Thermal Stress in MEMS

Overview

Microelectromechanical systems (MEMS) devices are sensors and actuators with the mechanical movement as a major performance measure. Therefore, the performance of a MEMS device can be strongly affected by thermal stresses resulting from constraining interactions among device’s multiple layers and between the package and the device. Such an effect is a main design and manufacturing consideration for almost every MEMS device and package. However, thermal stress is not always an undesirable effect. Thermal stress can be used to create novel configurations during MEMS fabrication and assembly. Various three-dimensional shapes can be formed by stress-induced deformations and displacements. Thermal stress can be used to generate mechanical movements for a MEMS actuator. The MEMS thermal actuator is commonly used with actuation controlled by heating and cooling configurations with asymmetric materials or temperature distributions or asymmetric geometry. Thermal stress can also be used for MEMS sensors. The stress-induced displacements of the sensor can be related to temperature, electromagnetic heating or other physical, chemical or biological loadings. As a result, the displacements can be measured to sense temperature, electromagnetic field or other environmental changes. This chapter will review the thermal stress effects on MEMS devices and thermal stress-enabled MEMS fabrication, actuation and sensing.

Introduction

MEMS sensors and actuators have been well known to the general public after the introduction of smartphones and game consoles in 2007. The amazing user interface resulting from the detection of user’s mechanical movements created a paradigm shift in these applications. These two examples are only the representatives of MEMS diverse applications, which include: pressure sensors, tilt sensors, accelerometers, gyros, chemical micro sensors, lab-on-a-chip, micro biomedical devices, resonators, displays, optical switches, radio frequency (RF) switches and passive components, printer heads, energy harvesting and storage, and data storage. Military and commercial applications will increasingly benefit from the use of MEMS, especially when anticipated significant improvements in performance are realized via the integration of MEMS with nanoelectromechanical systems (NEMS). The societal impact resulting from MEMS is already very visible, and the impact is expanding fast.

Figure 1 illustrates the main features in MEMS design, fabrication and wafer-level capping [1, 2]. MEMS pressure sensors, accelerometers, micro-mirrors, radio frequency (RF) switches and many other sensors and actuators can be represented by the cantilever beam as shown. MEMS design requires solid or fluid modeling and simulation considering electro-thermal-fluid-mechanical behaviors. MEMS fabrication involves deposition and removal of micron-thick layers with controlled mechanical, electrical, optical, chemical and biological properties [3, 4]. After fabrication, the sacrificial materials are removed in order to release the device for mechanical movements. Before or after release, all such micro-scale, movable devices are protected by wafer-level capping, so the devices can survive through regular packaging and testing processes.

Fig. 1

MEMS design, fabrication and wafer-level capping

As illustrated above, the micro-scaled mechanical movement is a major performance measure, which demands special design and manufacturing considerations. Such movements can be affected by thermal stresses resulting from constraining interactions among device’s multiple layers with different coefficients of thermal expansion (CTEs) and residual stresses. The thermal stresses can also be generated due to the CTE mismatch between the package and the device. As a result, thermal stress effect on device performance during manufacturing and operation is a main consideration for almost every MEMS device and package.

Thermal stress is in fact not undesirable if we know how to control the stress. Thermal stress can be used to create novel configurations during MEMS fabrication and assembly. For examples, the seal between a check valve and the substrate can be enhanced with a pre-stressed configuration. An array of probes with out-of-plane configurations can be formed by stress-induced bending or curling.

Thermal stress is commonly used in thermal actuators with asymmetric materials or temperature distributions or asymmetric geometry. With an electrical current passing through the structure, a thermal actuator can achieve the designed movements controlled by heating power and frequency. Energy consumption is a major concern on thermal actuators; however, they are widely used because of the following advantages: large forces and displacements, easy fabrication, and less susceptible to adhesion failure.

Thermal stress can also be used for MEMS sensors. The stress-induced mechanical displacements of the sensor can be related to temperatures. As a result, the displacements can be measured to detect temperatures in a harsh environment, e.g., on a 600 °C engine block. Similarly, a probe can be heated by an electromagnetic field and its nano-scaled displacements can be measured to detect the field strengths.

