Microstrip antenna tuning using variable reactive microelectromechanical systems

Due to their small size, low loss, and compatibility with integrated circuits, microelectromechanical systems (MEMS) have been the focus of recent efforts at developing frequency reconfigurable patch antennas able to operate at a number of different user-selected frequencies. Although reactive loading with MEMS variable capacitors is a well-documented method in reconfigurable antenna design [1], MEMS variable inductors have not received the same attention in that role. Designs for frequency tunable patch antennas using MEMS variable inductors and capacitors are presented here. Variations on previous thermally actuated MEMS variable inductors [2] and electrostatically actuated MEMS variable capacitors fabricated through a multi-user surface micromachining process are tested and shown to successfully vary reactance when actuated. Microstriped patch antennas loaded with discrete surface-mount reactive components were successfully tested to demonstrate the principle of this MEMS application.


Introduction
Microstrip patch antennas are desirable due to their ease of production, ruggedness, and conformability. However, such resonant antennas suffer from narrow bandwidths and fixed operating frequencies. Tuning the operating frequency of a patch antenna can overcome these limitations, allowing a single antenna design to serve a communications system operating on multiple channels.
Recent reconfigurable planar antenna designs have capitalized on the low power, small geometries, and high performance characteristics of microelectromechanical systems (MEMS) to develop frequency tunable patch antennas. Discrete frequency tuning has been achieved using MEMS switches that reconfigure antenna geometries [3], and continuous frequency tuning has been achieved by loading antenna structures with MEMS variable capacitors. Although reactive loading with MEMS variable capacitors is a welldocumented method in reconfigurable antenna design [1], MEMS variable inductors have not received the same attention in that role.
We present a continuous tuning method for a microstrip patch antenna using MEMS tunable inductors and capacitors. The tunable antennas studied in this research are designed for the 4-7 GHz range a 5.5 GHz unaltered operating frequency. The thermally actuated MEMS variable inductors and electrostatically actuated variable capacitors are variations of prior work on RF MEMS components [1]- [2]. The components are fabricated through the MetalMUMPs multiuser electroplated nickel micromachining process and are ultimately intended for use in a coplanar waveguide-fed patch antenna design described later in this paper.

Antenna Tuning Concept
Microstrip patch antennas are resonant antennas, operating best at frequencies tied to the antenna geometry. From a transmission line model perspective, the input impedance fluctuates between inductive and capacitive states with respect to frequency and resonates at that frequency at which the impedance is purely real. Theoretically, introducing lumped inductances and capacitances into an antenna circuit can offset reactances, making the input impedance of a patch antenna purely real at a frequency other than the original resonant frequency.
To demonstrate this concept, a 5.5 GHz microstrip-fed rectangular patch antenna was designed using well-known equations [4]. One variation of the antenna, shown in Fig. 1, is fed by a single 50- microstrip feedline matched with a quarter wave transformer. The actual antenna operating frequency was measured to be 5.53 GHz. A second variation, shown in Fig. 2, introduces a 12.5-mil gap and 20-mil by 29-mil contact pads in the 50- feedline to accommodate 0302 size surface mount inductors and 0402 size surface mount capacitors. Simulated reflection coefficient (S11) characteristics for various loaded and unloaded antenna were obtained using Sonnet High Frequency Electromagnetic Software and are shown below in Fig. 3. The S11 characteristic typically indicates an antenna's effective operating frequency. The simulated inductive and capacitive loads are modeled in software as ideal components. The simulation results shown in Fig. 3 suggest that tuning can be achieved by reactive loading along a microstrip feedline. Loading with a 6.8-pF capacitor shifts the apparent effective operating frequency of the antenna slightly lower than the unloaded operating frequency while inductive loads increase the effective operating frequency. A 7.4-nH inductance results in a simulated operating frequency of nearly 6 GHz.
However, it is notable that increasing inductance decreases the operating frequency, bringing the effective operating frequency closer to the unloaded operating frequency. This phenomenon may be a result of unaccounted for impedance transformation due to the transmission lines and quarter wavelength transformer in the feedline to the antenna.

