MEMS Device for Quantitative In Situ Mechanical Testing in Electron Microscope

In this work, we designed a micro-electromechanical systems (MEMS) device that allows simultaneous direct measurement of mechanical properties during deformation under external stress and characterization of the evolution of nanomaterial microstructure within a transmission electron microscope. This MEMS device makes it easy to establish the correlation between microstructure and mechanical properties of nanomaterials. The device uses piezoresistive sensors to measure the force and displacement of nanomaterials qualitatively, e.g., in wire and thin plate forms. The device has a theoretical displacement resolution of 0.19 nm and a force resolution of 2.1 μN. The device has a theoretical displacement range limit of 5.47 μm and a load range limit of 55.0 mN.


Introduction
Mechanical properties of materials are influenced strongly by their microstructures. Materials often show special mechanical properties compared to their bulk counterparts when the grain size or physical dimension is reduced to the nanometer scale [1][2][3][4][5][6][7][8]. For example, brittle Si and SiC show super-plasticity when their sizes are reduced to below~100 nm [9][10][11][12][13]; the elasticity of nanosized copper can approach the theoretical elastic strain limit [14]; and the martensitic transformation can be completely suppressed in NiTi shape memory alloys when the sample thickness is below 50 nm [15,16]. In this regard, it is important to develop capabilities to allow quantitative studies of the structure-property correlations of materials at the nano and atomic scales.
Many techniques and devices, adapted to scanning electron microscopes (SEM) or transmission electron microscopes (TEM), have been designed to allow in situ studies of structure-property Figure 1 shows a schematic of the testing system design. Its operation control is given in Figure  1a. The system comprises an actuation system, a MEMS device, two single-system-power suppliers, and two digital multimeters (6½ digits). The two single-system-power suppliers provide a DC operating voltage for the two piezoresistive sensors. The beams on the device are driven by a piezoceramic actuator with a travel distance of 100 μm and a minimum step of 7 nm. Two digital multimeters were used to collect the sensor output voltages. The MEMS device was fabricated using bulk silicon microfabrication process, and the design is illustrated in Figure 1b. Two sensors, A and B, were placed on two beams, A and B, to measure the displacement and force. Sensor A was placed at the root of beam A to measure the deflection of its center. To maintain the stability of sensor A, two more clamped beams parallel to beam A were connected with it by a shuttle, as can be seen in Figure 1b. Sensor B was placed at the root of beam B, which is longer than beam A, to measure the deflection of beam B. The deformation strain of the specimen is calculated using Equation (1): where xa and xb are the center deflections of beams A and B, respectively, and l0 is the initial length of the specimen. The load applied on the specimen approximately equals the driving force acting on the beam center, which was calculated based on the deflection of beam B. The stress is given as follows: The MEMS device was fabricated using bulk silicon microfabrication process, and the design is illustrated in Figure 1b. Two sensors, A and B, were placed on two beams, A and B, to measure the displacement and force. Sensor A was placed at the root of beam A to measure the deflection of its center. To maintain the stability of sensor A, two more clamped beams parallel to beam A were connected with it by a shuttle, as can be seen in Figure 1b. Sensor B was placed at the root of beam B, which is longer than beam A, to measure the deflection of beam B. The deformation strain of the specimen is calculated using Equation (1): where x a and x b are the center deflections of beams A and B, respectively, and l 0 is the initial length of the specimen. The load applied on the specimen approximately equals the driving force acting on the beam center, which was calculated based on the deflection of beam B. The stress is given as follows: where F b is the driving force acting on beam B and S is the cross-sectional area of the specimen. Using a push-to-pull structure, the external compressive stress on the specimen was transferred to a tensile stress, as shown in Figure 1b. Figure 2 shows the MEMS device with two piezoresistive sensors. The overall dimensions of the device are 1.3 mm × 2.4 mm × 0.44 mm, thus they can be easily installed on the sample stage at the head of a TEM hold. Four piezoresistors indicated by green were prepared on beams A and B, as shown in the insets of Figure 2. The four sensors form a half Wheatstone bridge, to allow a precise measurement of the resistive variance. The two piezoresistors, R 1 and R 2 , located at the roots of the beams A and B, measure the beam deflection. The resistances of the other two resistors, R 3 and R 4 , attached to the substrate, remain constant during the beam deflection. The four resistors are electrically connected with aluminum (Al) interconnects. To improve the consistency of the process and to partially compensate the influence of temperature, the resistors on the beam and those on the substrate were placed as close as possible to each other [52]. The beams and the piezoresistors are aligned along the <110> direction on the (100) plane of a silicon wafer substrate for obtaining a better sensor sensitivity [53].

