sensors Article
Research on a 3D Encapsulation Technique for Capacitive MEMS Sensors Based on Through Silicon Via † Meng Zhang 1,2 , Jian Yang 1,2 , Yurong He 2 , Fan Yang 1,2 , Fuhua Yang 1,3 , Guowei Han 1, *, Chaowei Si 1, * and Jin Ning 1,4,5, * 1
2 3 4 5
* †
Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China;
[email protected] (M.Z.);
[email protected] (J.Y.);
[email protected] (F.Y.);
[email protected] (F.Y.) College of Materials Science and Opto-Electronic Technology, University of Chinese Academy of Sciences, Beijing 100049, China;
[email protected] School of Microelectronics, University of Chinese Academy of Sciences, Beijing 100049, China School of Electronic, Electrical and Communication Engineering, University of Chinese Academy of Sciences, Beijing 100049, China State Key Laboratory of Transducer Technology, Chinese Academy of Sciences, Beijing 100083, China Correspondence:
[email protected] (G.H.);
[email protected] (C.S.);
[email protected] (J.N.); Tel.: +86-10-8230-4492 (J.N.) This paper is an extended version of an earlier conference paper: Zhang, M., Yang, J., He, Y., Yang, F., Han, G., Si, C., Ning, J. Research on a 3D encapsulation technique for capacitive MEMS sensors based on through silicon via. In Proceedings of the 9th International Conference of the Chinese Society of Micro-Nano Technology, Zhengzhou, China, 19–22 September 2018.
Received: 18 November 2018; Accepted: 21 December 2018; Published: 28 December 2018
Abstract: A novel three-dimensional (3D) hermetic packaging technique suitable for capacitive microelectromechanical systems (MEMS) sensors is studied. The composite substrate with through silicon via (TSV) is used as the encapsulation cap fabricated by a glass-in-silicon (GIS) reflow process. In particular, the low-resistivity silicon pillars embedded in the glass cap are designed to serve as the electrical feedthrough and the fixed capacitance plate at the same time to simplify the fabrication process and improve the reliability. The fabrication process and the properties of the encapsulation cap were studied systematically. The resistance of the silicon vertical feedthrough was measured to be as low as 263.5 mΩ, indicating a good electrical interconnection property. Furthermore, the surface root-mean-square (RMS) roughnesses of glass and silicon were measured to be 1.12 nm and 0.814 nm, respectively, which were small enough for the final wafer bonding process. Anodic bonding between the encapsulation cap and the silicon wafer with sensing structures was conducted in a vacuum to complete the hermetic encapsulation. The proposed packaging scheme was successfully applied to a capacitive gyroscope. The quality factor of the packaged gyroscope achieved above 220,000, which was at least one order of magnitude larger than that of the unpackaged. The validity of the proposed packaging scheme could be verified. Furthermore, the packaging failure was less than 1%, which demonstrated the feasibility and reliability of the technique for high-performance MEMS vacuum packaging. Keywords: 3D encapsulation; capacitive; MEMS; vertical interconnect; glass reflow
1. Introduction Encapsulation is crucial for microelectromechanical systems (MEMS) sensors, since it can shield the sensitive and fragile structures from the influence of the external environment to obtain a high Sensors 2019, 19, 93; doi:10.3390/s19010093
www.mdpi.com/journal/sensors
Sensors 2018, 18, x FOR PEER REVIEW Sensors 2019, 19, 93
2 of 10 2 of 10
performance. On account of the feature of customization, that the packaging requirements for various MEMS sensorsOn areaccount different, cost accounts for of the total fabrication [1]. performance. of the packaging feature of customization, thatover the 50% packaging requirements forcost various Therefore, choosing a suitablethe packaging solution is significant in order tothe obtain performance MEMS sensors are different, packaging cost accounts for over 50% of totalbetter fabrication cost [1]. and lower choosing costs. Thea 3D wafer-level packaging with a wafer and Therefore, suitable packaging solutiontechnology, is significantcombined in order to obtain betterbonding performance vertical interconnection technique,packaging has greattechnology, potential in realizing a smaller size, lower power and lower costs. The 3D wafer-level combined with a wafer bonding and vertical consumption, and lower fabrication [2–6]. Due to the fact that Pyrex glass haspower a similar coefficient interconnection technique, has greatcost potential in realizing a smaller size, lower consumption, of thermal to silicon, between silicon and glass is usually and lowerexpansion fabrication(CTE) cost [2–6]. Duethe to anodic the factbonding that Pyrex glass has a similar coefficient of applied thermal to the hermetic packaging to between obtain high packaging reliability. Considering expansion (CTE) to silicon,for theMEMS anodicdevices bonding silicon and glass is usually applied to the the cost, reliability, and efficiency, anodic bonding is optimal for MEMS sensors compared with a hermetic packaging for MEMS devices to obtain high packaging reliability. Considering the cost, commercially available eutectic bonding technique. Thus, it is significant to design the packaging reliability, and efficiency, anodic bonding is optimal for MEMS sensors compared with a commercially process ofeutectic MEMS sensors on anodic for lower available bondingbased technique. Thus,bonding it is significant to cost. design the packaging process of MEMS In addition to packaging hermeticity, the electrical interconnection is also indispensable to sensors based on anodic bonding for lower cost. transmit an electrical signal which responds external physical