Jin-Woo
Han
*,
Myeong-Lok
Seol
and
M.
Meyyappan
NASA Ames Research Center, Moffett Field, CA 94035, USA. E-mail: jin-woo.han@nasa.gov
First published on 1st February 2021
A nanoscale field emission vacuum channel gated diode structure is proposed and a tungsten cathode with an umbrella-like geometry and sharp vertical edge is fabricated. The edge of the suspended cathode becomes the field emission surface. Unlike in the traditional transistor with the gate typically located between the source and the drain, the bottom silicon plate becomes the gate here and the anode terminal is located between the umbrella cathode and the gate. The fabricated devices show excellent diode characteristics and the gated diode structure is attractive for extremely low gate leakage.
In this work, a gated diode structure with a 100 nm anode–cathode gap is experimentally demonstrated. The device features an inverse vertical channel with two major distinctions. First, an innovative fabrication approach makes the cathode have an umbrella-like shape with a sharp-edged rim. Inherent deposition and etching process steps alone allow producing the sharp cathode without any additional dedicated processes to sharpen the corner. Second, the gate is separated from the source (or the cathode) farther than the drain (or the anode), which results in great reduction of the gate leakage. The fabrication details and the measured device characteristics are provided and complemented by multi-physics simulations to gain insight.
The dimensions of the device are labeled in the step iv panel of Fig. 1a. The thicknesses of the gate, anode and cathode are 100 nm, 20 nm and 100 nm respectively. The spacer distances are 40 nm between the gate and anode and 100 nm between the anode and cathode. The separation of the cathode and gate is 120 nm and the overhang of the cathode over the anode is 100 nm. These dimensions are not necessarily optimal, as no attempt was made to optimize the performance with an extensive parametric study.
Fig. 2 shows the results from the electron propagation trajectory simulation of the umbrella cathode VFEGD with conventional and proposed terminal assignments. A finite element solver, COMSOL Multiphysics was used for the simulation. The simulated device dimensions are the same as in the fabricated device shown in Fig. 1. All the material and model parameters were set at the default values in the Multiphysics solver. Fig. 2a shows the traditional VFET where the gate is sandwiched between the source and the drain as in any transistor. Fig. 2b is the proposed VFEGD with the gate placed under the anode. At the off-state (Vg = 0 V), the off-state leakage electrons in both devices are collected on the drain or anode terminal. The trajectory in Fig. 2a is strongly bent undesirably towards the gate electrode as the gate voltage Vg increases with an increased electric field in that direction consequently attracting electrons. As a result, the gate leakage can significantly increase with Vg and the gate current can even be greater than the drain current when the gate is located in the middle of the channel. In Fig. 2b, however, the electron trajectory stays towards the anode direction even when Vg increases. The gate leakage current can be minimal when the gate is located under the anode; the anode captures most of the electrons before they reach the gate, which is reflected in the current–voltage (I–V) behavior discussed below.
