MEMS VARIABLE CAPACITOR WITH SUPERIOR LINEARITY AND LARGE TUNING RATIO BY MOVING THE PLATE TO THE INCREASING-GAP DIRECTION Chang-Hoon Han, Dong-Hoon Choi, Seon-Jin Choi and Jun-Bo Yoon Department of Electrical Engineering, KAIST 373-1 Guseong-dong, Yuseong-gu, Daejeon 305-701, Republic of Korea response to the noise signal whose frequency exceeds the mechanical resonant frequency of the moving actuator. Thanks to the electrostatic actuation property, we do not need to consider the forward biasing problem which is critical supply voltage limit of p-n junction varactors [3]. However, common parallel-plate MEMS variable capacitors show non-linear C-V response that most of the capacitance change occurs near the highest control voltage region because they increase the capacitance value by closing the gap between two parallel plates. This inherent non-linearity is one of the main factors which result in non-constant phase noise over the frequency tuning range in VCOs [4] and loop bandwidth variation in LC-VCO based PLL [5]. These properties make it difficult to set the voltage for achieving a desired capacitance value. Recently, a few MEMS variable capacitors showed high linearity factor (98.5 ~ 99.6%) and larger capacitance tuning ratio than the conventional parallel-plate type devices [6-7]. However, it was difficult to design capacitance value and control voltage range since their main approach was increasing the mechanical spring constant as the applied voltage closes to the pull-in point. Therefore, they needed to get through several trial and errors for optimization. In this work, we paid our attention to the direction of the moving plate. Opposing to the common closing-gap direction, we found that by moving the plate simply to the increasing-gap direction, we could achieve excellent C-V linearity and large tuning ratio simultaneously. ABSTRACT This paper reports an innovative and simple method to achieve highly linear capacitance vs. voltage response and large tuning ratio in a parallel-plate MEMS variable capacitor by moving the plate to the increasing-gap direction. The proposed structure adopted a levering actuator to achieve the opposite moving direction and large displacement of the top plate. The proposed MEMS variable capacitor, which was made by metal surface micromachining, showed excellent linearity factor of 99.1% in C-V response and the capacitance tuning ratio of 178% were achieved at 2GHz. INTRODUCTION Capacitance Displacement Capacitance In RF applications, a parallel-plate MEMS variable capacitor has been widely studied for its simple fabrication and relatively high quality factor compared to other MEMS variable capacitors and p-n junction varactors, which were used in voltage-controlled oscillators (VCOs) and phase-locked loop circuits (PLL) [1-2]. The parallel-plate MEMS variable capacitor is generally made of metal materials which have low electrical resistivity as a signal path structure and air gap as an insulator, which leads to high Q-factor. In hence, low loss and low phase noise of RF signal are easily achievable. Additionally, since the MEMS variable capacitors change its capacitance value by mechanical movement, the moving parts does not sensitively Voltage Displacement < Actuator > < Capacitor > Port 2 Voltage Port 1 Capacitance Displacement Capacitance (a) Voltage Displacement < Actuator > < Capacitor > b t < C-V response> L Ground Voltage Bias Vctrl < C-V response> (a) (b) (b) Figure 2: The proposed structure (a) Perspective view of initial and pull-in state, (b) Magnified view of the actuator part Figure 1: Conceptual graphs of capacitance vs. voltage response characteristics for (a) conventional MEMS variable capacitor and (b) newly proposed device 978-1-4244-9633-4/11/$26.00 ©2011 IEEE W 772 MEMS 2011, Cancun, MEXICO, January 23-27, 2011 coefficient equation (inset of Figure 3) [8]. The LF value lets us know how well the curve is close to the straight line. When a curve is straight line, the LF value is 100%. The LF value is independent from the number of samples. The beam length ratio from the levering beam to each end of the actuator beam was 7:1 (560μm: 80μm). The gap between two parallel plates was initially 1.25μm and it was supposed to be increased to 3.53μm when the applied voltage was 50V. CONCEPT AND SIMULATION Concept As depicted in Figure 1 (a), a conventional closing-gap parallel-plate MEMS variable capacitor shows non-linear C-V response because as the voltage increases, gap decreases fast, and as the gap decreases, capacitance increases faster. On the other hand, if the gap