quantities. interconnection In addition to packaging hermeticity, the to electrical interconnection is also Vertical indispensable to transmit techniques reduce delay and power consumption to shorter interconnection lengths [7] and an electricalcan signal which responds to external physicaldue quantities. Vertical interconnection techniques improve the integrated level, space efficiency, and design simplicity [8]. Compared with lateral can reduce delay and power consumption due to shorter interconnection lengths [7] and improve the feedthrough withspace the problems stepdesign coverage, it is easier to realize good hermetic encapsulation integrated level, efficiency,ofand simplicity [8]. Compared with lateral feedthrough with with the vertical [9].itThe vertical interconnection structures fabricated by a Pyrex glass the problems of feedthrough step coverage, is easier to realize good hermetic encapsulation with the vertical reflow process, where the low-resistivity silicon pillars surrounded by the insulated glass act as the feedthrough [9]. The vertical interconnection structures fabricated by a Pyrex glass reflow process, medium transmit an electrical signal, has the advantages of good isolation, negligible where thetolow-resistivity silicon pillars surrounded by the insulated glass act as the medium parasitics, to transmit and low crosstalk [10], in addition to good compatibility with anodic bonding. The technology has an electrical signal, has the advantages of good isolation, negligible parasitics, and low crosstalk [10], been widely to applied to MEMS pressure sensorsbonding. [11,12], radio [12–14], and in addition good compatibility with anodic The frequency technology(RF) has resonators been widely applied to gyroscopes [15]. MEMS pressure sensors [11,12], radio frequency (RF) resonators [12–14], and gyroscopes [15]. However, the patterned metal layer or polysilicon in However,in inplane-parallel plane-parallelcapacitive capacitiveMEMS MEMSsensors, sensors, the patterned metal layer or polysilicon an etched glass cavity areare generally used in an etched glass cavity generally usedasasthe thefixed fixedcapacitance capacitanceplates plates [11,12,16]. [11,12,16]. The The issues issues regarding the adhesion between the metal and etched glass surface, and the thermal stress caused by regarding the adhesion between the metal and etched glass surface, and the thermal stress caused CTE mismatch are inevitable. Furthermore, the size of the fixed capacitance plate is often larger than by CTE mismatch are inevitable. Furthermore, the size of the fixed capacitance plate is often larger wanted due to theto design margin for for lateral leads, which will than wanted due the design margin lateral leads, which willdegenerate degeneratethe theperformance performanceand and reliability of the MEMS sensors. reliability of the MEMS sensors. In this paper, paper,anan improved 3D hermetic packaging scheme based on bonding anodic for bonding for In this improved 3D hermetic packaging scheme based on anodic capacitive capacitive MEMS sensors is studied, as shown in Figure 1. We use the embedded silicon pillars, MEMS sensors is studied, as shown in Figure 1. We use the embedded silicon pillars, instead of metal instead of metalas orthe polysilicon, as the fixed capacitance plate simplify processthe and improve the or polysilicon, fixed capacitance plate to simplify the to process andthe improve reliability. It is reliability. It is the of third of thecap, encapsulation cap, except for realizing sealing and the third function the function encapsulation except for realizing hermetic sealinghermetic and vertical electrical vertical electrical The interconnection. wafer with sensing structures is anodically bonded to interconnection. silicon waferThe withsilicon sensing structures is anodically bonded to the encapsulation the encapsulation cap at the wafer level to form the sensing capacitor. The sensing structures deform cap at the wafer level to form the sensing capacitor. The sensing structures deform with the change of with the physical change of external physical resulting the variation in the value. external quantities, resulting quantities, in the variation in theincapacitance value. Thecapacitance packaging scheme The packaging be designed flexibly for sensors, various such capacitive MEMS sensors, such as can be designedscheme flexibly can for various capacitive MEMS as pressure sensors, accelerators, pressure sensors, accelerators, gyroscopes, transducers, and so on.of The process gyroscopes, ultrasonic transducers, and soultrasonic on. The fabrication process flow thefabrication encapsulation cap flow of the encapsulation cap based on a GIS process is introduced in detail. The surface morphology based on a GIS process is introduced in detail. The surface morphology and the electrical property of and thevertical electrical propertywere of silicon vertical feedthrough were alsogyroscope characterized. Finally, a silicon feedthrough also characterized. Finally, a capacitive was successfully capacitive was successfully fabricated proposed packaging scheme. The quality fabricatedgyroscope with the proposed packaging scheme.with The the quality factors of the unpackaged gyroscope factors of the unpackaged gyroscope and packaged gyroscope were both measured to evaluate the and packaged gyroscope were both measured to evaluate the validity and reliability of the proposed validity and reliability of the proposed encapsulation scheme. encapsulation scheme.
Figure 1. Cross-sectional schematic for the packaged capacitive microelectromechanical systems Figure 1. Cross-sectional schematic for the packaged capacitive microelectromechanical systems (MEMS) sensor. (MEMS) sensor.