The thermal stress effects on MEMS devices and thermal stress-enabled MEMS fabrication, actuation and sensing will be reviewed in the following sections. Most of analytical and numerical models on thermal stress in MEMS are not different from those developed for electronic devices, packages and systems. Therefore, we will not repeat such thermo-mechanical models. If interested, these models can be found in the references cited and other chapters in this book.

Thermal Stress Effects on MEMS Devices

Thermal stress effect on micro-scaled movements should be considered during design and manufacturing of almost every MEMS device and package. The main challenge is to understand how to correlate MEMS device performance with thermal stress. With the understanding, we would then be able to design the MEMS device and package to control the thermal stress effect. We will use two examples to illustrate this challenge.

Figure 2 shows an example with a MEMS accelerometer in a plastic-encapsulated package soldered on a printed circuit board (PCB) [5]. Thermal stresses could be generated due to different temperature effects on the composite structure during manufacturing and operation. These stresses could deform a MEMS chip and cause offset shifts. Zhang et al. improved their modeling accuracy by correlating finite-element models using measured material properties and package warpage. In addition, they applied a reduced-order sensor and package interaction model to calculate the device offsets and compared the results with experimental data collected from a three-axis accelerometer, which used a single mass for all three axes sensing. The accelerometer achieved very low offset (< 10⁻³ g/°C) in all three axes over a temperature range from −40 °C to +80 °C.

Fig. 2

MEMS sensor output offset induced by packaging stresses in a plastic package (© 2007 IEEE. Reprinted, with permission, from Accurate Assessment of Packaging Stress Effects on MEMS Sensors by Measurement and Sensor–Package Interaction Simulations by Xin Zhang, Seungbae Park, and Michael W. Judy, in IEEE Journal of Microelectromechanical Systems, June 2007)

The 3-axis MEMS accelerometer is shown in Fig. 3. The optimization study focusing on thermal stress considered: (a) the shape and dimension of the sensing and referencing capacitors, (b) the location of the spring anchors, and (c) the length of the support arms. The optimum configuration was reached when the sensing capacitance and the referencing capacitance had negligible changes from their nominal values over the temperature range. They optimized the configuration with the Z-axis output signal being made insensitive to package stress and warpage. Then, they optimized the configuration further to minimize the thermal stress effects on the X-axis and Y-axis output offsets. The output offset was reduced by at least five times better than that of the device measured prior to the optimization study.

Fig. 3

(a) MEMS three-axis accelerometer fabricated in the iMEMS process from Analog Devices. (b) Magnified view of a quarter of the MEMS structure. Two fixed fingers and one sensing finger are interdigitated for the differential capacitive measurements during accelerations in the X-axis or Y -axis. Z-axis sensing relies on direct sensing of capacitance change between the movable polysilicon mass and the ground plane fabricated on top of the silicon substrate (© 2007 IEEE. Reprinted, with permission, from Accurate Assessment of Packaging Stress Effects on MEMS Sensors by Measurement and Sensor–Package Interaction Simulations by Xin Zhang, Seungbae Park, and Michael W. Judy, in IEEE Journal of Microelectromechanical Systems, June 2007)

Figure 4 illustrates another example on the control of the thermal stress effect on MEMS device performance [6]. Fixed-fixed beam capacitive RF MEMS switches with rectangular configurations were sensitive to residual bi-axial stress, which affected beam’s stiffness. When the residual stress was reduced by the CTE mismatch between the device and the substrate, it could lead to excessively varying actuation voltages and up-state capacitances when the environmental temperatures changed. Reines et al. solved this problem by using a circular switch. The suspended plate was anchored symmetrically in four locations with arc-angled cutouts placed next to each anchor. When the switch was subjected to a temperature increase, the fixed anchors would induce a compressive stress in the switch and reduced the initial residual stress. The switch’s stiffness could be designed to achieve excellent performance over a wide temperature range, e.g., −15 °C to 125 °C. As shown in Fig. 4, four different designs were simulated. Case 1 represented a flat circular membrane. Case 2 was the switch consisting of both the 0.5-μm-thick bottom metal and 0.28-μm dielectric layers with a sidewall angle θx = 90 ° and a total vertical step height of s = 0.78 μm. Case 3 considered the anchor area with the side-wall angle θs = 30 °. Case 4 was the addition of case 2 and 3 and represented the real topography of the circular switch to be fabricated and characterized. These switches’ changes in the vertical displacements at 25 °C and 95 °C are also presented in the figure. These results demonstrated that the switch’s topography significantly affected the vertical beam displacements versus both initial in-plane biaxial stress and temperature. It was essential to minimize the effect; otherwise, thermal stress in poor designs could lead to excessively varying up-state capacitance and pull-in voltage.

