Nano-Particle VO₂ Insulator-Metal Transition Field-Effect Switch with 42 mV/decade Sub-Threshold Slope

Abstract

The possibility of controlling the insulator-to-metal transition (IMT) in nano-particle VO₂ (NP-VO₂) using the electric field effect in a metal-oxide-VO₂ field-effect transistor (MOVFET) at room temperature was investigated for the first time. The IMT induced by current in NP-VO₂ is a function of nano-particle size and was studied first using the conducting atomic force microscope (cAFM) current-voltage (I-V) measurements. NP-VO₂ switching threshold voltage (V_T), leakage current (I_leakage), and the sub-threshold slope of their conductivity (S_c) were all determined. The cAFM data had a large scatter. However, V_T increased as a function of particle height (h) approximately as V_T(V) = 0.034 h, while I_leakage decreased as a function of h approximately as I_leakage (A) = 3.4 × 10⁻⁸e^−h/9.1. Thus, an asymptotic leakage current of 34 nA at zero particle size and a tunneling (carrier) decay constant of ~9.1 nm were determined. S_c increased as a function of h approximately as S_c (mV/decade) = 2.1 × 10⁻³e^h/6 and was around 0.6 mV/decade at h~34 nm. MOVFETs composed of Pt drain, source and gate electrodes, HfO₂ gate oxide, and NP-VO₂ channels were then fabricated and showed gate voltage dependent drain-source switching voltage and current (I_DS). The subthreshold slope (S_t) of drain-source current (I_DS) varied from 42 mV/decade at V_G = −5 V to 54 mV/decade at V_G = +5 V.

1. Introduction

Vanadium dioxide (VO₂) is a model insulator–metal transition (IMT) material that displays a first-order transition from a monoclinic insulating phase to a tetragonal metallic phase at a critical temperature (T_C) of 341 K in its bulk form [1]. This transition in addition to the semiconductor field-effect in VO₂ provides an opportunity to potentially realize switches with steep subthreshold slope of better than 60 mV/decade of silicon for energy efficient devices and applications.

In addition to heat (temperature), the IMT in VO₂ can be induced by charge injection, light, mechanical stress, terahertz signals, and many other stimuli. In early reports, the charge-induced IMT was ascribed to local heating of the VO₂ material over the critical temperature (T_IMT) by the current flowing through the device [2]. Recently, however, other non-thermal mechanisms of the voltage-triggered IMT were proposed [3,4,5]. Specifically, the electric field, rather than local dissipated power due to Joule heating, was suggested to be the origin of the IMT in some cases [2,4,5,6]. Charge injection induced IMT in VO₂ was reported in the past and was used to realize negative differential gate capacitance to improve the subthreshold current slope (S_t) [4]. Additionally, the investigation of electric field effect on the IMT has gained importance as a field-controlled IMT effect in fast and reliable electronic devices.

Our main motivation in this study is to develop field-effect transistor (FET) devices with very steep sub-threshold current slopes for energy efficient switches. In crystalline VO₂ the insulating phase carrier concentration is quite high on the order of 10²¹ cm⁻³. Therefore, the effect of gate electric field in modulating the I_DS is small since the channel has large conductivity to begin with. In VO₂ nano-particles and polycrystalline films, small crystalline regions are separated by grain boundary regions that have lower conductivity (n~ 10¹⁷–10¹⁸ cm⁻³) [7]. In these materials, the electric field-effect in the boundary regions appears to control the channel turn-on voltage and the sub-threshold current slope as discussed here. The electric field effect and the transition temperature can be related to each other. We have recently shown that the transition temperature in VO₂ reduces at higher applied gate voltages, and the relationship is approximately given by: T_transition = 1.2 V_G + 63.8 °C where V_G < 0 [7].