Sensors 2019, 19,18, 93x FOR PEER REVIEW Sensors 2018,
33ofof1010
2. Fabrication of Encapsulation Cap 2. Fabrication of Encapsulation Cap The fabrication process flow of the encapsulation cap with through silicon vias (TSVs) is shown The fabrication process flow of the encapsulation cap with through silicon vias (TSVs) is shown in Figure 2a, which is presented in detail as follows. in Figure 2a, which is presented in detail as follows. (a) A 4-inch silicon wafer with a low resistivity (0.0009 Ω·cm) was used to realize a low loss electrical (a) Ainterconnection. 4-inch silicon wafer a low resistivity (0.0009 Ωby ·cm)photolithography was used to realizeand a lowanloss electrical Thewith substrate was patterned 8-μm thick interconnection. The substrate was patterned by photolithography and an 8-µm thick photoresist photoresist acted as the mask for deep silicon etching. as the mask deep silicon (b) acted The mature deep for reactive ion etchetching. (DRIE) process was utilized to etch silicon of 350 μm to form (b) The mature deep reactive ion etch (DRIE) process utilized to etchremoved silicon ofand 350 the µm chemical to form silicon molds for the glass reflow process. Thewas photoresist was silicon molds forwafer the glass process. The photoresist was removed and the chemical cleaning cleaning of the wasreflow carefully conducted for the following anodic bonding. thesilicon wafer substrate was carefully for Pyrex the following anodic (c) ofThe was conducted bonded to the glass wafer in abonding. vacuum environment. The silicon substrate the glass wafer had same thickness a similar coefficientThe of thermal (c) The silicon and substrate was bonded to thethe Pyrex glass wafer inand a vacuum environment. silicon expansion (CTE) to minimize distortion the bonded wafer. It should be noted substrate and the glass wafer hadthe thewarp sameand thickness and aof similar coefficient of thermal expansion that itto is minimize necessary to theand vacuum conditions 10−2 mTorr (1 Torr ≈ 133 Pa) for enough (CTE) thehold warp distortion of theofbonded wafer. It should be noted that ittime is − 2 to achievetovacuum which can generating voids during the glass reflow necessary hold theoutgassing, vacuum conditions of prevent 10 mTorr (1 Torr ≈ 133 Pa) for enough time to process.vacuum outgassing, which can prevent generating voids during the glass reflow process. achieve ◦ C for°C (d) The Thebonded bondedwafer wafer was then placed in furnace the furnace at 1050 h to ensure themolds silicon (d) was then placed in the at 1050 2 hfor to 2ensure that thethat silicon molds were fully filled with the fused glass. the temperature wasabove elevated °C, were fully filled with the fused glass. When theWhen temperature was elevated 750 ◦above C, the 750 fused the fused glass started to be pushed into the silicon molds under the large pressure difference glass started to be pushed into the silicon molds under the large pressure difference between between and outside thecavity. hermetic cavity. Subsequently, the glass by process. a cooling inside andinside outside the hermetic Subsequently, the glass was cooledwas by cooled a cooling process. (e) The reflowed wafer was lapped and thinned on double sides until the formation of the silicon (e) vertical The reflowed wafer A was lappedmechanical and thinnedpolishing on double sidesprocess until the formation silicon feedthrough. chemical (CMP) was exploitedoftothe provide vertical feedthrough. A later chemical good surface quality for wafermechanical bonding. polishing (CMP) process was exploited to provide good surface quality for later wafer bonding. (f) To form the encapsulation cavity, the silicon was etched by an Inductively Coupled Plasma (ICP) (f) process To formtothe encapsulation cavity, thetosilicon etched by anThe Inductively Coupled a certain depth according designwas requirements. cavity acted as thePlasma gap of(ICP) the process to a certain depth according to design requirements. The cavity acted as the gap of the sensing capacitor when the following anodic bonding was accomplished. sensing capacitor when the following anodic bonding was accomplished.
(a)
(b)
Figure 2. (a) Process flow for the encapsulation cap; (b) schematic and the sectional view of the encapsulation Figure 2. (a) cap. Process flow for the encapsulation cap; (b) schematic and the sectional view of the
encapsulation cap.
The schematic of the encapsulation cap is shown in Figure 2b. The silicon feedthrough structure runs through the wafer from top to bottom just its name implies. The silicon silicon feedthrough vertical feedthrough The schematic of the encapsulation cap is as shown in Figure 2b. The structure minimizes the routing lengths for electrical signals between the substrates, can realize a runs through the wafer from top to bottom just as its name implies. The siliconwhich vertical feedthrough smaller footprint and lower power [12].between In particular, some ofwhich the silicon pillars a minimizes the routing lengths for consumption electrical signals the substrates, can realize surrounded by glass serve as the fixed capacitance plates and are etched to form the packaging smaller footprint and lower power consumption [12]. In particular, some of the silicon pillars cavity after anodic bonding. electrical feedthrough with a certain thickness may become surrounded by glass serve asThe thelateral fixed capacitance plates and are etched to form the packaging cavity the leakage path throughThe the lateral bonding interface. It is easier to realize hermetic encapsulation with the after anodic bonding. electrical feedthrough with a certain thickness may become the
leakage path through the bonding interface. It is easier to realize hermetic encapsulation with the
vertical feedthrough. This is the reason why the vertical electrical interconnects are adopted in our packaging scheme. 3. Characterization and Measurement Sensors 2019, 19, 93
4 of 10
The accomplished capping wafer is shown in Figure 3a. It can be seen that the silicon feedthrough structures, which have been fabricated in different sizes and shapes, are surrounded by glass. Figure 3b This shows electron micrograph (SEM)interconnects for the cross-section of 4silicon vertical feedthrough. is the the scanning reason why the vertical electrical are adopted Sensors 2018, 18, x FOR PEER REVIEW of in 10 our feedthrough embraced by reflowed glass. No voids are found in the reflowed glass and even at the packaging scheme. silicon-glass interface, which means that the the glass reflowelectrical process parameters areare reasonable the vertical feedthrough. This is the reason why vertical interconnects adopted and in our reflowed glass can provide good insulation. In addition, the silicon etching sidewall is nearly packaging scheme. 3. Characterization and Measurement perpendicular to the substrate with an angle of 88.6°. The good etching profile is attributed to mature The accomplished capping wafer shown in with Figure It canvia, be the seen that the silicon feedthrough silicon etching technology. Iniscomparison the3a. tapered thermal stress concentration 3.deep Characterization and Measurement as a result nonuniformity can bein alleviated in the verticalare viasurrounded [17], obtaining structures, whichofhave been fabricated differentgreatly sizes and shapes, by good glass.thermal Figure 3b The accomplished capping wafer is shown in Figure 3a. It can be seen that the silicon stability and high reliability. showsfeedthrough the scanning electron which micrograph (SEM) for thein cross-section of and silicon feedthrough embraced structures, have been fabricated different sizes shapes, are surrounded by by To characterize the electrical property of the silicon vertical feedthrough directly and effectively, reflowed glass. No voids are found in the reflowed glass and even at the silicon-glass interface, which glass. Figure 3b shows the scanning electron micrograph (SEM) for the cross-section of silicon the resistances were measured by a four-probe method which could eliminate the parasitic effect of meansfeedthrough that the glass reflowbyprocess are reasonable the reflowed glass canatprovide embraced reflowedparameters glass. No voids are found in and the reflowed glass and even the the testing system. The sample was placed on the probe station and the sweep signals from B1500A good insulation. addition, the means siliconthat etching sidewall nearlyparameters perpendicular to the substrate silicon-glass In interface, which the glass reflowisprocess are reasonable and thewith were imposed on the metal electrodes by probes. The testing schematic of the experimental sample ◦ . The reflowed glass cangood provide goodprofile insulation. In addition, the silicon etching sidewall nearly an angle of 88.6 etching is attributed to mature deep silicon etchingistechnology. is shown in Figure 4a. Two testing loops were obtained, in which one loop offered the sweep current perpendicular to the substrate with an angle of 88.6°. The good etching profile is attributed to mature In comparison with loop the tapered the thermal stress concentration as a resultbetween of nonuniformity and the other worked via, equivalently to a voltmeter. Thus, the relationship the voltagecan deep silicon etching technology. In comparison with the tapered via, the thermal stress concentration be alleviated greatly in the vertical via [17], obtaining good thermal stability and high reliability. (V) and the sweep current (I) across the testing samples could be obtained. as a result of nonuniformity can be alleviated greatly in the vertical via [17], obtaining good thermal stability and high reliability. To characterize the electrical property of the silicon vertical feedthrough directly and effectively, the resistances were measured by a four-probe method which could eliminate the parasitic effect of the testing system. The sample was placed on the probe station and the sweep signals from B1500A were imposed on the metal electrodes by probes. The testing schematic of the experimental sample is shown in Figure 4a. Two testing loops were obtained, in which one loop offered the sweep current and the other loop worked equivalently to a voltmeter. Thus, the relationship between the voltage (V) and the sweep current (I) across the testing samples could be obtained.
