Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.

To obtain low insertion loss in the switch’s “on” state, a cascaded electrothermal buckle-beam actuator was used. This actuator can provide large displacements and contact forces, achieving the lower contact resistance which has a close relationship with the insertion loss in the switch’s “on” state. Also, a deep trench was etched below the contact area of the switch, reducing the coupling capacitance when a signal goes through the substrate; this further increases the isolation of the switch. Simultaneously, the contact area used sputtered gold as the contact metal to further reduce the insertion loss of the switch.

MEMS Switch Manages Millimeter-Wave Signals, Fig. 3

Figure 3 shows structure-based lumped-circuit models for the MEMS series switch. The contacts are modeled as resistors or capacitors, depending on the state of the switch. The lateral contact structure has two contact areasbetween the input and output ports. The coupling capacitance contains two series of capacitance Cc in parallel with capacitance Cg. Capacitance Cc is the “off”-state capacitance due to the distance between the overlapping part between the switch’s contact plate and each broken signal line.

Capacitance Cg refers to the signal-line coupling capacitance due to the gap between each broken signal line. Resistance Rc is one-half of the contact resistance when the switch is in its “on” state. The switch’s “on” and “off” states are equivalent to switching between Rc and Cc. The isolation of the switch’s “off” state increased with an increase in the contact gap between the switch’s RF contact plate and the signal line, as shown in Fig. 4.

MEMS Switch Manages Millimeter-Wave Signals, Fig. 4

Switching between the two states is done by means of a cascaded electrothermal buckle-beam actuator. Figure 5 shows the actuator design, where three V-shaped actuator beams of the same dimension are cascaded. The width of the contact beam (W1) is 60 μm. Eachsingle bent beam has a length Lt of 400 μm, a width W2 of 10 μm, and beam angles α equal to 7deg.These parameters were chosen based on optimization of a mathematical model.

MEMS Switch Manages Millimeter-Wave Signals, Fig. 5

The finite-element method (FEM) method with HFSS software was used to analyze the thermal actuator. The effects of temperature-dependent material properties and thermal radiation were ignored. Figure 6 shows the simulation results of displacement and temperature versus driven voltage for the thermal actuator. Figure 7 shows that the contact point can achieve a displacement of more than 12 μm by means of a voltage of 0.12 V.

MEMS Switch Manages Millimeter-Wave Signals, Fig. 6

MEMS Switch Manages Millimeter-Wave Signals, Fig. 7

The simulated MEMS switch was manufactured using MetalMumps technology.4 The switch was fabricated on a high-resistivity silicon substrate (with resistivity of greater than 4000 Ω-cm). Figure 8 shows the process flow for the fabricated switch. In this flow, 2-μm-thick oxide was grown on the surface of the starting n-type (100) silicon wafer (a). This is followed by deposition of a 0.5-μm-thick sacrificial phosphosilicate glass (PSG) layer (Oxide 1). Wet chemical etching was used to remove the unwanted sacrificial PSG. Oxide 1 also defines the regions in which the silicon trench will be formed.

MEMS Switch Manages Millimeter-Wave Signals, Fig. 8

In this process flow, two 0.35-μm-thick layers of silicon nitride were deposited as mechanical support to link the actuator and the contact plate (b). Then, RIE etching was performed to remove nitride from the patterned areas where not needed (c). Following this, a second sacrificial layer (Oxide 2), 1.1-μm-thick PSG, was deposited and wet etched and a thin metal layer  was deposited (d). In the next step (e), the plating base layer, consisting of 500-nm Cu + 50-nm Ti, was deposited (not shown in Fig. 8). 

The wafers were coated with a thick layer of photoresist and patterned to form a stencil for the electroplated metal layer. Nickel was then electroplated to a nominal thickness of 20 μm into the patterned resist stencil (f). The photoresist stencil was then chemically removed (g). A 1 to 3 μm gold layer (sidewall metal) was electroplated to provide a low resistance contact (h). Then, a 49% HF solution was used to remove the PSG sacrificial layers (Oxides 1 and 2) and the oxide layer over the trench areas (i).

Finally, a KOH silicon etch was used to form a 25-μm-deep trench in the silicon substrate in the areas defined by the Oxide 1 (j). The SEM image for the fabricated switch is shown in Fig. 9, with a gap between the signal lines of 100 μm. The entire switch measures about  0.7 x 0.8  mm2.

MEMS Switch Manages Millimeter-Wave Signals, Fig. 9

Gold (Au) was chosen in this process as the contact metal because of its low resistivity, good stability, and high efficiency for RF signal propagation. The main problem with Au is its propensity toward high adhesion, which may lead to failure if restoring forces are not large enough to break a contact.5 Due to the large restoring force of the MEMS switch, no adhesion-related problems should be experienced.

Download this article in .PDF format
This file type includes high resolution graphics and schematics when applicable.