Sensor Design
where Fb is the driving force acting on beam B and S is the cross-sectional area of the specimen. Using a push-to-pull structure, the external compressive stress on the specimen was transferred to a tensile stress, as shown in Figure 1b. Figure 2 shows the MEMS device with two piezoresistive sensors. The overall dimensions of the device are 1.3 mm × 2.4 mm × 0.44 mm, thus they can be easily installed on the sample stage at the head of a TEM hold. Four piezoresistors indicated by green were prepared on beams A and B, as shown in the insets of Figure 2. The four sensors form a half Wheatstone bridge, to allow a precise measurement of the resistive variance. The two piezoresistors, R1 and R2, located at the roots of the beams A and B, measure the beam deflection. The resistances of the other two resistors, R3 and R4, attached to the substrate, remain constant during the beam deflection. The four resistors are electrically connected with aluminum (Al) interconnects. To improve the consistency of the process and to partially compensate the influence of temperature, the resistors on the beam and those on the substrate were placed as close as possible to each other [52]. The beams and the piezoresistors are aligned along the <110> direction on the (100) plane of a silicon wafer substrate for obtaining a better sensor sensitivity [53]. When the clamped beams deform upon an in-plane force, stress will concentrate at the beam roots. The stress is free at the neutral plane of the beam and increases gradually with increasing distance from the neutral plane. The variance of the resistance is given in Equation (3) [54]:

Sensor Design
where σl and σt are the longitudinal and transverse stresses, and πl and πt are the piezoresistance coefficients along the longitudinal and transverse directions. Since σt is much smaller than σl, the term t t π σ can be neglected [51]. Figure 3 shows the positions of the piezoresistors on the beam and the parameters used in Equations (4) and (5). In the case of small deformation, the variance of the resistance can be written as: When the clamped beams deform upon an in-plane force, stress will concentrate at the beam roots. The stress is free at the neutral plane of the beam and increases gradually with increasing distance from the neutral plane. The variance of the resistance is given in Equation (3) [54]: where σ l and σ t are the longitudinal and transverse stresses, and π l and π t are the piezoresistance coefficients along the longitudinal and transverse directions. Since σ t is much smaller than σ l , the term π t σ t can be neglected [51]. Figure 3 shows the positions of the piezoresistors on the beam and the parameters used in Equations (4) and (5). In the case of small deformation, the variance of the resistance can be written as: where E = 169 GPa [55] is the Young's modulus of silicon in <110> direction, L is the beam length, l p is the piezoresistor length, ω is the beam center deflection, and d is the distance between the resistor centerline and neutral plane of the beam. The value d can be expressed as: where D is the beam thickness (along the force direction), w p is the width of the piezoresistor, and d" is the distance between the outer edge of the piezoresistor and the beam. Figure 3 shows all the parameters in Equation (5).
where D is the beam thickness (along the force direction), wp is the width of the piezoresistor, and d″ is the distance between the outer edge of the piezoresistor and the beam. Figure 3 shows all the parameters in Equation (5).
When a tensile stress is applied to the specimen, assuming that the piezoresistors R3 and R4 have an equal resistance variance ΔR, the bridge output voltage is given as: where VB is the bridge bias voltage. The displacement sensitivity ( D S ) of the piezoresistive sensors is defined as the ratio of the output voltage to the beam deflection, and is given by combining Equations (4) and (6): The piezoresistive sensor noise arises from both intrinsic and extrinsic sources. Johnson noise and 1/f noise are the two main intrinsic noises, and their noise power spectral densities are given in Equations (8) and (9), respectively [56]: where kB is the Boltzmann constant, T is the absolute temperature, Np is the dopant concentration, q is the amount of the carrier charge, μp is the hole mobility, dp is the piezoresistor thickness, α is a non-dimensional fitting parameter depending on the annealing conditions, and fmax and fmin are the upper and lower measurement frequency limits. The displacement resolution of the sensors is defined as the ratio of the noise to the displacement sensitivity (only considering Johnson and 1/f noise), and can be written as: When a tensile stress is applied to the specimen, assuming that the piezoresistors R 3 and R 4 have an equal resistance variance ∆R, the bridge output voltage is given as: where V B is the bridge bias voltage. The displacement sensitivity (S D ) of the piezoresistive sensors is defined as the ratio of the output voltage to the beam deflection, and is given by combining Equations (4) and (6): The piezoresistive sensor noise arises from both intrinsic and extrinsic sources. Johnson noise and 1/f noise are the two main intrinsic noises, and their noise power spectral densities are given in Equations (8) and (9), respectively [56]: where k B is the Boltzmann constant, T is the absolute temperature, N p is the dopant concentration, q is the amount of the carrier charge, µ p is the hole mobility, d p is the piezoresistor thickness, α is a non-dimensional fitting parameter depending on the annealing conditions, and f max and f min are the upper and lower measurement frequency limits. The displacement resolution of the sensors is defined as the ratio of the noise to the displacement sensitivity (only considering Johnson and 1/f noise), and can be written as: Optimized design can effectively decrease the noise power and improve the displacement resolution, i.e., the smallest signal that can be detected. According to Equation (10), the density of the noise power decreases with increasing piezoresistor width (w p ). For a nearly constant d", depending on the MEMS technology, the d will become shorter when w p is increased, leading to a decrease in the displacement sensitivity. To obtain an optimal width (w p-optimal ), the resolution R D can be partially differentiated against the piezoresistor width. The optimal width thus obtained is: According to Equations (8) and (9), increase in the length, l p , of the piezoresistor has an opposite influence on the power density of the Hooge and Johnson noises, i.e., a decrease in the Hooge noise but an increase in the Johnson noise. It is further noted from Equation (7) that increase in l p can also lead to a loss in the displacement sensitivity. Because of the complicated effect of l p on the displacement resolution and sensitivity, the resolution was partially differentiated against l p to obtain an optimal length contributing to a high resolution. For each clamped beam length, L, an optimal l p-optimal is obtained. Figure 4 shows the evolution of l p-optimal /L as a function of the clamped beam length. The results show that the l p-optimal /L decreases with increasing L.
Optimized design can effectively decrease the noise power and improve the displacement resolution, i.e., the smallest signal that can be detected. According to Equation (10), the density of the noise power decreases with increasing piezoresistor width (wp). For a nearly constant d″, depending on the MEMS technology, the d′ will become shorter when wp is increased, leading to a decrease in the displacement sensitivity. To obtain an optimal width (wp-optimal), the resolution RD can be partially differentiated against the piezoresistor width. The optimal width thus obtained is: According to Equations (8) and (9), increase in the length, lp, of the piezoresistor has an opposite influence on the power density of the Hooge and Johnson noises, i.e., a decrease in the Hooge noise but an increase in the Johnson noise. It is further noted from Equation (7) that increase in lp can also lead to a loss in the displacement sensitivity. Because of the complicated effect of lp on the displacement resolution and sensitivity, the resolution was partially differentiated against lp to obtain an optimal length contributing to a high resolution. For each clamped beam length, L, an optimal lp-optimal is obtained. Figure 4 shows the evolution of lp-optimal/L as a function of the clamped beam length. The results show that the lp-optimal/L decreases with increasing L. Setting one of the supporting points to be an original point, the deflection/displacement of a position at a distance x from the original point can be written as Equation (12) according to mechanical analysis [57]: where W is the beam width, and F is the force applied on the beam center. The longitudinal stress of each point can be expressed as [57]: Substituting Equations (12) into (13), we obtain Equation (14): Setting one of the supporting points to be an original point, the deflection/displacement of a position at a distance x from the original point can be written as Equation (12) according to mechanical analysis [57]: where W is the beam width, and F is the force applied on the beam center. The longitudinal stress of each point can be expressed as [57]: Substituting Equations (12) into (13), we obtain Equation (14): According to Equation (14), the maximum longitudinal stress is located on the beam surface (d = D/2), roots (x = 0), and center (x = L), which can be written as: The sensors will fail when σ l−max equals to the bending strength (3.70 GPa, [3]) of (100) single crystal silicon. According to Equation (15), the theoretical maximum forces that can be applied on the beam center and measured by the two sensors are calculated to be 888 mN for sensor A and 55.0 mN for sensor B, as listed in Table 1. According to Equation (12), the beam center deflection is given by: Then, the maximum displacements of the two sensors are calculated to be 5.47 µm for sensor A and 175.14 µm for sensor B, as listed in Table 1. Therefore, the theoretical range limits of load and displacement that can be measured by the MEMS device are 55.0 mN and 5.47 µm, respectively.