Figure 3. (a) Photograph of the capping wafer; (b) Cross-sectional SEM view of the silicon vertical Figure 3. (a) Photograph of the capping wafer; (b) Cross-sectional SEM view of the silicon vertical feedthrough structure. feedthrough structure.
To characterize the electrical property of the silicon vertical feedthrough directly and effectively, the resistances were measured by a four-probe method which could eliminate the parasitic effect of I_sweep the testing system. The sample was placed on the probe station and the sweep signals from B1500A were imposed on the metal electrodes by probes. The testing schematic of the experimental sample is V shown in Figure 4a. Two testing loops were obtained, in which one loop offered the sweep current and the other loop worked equivalently to a voltmeter. Thus, the relationship between the voltage (V) and Figure 3. (a) Photograph of the capping wafer; (b) Cross-sectional SEM view of the silicon vertical the sweepfeedthrough current (I)structure. across the testing samples could be obtained.
Silicon
I_sweep
Glass
Ti
(a) V
Silicon
(b)
Glass
(a)
Ti
(b)
Figure 4. (a) Testing schematic using a four-probe method; (b) The voltage–current testing results using B1500A. The black curve was acquired before annealing and the red curve was acquired after annealing.
Sensors 2018, 18, x FOR PEER REVIEW
5 of 10
Figure 4. (a) Testing schematic using a four-probe method; (b) The voltage–current testing results using B1500A. Sensors 2019, 19, 93 The black curve was acquired before annealing and the red curve was acquired after
5 of 10
annealing.
It shouldbe be noted noted that Schottky junction exists between metal and silicon, It should thatthe the Schottky junction exists between metal and which silicon,is unfavorable which is for resistance measurement. The thermal annealing is generally performed for the purpose obtaining unfavorable for resistance measurement. The thermal annealing is generally performedoffor the the ohmic contact. However, ohmic contact can beohmic also achieved anodic bonding because purpose of obtaining the ohmicgood contact. However, good contact during can be also achieved during the wafer canbecause be simultaneously thermally annealed by the highannealed bonding by temperature of 400 ◦ C. anodic bonding the wafer can be simultaneously thermally the high bonding It means that good ohmic contact be realized withoutcan extra process and that the packaging temperature of 400 °C. It means thatcan good ohmic contact be realized without extra process design and is the totally compatible withisthe fabrication flow of MEMS Theflow measurement obtained that packaging design totally compatible with the sensors. fabrication of MEMS results sensors. The before and after annealing are shown in Figure 4b. The results indicate clearly that good ohmic contact measurement results obtained before and after annealing are shown in Figure 4b. The results indicate was achieved, thecontact resistance 263.5 from thecalculated slope of the clearly that goodwith ohmic wascalculated achieved,as with themΩ resistance as voltage–current 263.5 mΩ from curve, the which was almost as low as the resistance of copper feedthrough [18]. Good electrical interconnection slope of the voltage–current curve, which was almost as low as the resistance of copper feedthrough was realized and the feasibility to use an realized embedded silicon feedthrough capacitance plate [18]. Good electrical interconnection was and the feasibility to as usethe anfixed embedded silicon was also validated at the same time. feedthrough as the fixed capacitance plate was also validated at the same time. The encapsulation cap and silicon wafer need to be bonded together realize wafer-level The encapsulation cap and silicon wafer need to be bonded together to to realize 3D3D wafer-level encapsulation. Prior to the wafer bonding, the surface roughness after polishing was characterized encapsulation. Prior to the wafer bonding, the surface roughness after polishing was characterized to to evaluate feasibility of wafer bonding. roughness affect packaging hermeticity. During evaluate thethe feasibility of wafer bonding. TheThe roughness willwill affect thethe packaging hermeticity. During ◦ C to anodic bonding, the plastic deformation occurs to the glass at the bonding temperature of 400 anodic bonding, the plastic deformation occurs to the glass at the bonding temperature of 400 °C to ensure close contact between glass and silicon interface. Since plastic deformation ensure thethe close contact between glass and silicon at at thethe interface. Since thethe plastic deformation of of glass is limited, the hermeticity will become worse with the increase of surface roughness in a certain glass is limited, the hermeticity will become worse with the increase of surface roughness in a certain range. However, anodic bonding more tolerant roughnessononthe thebonding bondingsurface surfacecompared comparedtoto a range. However, anodic bonding is is more tolerant ofof roughness fusionbonding bondingtechnique technique[19]. [19]. a fusion The surface roughness measured under a scanning range of 10 × μm 10 µm atomic The surface roughness waswas measured under a scanning range of 10 μmµm × 10 by by an an atomic force microscope (AFM) in tapping mode. scanning results of glass silicon respectively force microscope (AFM) in tapping mode. TheThe scanning results of glass andand silicon areare respectively exhibited Figure 5a,b. RMS roughnesses of glass silicon were 1.120 0.814 exhibited in in Figure 5a,b. TheThe RMS roughnesses of glass andand silicon were 1.120 nmnm andand 0.814 nm,nm, respectively, values which were much smaller than the surface roughness requirement of 5 nm respectively, values which were much smaller than the surface roughness requirement of 5 nm forfor anodic bonding [20] and surface was smooth enough anodic bonding. anodic bonding [20] and thethe surface was smooth enough forfor anodic bonding.