Fig. 4

(a) Top view and cross sections of the four cases of switch topography, and (b) the change in the vertical displacement (ΔZ) of the beam versus x-axis position at both 25 °C and 95 °C, with residual stress = 60 MPa and without a stress gradient across the beam (© 2011 IEEE. Reprinted, with permission, from by Isak Reines, Brandon Pillans, and Gabriel M. Rebeiz, in IEEE Journal of Microelectromechanical Systems, Feb. 2011)

With the optimum design, the average pull-in voltage slope versus temperature for the two wafers was −55 mV/°C for the circular design, compared to −310 mV/°C for the standard, rectangular switch. The pull-in voltage of a standard switch measured was about 42 V at 25 °C and 12 V at 100 °C. The pull-in voltage of a circular switch was about 37 V at 25 °C and 35 V at 100 °C. The performance of a RF MEMS switch could be less dependent on the temperatures with the thermal stress effect controlled. The thermal stress effect is very important; the design procedure illustrated by the aforementioned two examples for the MEMS accelerometers and switches should be applied to design almost every MEMS device and package.

Thermal Stress-Enabled MEMS Fabrication

Thermal stress is not always undesirable for MEMS devices. Sometimes, thermal stress can be used to create specific pre-stressed configurations or stiffness levels. We will illustrate this thermal stress-enabled fabrication with two examples. Figure 5 shows a MEMS check valve [7]. The valve’s cracking pressure, i.e., the pressure required to open the valve, was an important performance measure for micro-fluidic applications. The surface-micromachined parylene check valve used residual thermal stress in the parylene to control the cracking pressure. The check valve was thermally annealed at predetermined temperature after the sacrificial photoresist was released. After thermal annealing, it was quenched down. Since the residual tensile stress of the thermally annealed parylene C could reach 34 MPa at 250 °C, the parylene-based slanted tethers could provide a high downward force to assure good sealing.

Fig. 5

Schematic of cracking-pressure controlled parylene check valve using the residual tensile stress in parylene after thermal annealing (© 2010 IEEE. Reprinted, with permission, from Cracking Pressure Control of Parylene Check Valve Using Slanted Tensile Tethers by J. Lin, F. Yu, Y.-C. Tai, in Tech. Digest 23rd IEEE International Conference on MicroElectroMechanical Systems, Jan. 24–28, 2010)

If the cracking pressure was not high enough, there would be flow leakage. If the cracking pressure was too high, it would introduce undesirable pressure loss and the correspondingly reduced flow rate. In this thermal stress-enabled sealing, the annealing temperature could be adjusted in order to control the cracking pressure. In addition, a design could change tethers geometry, parylene thickness, and tethers’ slope angle. If needed, different materials could be used to change the CTE mismatch between the structure and the substrate.

Thermal stress-induced deformations have a potential problem: the deformations or the corresponding loadings can be temperature dependent. This problem can be solved by introducing stresses through other means. Figure 6 shows an example of a stress-enabled MEMS fabrication without the temperature effect [8]. There were 4 rows with 200 springs in each row for interconnect and testing. The springs could be fabricated with standard wafer-scale thin-film deposition processes on a BICMOS wafer, Corning 1,737 glass or any other substrates. The main structure of each spring consisted of multiple layers to provide a stress gradient with tensile layers on top. The spring metal layers could be sputtered MoCr layers which were deposited at different pressures. In addition, the structure could be covered by electroplated nickel hardened gold. This alloy provided mechanical strength and stiffness, and the gold protected it from oxidation and increased its electrical conductivity.