Fabrication Process
Using bulk silicon micromachining process, the MEMS device was fabricated on a 4 inch (100) SOI (Silicon-On-Insulator) wafer, consisting of a 380-µm-thick handle layer, a 60-µm-thick n-type device layer with resistivity of 1-2 Ω·m, and a 0.5-µm-thick buried oxide layer. The minimum feature size of the device structure is 6 µm. The device has an aspect ratio of ten. Figure 5 shows the main steps of the fabrication process. First, a 300-µm-thick SiO 2 insulation layer was grown on each side of the wafer by thermal oxidation (Figure 5a). The thermal oxidation SiO 2 layer on the device side was patterned and etched as a mask for subsequent ion implantation. The piezoresistors, with resistivity of 1.17 × 10 −2 Ω·cm, were prepared by boron doping by means of ion implantation at 100 keV with a dose of 10 15 cm −2 (p-doped, Figure 5b). Using the same process, electrical contacts were created with an implantation energy of 100 keV and a dose of 3 × 10 15 cm −2 (p+ doped, Figure 5c). A 1-µm-thick aluminum film was then sputtered on the surface and then etched using potassium hydroxide (KOH) solution to form interconnects and pads (Figure 5d). The device layer was then etched by inductively coupled plasma (ICP) etching (Figure 5e). Finally, ICP was used to etch out the handle and buried oxide layers from the backside to create a movable structure (Figure 5f). Figure 6a shows a SEM image of a MEMS fabricated device. Figure 6b shows a magnified view of sensor B. The corresponding lithography maps are shown in Figure 6c,d. Comparison between the fabricated device and the lithography maps shows that the lateral undercutting of both ICP etching and aluminum film wet etching was less than 1 µm, indicating a well-controlled etching processes.
100 keV with a dose of 15 10 2 cm -(p-doped, Figure 5b). Using the same process, electrical contacts were created with an implantation energy of 100 keV and a dose of 15 3 10 × 2 cm -(p+ doped, Figure   5c). A 1-μm-thick aluminum film was then sputtered on the surface and then etched using potassium hydroxide (KOH) solution to form interconnects and pads (Figure 5d). The device layer was then etched by inductively coupled plasma (ICP) etching (Figure 5e). Finally, ICP was used to etch out the handle and buried oxide layers from the backside to create a movable structure ( Figure  5f). Figure 6a shows a SEM image of a MEMS fabricated device. Figure 6b shows a magnified view of sensor B. The corresponding lithography maps are shown in Figure 6c,d. Comparison between the fabricated device and the lithography maps shows that the lateral undercutting of both ICP etching and aluminum film wet etching was less than 1 μm, indicating a well-controlled etching processes.

Device Calibration and Quantitative Tensile Testing
A focused ion/electron dual-beam system (FIB/SEM, FEI Helios Nanolab 600 i, FEI, Hillsboro, USA) was used to calibrate the displacement of the MEMS device in a high vacuum environment ( 100 keV with a dose of 10 cm (p-doped, Figure 5b). Using the same process, electrical contacts were created with an implantation energy of 100 keV and a dose of 15 3 10 × 2 cm -(p+ doped, Figure   5c). A 1-μm-thick aluminum film was then sputtered on the surface and then etched using potassium hydroxide (KOH) solution to form interconnects and pads (Figure 5d). The device layer was then etched by inductively coupled plasma (ICP) etching (Figure 5e). Finally, ICP was used to etch out the handle and buried oxide layers from the backside to create a movable structure ( Figure  5f). Figure 6a shows a SEM image of a MEMS fabricated device. Figure 6b shows a magnified view of sensor B. The corresponding lithography maps are shown in Figure 6c,d. Comparison between the fabricated device and the lithography maps shows that the lateral undercutting of both ICP etching and aluminum film wet etching was less than 1 μm, indicating a well-controlled etching processes.