Figure 5. The roughness measurement results obtained using an atomic force microscope (AFM) after Figure 5. The roughness measurement results obtained using an atomic force microscope (AFM) after polishing (a) the measurement result for glass surface; (b) the measurement result for silicon surface. polishing (a) the measurement result for glass surface; (b) the measurement result for silicon surface.
4. Application and Discussion 4. Application and Discussion Finally, the encapsulation scheme was applied to a capacitive gyroscope to evaluate the validity. Finally, the encapsulation scheme was applied to a capacitive gyroscope to evaluate the validity. The silicon wafers prefabricated with sensing structures were anodically bonded in a vacuum at the The silicon wafers prefabricated with sensing structures were anodically bonded in a vacuum at the wafer level with two encapsulation caps to form a sandwich structure. Figure 6a shows two packaged wafer level with two encapsulation caps to form a sandwich structure. Figure 6a shows two packaged gyroscopes with the size of 3 mm × 3 mm and the enlarged view of the device is also given in Figure 6b. gyroscopes with the size of 3 mm × 3 mm and the enlarged view of the device is also given in Figure 6b. The sensing capacitor consisted of the movable sensing structure and two silicon fixed plates The sensing capacitor consisted of the movable sensing structure and two silicon fixed plates embedded in the caps while the encapsulation cavities served as the capacitance gaps. The input angle embedded in the caps while the encapsulation cavities served as the capacitance gaps. The input velocity caused the displacement of the movable sensing structure owing to the Coriolis Effect and angle velocity caused the displacement of the movable sensing structure owing to the Coriolis Effect was exhibited as the variation in the capacitance. and was exhibited as the variation in the capacitance.
Sensors 2019, 19, 93 Sensors 2018, 18, x FOR PEER REVIEW
6 of 10 6 of 10
Figure6.6.(a)(a)The Thevacuum vacuumpackaged packagedgyroscopes; gyroscopes;(b) (b)close-up close-upview viewofofa agyroscope. gyroscope Figure
Vacuumpackaging packagingisisdone donefor forthe theair airdumping dumpingcontrol controlwhich whichisiscritical criticalfor foraahigh-performance high-performance Vacuum gyroscope. The gyroscope could be equivalent to a second-order underdamped system, whose whose motion gyroscope. The gyroscope could be equivalent to a second-order underdamped system, equation is givenisas follows [1]. motion equation given as follows [1].
𝑚 m
𝑑 𝑑𝐴(𝑡) dA(t) d2 𝐴(𝑡) A(t) + ++𝑘𝐴(𝑡) +𝑐c kA(t)==00 𝑑𝑡 𝑑𝑡 dt dt2
(1)(1)
where where 𝐴(𝑡) A(t) is is the the instantaneous instantaneous displacement displacementof ofthe thesensing sensingstructure structurefrom fromthe theequilibrium equilibriumposition, position, 𝑚m is the mass, 𝑘 is the spring stiffness, and 𝑐 is the damping factor which describes the influence is the mass, k is the spring stiffness, and c. is the damping factor which describes the influenceofof damping dampingtotothe thevibration vibrationof ofthe thesensing sensingstructure. structure. 𝐴(𝑡) can be derived as Equation A(t) can be derived as Equation(2) (2)[1] [1]by bysolving solvingthe theabove abovesecond-order second-orderdifferential differentialequation, equation, which describes the vibration of the sensitive structure. which describes the vibration of the sensitive structure. ≫ λ ) 𝐴(𝑡) = 𝐴 𝑒 sin q 𝑤 − λ 𝑡 + 𝜑 (𝑤 A(t) = A0 e−λt sin wr 2 − λ2 t + ϕ0 wr 2 λ2
(2) (2)
𝑘 (3) 𝑤 =r 𝑚k (3) wr = m where 𝐴 is the initial amplitude, 𝜑 is the initial phase, λ is the ratio of 𝑐/2𝑚, whose inverse where A0 is the initial amplitude, ϕ0 is the initial phase, λ is the ratio of c/2m, whose inverse 1 stands stands for the relaxation time, and 𝑤 is the resonant angle frequency of the gyroscope andλcan be for the relaxation time, and wr is the resonant angle frequency of the gyroscope and can be calculated calculated according to Equation (3) [1]. From Equation (2), it can be seen that the vibrating according to Equation (3) [1]. From Equation (2), it can be seen that the vibrating amplitude, A0 e−λt , amplitude, 𝐴 𝑒 , decreases exponentially along with the time caused by the damping effect. If the decreases exponentially along with the time caused by the damping effect. If the damping factor c is damping factor 𝑐 is zero, the gyroscope can make periodic back-and-forth motions at a constant zero, the gyroscope can make periodic back-and-forth motions at a constant amplitude, which means amplitude, which means the energy stored in the gyroscope remain unchanged. Thus, the damping the energy stored in the gyroscope remain unchanged. Thus, the damping effect, especially air effect, especially air damping, is the main factor that causes energy dissipation. damping, is the main factor that causes energy dissipation. To evaluate the encapsulation performance, mainly the damping effect, the quality factor (Q) of To evaluate the encapsulation performance, mainly the damping effect, the quality factor (Q) of the gyroscope was characterized, which is defined as the ratio of the energy stored in a device to the the gyroscope was characterized, which is defined as the ratio of the energy stored in a device to the energy dissipated per cycle of resonance [21], as shown in Equation (4). A high quality factor indicates energy dissipated per cycle of resonance [21], as shown in Equation (4). A high quality factor indicates low rate of energy dissipation, namely, low damping. low rate of energy dissipation, namely, low damping. 