應用於無線電接收器與慣性元件表面塗層監測之CMOS-MEMS兩端固定樑式共振開關

Abstract

本論文利用CMOS-MEMS製程平台，製作兩端固定樑式共振開關。微機電共振開關可應用於感測器接收器中，透過頻率選擇性質以達到濾波的功能，簡化傳統複雜的訊號接收系統。更重要的是，此開關無靜態功耗，適合用於超低功耗的感測器應用。文獻上所發表的共振開關研究皆為使用特殊製程製作，本研究首次以台積電的CMOS製程製作，以達成低成本且具有商業競爭力之元件。 此次製作的微機電共振式開關結構為一兩端固定樑，驅動電極位於樑的兩側，輸出電極位於樑的中間。當輸入於驅動電極的交流訊號頻率為樑的自然共振頻率時，樑的位移會有放大的效果，使樑位移最大位置(中間)撞擊輸出電極而測得電訊號，為了避免樑撞擊輸入電極，輸入端位移必須小於輸出端。若定義樑在輸出電極位移與輸入電極上最大位移的比值為位移增益，亦及共振開關需要大於一的位移增益。當輸入電極長度縮短時，能使位移增益增大，但將付出受力面積變小的代價。此兩端固定樑長為110 μm，寬為4 μm，輸入電極長度為33 μm，位移增益為1.35。自然共振頻率約為2.1 MHz。 製程方面，元件透過國家晶片中心下線，取得晶片後，以特殊蝕刻液來釋放元件。兩端固定樑結構為金屬及氧化層堆疊：第一層金屬(METAL1)至第三層金層(METAL3)，氧化層周圍以VIA層柱(鎢)包圍，避免蝕刻液侵蝕內部氧化物。在掃描式電子顯微鏡下可觀察出結構及電極上的VIA層相較於METAL(鋁)層會較凸出(其原因應為製程缺陷)，形成鎢與鎢的碰撞。晶片在客製化真空腔裡進行量測(10-5 Torr)，使用網路分析儀量測之共振頻率約為2.12 MHz。量測撞擊訊號時，輸入訊號採用掃頻的方式，使輸入頻率慢慢接近其共振頻率，當輸入訊號到達共振頻率時，示波器會採集到輸出端撞擊的脈衝訊號，輸入訊號為66 V的偏壓和3.52 Vamp的交流電壓，樑上的偏壓為4 V，負載電阻為470 Ω，由量測到的電訊號可推得平均功率增益為18.0 dB。然而，有時候可以觀察到輸出碰撞訊號有許多週期出現遺失的現象，原因應該為較硬的電極結構及剛性碰撞材料，在碰撞時產生的squegging effect，使得碰撞不會發生在每次位移的循環中。 另一方面，此樑式共振開關亦可以應用於監測表面塗層的品質。當微機電慣性感測器如加速度計受到一相當大的加速度時，質量塊(Proof mass)容易與擋塊(Stopper)碰撞，當表面黏滯力過大時，彈簧彈性的力無法使質量塊回復到原本位置，此時加速度計將喪失感測功能。然而現今已有非常多研究試圖防止此事件發生，最常見的方法為在結構表面沉積防黏滯層，但如何有效監測表面塗層品質仍為一項待解決的問題，本研究嘗試使用此微機電式共振開關來監測表面塗層之品質，概念來自於原子力顯微鏡(Atomic force microscope)操作在輕碰模式(Tapping mode)，此元件可以嵌入微機電慣性感測元件中，但相較於原子力顯微鏡，微機電共振開關具有嵌入於慣性感測元件的可能性，即使是封裝元件也能夠做到即時監測。 為預測此碰撞行為，本研究使用較適合描述硬質材料碰撞行為的DMT碰撞模型，此模型的運動方程式包含靜電驅動力、吸引力和排斥力，當驅動力足以讓樑碰撞輸出電極時，在頻域上會出現平坦的區域，且曲線的兩側會出現不連續點，然而頻寬長短取決於吸引力與排斥力的大小。舉例來說，鎢與鎢碰撞之Hamaker常數為8.9×10-19，若將其減為5×10-19來模擬，此時頻寬將出現往高頻處增加的現象，可進而判斷出結構表面塗層的狀況。 本研究使用FDTS沉積在共振開關表面使表面能降低，導致吸引力減少，此時吸引力無法和排斥力抗衡，頻寬表現出向高頻處拉寬的現象，實驗結果中，經FDTS沉積過後，頻寬從 增加為 ，因此藉由量測頻寬的大小來監測FDTS塗層之品質。在另一項實驗中，改變元件溫度來觀察頻寬的變化，當溫度從35°C增加為80°C時，頻寬漸漸下降，解釋為當溫度升高時材料變軟，使黏滯力上升。