Fig. 6

SEM photograph a spring chip (© 2006 IEEE. Reprinted, with permission, from Pressure Contact Micro-Springs in Small Pitch Flip-Chip Packages by Chow EM, Chua C, Hantschel T, Van Schuylenbergh K, and Fork DK, in IEEE Transactions on Components and Packaging Technologies, 2006)

There were 800 springs on a 3×10 mm chip with each spring 180 μm long, 14 μm wide, and 5 μm thick. The spring tips were 57 μm tall, with +/− 5 μm height variations. The springs were interleaved with a pitch of 40 μm. The pitch of the pads to be probed was 20 μm. As shown in Fig. 6, stress-formed shapes did not vary much. This fabrication process could produce a large number of three-dimensional MEMS devices uniformly distributed across a chip or a wafer. It should be noted that these shapes are formed by a controlled stress gradient resulting from the fabrication pressures instead of temperatures. As a result, these shapes are temperature independent. For thermal stress-enabled fabrication, the configurations or the loadings are usually affected by temperature. As shown in this example, this effect can be eliminated if needed.

Thermal Stress-Enabled MEMS Actuators

Thermal stress-enabled MEMS actuators can generate displacements or forces for a MEMS device by constraining differential thermal expansions. Thermal actuators require considerable power consumption since continuous heating is needed to maintain the thermal stress-induced displacements or forces. Their switching speeds can reach only in the KHz range, which is limited by cooling rates. In addition, due to high operating temperatures they are prone to overheating and permanent structural deformation. However, thermal actuators possess many advantages, including high force generation, large static displacement, low operating voltages, easy of fabrication, and less susceptible to adhesion failures. Therefore, thermal actuators have been widely studied [9]. Over the decades, various microsystems utilizing thermal actuators have been designed and fabricated, including microswitch and relay, microprobe, micromirror, microgripper and tweezers, micropositioner and tester, microleg and cilia, step motors, and micro pump and valve.

Thermal actuators can be categorized by how the displacements are generated through asymmetric distributions of materials, temperatures and geometries. The basic structures of these three categories are to be described as follows.

Asymmetric Materials Distribution: Bimorph is the most popular approach for thermal actuators with asymmetric materials distribution. It is shown in Fig. 7 with two materials with different CTEs laminated and anchored at one end. Upon heating by passing electrical current through a conductor layer, one material expands more than the other and introduces large thermal stresses at the interface. To release the stresses, the actuator bends or curls along the direction normal to the plane of fabrication. The curvature can be determined by geometric parameters, material properties and temperature distributions of the beam. To increase the curvature, we can increase the CTE mismatch and temperature change, or decrease the thickness of the cantilever. The tip deflection can be increased by increasing beam length. To further amplify the displacement of a single bimorph, cascaded bimorph actuators can be designed [10].

Fig. 7

Schematic of a bimorph-based thermal actuator

Asymmetric Temperature Distribution: Thermal actuators can generate displacements or forces by developing a non-uniform temperature distribution along a single layer of material. Figure 8 illustrates the use of cold and hot arms to move the free end in plane or out of plane. The hot and cold arms are electrically connected in series; therefore, an electrical current would pass through both arms. The hot arm’s cross-sectional area is much smaller than that of the cold arm. As a result, the temperatures in the hot arm would be much higher than those in the cold arm due to different current densities and the corresponding heat sources. Figure 9 illustrates a polysilicon-based thermal actuator with an out-of-plane movement. The arm’s thickness is 2 μm and its width is 20 μm. With 3 V applied, the maximum temperature in the hot arm increases to about 600 °C while that in the cold arm reaches only 300 °C. Without the voltage applied, the temperatures would drop to room temperatures through heat conduction to the substrate through the arms, the flexure and the anchors. With micro-scaled size, the actuator can be heated or cooled in msecs or quicker.

Fig. 8

Schematics of thermal actuators driven by asymmetric temperature distributions

Fig. 9

Asymmetric temperature distribution of a thermal actuator simulated

Asymmetric Geometry: Thermal actuator can also operate with an asymmetric geometry. As shown in Fig. 10a, for an actuator with both ends fixed, thermal stresses resulting from the constraining thermal expansion would move the apex outward. This Chevron-like structure is the most popular approach for thermal actuators with asymmetric geometry. To increase the displacement, we should increase the leg length or decrease the leg angle. A meandering, i.e., compliant, actuator as shown in Fig. 10b is a good approach [11].