𝑈 𝑈 Q = 2π = 2π (4) U UStore 𝑈 𝑈 −t= 𝑈t0 Q = 2π = 2π (4) UDisspate per cycle Ut=t0 − Ut=t0 +T where 𝑈 stands for the energy, 𝑡 is the time, and 𝑇 is the vibration period. The energy 𝑈 can be calculated by the Formula (5) [1]. 𝐴 is the vibration amplitude of the sensing structure, which is the maximum displacement from the equilibrium position. 1 𝑈 = 𝑘𝐴 2
(5)
Sensors 2019, 19, 93
7 of 10
where U stands for the energy, t is the time, and T is the vibration period. The energy U can be calculated by the Formula (5) [1]. A is the vibration amplitude of the sensing structure, which is the maximum displacement from the equilibrium position. Sensors 2018, 18, x FOR PEER REVIEW
7 of 10
1 U = kA2 (5) 2 the Equation (4), Q can be derived as shown in By substituting Equations (2) and (5) into Equation (6) [21], where 𝑓 is the natural resonant frequency of the gyroscope. Thus, the relaxation By substituting Equations (2) and (5) into the Equation (4), Q can be derived as shown in 1 and the resonant 𝑓 resonant are necessary to determine the qualityThus, factorthe of the gyroscope. time Equation (6) [21], where f r isfrequency the natural frequency of the gyroscope. relaxation 1 time λ and the resonant frequency f r are necessary to determine the quality factor of the gyroscope. 𝜋 𝜋 1𝑤 1 𝑤 −λ (6) Q= = p∗ ≈𝜋 =𝜋 𝑓 λT λ 2 2𝜋2 λ 2𝜋 λ π wr − λ π 1 wr 1 Q = was=determined ∗ ≈ π the=free π vibrations fr Here, the quality factor of the gyroscope(6) in the λT λ 2πby studying λ 2π λ time domain. The sample frequency was 31.25 kHz. The signal intensity during free vibration as a Here, theofquality factor was determined by studying free vibrations of the gyroscope in the the function time was measured, see the blue curve inthe Figure 7. The signal intensity reflected timevibration domain.ofThe sample frequency was 31.25 kHz. The signal intensity during free vibration as of the sensing structures. The green curve shows the vibration amplitude as a function a function of time was measured, see the blue curve in Figure 7. The signal intensity reflected the time obtained by the upper envelope fitting. It could be seen that the signal intensity gradually vibration of the sensing structures. The green curve shows the vibration amplitude as a function decreased over time, in accordance with the theory. The relaxation time of 6.3365, can be acquired of time obtained by the upper envelope fitting. It could be seen that the signal intensity gradually by an exponential fit of the vibration amplitude. decreased over time, in accordance with the theory. The relaxation time λ1 of 6.3365, can be acquired by an exponential fit of the vibration amplitude.
Figure 7. The time–domain dynamic response measured at the drive mode of the gyroscope. Figure 7. The time–domain dynamic response measured at the drive mode of the gyroscope.
To obtain the resonant frequency f r of the gyroscope, the Fast Fourier Transform (FFT) was performed dynamic response in the the gyroscope, time domain withFourier the MATLAB software. To obtainforthethe resonant frequency 𝑓 of the Fast Transform (FFT) was The performed frequency–domain characteristics caninbe calculated, as shown in Figure 8. The resonant frequency for the dynamic response the time domain with the MATLAB software. The frequency– point can becharacteristics easily determined tocalculated, be 11.193 kHz by observing amplitude–frequency domain can be as shown in Figurethe 8. The resonant frequencycurve pointand can be phase–frequency curve.to be 11.193 kHz by observing the amplitude–frequency curve and phase– easily determined frequency curve.
Sensors 2018, 18, x FOR PEER REVIEW Sensors 2018, 18, x FOR PEER REVIEW Sensors 2019, 19, 93
8 of 10 8 of 10 8 of 10
Figure 8. The frequency–domain characteristics of the gyroscope obtained by Fast Fourier Transform Figure 8. in The frequency–domain characteristics of the gyroscope obtained byphase–frequency Fast Fourier Transform (FFT), which the black is thecharacteristics amplitude–frequency curve and the red by is curve. Figure 8. The frequency–domain of the gyroscope obtained Fast Fourier Transform (FFT), in which the black is the amplitude–frequency curve and the red is phase–frequency curve. (FFT), in which the black is the amplitude–frequency curve and the red is phase–frequency curve.