increases fast as the voltage increases, capacitance decreases linearly to the voltage since capacitance decreases slow as the gap increases, as shown in Figure 1 (b). This concept is presented for the first time to our knowledge. To realize this concept, we need an actuator that can increase the gap between two parallel plates as the actuation voltage increases. Figure 2 shows a schematic view of the proposed structure which transforms the common closing-gap electrostatic actuation in the actuator part into the increasing-gap actuation in the central parallel plates by means of the levering beam. When the voltage is applied between the actuator plate and the bottom DC pad which is located underneath the actuator plate, the actuator plate moves down but the opposite side of the actuator beam is lifted up due to the torsion of levering beam. As a result, we can achieve increasing-gap actuation of the parallel plates as the applied voltage increases. Additionally, this structure makes it possible to separately design the capacitance value, actuation voltage range and the linearity in C-V response. The capacitor plate’s size and the length of ‘b’ in actuator part determine the capacitance value and actuation voltage range respectively. The beam length ratio from the levering beam position to the capacitor plate connector and to the end of actuator plate is supposed to determine the linearity solely. FABRICATION Figure 4 represents the fabrication process of the actuator part of the proposed MEMS variable capacitor. The process started with a glass substrate for low substrate loss of RF signal. Thermal evaporated Au (1000Ǻ) made the bottom capacitor plate and bias pad pattern. To prevent DC electrical short, silicon nitride was deposited on the bottom layer using PECVD and patterned by plasma etcher. The sacrificial layer was then formed using sputtered Cr (1500Ǻ), Al (4000Ǻ) and Cr (6500 Ǻ) in order (step A). Thick photoresist was spin-coated and patterned to form the 1st mold. 6μm thick Cu was electroplated to form the levering beam and the connector for the capacitor plate (step B). The photoresist used in step B was totally removed and thicker 2nd photoresist was patterned for 25μm thick electroplated Cu (step C). In the last step, the 2nd photoresist was removed and the sacrificial layer was wet-etched. Finally, the structure was A Simulation The proposed concept was verified using CoventorWareTM and HFSSTM, and the result was shown in Figure 3. An excellent linearity factor of 99.9% was anticipated in the voltage range of 10 ~ 50V. The linearity factor (LF) was defined by the correlation Capacitance (pF) 2.0 1.6 B Cu 6μm Si3N4 4000Ǻ Cu 25μm LF = 0.992 (total) 0.999 (10~50V) 1.4 Au 1000Ǻ PR mold 1GHz 2 400x400μm , b=300μm W=10μm, L=20μm, t=6μm 1.8 Cr 1500Ǻ / Al 4000Ǻ / Cr 6500Ǻ C 1.2 1.0 0.8 0.6 0 LF = n∑ CiVi − ∑ Ci ∑ Vi [n∑ C − (∑ Ci ) 2 ][n∑ Vi 2 − (∑ Vi ) 2 ] 2 i 10 20 30 Voltage (V) 40 D 50 Figure 3: Simulated capacitance vs. voltage response using CoventorWareTM and HFSSTM Glass substrate Au Cu Cr / Al / Cr layer Actuator part Si3N4 PR Figure 4: Simplified fabrication process of the proposed structure 773 at 0V. This is relatively low compared to the conventional parallel-plate MEMS variable capacitors. Because the RF signal should pass through the long actuator beam which acts as an inductive component in the device. Figure 7 shows measured Q-factor with respect to the frequency sweep. The values were between 7.3 and 13.5 at 1GHz. These values are very low at this moment because the RF signal should pass through the narrow torsional spring structures (connector between capacitor plate and actuator beam, levering beam). However, these problems as well as the low SRF can be solved by adopting the split bottom plate structure developed in our group previously [10]. 400x400μm2 Capacitor Plate 20 μm Actuator Beam Levering Beam Actuator Plate CONCLUSION 50 μm A new MEMS variable capacitor with highly linear C-V response and large tuning ratio by only changing the Figure 5: SEM photographs of the fabricated MEMS variable capacitor. Over-view (large), magnified view (small) released using critical point dryer (CPD) to prevent stiction between parallel capacitor plates. Figure 5 shows the scanning electron microscope (SEM) photographs of the fabricated MEMS variable capacitor. All four levering actuators were successfully fabricated without any bending which can critically affect the C-V response characteristic. MEASUREMENT Capacitance (pF) 200 μm 0V 5V 10V 15V 20V 25V 30V 35V 40V 45V 1.4 1.2 1.0 0.8 0.6 0.4 0.5 Figure 6 shows