temperature-controlled probe station. The conduction mechanisms in the composite medium can be analyzed by considering the temperature and voltage dependence of the leakage current through the capacitor.

Poole-Frenkel (P-F) and Schottky emission models were employed to determine the conduction mechanism of the different dielectric materials. Specifically, the current density resulting from P-F emission can be written as³⁷ 

while Schottky emission can be expressed as³⁸ 

In Eqs. (2) and (3),

J is the current density, C is the Faraday constant, A is the effective area of capacitor, e is the charge of an electron, $μ$ is the carrier mobility, $n o$ is the carrier concentration, E is the electric field, k is the Boltzmann constant, T is the temperature in Kelvin, $ϕ 0$ is the Schottky barrier height between the metal and semiconductor dielectric. $m *$ is the effective mass of carriers, and h is the Planck constant.

The slopes of ln(J) vs V^1/2 and ln(J/V) vs V^1/2 plots should be constant for Schottky emission and the Poole-Frenkel effect, respectively.

X-ray diffraction (XRD) patterns were collected using a PANalytical MRD X'Pert Pro High Resolution XRD instrument. X-ray photoelectron spectroscopy (XPS) was carried out on the composite thin films using a Kratos spectrometer with an Al Kα of 1.4866 keV as the x-ray source. The carbonaceous C_1s line at 284.6 eV was used as a reference to calibrate the binding energies. The reflectance of the patterned samples in the mid-infrared wavelength range was investigated via near-normal incidence measurements using a mid-infrared microscope (Hyperion 2000) with a Cassegrain objective (NA = 0.4), attached to a Fourier transform infrared (FTIR) spectrometer (Bruker Vertex 70). The mid-infrared source in our FTIR setup is a Globar (a heated silicon carbide rod).

High-angle annular dark field-scanning transmission electron microscope (HAADF-STEM) images were collected using FEI TALOS F200X TEM operated at 200 kV. The high-resolution electron energy loss spectroscopy (EELS) maps and energy dispersive spectroscopy (EDS) mapping results were collected using an FEI Titan³ G2* TEM operating at 80 kV, equipped with a built-in Super X-EDS system, and Gatan Quantum 966 instrumentation. The cross-sectional transmission electron microscope (TEM) samples were prepared using a Helios NanoLab 660 field emission scanning electron microscope (FESEM) with a focused ion beam (FIB). The preparation procedure included ion-beam gas-assisted chemical vapor deposition of a 2- $μ$m-thick carbon layer to protect the sample during milling.

Figure 2(a) displays the cross-sectional STEM image taken in HAADF mode from a composite capacitor device, which clearly shows the interface between dielectric layers (SiO₂/composite/SiO₂) and underlayers (Pt/Ti/SiO₂/Si). The corresponding EDS mapping shows the elemental composition of each layer [Fig. 2(b)]. High resolution TEM-EELS and EDS elemental mapping of the composite layer is collectively shown in Fig. 2(c). As expected from co-sputtering, the Si and V elements are uniformly mixed and dispersed throughout the film. The particle size of VO₂ clusters is smaller than 5 nm. The temperature-dependent normalized in-plane resistance evolution (during heating) of a VO₂ single layer is shown in Fig. 3(a). The film resistance decreases by more than two orders of magnitude from 25 °C to 100 °C, with the MIT temperature at around 72 °C. The VO₂ single layer shows the characteristic (⁠ $2 ¯$11) diffraction peak at 36.22°, which is ascribed to the vanadium dioxide (VO₂) phase [Fig. 3(b)].³⁹ By comparison, neither a sharp conductance transition nor significant VO₂ XRD peaks could be found on the composite sample. However, high-resolution XPS analysis of the pure VO₂ film and composite films with different V/Si ratios [Fig. 3(c)] clearly shows the V_2p3/2 peak at ∼516 eV which we attribute to V⁴⁺ in VO₂.^30,40–42 The combination of XRD and XPS characterization indicates the presence of small VO₂ particles dispersed throughout the composite layer, which is consistent with the STEM results. The ultra-small particle size is likely responsible for the graded switching behavior.

