Mechanical strength study of SiO₂ isolation blocks merged in silicon substrate

The fabrication of inertial sensors with bulk micromachining has the advantages of large proof mass, high sensitivity, less residual stress and good mechanical stability. The integration of a bulk sensor with an electronic circuit demands the mechanically robust electric isolation of microelectromechanical systems (MEMS) structures from each other and circuits. Different ways of isolating moveable microstructures from fixed ones have been reported [1–5].

Silicon micromachining methods that are able to create silicon dioxide (SiO₂) blocks at selected locations deep in a silicon substrate have been proposed [6–9]. The process combines deep-reactive-ion etching (DRIE), thermal oxidation, deposition of SiO₂ and optional planarization. This isolation block (IB) serves as an underlying large-area isolation base for capacitors, transmission lines, coplanar waveguides and transformers [6]. Recently, this technology has been used for the fabrication of MEMS devices with moveable parts, where the IB is also a bearing structure for the suspended elements. These devices are the comb drive, accelerometer and gyroscope [10]. The IB is used for mechanical connection and bearing, but it electrically isolates two structures in the monolithic bulk integration of MEMS devices. The SiO₂ films and thick silicon oxide layer electrical properties have been studied [8, 11] and are well defined. However, the mechanical properties of IB have not been studied at all, because in previous works, the IB was merged into silicon substrate and did not bear sufficient load. The mechanical characterization of the IB as a part of MEMS is particularly important for their design, performance realization and reliability analysis.

In order to ensure reliability, we need to have precise knowledge of the materials' properties. The mechanical properties of construction materials should be measured on the same scale as microdevices, since they are different from those of bulk materials. The reason for the differences is the size effects of thin-film materials. Thin-film materials often have a different composition phase and microstructure from those of bulk materials, even if they have the same material names. The formation processes, such as deposition, thermal treatment, implantation and oxidation, are inherent methods for thin film materials. The stiffness of the thin film structure additionally depends on the internal stress, which changes by an order of magnitude, and the effects of the elastic properties on the device performance should be considered [12]. Indentation is perhaps the most commonly applied means of testing the mechanical properties of materials. The nanoindentation technique was developed in the mid-1970s to measure the hardness of small volumes of material. Nanoindentation can be used to measure the mechanical parameters of thin layers, such as for example the Young's modulus (E), and it can be used to determine MEMS characteristics, such as the stiffness of beams and bridges [13]. The principle is simple: a tip, which is assumed to be non-shrinking, is pressed against the test materials. Conventional nanoindentation methods for the calculation of the elasticity modulus (based on the unloading curve) are limited to linear, isotropic materials. Problems associated with the 'pile-up' or 'sink-in' of the material on the edges of the indenter during the indentation process remain under investigation. The nanoindentation technique could hardly be applied directly to study the elastic properties of IBs. In fact the IB is not just a small piece of material but rather a set of constructive elements (beams) of fully and/or partially oxidized silicon. The mechanical characteristics could be determined from its load-deflection behavior. A cantilever beam can be bent by pressing on the free end with a nanoindenter stylus. The nanoindenter can monitor both the force applied and the resulting displacement. A simple beam theory can convert the displacement into strain to obtain strength and Young's modulus [14]. A similar technique [15] involves pulling downward on a cantilever beam by means of the electrostatic force. An electrode is fabricated into the substrate beneath the cantilever beam, and voltage is applied between the beam and the bottom electrode. The force acting on the beam is equal to the corrected electrostatic force that includes the effects of fringing fields acting on the sides of the beam. Another measurement besides the stress–strain behavior that can reveal the Young's modulus of a material is the determination of the natural resonance frequency [16], provided that the geometry and density are known. The cantilever can be vibrated by a number of techniques, including a laser, loudspeaker or piezoelectric shaker. The frequency that produces the highest amplitude of vibration is the resonance frequency.

Any of the techniques mentioned above that strain specimens in order to measure mechanical properties can also be used to measure fracture strength. The fracture strength can be obtained from the load displacement curves if the geometry of the specimen is known, using analytical analysis or the finite element approach (FEA).