Recently, there has been an increased interest in developing high energy efficiency electronic switches using transition metal dichalcogenides [8] and IMT materials [9,10]. The grain size in polycrystalline VO₂ affects the IMT temperature as discussed in [11]. We note that to realize efficient transistors, IMT materials are great candidates, since in addition to providing the usual semiconducting carrier control mechanism, they also provide very steep IMT transitions. VO₂ has a transition temperature of 63 °C and is used here to demonstrate the feasibility of metal-oxide-VO₂ field-effect transistor (MOVFET). A better channel material is Cr-doped V₂O₃ with possible transition temperature of above 150 °C [12].

2. Nano-Particle Studies

VO₂ nano-particles, obtained from Strem Chemicals (Newburyport, MA, USA) [13], were deposited on a gold covered oxidized silicon chip for atomic force microscope (cAFM) studies. To fabricate the chips, we started with a p-type silicon wafer that was cleaned using piranha and buffered oxide etches. Subsequently, the silicon was oxidized (~90 nm) using wet thermal oxidation and then it was coated with 100 nm of Au on 20 nm of Ti adhesion layer. Finally, the wafer was diced into 1 cm² square chips suitable for cAFM scans and measurements (Figure 1). The VO₂ powder was mixed with ethanol and ultra-sonicated for 3 min to prevent agglomeration. The mixture was poured over the central region of the gold-covered silicon chip (Figure 1b) and allowed to dry to obtain evenly spread VO₂ particles over the sample. Conductive atomic force microscopy (cAFM) measurements were then carried out on a Multimode AFM using a Pt conducting tip connected to a semiconductor parameter analyzer for I-V measurements (Figure 1a). To measure the thickness of each particle, we used the metrology capability of the AFM system that provided a numerical value for particle height. We assumed that the particles are spherical and used the particle height as its diameter. The I-V measurements were carried out by locating a VO₂ particle using the cAFM probe, which also acted as the top contact to the particle with the gold substrate as the bottom contact. All experiments were done at room temperature.

Figure 1c shows a representative AFM scan performed over the Au substrate containing VO₂ particles. A wide distribution of particle heights was obtained during a single scan. Particle heights were determined using AFM line scans shown by dashed lines in Figure 1c and displayed with numerical values by the AFM software in Figure 1d.

Figure 2 shows I-V measurement results obtained for 48 nm and 35 nm particles. Sharp transitions in I-Vs were observed in the NP-VO₂ ranging in size from 13 nm to 53 nm with turn-on voltages (V_T) ranging from 0.5–2.0 V. The leakage current and the sub-threshold slopes were calculated for each nano-particle at V = V_T/2 and V = V_T, respectively, as shown in Figure 2b,d.

Figure 3a shows V_T (defined in Figure 2) as a function of the particle size. Assuming that V_T = 0 V for h = 0 nm, the scattered data tentatively fits a line given by V_T (V) = 0.034 h with “h” in nm. Therefore, the critical transition field (E_c = V_T/h) is E_c ~ 3.44 × 10⁵ V/cm. The leakage current as a function of particle size is shown in Figure 3b and has an approximate dependence given by I_leakage(A) = 3.4 × 10⁻⁸e^−h/9.1. This indicates an asymptotic leakage current of 34 nA at zero particle size and tunneling (carrier) decay constant of 9.1 nm inside the NP-VO₂.

Next, we examined the slope of the I-V at V_T/2 (see Figure 2b). This slope can be viewed as the switching current slope denoted by S_c. S_c increased as a function of particle size and was approximately given by S_c (mV/decade) = 2.1 × 10⁻³e^h/6 as shown in Figure 3c and was around 0.6 mV/decade for particle size of 34 nm. The slope of I-V curve indicated by S_c is different than the slope of the MOVFET I_DS versus V_G curve indicated by S_t. The exponential dependence of S_c on particle size indicates that as the particle size becomes larger, the conduction tunnel paths and energies through the particle become more numerous leading to shallower I-V switching curve.