Antenna Tuning Results
Various surface mount components were soldered onto the aforementioned antennas, which were subsequently tested. The measured reflection coefficient characteristics of the unloaded antenna and loaded antenna are shown in Fig. 4. Using an Agilent E5071A network analyzer, the S11 minimum is observed at 7.27 GHz when the antenna is loaded with a 6.5-nH inductor and at 5.50 GHz when loaded with a 6.8-pF capacitor. Except for when loaded with the 6.5-nH inductor, the antenna S11 characteristic matched simulated results within 5 dB and demonstrated similar qualitative behavior.
When loaded with a 6.5-nH inductor, the antenna minimum S11 differs in frequency from the expected 6 GHz operating frequency by as much as 1.27 GHz. Also, the simulated weaker resonance around 5.35 GHz in the 6.5-nH loaded antenna is reflected in the measurement of a more pronounced resonant frequency at 5.4 GHz. The difference between simulated and measured S11 characteristics may be a result of oversimplified models of the surface mount inductors and capacitors. The ideal inductances and capacitances used to model the loading components contain resistances and parasitic reactances that can offset matching results.
The loaded and unloaded antennas were also tested in an anechoic chamber, and radiation patterns were recorded. The Figure 4: Measured S11 of loaded and unloaded antennas results from the radiation patterns more closely align with simulated results. The azimuth-oriented radiation patterns for the unloaded, 6.5-nH loaded, 7.4-nH loaded, and 6.8-pF loaded antennas at their frequencies of greatest gain are shown in Fig. 5, 6, 7, and 8, respectively. The unloaded antenna was found to resonate best at 5.54 GHz. The 7.4-nH and 6.8-pF loaded antennas resonated at 5.99 GHz and 5.49 GHz, respectively, as predicted. The 6.5-nH antenna maintained a largely resonant antenna pattern in the 5.5 to 7.25 Ghz range, achieving its best performance at 5.53 GHz and 6.3 GHz.  As observed through the radiation pattern measurements, the resonant frequencies of the loaded antennas more closely align with simulated results. In particular, the observed resonance of the 6.5-nH loaded antenna at 6.36 GHz represents only a 6% difference between simulated and measured resonant frequency. The 7.27 GHz resonance suggested by the S11 characteristic represents a 21% difference between simulation and measurement. At resonance, each antenna radiation pattern matches the general expected behavior for rectangular patch antennas.
Overall, the measured antenna results support the concept of antenna tuning using inductive and capacitive loads. Using discrete surface mount inductors and capacitors between 6.8 pF and 7.4 nH, a tuning range between 5.4 GHz and 6.3 GHz was achieved. The objective of the MEMS components presented in this paper is to replace the aforementioned discrete reactive components with variable ones in order to achieve a tunable patch antenna design.

MEMS Variable Capacitors Design
The MEMS capacitors reported here feature nickel and polysilicon layers as the movable plates of a variable parallel plate capacitor. The polysilicon layer is embedded in silicon nitride for electrical isolation and suspended over a trench in the silicon substrate. The nickel capacitor plate is suspended over and perpendicular to the polysilicon layer. This nickel bridge also serves as the RF signal line for the capacitor.
To achieve electrostatic actuation, the polysilicon-silicon nitride membranes include embedded, isolated patches of polysilicon at each end of the membrane bridge. These patches lie beneath the suspended edge of grounded nickel structures that flank each side of the nickel capacitor bridge as shown in Fig. 9. When a DC voltage is applied to the embedded, isolated patches of polysilicon, an electrostatic force pulls the entire polysilicon-silicon nitride bridge up, decreasing the distance between the nickel capacitor bridge and the center polysilicon plate. This variable distance between the nickel bridge and the center polysilicon plate results in a change in capacitance seen by an RF signal passing through the nickel bridge. The design choice for electrically isolated polysilicon electrodes in the underlying membrane was driven by the desire to not overlay the DC control voltage on the signal line and the need for a conductive signal line material. As this project was constrained to use of the MetalMUMPS process, nickel was chosen as the signal line material, and the isolated control electrodes in the underlying membrane allow variability without directly affecting any of the structures that comprise the capacitor except the distance between them.
The capacitance C of the described MEMS device was derived from the parallel plate capacitor model and is given by the equation: where ε r is the relative dielectric constant between the nickel signal line and the embedded center polysilicon plane, and t d is thickness of the insulating silicon nitride layer between the polysilicon and the nickel. This design equation does not account for fringing capacitance, which will increase the overall capacitance of the component. The electrostatic force that pulls up the polysilicon-silicon nitride membrane is implemented by applying a DC voltage to the flanking polysilicon plates and grounding the flanking nickel structures that overhang them. To hold the membrane up, the electrostatic force must exceed the mechanical restoring force, requiring a pull-up voltage V pu given by: where ε f accounts for capacitance losses due to roughness on the nickel and silicon nitride surfaces; g represents the distance between the plates, and g 0 is the non-actuated distance between the plates. A a is the effective area of the overlapping nickel and polysilicon actuating structures. The expression for the restoring force, k e (g 0 -g), is given by: where k is the partial spring constant of a fixed-fixed beam given by: where E is Young's modulus of the polysilicon membrane. w is the width of the membrane. l is the length, and t is the thickness of the membrane. x is half the length over which the actuating force is applied. σ represents the biaxial residual stress, and ν is Poisson's ratio.
The constant for the non-linear term in the restoring force is given by: The equivalent Young's modulus, Poisson's ratio, and residual stress for the polysilicon-nitride membrane is the weighted volumetric average of the various layers. Thus, Young's modulus is given by : where E n and tn are the Young's modulus and thickness of each individual layer.
The dimensions for the capacitors featured in this paper were based on equations 1-6, all of which are documented in [5]. The targeted tunable capacitance range for the capacitors is between 0.75 pF to 1.5 pF for actuation voltages up to 6 V. While a wider capacitance range is possible by considering flattening effects after pull-up, this operation has the disadvantage of non-continuous tuning and the possibility of the polysilicon-silicon nitride bridge acting as a capacitive switch, blocking RF transmission across the nickel signal line. The intent is to operate with actuation voltages below the pull-up voltage that will cause the polysilicon plate to snap into contact with the nickel signal line.