The disadvantages of the electrostatic loading technique are that it is low in force—a few millinewtons—and it demands electrodes. The resonant method, in the case being considered, demands frequencies above 100 kHz, consequently requiring a vacuum environment and good mechanical contact between the exciter and the specimen under study.

The method with the most potential is seemingly integrated on-chip testing, in which the specimen, actuator and displacement-force sensors are fabricated on a chip [17]. There is little point in implementing this method for the device level tests.

This article describes a new, successful methodology of the fracture testing approach which incorporates an accurate experimental system with the finite element method to determine the fracture strength of the IB by bend and tensile tests. The tests were performed at the device level. The bending strength and the fracture tensile strength of the IBs were achieved by means of testing. Compressive and fatigue strength properties were estimated.

IB technology for MEMS devices was described in [8, 18]. Accelerometer design based on IB technology is able to substitute MEMS capacitive accelerometers fabricated with silicon-on-insulator bulk technology [2, 19]. Figure 1 presents a model of an in-plane IB capacitive accelerometer and an SEM image of the structures under study, which are parts of a microaccelerometer. The cross-section plane at the anchors of the supporting spring shows the membrane profile. The IBs supporting the set of stationary electrodes both broken after the bending test and untouched are shown in figure 1(b).

Zoom In Zoom Out Reset image size

Handling a specimen that is only a few microns thick and a few millimeters in planar dimension is a challenge; one cannot simply pick it up and place it in a testing machine. It means that one part of the specimen must remain fastened to a substrate that can be handled and another part is attached to a mechanism that deforms it and measures both the forces and the resulting displacements or strains. The test principle is simple: a tip assumed to be non-shrinking is pressed against the test structures. After the loading-unloading cycle, a deflection versus force curve is plotted. The stress is evaluated from this plot for the tested structure. We designed and calibrated a stress test apparatus comprising a low friction actuator based on the voice coil actuator (VCA) found in generic hard disk drives (HDD). These actuators are well known as direct-drive limited-motion devices that utilize a permanent magnetic field and coil winding to produce a force on the winding proportional to the current in the coil. The VCA in the HDD generally has no 'natural position'. All movement and positions are equal.

The advantages of the VCA are zero hysteresis, linearity (i.e. a linear relationship between force and current), high force, high precision and direct drive (i.e. there are no gears, cogs or screws, thus there is no backlash). The loading equipment shown in figure 2 was used for the bending tests.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Particular attention has been given to the rigidity of the test frame and to making specimen mounting easier. A specimen is bonded to a fixed sample holder mounted on the 3D moving platform with strong fixation after aligning. We have enhanced the voice coil leads to avoid generic ribbon cable stiffness. A loading force is applied to the structure under study by means of a tungsten wire or piece of silicon of special geometry. The probe tip is firmly held to the actuator arms where read/write (R/W) heads are mounted. The R/W head support arm of the VCA is very light, but also stiff. The deflection of the gauge, which is the specimen part, is recorded during the test at different values of the applied load. The microscope of the test apparatus is equipped with a charge-coupled device (CCD) video camera and the image size is 1024 × 768 pixels. The zoom setting in the microscope and the image-forming lens combined with the objective give magnification to the CCD, so that each pixel is equal to 0.3 µm. During loading, sequential digital images are taken through the microscope and saved individually.

The strength properties of the actuator with the ready-assembled tip were tested. The actuator was supplied with the current to develop a force of 0.5 N with its tip against a stainless steel block. The visual inspection of the lever and tip with the microscope during load did not reveal any noticeable bending of the arm or tip on the scale of 0.5 µm. There was no evidence of tip damage after the test.