Device Calibration and Quantitative Tensile Testing
A focused ion/electron dual-beam system (FIB/SEM, FEI Helios Nanolab 600 i, FEI, Hillsboro, USA) was used to calibrate the displacement of the MEMS device in a high vacuum environment (

Device Calibration and Quantitative Tensile Testing
A focused ion/electron dual-beam system (FIB/SEM, FEI Helios Nanolab 600 i, FEI, Hillsboro, USA) was used to calibrate the displacement of the MEMS device in a high vacuum environment (10 −5 Pa) with a purpose of avoiding the disturbances arising from variations in temperature, humidity, electromagnetic radiation, mechanical vibration, etc. [58,59], as shown in Figure 7a. Figure 7b shows the design of the testing system setup for calibration prior to tensile testing. The MEMS device was first glued on to a custom-made printed circuit board (PCB) and then connected with the PCB by ultrasonic bonding. The PCB was then fixed onto a three-dimensional micropositioner. The shuttle beam of the device was precisely aligned with the probe fixed on the piezo nanopositioner. The piezo nanopositioner moved in a step of 7 nm to drive the shuttle through the probe,~10 µm in diameter.
For each batch of MEMS devices fabricated on the same wafer, a few MEMS devices were selected for calibration. The displacements of the beams A and B on the selected calibration device were measured from the SEM images taken. The corresponding output voltages of the displacement sensors were simultaneously measured by a digital multimeter and recorded in a computer. A sensor voltage-image displacement curve is then obtained for each of the two sensors. During testing, the displacement on the sample is read out from the sensor voltage output of the two beams (x a and x b ) based on the calibration. The deformation strain is then calculated according to Equation (1) by knowing the original length (l 0 ) of the specimen prior to testing. The force applied on the specimen (F b ) is calculated based on the displacement of beam B, as per Equation (16). The stress can then be calculated using Equation (2) by measuring the cross-sectional area (S) of the specimen prior to tensile testing.
7b shows the design of the testing system setup for calibration prior to tensile testing. The MEMS device was first glued on to a custom-made printed circuit board (PCB) and then connected with the PCB by ultrasonic bonding. The PCB was then fixed onto a three-dimensional micropositioner. The shuttle beam of the device was precisely aligned with the probe fixed on the piezo nanopositioner. The piezo nanopositioner moved in a step of 7 nm to drive the shuttle through the probe, ~10 μm in diameter. For each batch of MEMS devices fabricated on the same wafer, a few MEMS devices were selected for calibration. The displacements of the beams A and B on the selected calibration device were measured from the SEM images taken. The corresponding output voltages of the displacement sensors were simultaneously measured by a digital multimeter and recorded in a computer. A sensor voltage-image displacement curve is then obtained for each of the two sensors. During testing, the displacement on the sample is read out from the sensor voltage output of the two beams (xa and xb) based on the calibration. The deformation strain is then calculated according to Equation (1) by knowing the original length (l0) of the specimen prior to testing. The force applied on the specimen (Fb) is calculated based on the displacement of beam B, as per Equation (16). The stress can then be calculated using Equation (2) by measuring the cross-sectional area (S) of the specimen prior to tensile testing. Using this testing system, uniaxial tensile tests were conducted on aluminum thin films in the FIB/SEM system. During the testing process, the piezocontroller, power suppliers, and multimeters, which were placed outside the FIB/SEM system, were connected with the device and the nanopositioner inside the system with shielded cables through the flange of the SEM.

Specimen Preparation
The accuracy of the testing system was tested by measuring the stress-strain curve of aluminum film samples with different thicknesses. The aluminum films for tensile testing were prepared by DC magnetron sputtering. The purity of the aluminum target was 99.99% and the sputtering parameters included a working power of 400 W, vacuum level of 6 1.316 10 × -Pa and argon pressure of 2-3 Pa. An aluminum film with a thickness of ~1 μm was deposited on a thermally oxidized silicon wafer. Specimens for tensile testing in SEM were cut by FIB in the following sequence. First, a piece of the film was cut out with an ion beam using a current of 9.4 nA, lifted out and moved to the specimen stages on the MEMS device with a W probe, and fixed by Pt deposition Using this testing system, uniaxial tensile tests were conducted on aluminum thin films in the FIB/SEM system. During the testing process, the piezocontroller, power suppliers, and multimeters, which were placed outside the FIB/SEM system, were connected with the device and the nanopositioner inside the system with shielded cables through the flange of the SEM.