The quality factor of the gyroscope at the driving frequency of 11.193 kHz can be calculated as The quality factor of the gyroscope at the driving frequency of 11.193 kHz can be which calculated as 222,815 according to of the Formula (6). The high qualityfrequency factor means low air damping, indicates The quality factor the gyroscope at the driving of 11.193 kHz can be calculated as 222,815 according to the Formula (6). The high quality factor means low air damping, which indicates a high vacuum to degree of the packaging cavity. 222,815 according the Formula (6). The high quality factor means low air damping, which indicates a high vacuum degree of the the packaging cavity. To further evaluate pressure level a high vacuum degree of the packaging cavity. in the encapsulation cavity and the validity of the To furthermethod, evaluate the the pressure level in the encapsulation cavity and the validity of the packaging packaging of the gyroscopes packaging also To further evaluate theresonant pressurecharacteristics level in the encapsulation cavitybefore and the validitywere of the method, the resonant characteristics of the gyroscopes before packaging were also measured by measured by a Laser Signal Analyzer. The gyroscope to be tested was fixedalso in athe packaging method, theDoppler resonantDynamic characteristics of the gyroscopes before packaging were Laser Doppler Dynamic Signal Analyzer. The gyroscope to be tested was fixed in the chamber of the chamberbyofa Laser the vibration The motion of the The sensitive structures the periodic chirp measured Dopplerstation. Dynamic Signal Analyzer. gyroscope to be under tested was fixed in the vibration station. The motion of the sensitive structures under the periodic chirp excitation could be excitation could be recorded by the laser. chamber of the vibration station. The motion of the sensitive structures under the periodic chirp recorded by the laser. excitation could be recorded by the laser. fr 𝑓 (7) Q =Q = ∆𝑓 (7) ∆𝑓f 3dB Q= (7) ∆𝑓 The resonant characteristics gyroscope were obtained, shown Figure 9, when The resonant characteristics of of thethe gyroscope were obtained, as as shown in in Figure 9, when thethe airair pressure in the chamber was stabilized at 3 Pa. The quality factor for the unpackaged gyroscope pressure in the chamber was stabilized at 3 Pa. Thewere quality factor as forshown the unpackaged The resonant characteristics of the gyroscope obtained, in Figure 9,gyroscope when thewas airwas calculated as 13,313 the Formula which is much smaller than quality factor 220,000 calculated 13,313 byby the Formula (7)(7) [21], which is much smaller thethe quality factor of of 220,000 pressure in as the chamber was stabilized at 3[21], Pa. The quality factor forthan the unpackaged gyroscope was is the 3-dB bandwidth, namely, the corresponding bandwidth for the packaged gyroscope. ∆𝑓 for the packaged gyroscope. ∆ f is the 3-dB bandwidth, namely, the corresponding bandwidth when 3dB (7) [21], which is much smaller than the quality factor of 220,000 calculated as 13,313 by the Formula the is 3than dB less gain. thewhen gainpackaged is 3 gain dB less the than mid-band gain. the 3-dB bandwidth, namely, the corresponding bandwidth for the gyroscope. ∆𝑓 theismid-band when the gain is 3 dB less than the mid-band gain.
Figure 9. The resonant characteristics at the drive mode of the unpackaged gyroscopes at a pressure of Figure 9. The resonant characteristics at the drive mode of the unpackaged gyroscopes at a pressure 3 Pa measured by a Laser Doppler Dynamic Signal Analyzer. of 3 9. PaThe measured bycharacteristics a Laser Doppler Dynamic Signalof Analyzer. Figure resonant at the drive mode the unpackaged gyroscopes at a pressure of 3 Pa measured by a Laser Doppler Dynamic Signal Analyzer.
Sensors 2019, 19, 93
9 of 10
Compared with the quality factor for the unpackaged gyroscopes, the quality factor increased by one order of magnitude at least by adopting the proposed packaging scheme. The pressure in the packaged cavity could be evaluated below 1 Pa (10−2 mbar) according to the theoretical model for quality factor [22], which lies in the required vacuum range from 10−1 to 10−4 mbar for MEMS gyroscopes [23]. The validity of the proposed packaging scheme can be verified. Meanwhile, the feasibility and reliability of the packaging scheme were also demonstrated by the fact that the packaging failure is less than 1% and the air pressure in the cavity remained nearly unchanged even after 700 days. 5. Conclusions A 3D hermetic packaging technique suitable for capacitive MEMS sensors is studied. The encapsulation cap with a silicon vertical feedthrough and good surface quality was successfully fabricated and applied to the encapsulation of a capacitive gyroscope. Good electrical interconnection was realized by using the silicon vertical feedthrough. The feasibility to use the silicon pillars embedded in the cap, instead of metal or polysilicon, as the fixed capacitance plate was also validated, which could simplify the process and improve the reliability. The quality factor was enhanced by one order of magnitude by adopting the proposed encapsulation scheme, which proved the validity of the scheme. The feasibility and reliability of the 3D vacuum packaging technique were demonstrated by the successful application to MEMS gyroscopes. The packaging scheme can be designed flexibly to control the vacuum degree for various capacitive MEMS sensors to realize high performance. Author Contributions: M.Z., C.S., G.H., F.Y. (Fan Yang) and J.N. conceived and designed the experiments; M.Z. and J.N. performed the experiments; M.Z., J.Y., C.S. and G.H. analyzed the data; M.Z. wrote the paper; M.Z., F.Y. (Fuhua Yang) and Y.H. revised the paper. All authors read and approved the final manuscript. Funding: This research was funded by National Key R&D Program of China (No. 2018YFF01010504) and Chinese National Science Foundation (Contract No. 61504130, No. 61704165 and No. 5172015004) and the Science and Technology on Analog Integrated Circuit Laboratory Stability support project (Contract No. 6142802WD201806). Conflicts of Interest: The authors declare no conflict of interest.
References 1. 2. 3. 4.
5. 6.
7. 8.