The current through the voice coil defines the force applied to the tip. In order to take a measurement with the actuator, its sensitivity in Newtons per Ampere has to be known as precisely as possible. To ensure reliable experimental data one needs to calibrate the apparatus with the well-defined simple system of the same scale as the structure under investigation. A useful technique to calibrate the actuator involves the use of a reference cantilever as a reference force [20]. The MEMS element described in [21] was chosen as the reference specimen. The element comprises a silicon frame with a silicon membrane suspended on two consoles. This structure has a simple geometry and its mechanical properties could easily be evaluated with the FEA model. One part of the silicon frame is removed to allow visual inspection of the suspended membrane deflection under load. The test procedure is performed as follows. Before testing, the image of the control specimen surface, where the load will be applied, is captured with a CCD camera through the microscope. A silicon frame is bonded to the base of the 3D platform and adjusted to be viewed in the microscope field. A microscope X-Y stage with a VCA attached is positioned so that the tip touches the test sample. The relative position of the tip on the surface is controlled with the help of a horizontal microscope. The tip is aligned normally to the sample surface. The deflection of the structure under load/unload is captured by the CCD camera. The current-displacement curves of the system are shown in figure 4. After the experiment, the sample surface was inspected to define the tip position and to ensure the load remained at a fixed point during the entire testing procedure.

Zoom In Zoom Out Reset image size

As can be seen in figure 4, the load and unload points do not match each other. The most important source of the experimental curve mismatch in these measurements is friction in the arm shaft. The value of the current-deflection constant k is determined from the linear regression of the data averaged from the number of load-unload cycles (dashed dotted line in figure 4). The loads were sequentially applied to both sides of the sample to take into account the internal stress in silicon, if any. A three-dimensional finite element model (FEM) of the reference mechanical structure has been created to calculate the quantitative mechanical response of the reference structure to the applied force. The isotropic Young's modulus of bulk silicon was 165 GPa and the specimen geometry was determined in the work mentioned above. The deflection found in the model is compared to the measured data in figure 4. From these experiments the actuator resolution is found to be ± 80 µN. The calibration coefficient that maps the measured gauge deflection and FEM-calculated deflection caused by the applied force was (0.002 ± 0.000 15) N mA⁻¹. The force calibration is linear in the range 0–0.8 N.

The presented methodology combines experimental techniques and numerical modeling to determine the calibration coefficient of the actuator. A precise experimental system is designed to bend-fracture the structure under study with controllable load and to collect force-deflection-fracture characteristics. The stress at fracture is then derived from the maximum stress in the model. With this methodology, the mechanical properties and fracture strength of the IB can be obtained.

The second part of the investigation considers the FEM application as a method for analysis of the force-deflection behavior observed in the microstructure under loading by the actuator. The FEM is needed to account for large membrane and beam deflections, twist, material anisotropy, the non-ideal geometry at the base of the bearing membrane and the complex structure of beams with silicon inclusions. The torsional deformation or twist is an additional deflection caused by a point load that is applied at a distance from the center line of the structure. The experimental fracture forces are used to estimate the fracture stresses with values derived from the validated FEM. The linear static analysis allows calculation of the deflections, stresses and strains produced by the applied loading. We did not use the nonlinear analysis because brittle materials like silicon and SiO₂ fail in an elastic manner. This was supported by the fracture tests.