the measured frequency characteristics of the capacitance (a) and C-V response extracted at specific frequencies (b). The linearity factor was 98.4% at 1GHz and 99.1% at 2GHz in the total range without any abrupt changes. Capacitance tuning ratio was 134% at 1GHz and 178% at 2GHz. Since there is little capacitance change in the first few voltages (0 ~ 10V), the device is expected to be used in the region from 10V to 45V. In this actual usage region, the linearity was 99.5% at 1GHz and 99.9% at 2GHz. These are extremely improved values compared to the conventional closing-gap MEMS variable capacitor whose linearity factor is typically about 85%. The capacitance tuning ratio of the highly linear region was 125% at 1GHz and 164% at 2GHz (10 ~ 45V), respectively, and these values are remarkably higher than 50% which is the highest value that can be achievable from the conventional closing-gap parallel-plate MEMS variable capacitors. Suppose that the conventional parallel-plate MEMS variable capacitor with separate actuator is used to achieve the same amount of the capacitance tuning ratio as in the proposed device, the linearity factor becomes lower than the conventional 85% since, in order to increase the capacitance tuning ratio, the top plate moves down lower than two thirds of the initial gap deteriorating linearity in C-V response severely as shown in [9]. Self-resonant frequency (SRF) was measured to 3.5GHz 1.0 1.5 2.0 Frequency (GHz) 2.5 Capacitance (pF) (a) 1.4 1GHz 2GHz 2 400x400μm , b=300μm W=10μm, L=20μm, t=6μm 1.2 LF = 0.991 (total) 0.999 (10~45V) 1.0 0.8 LF = 0.984 (total) 0.995 (10~45V) 0.6 0.4 0 5 10 15 20 25 30 35 40 45 Voltage (V) (b) Figure 6: Measured results (a) Capacitance vs. frequency sweep, SRF is 3.5GHz at 0V (b) Capacitance vs. voltage response at 1GHz and 2GHz 774 25 20 Q-factor Voltage-Controlled Oscillators”, IEEE Trans. Microwave Theory Tech., VOL. 48, NO.11, November 2000, pp. 1943-1949 [2] M. Yellepeddi, K. Mayaram, “Issues in the Design and Simulation of a MEMS VCO based Phase-Locked Loop”, IEEE international symposium on circuits and systems, 2007, pp. 1553-1556 [3] A. –S. Porret, T. Melly, C. C. Enz, E. A. Vittoz, “Design of High-Q Varactors for Low-Power Wireless Applications Using a Standard CMOS Process”, IEEE J. Solid-State Circuits, VOL. 35, NO. 3, March 2000, pp. 337-345 [4] John W. M. Rogers, J. A. Macedo, C. Plett, “The Effect of Varactor Nonlinearity on the Phase Noise of Completely Integrated VCOs”, IEEE J. Solid-State Circuits, VOL. 35, NO. 9, September 2000, pp. 1360-1367 [5] T. Wu, P. K. Hanumolu, K. Mayaram, U.-K. Moon, “Method for a Constant Loop Bandwidth in LC-VCO PLL Frequency Synthesizers”, IEEE J. Solid-State Circuits, VOL. 44, NO. 2, February 2009, pp. 427-435 [6] M Shavezipur, P Nieva, A Khajepour, S M Hashemi, “Development of Parallel-plate-based MEMS Tunable Capacitors with Linearized Capacitance-voltage Response and Extended Tuning Range”, J. Micromech. Microeng. 20, December 2009 [7] M. Bakri-Kassem, R. R. Mansour, “Linear Bilayer ALD Coated MEMS Varactor with High Tuning Capacitance Ratio”, IEEE J. Microelectromechanical Systems, VOL. 18, NO. 1, February 2009, pp. 147-153 [8] E Weisstein, Correlation Coefficient, Math World, A Wolfram Web Resource [9] J. Zou, C. Liu, J. Schutt-Aine, J. Chen, S.-M. Kang, “Development of a Wide Tuning Range MEMS Tunable Capacitor for Wireless Communication System”, IEEE International Electron Device Meeting (IEDM) 2000, Technical Digest, pp. 403-406 [10] Y. -J. Yoon, H.-S. Lee, J. -B. Yoon “MEMS Variable Capacitor Actuated with an Electrically Floating Plate”, IEEE International Electron Device Meeting (IEDM) 2007, Technical Digest, pp. 431-434 0V 45V 15 10 5 0 -5 1 2 3 4 Frequency (GHz) 5 Figure 7: Q-factor vs. frequency sweep, Q-factor is 7.3 (0V) and 13.5 (45V) at 1GHz direction of the moving plate is proposed and developed successfully to demonstrate its performance with the experimental results. The fabricated device showed over 99% linearity factor in C-V response and over 100% capacitance tuning ratio without any abrupt capacitance change region. The device is expected to be favorably used in MEMS VCOs and various electrical circuits with its remarkably high linearity and large tuning ratio. ACKNOWLEDGEMENT The authors would like to thank to the graduate students of 3D Micro-Nano Structures laboratory in KAIST. This work was partially supported by the Korea Science and Engineering Foundation (KOSEF) grant funded by the Korea government (MEST) (No.R11-2007-045-03003-0). REFERENCES [1] A. Dec, K. Suyama, “Microwave MEMS-Based 775