Fig. 10

Schematic of thermal actuators with asymmetric geometry with (a) chevron structure and (b) meandering structure

It should be noted that thermal actuators are not limited by the use of inorganic materials. A recent example was to embed silicon within SU-8 polymer [11]. In this design, the silicon microstructure with a high aspect ratio was meandering in the layout. The space surrounding the meandering silicon was filled by SU-8 polymer, forming a layer-like composite. The confinement of the polymer inside the silicon structure led to higher displacement, higher stiffness, and less out-of-plane motion. For example, a microgripper 440 μm long and 85 μm wide achieved single arm displacement of 34 μm at an input voltage of 4 V with an average temperature of approximately 170 °C [12].

MEMS Devices Driven by Thermal Actuators: Figure 11 presents a MEMS device driven by bimorph actuators. The device was a 1 mm², 2-axis single crystalline silicon-based, aluminum-coated scanning micromirror [13]. With two arrays of bimorph actuators, the micromirror could rotate up to 40°. When the mirror rotated 40°, the applied current was about 6.3 mA, and the power dissipation was about 95 mW with 15 V applied. Beyond this angle, the high thermal stress in the bimorph actuator resulted in mirror instability. The resonant frequency of the mirror was measured to be 445 Hz. When the frame rotated 25°, the current, voltage and power were 8 mA, 17 V and 135 mW, respectively. The resonant frequency of the frame actuator structures was measured to be 259 Hz.

Fig. 11

SEM of a bimorph-actuated 2-D mirror (© 2004 IEEE. Reprinted, with permission, from A Two-Axis Electrothermal Micromirror for Endoscopic Optical Coherence Tomography by A. Jain, A. Kopa, Y. Pan, G. K. Fedder, and H. Xie, in IEEE Journal of Selected Topics in Quantum Electronics, May/June 2004)

Figure 12 presents another example of thermal actuator-based MEMS device. A micro-mirror driven by thermal actuator was used for optical active alignment [14]. The micromirror steered a laser beam for active alignment without requiring a precision robot and fixture. The micromirror was suspended by four thermal actuators with asymmetric temperature distributions (see Fig. 8b). These structures were lifted up to 45° with mechanical locking mechanisms; thereafter the micromirror could be steered two-dimensionally by the actuators. Each actuator could move out-of-plane due to bending resulting from differential heating. Powering a pair of the actuators could rotate the mirror along a horizontal or vertical axis. Powering a single actuator can rotate the mirror along a diagonal axis. Beam steering with such a large degree of freedom could improve the optical coupling efficiency from 10 % to over 80 % in an experiment using a vertical-cavity surface-emitting laser (VCSEL) and a multi-mode fiber.

Fig. 12

(a) Schematic view of a laser-to-fiber coupling concept and (b) micrograph of the microstructure with four thermal actuators (© 2002 IEEE. Reprinted, with permission, from An Integrated Micro-Optical System for Laser-to-Fiber Active Alignment by Ishikawa K, Zhang J, Tuantranont A, Bright VM, and Lee YC, in Tech. Digest 15th IEEE International Conference on MicroElectroMechanical Systems, Jan 20–24, 2002)

Thermal Stress-Enabled MEMS Sensors

Thermal stress can be used for MEMS sensors. For a bimorph actuator mentioned above, its mechanical displacements are usually directly related to the temperatures. As a result, we can use the same device as a temperature sensor by detecting its displacements and calculate the corresponding temperature. Such thermal stress-enabled sensors can be used in applications that are beyond the range covered by the semiconductor sensors. Active silicon circuits usually do not operate far beyond 120 °C due to the small band-gap. Thermocouples, fiber optic sensors or SAW-based sensors are not appropriate when a wired connection and line-of-sight are not feasible. Two examples are to be presented to illustrate such applications.

Figure 13 shows a passive, capacitive temperature-sensing solution based on multimorph cantilevers [15]. The multimorph structure consisted of a 500 nm-thick layer of thermal oxide grown on a silicon wafer. A low-stress 500 nm-thick nitride film was deposited on the oxide. After patterning these two layers, a thin (30 nm) Ti adhesion layer and 500 nm-thick Au film were deposited and patterned. This temperature sensor was good for engine component health-monitoring applications with temperatures above 600 °C. As temperature increased, the cantilevers deflected due to the CTE mismatch. This deflection resulted in a change in the capacitance between the cantilever and the substrate. It was an effective sensor for harsh environment applications due to the passive implementation, silicon micro-fabrication, and robust materials used.