MEMS Variable Inductors Design
The MEMS inductors reported here are based on a design by I. Zine-El-Abidine and M. Okoniewski [2] and feature a pair of suspended pre-bent coplanar nickel beams that buckle away from each other under thermally actuated mechanical force to achieve a variable mutual inductance. The beams are electrically connected at the end closest to the actuating mechanism as shown in Fig. 10, and the RF signal travels through each beam. An array of differential thermal actuators provides a lateral force when a voltage difference is applied across the actuators. A silicon nitride bridge suspended over a trench serves as electrical isolation and mechanical connection between the actuating mechanism and the inductor beams. As the lateral force from the actuators is transferred to the inductor, the beams bow outwards, increasing the area of the loop they form and increasing their mutual inductance.

Fabrication
The antenna structures featured here are fabricated using a photolithography and wet etching process on 30-mil thick, 1oz. copper electrodeposited and rolled ULTRALAM ® 2000 high frequency laminate substrate (ε r = 2.5, tan δ = 0.0022). Positive resist is spin-coated onto the substrate, which is then selectively exposed to ultraviolet light using chrome-on-glass masks and developed. After etching with ferric chloride, the electrodeposited copper layer only remains on those areas of the substrate not exposed to the ultraviolet light. The MEMS components are fabricated through the multiuser nickel electroplated micromachining MetalMUMPs process [3]. The process allows for two conducting layers: a 21-m thick electroplated nickel layer and a 0.7-m thick polysilicon layer on an n-type silicon substrate. Two 0.35-m thick silicon nitride layers can be used for structural support and electrical isolation between the polysilicon and nickel layers. Three sacrificial oxide layers allow for trench etching in the silicon substrate and physical separation of the conducting layers. A microscope photograph of the featured capacitor is shown below in Fig. 11, and photographs of the inductor actuating mechanism and buckling beams are shown in Fig. 12 (a) and (b), respectively.

Measured Results
The MEMS capacitors and inductors were mounted onto a test circuit board using epoxy and then wire bonded to control lines on the circuit boards. The devices were then tested under varying DC control inputs using an Agilent E5071A network analyzer. The measured capacitance under actuation voltages from 0-110 V is shown in Fig. 13.
The capacitor demonstrated an increase in capacitance between 0.1 and 0.25 pF with increasing actuation voltage over the range of 0-20 V at frequencies greater than 4.75 GHz and less than 6 GHz. Beyond 20 V, the measured capacitance remained constant, indicating that the membrane and the overhanging nickel structure had made full contact.
The higher than designed capacitance values observed in Fig. 13 may result from fringing and from self-actuation caused by the RF signal voltage on the nickel capacitor plate. The latter also may contribute to the small variability in device capacitance. Another issue that came to light during testing was the need to discharge the control electrodes after actuation. Under the current configuration, no internal safe discharge method exists. An external resistor was used to help discharge the structures during testing.
Measurements for the MEMS inductor are shown below in Fig. 14 under no actuation and an actuating voltage of 25 V. When actuated with 25 V, the MEMS inductors drew 35 mA of current. In both the actuated and non-actuated states, the inductor appeared capacitive at frequencies lower than 3.25 GHz and higher than 5.5 GHz. It is notable that while the inventors of this inductor design reported an increase in inductance upon actuation [2], we have observed the opposite effect. The successful outward buckling of the inductor beams was observed using a probe station and a microscope, but the inductance measured using a network analyzer showed a clear decrease upon actuation.
Damage to the bridge structure coupling the inductor and actuator mechanically during fabrication is the most probable source of this unanticipated behavior. The bridge was never fully intact on any of the tested inductors. The reverse behavior upon actuation could be related to the introduction of the increasing capacitive effect of a nickel actuator bridge moving closer to the inductor structure. Damage to the bridge structure could also result in electrical connection between the inductor beams and the actuator array, further distorting results.
Also, the inductance change per actuating voltage is much smaller than originally anticipated. This phenomenon can also be traced to the damaged bridge structure. Because a direct mechanical connection is not immediately made at the bridge, a much larger displacement by the actuators is required before a lateral force can be brought to bear on the buckling beams.
Despite these issues, both the MEMS capacitor and inductor designs have successfully demonstrated the capability to vary reactance. Although more testing and an improved inductor bridge are necessary to factor out parasitic effects and achieve desired behavior, the devices described here will operate as variable reactive loads that may be able to successfully tune a patch antenna.

Conclusions
Microstrip-fed patch antennas loaded with discrete surface-mount reactive components were successfully tested to demonstrate the principle of tuning antennas by loading them with reactive MEMS components. A MEMS capacitor and inductor are presented and tested as potential variable components in loading and tuning an antenna. While the MEMS components are able to achieve changes in inductance and capacitance, damaged structures and the potentially strong parasitic effects resulting from suboptimal packaging have produced unanticipated results. Ultimately, a second design, fabrication, and test of the MEMS components with a focus on reliability, packaging, and integration with antenna designs are desired.