Specimen Preparation
The accuracy of the testing system was tested by measuring the stress-strain curve of aluminum film samples with different thicknesses. The aluminum films for tensile testing were prepared by DC magnetron sputtering. The purity of the aluminum target was 99.99% and the sputtering parameters included a working power of 400 W, vacuum level of 1.316 × 10 −6 Pa and argon pressure of 2-3 Pa. An aluminum film with a thickness of~1 µm was deposited on a thermally oxidized silicon wafer. Specimens for tensile testing in SEM were cut by FIB in the following sequence. First, a piece of the film was cut out with an ion beam using a current of 9.4 nA, lifted out and moved to the specimen stages on the MEMS device with a W probe, and fixed by Pt deposition on both ends. The film was then gradually thinned to hundreds of nanometers with the ion beam at a series of currents equal to 9.4 nA, 2.3 nA, 0.77 nA, and 7 pA to minimize the damage caused by the incident ion beam. A specimen with thicknesses of 510 nm is shown in Figure 8. The dimensions of the specimen are listed in Table 2. on both ends. The film was then gradually thinned to hundreds of nanometers with the ion beam at a series of currents equal to 9.4 nA, 2.3 nA, 0.77 nA, and 7 pA to minimize the damage caused by the incident ion beam. A specimen with thicknesses of 510 nm is shown in Figure 8. The dimensions of the specimen are listed in Table 2.

Length (nm) Width (nm) Depth (nm)
4880 620 510 Figure 9 shows the current-voltage characteristics of the two piezoresistive sensors. It is seen that both sensors exhibited perfect linear behavior within the range of −5-5 V. This also indicates a perfect contact between the sensors and the Al interconnects. The resistances of sensors A and B are determined to be 0.34 and 2.1 kΩ, respectively. Figure 10 shows the effect of the bias voltage on the zero-point output voltage of the two sensors. It is seen that the zero-point output also has a linear dependence on the bias voltage for both sensors. The zero-point deviation can be attributed to the uneven ion implantation among the four resistors, despite identical processing conditions used. Therefore, zero setting of the bridge output voltage was conducted on each sensor before calibration.    Figure 9 shows the current-voltage characteristics of the two piezoresistive sensors. It is seen that both sensors exhibited perfect linear behavior within the range of −5-5 V. This also indicates a perfect contact between the sensors and the Al interconnects. The resistances of sensors A and B are determined to be 0.34 and 2.1 kΩ, respectively. Figure 10 shows the effect of the bias voltage on the zero-point output voltage of the two sensors. It is seen that the zero-point output also has a linear dependence on the bias voltage for both sensors. The zero-point deviation can be attributed to the uneven ion implantation among the four resistors, despite identical processing conditions used. Therefore, zero setting of the bridge output voltage was conducted on each sensor before calibration. on both ends. The film was then gradually thinned to hundreds of nanometers with the ion beam at a series of currents equal to 9.4 nA, 2.3 nA, 0.77 nA, and 7 pA to minimize the damage caused by the incident ion beam. A specimen with thicknesses of 510 nm is shown in Figure 8. The dimensions of the specimen are listed in Table 2. Figure 8. SEM image of a specimen fabricated for tensile testing.

Length (nm) Width (nm) Depth (nm)
4880 620 510 Figure 9 shows the current-voltage characteristics of the two piezoresistive sensors. It is seen that both sensors exhibited perfect linear behavior within the range of −5-5 V. This also indicates a perfect contact between the sensors and the Al interconnects. The resistances of sensors A and B are determined to be 0.34 and 2.1 kΩ, respectively. Figure 10 shows the effect of the bias voltage on the zero-point output voltage of the two sensors. It is seen that the zero-point output also has a linear dependence on the bias voltage for both sensors. The zero-point deviation can be attributed to the uneven ion implantation among the four resistors, despite identical processing conditions used. Therefore, zero setting of the bridge output voltage was conducted on each sensor before calibration.    Figure 11 shows SEM images used to calibrate the displacements measured by the sensors A and B. For each sensor, the beam center deflection, on which the sensor is located, was obtained by measuring the distance variation between two reference points in SEM, i.e., points C and D for sensor A (Figure 11a-f), and E and F for sensor B (Figure 11g-l). The movements of beams A and B were driven by the piezo nanopositioner through the shuttle beam and probe with a movement step size of 7 nm. Beams A and B need to be calibrated separately. For this, beam A was calibrated prior to beam B until rupture. When beam B was calibrated, beam A was already in the broken state as shown in Figure 11g. Figure 12 shows the output voltages of the two sensors versus the measured beam center deflections under a bias voltage of 3 V. It is seen that the sensor outputs are practically linear against the beam deflection. Using linear fitting, the displacement sensitivity is determined to be 37.4 μV/nm for sensor A and 4.8 μV/nm for sensor B. Comparing with the theoretical values listed in Table 1, the deviations are 51.5% and 17.9% for the two sensors, respectively. This is commonly attributed to the non-uniformity of ion implantation and excess resistance in the bridge [60]. The linearities of the sensors A and B are calculated to be 1.92% and 1.94%, respectively. Both values are smaller than the standard of a good linearity, i.e., 2%, demonstrating good sensor performance.