It has been previously reported that the precipitation of VO₂ in the SiO₂ matrix also influences the optical properties of VO₂.^43,44 In our case, temperature-dependent near-infrared reflectance integrated over 1.5–2 $μ$m was measured to reveal the optical evolution of different thin films (Fig. 4). The VO₂ control layer shows an abrupt reflectance increase (33% to 64%) across the T_MIT. In comparison, the composite layer only shows a slight change of 0.6%. This small change is ascribed to the small particle size of VO₂, and is consistent with the characterization data shown in Figs. 2 and 3.

Analysis of the leakage current of composite samples with different VO₂ contents [Fig. 5(a)] at 1 V bias shows that the leakage current significantly decreases with the reduction of the VO₂ content. Figure 5(b) shows the I-V characteristics of the composite layer (50 mol. % VO₂ content) with temperature from 25 °C to 100 °C. The current density of the composite capacitor is lower than 10⁻⁶A cm⁻² at 1 V. By comparison, the current leakage of the VO₂ control layer reaches 10⁻¹A cm⁻² at 0.1 V, and the superlattice VO₂/SiO₂ multilayer shows leakage higher than 10⁻³A cm⁻² at 1 V (Fig. 6).

To evaluate the temperature-dependent current transport mechanism in the composite film, Poole-Frenkel (P-F) and Schottky emission models were employed. These models describe the typical bulk-limited and electrode-limited conduction mechanism in dielectric films, respectively. The P-F and Schottky fitting plots are shown in Figs. 5(c) and 5(d), respectively. Typically, current transport following the P-F mechanism relies more on the thermal emission of electrons from the conduction band of the dielectric.⁴⁵ In comparison, Schottky emission is used for explaining the electrons that are thermally activated to overcome the energy barrier at the metal-dielectric interface. Conduction in the VO₂ layer is not purely from the P-F mechanism especially when the temperature crosses T_MIT. This is because of the metallic state of VO₂ beyond the phase-transition temperature, which leads to ohmic conduction. A better linear fitting with a slope of 2.9 using the P-F model [ln (I/V) vs V^1/2] is observed in the composite sample when the temperature approaches or is above the transition temperature of VO₂ (∼65 °C–100 °C), suggesting the conduction is gradually dominated by the P-F mechanism. In comparison, better linear fitting of composite data using the Schottky model [ln(I) vs V^1/2] results in a slope of 5.1 when the temperature is below the transition temperature (∼25 °C–65 °C). Unlike the homogenous film of VO₂, conducting VO₂ particles embedded in SiO₂ at high temperature can be seen as a localized electron that can be excited under bias. Consequently, P-F emission is dominant. In comparison, at low temperature, the VO₂ is more insulating. Therefore, the conduction mainly relied on the transport of electrons from the electrode across the electrode-dielectric interface (following the Schottky model). When the V^(1/2) is in the extremely small range (< 0.1 V) in Fig. 5(c), the conduction of the thin film may follow the ohmic mechanism.⁴⁵ In this case, the ohmic conduction current intensity (I) is linearly dependent on the electric bias (V).

The temperature-dependent capacitance density for various capacitors from 25 °C to 100 °C is shown in Fig. 7. As a reference, a capacitor with only SiO₂ as the dielectric was prepared (Table I) and measured. Unlike the SiO₂ device, which has negligible capacitance variation (<0.13%), the capacitances of all VO₂-containing devices increase with increasing temperature. The distribution of the SiO₂ and VO₂ has a significant effect on the capacitance evolution. The capacitance density of the composite sample smoothly increases from 84.3 nF cm⁻² at 25 °C to 134.1 nF cm⁻² at 100 °C, an increase of ∼60% [Fig. 7(a)]. Such tunability is close to previous studies on ceramic/metal-polymer composite dielectrics.^46–48 Our VO₂-SiO₂ composite shows a smoothly varying thermal tunability of ∼60% spanning the T_MIT region. On the other hand, the VO₂ control layer displays a sharp capacitance density jump by a factor of 2 proximal to the transition region (563.1 nF cm⁻² at 70 °C vs. 1038.4 nF cm⁻² at 80 °C) [Fig. 7(b)].