The finite element model of the tested structure shown in figure 5 is comprised of the device silicon membrane, oxidized silicon beams, the IB and the silicon bar fastening the free ends of the oxidized beams. We used a Cartesian coordinate system with the X, Y and Z-axis in the beams' length, thickness and width directions, respectively. These axes coincide with the crystallographic axes [1 0 0], [0 1 0] and [0 0 1] of the cubic single crystal silicon. Due to this orientation, the anisotropy of the material is easily modeled by defining the Young's modulus, shear modulus and Poisson's ratio for the XYZ directions. These properties are listed in [22]. The length of all beams is 65 µm, and their widths vary from 3.5 to 10 µm. There are inclusions of silicon at the nodes of the beams. The length direction of the IB's beams is along the [1 0 0] orientation of the wafer. The IB was fabricated by etching wavy line trenches in silicon substrate with DRIE and thermal oxidation of the silicon walls between trenches. The oxidized walls of the non-uniform cross-section form beams with nodes of partially oxidized silicon islands. The silicon bar is 100 µm wide and its length is 400 or 660 µm. The thickness of the silicon bar and beams is 30 µm. The silicon beam, which is a fixed electrode of the accelerometer sensing capacitor, serves as a gauge. Its length is 500 µm. The membrane of the device is etched by DRIE in silicon wafer and has a non-uniform complex profile (figure 5). The thick side of the membrane is integral with the bulk of the wafer which is fixed. The arrows designated in figure 5 as 'Bending', 'Tension' and 'Compression' simulate the forces applied to the silicon bar to bend, tense and compress the IB. The force is applied as a distributed load acting on a small area of the bar with dimensions defined from the load tip imprint on the bar surface. The beam and membrane geometry were measured with a calibrated SEM at a magnification of 1000. Particular attention has been devoted to mesh the geometry of the oxidized silicon beams. The FEM model comprises elements with linear dimensions of 1 µm on average for both the beam and the silicon inclusions. In the regions where higher stress was expected, the density of the grid elements was increased, as shown in figure 6(a).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The structure of the oxidized beams of the IB was interpreted using an optical microscope. The left part of the optical image in figure 6(b) shows beams covered with an aluminum layer. SiO₂ is transparent, so one can see the beam structure with silicon inclusions at nodes of the beams on the right side of the image. The horizontal line is the cross-section plane that corresponds to the FEM model section pictured in figure 6(a). The arrows map one-to-one correspondence between the FEM model and physical object.

4.1. FEM model verification

It is important to verify that the stiffness (force deflection) determined from the finite element model agrees well with that from the experimental data. The two most important parameters to determine, excluding the structure geometry, are the position of the applied load and the IB material Young's modulus. The Young's modulus and the strength of SiO₂ have been measured by various test methods [23]. A variation from 68 to 73 GPa for the E value induces a deviation of 2.5% for the force-deflection curves modeled with FEM. For our model we took E equal to 70 GPa. The exact position of the applied load is defined with the microscope after testing. The location for the applied load in the model is specified to agree with this measurement. The uncertainty in this location for the top surface of 0.5 µm leads to a variation in the force-deflection curves below 0.3%.

However, there is the second and unpredictable source of uncertainty—the residual stresses in the oxidized silicon. These residual stresses induce residual forces in the device and make the response from each device unique. In order to eliminate this huge uncertainty in the measured response, the loads were applied on both sides (front and back) of the structure under study. Two experimental force-deflection data points along with the modeled curves are shown in figure 7. There is good agreement between the calculated and measured data for loads applied to the front and back sides. Because the model does not include internal stress, this agreement means that the stress is smaller than the experimental model uncertainty. The numerical model is hence validated.

Zoom In Zoom Out Reset image size

After verification of the finite element model, it is used to estimate the fracture stress from the measured fracture force and species geometry.

Failure of brittle materials occurs when a critical sized flaw exists in the region that is under tensile or bending stress. The distribution of flaws is often random, so the strength of a brittle material can only be properly characterized by statistical measures. It is reasonable to fracture numerous samples to derive a fracture stress for each sample [24]. This fracture stress strength is not a material parameter, it is a design parameter. A statistical analysis is then applied to the fracture stress data. The structure under investigation consists of 50–100 oxidized beams, which is why we have restricted experiments with several runs. The fracture forces are obtained for several samples loaded from both the front and the back side of the wafer. The specimens were subjected to bending, uniaxial compressive stress and tensile stress. By using a.c. instead of d.c. voltages, the microactuators could be used to investigate the fatigue properties of the IB's structural materials.

5.1. Bending test

The test structure was mounted on the test apparatus and fixed with adhesive to the sample holder as shown in figure 2. The sample size was chosen as large as 1 × 1 cm to increase adhesive force of the sample to the sample holder. Specifically, the sample was aligned using the microscope and then a bending force was applied to the silicon bar by means of the incompressible stylus mounted at the end of the VCA lever. The bending force was increased with the increase of the displacement until fracture occurred. In all cases, the measured force-displacement response was linear up to failure, indicative of brittle fracture. Care was taken to apply the force near the center of the silicon bar in every test. Several samples were analyzed for both front and back loading.