Fig. 3 and 4 show the measured current–voltage characteristics of the umbrella-like cathode device explained in Fig. 1 and 2. The same curve can be interpreted in two different ways, i.e. behavior of the gated-diode (VFEGD) or the transistor (VFET), depending on how the terminal is defined as illustrated in Fig. 2a and b. The umbrella-like cathode on top is always the electron emitter or source. When the device is considered as a VFEGD, the middle n+ polySi and the bottom n+ Si become the anode and the gate, respectively. When the device is read as VFET, the middle n+ polySi and the bottom n+ Si become the gate and the drain, respectively. Fig. 3a and b respectively show the anode current versus gate voltage (Ia–Vg) and anode current versus anode voltage (Ia–Va) characteristics measured from the VFEGD shown in Fig. 2b. When the same device and measured characteristics are read as VFET, Fig. 3a is then regarded as gate current versus drain voltage (Ig–Vd) and Fig. 3b is the gate leakage current versus gate voltage (Ig–Vg) as illustrated in Fig. 2a. The high current in Fig. 3 indicates large gate leakage for the traditional terminal mode and VFET, which has been reported before4,5,14 with gate leakage even as much as 10% of the drain current.5 In addition, when the device is considered as a VFET, Fig. 3a shows an incomplete shut-off and an on-state to off-state current ratio of 2 for Vd = 20 V and 50 for Vd = 10 V. Thus, these characteristics are not ideal for the transistor mode. However, when the device is considered as a diode, the observed anode current and its dependence on the anode voltage are as expected.9–13Fig. 3b shows a typical diode curve with no current flowing below the turn-on voltage. Electron emission occurs above the diode turn-on voltage through the Fowler–Nordheim tunneling mechanism.15 This switching feature can be utilized in diode–diode logic, rectifiers, clamps and circuit breakers. The turn-on voltage (defined as the voltage needed to generate 1 nA anode current) is approximately 7.8 V, which is determined by the work function of the cathode and device geometry. Further device miniaturization and use of low work function cathode16 can lower the turn-on voltage. The diode turn-on voltage is also modulated by the gate voltage. Comparing the diode curve at Vg = 10 V, the diode turn-on voltage is reduced by approximately 400 mV from Vg = 20 V. Such diode turn-on voltage modulation by gate voltage could be useful when fine-tuning is necessary in order to compensate for any process and temperature induced variability. Table 1 provides a comparison of the present results with vacuum diodes previously reported in the literature, indicating comparable performance of our fabricated device.
Fig. 3 Measured device characteristics. (a) Anode current versus gate voltage and (b) anode current versus anode voltage for the proposed terminal assignment shown in Fig. 2b. |
Fig. 4 Measured gate current versus gate voltage for the proposed terminal assignment shown in Fig. 2b. |
Cathode | Channel gap, nm | Turn on voltage | Anode current | Ref. |
---|---|---|---|---|
a Indicates availability of multiple values for the gap in the cited work. b Current was deliberately limited to this value to avoid overheating. | ||||
SiC | 100a | 3.2 V | 22.3 nA@5 V | 8 |
VO2 | 10 | 0.5 V | 53 μA@1 V | 9 |
Cu | 20a | 1.75 V | 32.5 nA@5 V | 10 |
Al–Si–Cu (0.995 Al) 4 × 4 μm | 60 | — | 24.6 μA@10 V | 11 |
SnSe anoflower | 100 | 0.4 V | 550 A cm−2@0.75 V | 12 |
Tungsten | 95 | 13.0 | 100 nAb | 4 |
Tungsten | 100 | 7.8 V | 100 nA@20 V | This work |
A slightly weak dependence of the anode current on the gate voltage in the VFEGD mode is due to the large separation between the cathode and the gate here. A noticeable benefit on the other hand is that the device is intrinsically free from gate leakage. Fig. 4 shows the gate leakage current versus gate voltage characteristics. This sample plot can also be considered as the drain current versus drain voltage characteristics for the traditional VFET mode. Very low gate leakage in Fig. 4 can be translated into no drain current occurring in the VFET mode as all the electrons are intercepted by the gate as illustrated in Fig. 2a. In other words, no gate current is seen in the proposed diode mode as all of the field emission current is dominantly collected by the anode, as illustrated in Fig. 2b. Furthermore, no static power consumption is involved for biasing the gate as evidenced in Fig. 4.
The extent of gate modulation of the diode current could be controlled by device optimization. For example, the overhang distance of the emission edge and its vertical extension length could be a knob. This will make the emission surface closer to the gate but not too aggressively for emission current to be intercepted by the gate. By doing so, the gate with zero volt will be efficient enough to shut-off the anode current; consequently the on- and off-anode current ratio would be several orders of magnitude higher, and then the same architecture could be used as a transistor.
This journal is © The Royal Society of Chemistry 2021 |