Next, we consider the frequency dependence of the capacitance. Figures 8 and 9 show the logarithmic plots of capacitance density vs. frequency for different devices and temperatures. The SiO₂ control shows no significant change in capacitance as frequency is increased [Fig. 8(a)], which is consistent with previous reports.⁴⁹ By comparison, the capacitance density of all VO₂-containing samples shows a clear reduction with increasing frequency. At 25 °C, the composite layer capacitance densities decrease from 160 nF cm⁻² at 10⁴ Hz to 82 nF cm⁻² at 10⁶ Hz [Fig. 8(b)]. This frequency dependence becomes more significant at higher temperature. Similarly, the capacitance density of the VO₂ control layer and the superlattice sample also decreased with increasing frequency (600 nF cm⁻² to 300 nF cm⁻² and 350 nF cm⁻² to 225 nF cm⁻² at 25 °C, respectively) (Fig. 9). This AC frequency-dependent permittivity evolution has also been noted in ceramic-polymer composites.^50–53 

The electronic properties of the materials comprising the composite are summarized in Table II.⁵⁴ The frequency-dependent effective permittivity of the composite layer is shown in Fig. 10, which is much higher than that of the SiO₂ control, especially at low frequency and high temperature (effective permittivity of 8.5 and 16 at 25 and 100 °C [at 10⁴ Hz)].

Compared to SiO₂, the enhanced dielectric constant of the composite has several potential explanations. First, the introduction of higher dielectric constant VO₂ phase (compared to SiO₂) may increase the effective dielectric constant of the medium, especially in the metallic state of vanadium dioxide.²⁵ The second reason could be the enhanced interfacial polarization due to difference in electrical conductivity between SiO₂ and VO₂. Generally, in electrically heterogeneous composite materials, the large amount of charge accumulated at interfaces between the two different phases can result in an enhanced interfacial polarization density.⁵⁹ This can be explained by the Maxwell–Wagner–Sillars (MWS) effect.^60–63 The MWS mechanism is suited to explain the enhanced permittivity in composite dielectrics due to the different conduction properties in the two systems [like VO₂ and SiO₂ (Table II)]. The composite layer contains many SiO₂/VO₂ interfaces. By comparison, the VO₂ control layer possesses only two interfaces with the top and bottom SiO₂ layers and the superlattices (SiO₂/VO₂) have nine parallel planar interfaces. The largest number of VO₂/SiO₂ interfaces in the composite layer can qualitatively explain its most sensitive response to frequency change among the three types of structures (Table III). As temperature increases, the VO₂ phase becomes metallic, leading to greater difference in conductivity between the two media, which further enhances the MWS effect.

We used co-sputtering to fabricate nanocomposite VO₂-SiO₂ thin films with a thermally tunable and frequency-dependent dielectric constant. The nanocomposite dielectric film is compared to a VO₂ control layer and a VO₂/SiO₂ superlattice to systematically investigate the influence of structure and geometry on the resulting capacitance. The nanocomposite dielectric film shows smoothly tunable dielectric properties, resulting from the insulator-metal transition in VO₂. While leakage through the VO₂ in its conducting state will set the upper bound for acceptable loss in capacitors in eventual applications, further optimization of such tunable capacitors could be possible by careful spatial control of the inclusions, such as by lithographic patterning in an insulating matrix such as silica or alumina. The smoothly varying phase change behavior may also find applications in areas requiring continuous tunability of electrical, optical, or thermal properties across a range of temperatures.

Y.S., K.N., and S.R. acknowledge the Office of Naval Research (No. N00014-16-1-2398) and the U.S. National Science Foundation (No. DMR-1609898) for the financial support. M.A.K. acknowledges support from the Office of Naval Research (No. N00014-16-1-2556). X.S. and H.W. acknowledge the support from the U.S. National Science Foundation (No. DMR-1565822) for the TEM work. S.M. and K.C. acknowledge Ke Wang and Shih-Ying Yu for their assistance and advice.