SEM photographs of the fractured beams comprising IB are shown in figures 8(a) and (b). A clear brittle fracture signature is observed in figure 8. Most of the beams failed on the thin section of fully oxidized silicon close to the supporting silicon membrane (arrows on the left of figure 8(a)). When the actuator is in action the dominant failure mode is cleavage initiated at the top surface near the membrane. The initial crack propagates downward. The morphology shown in the bottom right corner of the figure 8(b) indicates the straight edges of the broken oxidized segments of the beams. The second situation is that the fracture happened at both ends of the thin section (arrows on the right side of figure 8(a)). The third is that the fracture happened at the end of the beam close to the silicon bar (some beams in figure 8(a) and even in the silicon bar area (foreground of figure 8(b)). The two latter cases can be interpreted as bouncing of the supporting silicon membrane after the initial fracture of the majority beams under bending load. This means that after the initial crack of some beams on the membrane side, the flexible silicon membrane with stored elastic energy plays the role of the actuator and moves upward. The time constant of the silicon membrane is much smaller than that of the actuator; thus, the actuator tip could be considered as stationary. This time the ultimate stress in the beams is gained at the silicon bar which is at rest, supported by the tip, and beams are fractured on this side. The bending load applied normal to the top surface of the specimen results in tensile and compressive stresses on the upper and lower parts on the broken structure, respectively. A curved line pattern is found at the lower left end of the fractured beams due to compressive stress (figure 8(b)). This is supported by the fact that fractures occurred at the silicon bar as shown in figure 8(b) because the ultimate tensile strength of silicon is comparable with the compressive strength of SiO₂. The energy for these fractures could be released only from a short punch of the silicon membrane.

Zoom In Zoom Out Reset image size

The main results of the micromechanical strength tests are ultimate fracture stress presented in figure 3. The graph represents stress distribution along the center line on the top surface of the beam schematically shown below the line. The distance is measured from the membrane (fixed end) to the silicon bar (free end).

Most studies where the mechanical properties of SiO₂ are reported concern film materials, since they serve mostly as an insulator but not as a deformable structural element. The study [25] lists the results of mechanical testing of SiO₂ microspecimens using the image-based strain measurement technique along with the results from other studies of SiO₂ using a variety of techniques. The reported values for fracture strength of oxide films vary from 350 MPa to 600–1900 MPa. The ultimate bending strength of this study is550 MPa. This value agrees closely with that for thermally grown SiO₂ films listed in the aforementioned table. The factor of discrepancy could be attributed to the distributions and variations in the cracks on surfaces. The chance of having a critical sized flaw capable of causing the material to fail at a given stress is proportional to the amount of material at that stress [24]. Our specimens have a significantly larger volume in comparison with those studied in [25, 26]. There was no significant difference (within experimental errors) between the average front and back fracture strength values.

From this work it follows that a designer can comfortably use a Young's modulus of 70 GPa and a fracture strength of 550 MPa in the bending load for thermally grown SiO₂ without residual stress.

5.2. Compression test

The testing of oxides is difficult due to very high compressive strengths and low failure strains. In particular, it is important to ensure that a uniform load is applied along the length of the sample uniaxially. The apparatus used for this experiment is the same as that used in a bending test. However, the probe tip is replaced by the ceramic enclosure of HDD heads for these tests. The ceramic housing (brick) has sharp, plane edges. Ceramics are generally strong in compression and can tolerate high compressive loads. The hexaferrite isotropic Young's modulus was found to be E = 183 GPa [27]. The sample is held on a fixed platform and the ceramic brick is aligned so that its edge is parallel to the silicon bar side. The load is applied on the rectangular area as shown in figure 5. Finally, because oxides fail after only about 0.3% strain, the samples under test must be perfectly aligned, otherwise bending stresses will introduce additional complications. The sample buckles out of plane or deforms transversely to the load direction if it is not exactly aligned. Rigorous compressive treatment of IB beams requires an even load distribution along the external side of the silicon bar. This brings strong aligning requirements, and the lever of the actuator must withstand the load without deformation.