3. Device Studies

We next incorporated the NP-VO₂ in the channel of an FET that were fabricated on a glass substrate with a 100 nm Pt gate covered by a 50 nm atomic layer deposited (ALD) HfO₂ dielectric and a 100 nm Pt drain and source electrodes, as shown in Figure 4a,b. The fabrication process is discussed in [14] and was started with etching a 4” glass substrate by immersing it in buffered oxide etch (BOE) for 1 min to create 100 nm deep trenches for the gate metallization regions. After 100 nm Pt gate metal deposition and patterning, 50 nm-thick HfO₂ was deposited using atomic layer deposition technique as gate oxide. 100 nm of Pt was then sputter deposited and patterned to create the source-drain regions. Figure 4b,c show the SEM image of the fabricated open-channel FET device.

The open-channel device geometry enables any material to be deposited in the channel region of the FET. To form the NP-VO₂ channel, we mixed the VO₂ powder with n-butyl-acetate and a small amount of silver paint (1% to 10% in weight) and then ultrasonicated the mixture that was subsequently deposited using a fine brush in the MOVFET’s channel region, as shown in Figure 4d.

Figure 5a shows a typical I_DS-V_DS characteristic of the MOVFET. The transition from insulator to metal results in the very sharp increase in the I_DS as a function of increasing V_DS. I_DS transitions for decreasing V_DS are also very sharp. In these experiments the I_DS was limited to 0.5 mA by the instrument to prevent device breakdown. The I_DS-ON/I_DS-OFF in these devices, measured using pulsed voltages, were higher than 1000, but we limited the current in static measurements. Figure 5b shows the sub-threshold current as a function of V_G. The sub-threshold slope increased as a function of gate voltage. Figure 5c shows the I_DS at V_T/2 (= I_leakage) as a function of V_G, clearly indicating that the leakage current is lower at the negative gate voltages compared to zero and positive gate voltages. We also note that positive gate voltages increased the V_T (Figure 5d), while the negative gate voltage reduced it. The V_DS voltage step in these experiments was 100 mV.

4. Discussion

The main charge carriers in VO₂ are believed to be electrons [2,15]. It is also assumed that the IMT is triggered by the onset of a critical density of electrons in the channel [2,16]. Thus, one expects the threshold voltage in the MOVFET to become smaller at more positive gate voltages. However, we note that just the opposite is observed in Figure 5c. We also note in Figure 3a that in nano-particles, the transition voltage became smaller in smaller particles. Putting these two observations together, we conclude that the gate electric field in our MOVFET changes the current path between the drain and source, as schematically shown by the white and gray (red) arrows in Figure 4e. When the gate voltage is positive, the conduction path is much wider than when the gate voltage is negative. A negative gate voltage “pushes” the conduction path away from the channel‒gate interface region and confines it to a thinner layer at the top. Thus, the conduction occurs over the smaller channel cross section. The V_G dependence of all other parameters (S_t, S_c and I_leakage) agree with this observation.

We also note that other groups working with single crystal thin-film VO₂ reported difficulty demonstrating gate field effect [17]. This is attributed to the presence of high carrier (electron) concentration ~10¹⁸–10²¹ cm⁻³ [7] in the insulator phase of the VO₂ that is difficult to modulate and requires very high breakdown gate insulator material. In our devices, we used NP-VO₂ that also show the same levels of electron concentrations as calculated from the leakage currents of Figure 2. However, nano-particles couple to each other through their outer boundary layers with lower electron concentrations. The relatively weak coupling between NP-VO₂ enables the gate field effect to modulate the current path that gives rise to the modulation of the V_T, I_leakage, and S_t.

While the exact origin of the field-effect induced IMT is still under investigation, it is believed to be non-thermal. Very fast switching (10 ns) in optimized two-terminal devices with thin-film VO₂ active regions is also reported [18] indicating the possibility of non-thermal switching mechanisms in these materials.

5. Conclusions

The current-induced IMT in NP-VO₂ was investigated using cAFM measurements. The transition voltage in NP-VO₂ decreased while the leakage current increased as a function of the particle size. These measurements were followed by construction of NP-VO₂ FET devices and the related I_DS-V_DS/V_G measurements. We used open-channel FET devices and drop-casted NP-VO₂ to form their channels and measured the resulting MOVFET characteristics that showed gate voltage dependent I_DS-V_DS.