Hsu, T.-R. MEMS and Microsystems: Design, Manufacture, and Nanoscale Engineering; John Wiley & Sons: Hoboken, NJ, USA, 2004; pp. 91–360, ISBN 7-111-13226-2. Kühne, S.; Hierold, C. Wafer-level packaging and direct interconnection technology based on hybrid bonding and through silicon vias. J. Micromech. Microeng. 2011, 21, 085032. [CrossRef] Lee, S.W.; Lee, S.K.; Park, J.H. High-density through-wafer copper via array in insulating glass mold using reflow process. Jpn. J. Appl. Phys. 2015, 54, 4. [CrossRef] Kettner, P.; Kim, B.; Pargfrieder, S.; Zhu, S. In New Technologies for advanced high density 3D packaging by using TSV process. In Proceedings of the International Conference on Electronic Packaging Technology & High Density Packaging (ICEPT-HDP 2008), Shanghai, China, 28–31 July 2008; pp. 1–3. Wang, Z.Y. 3-D Integration and Through-Silicon Vias in MEMS and Microsensors. J. Microelectromech. Syst. 2015, 24, 1211–1244. [CrossRef] Zoschke, K.; Manier, C.A.; Wilke, M.; Oppermann, H.; Ruffieux, D.; Dekker, J.; Jaakkola, A.; Dalla Piazza, S.; Allegato, G.; Lang, K.D. Application of TSV Integration and Wafer Bonding Technologies for Hermetic Wafer Level Packaging of MEMS Components for Miniaturized Timing Devices. In Proceedings of the 2015 IEEE 65th Electronic Components and Technology Conference, San Diego, CA, USA, 26–29 May 2015; IEEE: New York, NY, USA, 2015; pp. 1343–1350. Tezcan, D.S.; Pham, N.; Majeed, B.; Moor, P.D. Sloped through Wafer Vias for 3D Wafer Level Packaging. In Proceedings of the Electronic Components & Technology Conference, Reno, NV, USA, 29 May–1 June 2007. Premachandran, C.S.; Nagarajan, R.; Yu, C.; Zhang, X.; Chong, S.C. A novel electrically conductive wafer through hole filled vias interconnect for 3D MEMS packaging. In Proceedings of the Electronic Components & Technology Conference, New Orleans, LA, USA, 27–30 May 2003.
Sensors 2019, 19, 93
9.
10.
11.
12. 13.
14.
15.
16. 17.
18.
19.
20.
21. 22.
23.
10 of 10
Esashi, M.; Ura, N.; Matsumoto, Y. Anodic bonding for integrated capacitive sensors. In Proceedings of the Micro Electro Mechanical Systems, MEMS’92, an Investigation of Micro Structures, Sensors, Actuators, Machines and Robot, Travemunde, Germany, 4–7 February 1992; pp. 43–48. Jin, J.Y.; Yoo, S.H.; Yoo, B.W.; Kim, Y.K. A Wafer-Level Vacuum Package Using Glass-Reflowed Silicon Through-Wafer Interconnection for Nano/Micro Devices. J. Nanosci. Nanotechnol. 2012, 12, 5252–5262. [CrossRef] [PubMed] Haque, R.; Wise, K. An intraocular pressure sensor based on a glass reflow process. In Proceedings of the Solid-State Sensors, Actuators, and Microsystems Workshop, Hilton Head Island, SC, USA, 6–10 June 2010; pp. 49–52. Haque, R.U.; Wise, K.D. A Glass-in-Silicon Reflow Process for Three-Dimensional Microsystems. J. Microelectromech. Syst. 2013, 22, 1470–1477. [CrossRef] Haque, R.M.; Serrano, D.E.; Gao, X.; N.-Shirazi, A.; Keesara, V.; Ayazi, F.; Wise, K.D. Hermetic packaging of resonators with vertical feedthroughs using a glass-in-silicon reflow process. In Proceedings of the Solid-State Sensors, Actuators and Microsystems Conference, Beijing, China, 5–9 June 2011; pp. 2303–2306. Zhao, J.C.; Yuan, Q.; Wang, F.X.; Kan, X.; Han, G.W.; Sun, L.; Sun, H.Y.; Yang, J.L.; Yang, F.H. Design and characterization of a 3D encapsulation with silicon vias for radio frequency micro-electromechanical system resonator. Chin. Phys. B 2017, 26, 5. [CrossRef] Kuang, Y.; Xiao, D.; Zhou, J.; Li, W.; Hou, Z.; Cui, H.; Wu, X. Theoretical model and experiments of glass reflow process in TGV for 3D wafer-level packaging. In Proceedings of the 2018 IEEE International Symposium on Inertial Sensors and Systems (INERTIAL), Moltrasio, Italy, 26–29 March 2018; pp. 1–4. Rödjegård, H.; Andersson, G.I.; Rusu, C.; Löfgren, M.; Billger, D. Capacitive slanted-beam three-axis accelerometer: I. Modelling and design. J. Micromech. Microeng. 2005, 15, 1989. [CrossRef] Benali, A.; Bouya, M.; Faqir, M.; El Amrani, A.; Ghogho, M.; Benali, A.; Benabdellah, A. Through Glass via Thermomechanical Analysis: Geometrical parameters effect on thermal stress. In Proceedings of the 2013 8th International Design and Test Symposium, Marrakesh, Morocco, 16–18 December 2013. Lee, J.-Y.; Lee, S.-W.; Lee, S.-K.; Park, J.-H. Through-glass copper via using the glass reflow and seedless electroplating processes for wafer-level RF MEMS packaging. J. Micromech. Microeng. 2013, 23, 085012. [CrossRef] Xiao, Z.; Yamaner, F.Y.; Oralkan, Ö. Fabrication of Vacuum-Sealed Capacitive Micromachined Ultrasonic Transducers with Through-Glass-Via Interconnects Using Anodic Bonding. J. Microelectromech. Syst. 2017, 26, 226–234. Yamaner, F.Y.; Zhang, X.; Oralkan, O.M. A three-mask process for fabricating vacuum-sealed capacitive micromachined ultrasonic transducers using anodic bonding. IEEE Trans. Ultrason. Ferroelectr. Freq. Control 2015, 62, 972–982. [CrossRef] [PubMed] Helfrick, A.D. Q Factor Measurement; CRC Press: Boca Raton, FL, USA, 2004; Volume 200, pp. 1–5. Bianco, S.; Cocuzza, M.; Ferrero, S.; Giuri, E.; Piacenza, G.; Pirri, C.F.; Ricci, A.; Scaltrito, L.; Bich, D.; Merialdo, A. Silicon resonant microcantilevers for absolute pressure measurement. J. Vac. Sci. Technol. B Microelectron. Nanometer Struct. 2006, 24, 1803–1809. [CrossRef] Chidambaram, V.; Ling, X.; Bangtao, C. Titanium-based getter solution for wafer-level MEMS vacuum packaging. J. Electron. Mater. 2013, 42, 485–491. [CrossRef] © 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