Before the compression test the actuator was supplied with a current to develop a force of 1.0 N with the ceramic brick against the stainless steel block. The visual inspection of both the lever and brick by the microscope during load did not reveal any noticeable bending of the arm or brick on the scale of 0.5 µm. Furthermore, upon relieving stress there was no evidence of brick damage. The rigidity of sample clamping was also qualified before the test.

The specimen did not fail under a load of 1 N provided the sample and brick edges were properly aligned. In this way the IB's lower limit of compression strength was estimated. The stress distribution of the specimen along the center line of the beam and uniaxial to the force is shown in figure 9. The magnitude of the stress is maximal at the load application area as one can see in figure 9. The maximal stress value of 1.5 GPa does not exceed the fracture strength of silicon [28]. Insufficient actuator force allows us to define only the lower limit for the compressive strength of the IB. One can believe that for IB fracture compressive strength exceeds 600 MPa under uniaxial compression loads. This value is somewhat less than the 690–1380 MPa reported in [29] for SiO₂ film.

Zoom In Zoom Out Reset image size

5.3. Tensile stress

Tensile test measurement is the real challenge on the device level. One side of the specimen is anchored to the silicon wafer that can be handled but the free end with the silicon bar is the grip end. Mechanical chucking was used to fix a specimen to the sample holder. The rigidity of chucking was tested by force application to the specimen for lateral shear. Some attempts were made to attach the tension grip to the loading mechanism. Epoxy adhesives have shown low strength compared to the IB containing 50 beams. Reduction of the number of beams dramatically decreases the strength of the specimen and causes severe problems with handling. It was possible to remove some beams on each side of the IB to form claws so as to grip the silicon bar with tungsten hooks or Kevlar thread. However, problems with alignment and synchronizing loads on both sides still remain. One solution is to apply loads with an appropriate tip on one side of the silicon bar as shown in figure 10. Another problem is the tip choice. Tip thickness is bounded by 50 µm and a correct shape is required to allow visual alignment. The sample with the broken IB has a silicon membrane of suitable size and shape. This sample could be clamped to the loading mechanism easily and the elastic properties of silicon could be well defined. The FEA model becomes more complex because the problem of contact between two bodies has to be resolved.

Zoom In Zoom Out Reset image size

The beams on one side of the IB have been removed to form a ledge for grip. The silicon wafer fragment with a tip at the end is bonded to the actuator arm. The silicon tip and silicon bar are brought into contact by precise positioning and adjustment. Attention must be paid to the testing operation to guarantee the setting strength of the specimens and to improve the precision of measurements. Tensile loads below the fracture limit were applied to the specimen to check the validity of the FEA model with the contact. The elastic strain was estimated using the silicon beam as a gauge. Deflection curves versus actuation force for specimens with different quantities of beams are shown in figure 11. To improve the repeatability of measurements a series of experiments has been carried out.

Zoom In Zoom Out Reset image size

The slope of the curves in figure 11 depends on the number of undamaged beams after formation of the ledge. Cracks formed in the beams during the process of removal are difficult to see, thus a preliminary elastic test helps to define the initial condition exactly. After the FEA model of the contact problem of the two bodies is verified, it is utilized to estimate the fracture stress from the measured fracture force, the number of beams and their geometry. The ultimate stress distribution along the center line of the two outer beams closest to the load end is depicted in figure 12. The distance is measured from the silicon bar to the membrane (wafer) in the positive direction of the Y axis. The lowest stress is at the silicon inclusions. The maximal tensile stress along the beams was found in two locations. The first is on the external beam on the membrane side within a short fragment of oxidized silicon (figure 12(b)). This location corresponds to the distance of 0 to 7 micrometers with the maximum stress of 230 MPa in the graph of figure 12(a). The second location is on the second beam end near the silicon bar within the oxidized fragment. The tensile stress along the second beam is plotted with the dashed line in figure 12(a). The ultimate stress reaches a value of 210 MPa in the vicinity of 65 micrometers distance. Another area of tensile stress close to ultimate is at the distance from 40 to 53 micrometers in the first external beam. The array of oxidized beams in real experiments may begin with the 'second' beam. Visual inspection of the fractured IB after the test reveals that fracture occurs at the points shown with arrows in figure 12(b). The external beam fragment near the membrane in most cases was intact under external stress. The lower volume of the latter fragment suggests that the fracture strength of this specimen has to be stronger than other fragments under stress above 200 MPa; this is in agreement with the Weibull theory.