This research aims to develop a clamped-clamped beam resonant switch, a.k.a. resoswitch, based on a 0.35-μm 2-poly-4-metal CMOS-MEMS platform. The resoswitch can achieve both signal filtering and amplification with zero quiescent power, which could greatly simplify the conventional RF front-end architecture towards ultra-low power sensors. Several resoswitch technologies have been demonstrated previously. However, the processes of those resoswitches require complicated steps that are not CMOS compatible. In this work, the resoswitch is for the first time realized by a CMOS technology with low cost and a high degree of fidelity on device performance. The main structure of the device is a clamped-clamped beam (CC-beam). The input electrodes are on both sides of the beam towards the anchors and the output electrode is placed at the center of the beam. To operate, the input electrodes are biased with a dc bias VP and an ac signal vi, and a supply voltage VDD is applied on the beam. When the driving frequency matches the resonance fre-quency of the CC-beam, the beam starts to vibrate transversely that makes the center of the beam impact the output electrode. The CC-beam resoswitch requires displacement gain to prevent input impacting. The displacement gain of the CC-beam resoswitch is derived by vibration mode shape directly. The displacement gain is defined as the displacement magnitude at the center of beam di-vide by maximum displacement along the input electrodes. To obtain a higher displacement gain, the input electrodes need to shrink towards the anchors, which results in a larger driving voltage to pre-serve the same driving force. The release step of the standard CMOS fabricated chip uses a commercially available Al-compatible HF (Silox Vapox III). In this work, the CC-beam structure is composed of layers from METAL1 to METAL3 while the electrodes are composed of layers from METAL1 to METAL4. The SEM photo of a released CC-beam resoswitch shows the VIA layers are extruded from METAL layers which achieves a preferred hard refractory metal W-to-W contact. The measurement of a CC-beam resoswitch embedded power amplifier is carried out in a custom-built vacuum chamber that allows pressure pumped to below 10-5 Torr with a turbo pump. The resoswitch exhibits a measured resonance frequency of 2.12 MHz. The switching waveform is captured when the input electrodes are biased with a dc bias of 66 V and an ac signal of 3.52 Vamp. A supply voltage of 4 V is applied on CC-beam. A discharge resistor of 470 Ω attaches to the out-put electrode. The CC-beam resoswitch embedded power amplifier finally yields an average power gain of 18.0 dB. In addition, the CC-beam resoswitches could be applied to anti-stiction coating condition monitoring for MEMS motion sensors. Stiction is a common failure mechanism in MEMS devices such as accelerometers, which are usually composed of a proof mass, springs, sensing electrodes and stoppers. When the restoring force is not enough to resist the adhesion force, the proof mass may stuck permanently. To prevent this from happening, various solutions have been demonstrated. The most universal way is to use anti-stiction coating to reduce the surface energy and therefore the ad-hesion force. In this technique, the anti-stiction coating requires a surface condition monitoring to ensure the coating quality. To address this, the work utilizes the tapping bandwidth of the resoswitch to gauge the quality of FDTS coating. In addition, the MEMS resoswitches could be embedded in MEMS motion sensors, which is able to achieve in-situ surface condition monitoring. In order to predict the tapping bandwidth of the resoswitch after FDTS coating, this work uses a DMT contact model which better describe the contact behavior of the hard material. The simula-tion result shows the tapping bandwidth increases as the Hamaker constant decreases from the nominal 8.9×10-19 of W-to-W contact to a lower 5×10-19. However, the magnitude of the tapping bandwidth depends on the net force of the repulsive and attractive forces—the lower the attractive force, the wider the tapping bandwidth. The measurement result shows the tapping bandwidth increases from 17.37 kHz to 29.75 kHz after FDTS coating. This research also compares the tapping bandwidth at different tempera-tures. As the temperature increases from 35°C to 80°C, the tapping bandwidth decreases. The de-creased tapping bandwidth at elevated temperatures is likely due to material softening, which in turn increases the adhesion force between the contact surfaces.