Fig. 13

The multimorphs operate similarly to previous devices, with a static capacitance pad and a temperature-sensitive multimorph. The capacitance is changed between the metal on the beam and the low-resistivity silicon substrate (© 2012 IEEE. Reprinted, with permission, from A 600 °C wireless multimorph-based capacitive MEMS For component health monitoring by S. Scott, M. Scuderi and D. Peroulis, in Tech. Digest 25th IEEE International Conference on MicroElectroMechanical Systems, Jan/Feb 2012)

For such a temperature sensor, a large range of the temperature-dependent displacements is desirable for the measurement of a wide range of temperatures. Sensitivity is another important consideration. For high sensitivity, the capacitance change under a specific temperature change should be maximized. Dynamic response is another consideration for real time monitoring. System response time is usually determined by the thermo-mechanical response. Therefore, it is essential to fabricate micro-scaled sensors by using MEMS technologies.

Figure 14 illustrates another example of the thermal stress-enabled MEMS sensor. It was a multimaterial probe for sensing the magnetic field component by use of a calorimetric approach [16]. The sensor converted thermal power absorbed from the high-frequency magnetic field component to mechanical deflection. For reducing eddy current contribution, dielectric materials were needed. For the optical measurement, the cantilever should reach a tip deflection of less than 50 μm or a radius of curvature of more than 2.5 mm. This specification could only be achieved by utilizing an asymmetrical tri-layer design. Specifically, the cantilever structure was made of nitride–oxide–nitride, with different thicknesses for the top and bottom nitride layers.

Fig. 14

Schematics of (a) the basic configuration of the MEMS probe and (b) misalignment problem due the curled cantilever (© 2007 IEEE. Reprinted, with permission, from Dielectric Asymmetric Trilayer Cantilever Probe for Calorimetric High Frequency Field Imaging by S. Lee, T. M. Wallis, J. Moreland, P. Kabos, and Y. C. Lee, Journal of Microelectromechanical Systems, 2007)

As shown in Fig. 14, a gold ring was patterned at the tip of the cantilever as an actuation/detection sensor. A time-varying magnetic flux from the RF circuit passing through the gold ring induced an electromotive force, according to Faraday’s law. Induced current was linearly proportional to the magnitude of the magnetic field. This current generated joule heating in the gold ring, resulting in a temperature gradient across the body of the cantilever. The cantilever deflected due to the CTE mismatch among the dielectric materials. Tip deflection was measured with the laser beam bounce technique routinely employed in Atomic Force Microscope (AFM) systems. A laser beam was focused onto the tip of the cantilever, and was reflected back to a quad-photodiode detector. The change in the photodiode differential signal was a proportional measure of the deflection of the cantilever. An oxide-nitride bimorph structure was considered first. Its tip deflection was higher than 50 μm, and this undesirable curling of the cantilever had to be reduced in order to use the laser beam measurement system. This problem was solved by using a tri-layer cantilever beam with 500 nm oxide sandwiched by a 190 nm thick top nitride layer and a 150 nm thick bottom nitride layer. With such a tri-layer fabricated, magnetic field strengths of a microstrip RF resonator were detected at a distance of 300 μm away from the resonator and a RF power down to 10 μW. With such a very weak field strength, the tip displacement detected was only around 3 nm. Thermal stress-enabled MEMS sensor could be a powerful probe to detect electromagnetic field strength. Many other similar probes have been developed to characterize physical, chemical and biological loadings [17].

Summary

MEMS sensors and actuators are making a major societal impact. Thermal stress in MEMS is an essential design and manufacturing consideration. We have reviewed thermal stress effects on MEMS device and thermal stress-enabled fabrication, actuation and sensing. With a good understanding of the thermal stress effects, we are able to design a device and its package for reliable performance over a large range of temperatures. We are also able to control the thermal stresses to design and manufacture novel device configurations with superior actuating and sensing capabilities.