Zoom In Zoom Out Reset image size

As can be seen, the strength of SiO₂ under tensile load is 200 MPa. Tsuchiya's tensile results [30] show a very high value (1.9 GPa) for the plasma-enhanced chemical vapor deposition SiO₂ film. According to Sharpe et al [25], a SiO₂ film 950 nm thick was deposited using plasma-enhanced chemical vapor deposition onto silicon (1 0 0) substrates and a tensile fracture strength of 365 MPa was measured. This is one and a half times stronger than in our measurements. The explanation of the measured value in our tests is that larger specimens of brittle materials can have larger and more flaws, which lower the strength. Size effects for geometries other than straight specimens (e.g. at stress concentrations) can influence determination of the fracture strength and should be considered in the design of microelectronics and MEMS. Allowing the rule of thumb that compressive fracture strength is 10–15 times greater than tensile fracture strength [31], one can admit a compressive strength value of SiO₂ in the range 2–3 GPa, which is consistent with [29]. The tensile tests provided a mean stress to failure that was one third of that found in the bending tests.

5.4. Fatigue test

Fatigue is observed as a change in elastic constants, plastic deformation and strength decrease through the application of a cyclic or constant stress for a long time. Plastic deformation and changes in elastic constants cause sensitivity changes and offset drift in devices while the fatigue fracture causes sudden failure of the device functions. These should be avoided in order to produce highly reliable devices. The fatigue test measures the number of loading cycles that need to be applied by a specific load until failure. A fatigue test that imparts cyclic stresses in the form of tension or bending is suitable for MEMS specimens. Fatigue tests were conducted under the bending mode to evaluate the effect of cyclic loading on fractures in the beams comprising IB. The microactuator of a CD pickup voice coil has been used to investigate the fatigue properties of the structural materials of IB. A tungsten tip was attached to a lens holder and adjusted to touch a silicon bar. The cyclic bending load was applied to the specimen at a constant frequency of 5 Hz with a sinusoidal waveform. The bending stress amplitude was 60 MPa determined from gauge deflection. This value corresponds to the acceleration of 2 × 10⁶ g applied normally to the device surface. The specimens were subjected to half a million stress cycles and then the load was released. After this, the specimen was subjected to a bend load and the force-deflection curve was determined. The bending test results of the cycled specimen was then compared with the force-deflection behavior of the same specimens before cyclic stress. Figure 13 shows the results for modeled and experimental force-deflection curves before and after the test. The load-deflection experimental data obey elastic deformation behavior of the FEA model. None of the three specimens tested under these conditions was broken when cyclic stress was applied. Even though the test did not show a weakening effect, to ensure fatigue strength, a test with a greater number of stress cycles has to be performed.

Zoom In Zoom Out Reset image size

This article describes a methodology to combine experimental techniques and numerical modeling methods to determine the fracture strength of the IB by bending, tension and compression. An accurate experimental system is specially designed to bend the beams and acquire the resulting load-deflection characteristics. The force-deflection characteristics of the FEA model are validated with data from the experimental system. The stress at fracture is then derived from the maximum stress in the model. With this system, the fracture force is obtained for samples loaded from both the front and the back sides of the structure. The measured fracture strength under bending is 550 MPa, while under tensile load it is just above 200 MPa. The lower bound for the IB compressive strength is estimated as 600 MPa. The measured fracture strength values are reasonably coherent with other reports. The preliminary fatigue characterization of the IB provides information crucial to the design of reliable and stable MEMS devices. This study demonstrated the viability of merged isolation and its applicability for the design of MEMS devices.

This work was performed at the Facilities Sharing Centre 'Micro- and Nanostructures Diagnostics' and is supported by the Ministry of Science and Education of the Russian Federation.