Sensors Performance
The maximum displacement of the beam center achieved in this calibration is <5 μm. This is far smaller than the beam length, thus the deflection can be treated as a linear system and described by the small deflection theory [61]. Therefore, we take the stress defined by Equation (16) as the load applied on the clamped beam. Based on the displacement sensitivities of the sensors A and B and Equation (10), the corresponding displacement resolutions are determined to be ~0.19 nm for sensor A and ~6.8 nm for sensor B. The sample elongation resolution is then determined to be 6.8 nm. The force resolution of sample/sensor B is then calculated to be 2.1 μN using Equation (16). Since the displacements of the sensors are measured by imaging method, the resolution that can be experimentally determined depends on that of the SEM or TEM used. Further work on TEM is required to determine more accurately, the displacement resolution.
It is well known that the electron beam inside an electron microscope can lead to electron accumulation in the device, which may interfere with the sensor behavior. To assess the influence of the electron beam on the sensor displacement sensitivity, sensors A and B were irradiated under the electron beam at an accelerating voltage of 2 kV (operating voltage) inside a SEM under magnifications of 2000×, 8000× and 16,000×. Table 3 shows the sensitivities of the sensors under irradiation of e-beam in SEM. The results show that no apparent changes of sensitivity can be detected, indicating a neglected influence of electron beam on sensor sensitivity.  Figure 11 shows SEM images used to calibrate the displacements measured by the sensors A and B. For each sensor, the beam center deflection, on which the sensor is located, was obtained by measuring the distance variation between two reference points in SEM, i.e., points C and D for sensor A (Figure 11a-f), and E and F for sensor B (Figure 11g-l). The movements of beams A and B were driven by the piezo nanopositioner through the shuttle beam and probe with a movement step size of 7 nm. Beams A and B need to be calibrated separately. For this, beam A was calibrated prior to beam B until rupture. When beam B was calibrated, beam A was already in the broken state as shown in Figure 11g. Figure 12 shows the output voltages of the two sensors versus the measured beam center deflections under a bias voltage of 3 V. It is seen that the sensor outputs are practically linear against the beam deflection. Using linear fitting, the displacement sensitivity is determined to be 37.4 µV/nm for sensor A and 4.8 µV/nm for sensor B. Comparing with the theoretical values listed in Table 1, the deviations are 51.5% and 17.9% for the two sensors, respectively. This is commonly attributed to the non-uniformity of ion implantation and excess resistance in the bridge [60]. The linearities of the sensors A and B are calculated to be 1.92% and 1.94%, respectively. Both values are smaller than the standard of a good linearity, i.e., 2%, demonstrating good sensor performance.
The maximum displacement of the beam center achieved in this calibration is <5 µm. This is far smaller than the beam length, thus the deflection can be treated as a linear system and described by the small deflection theory [61]. Therefore, we take the stress defined by Equation (16) as the load applied on the clamped beam. Based on the displacement sensitivities of the sensors A and B and Equation (10), the corresponding displacement resolutions are determined to be~0.19 nm for sensor A and~6.8 nm for sensor B. The sample elongation resolution is then determined to be 6.8 nm. The force resolution of sample/sensor B is then calculated to be 2.1 µN using Equation (16). Since the displacements of the sensors are measured by imaging method, the resolution that can be experimentally determined depends on that of the SEM or TEM used. Further work on TEM is required to determine more accurately, the displacement resolution.
It is well known that the electron beam inside an electron microscope can lead to electron accumulation in the device, which may interfere with the sensor behavior. To assess the influence of the electron beam on the sensor displacement sensitivity, sensors A and B were irradiated under the electron beam at an accelerating voltage of 2 kV (operating voltage) inside a SEM under magnifications of 2000×, 8000× and 16,000×. Table 3 shows the sensitivities of the sensors under irradiation of e-beam in SEM. The results show that no apparent changes of sensitivity can be detected, indicating a neglected influence of electron beam on sensor sensitivity.

Stress-Strain Curve
An Al thin film of 510 nm in thickness was subjected to tensile deformation on the device in the FIB/SEM system with an actuation step size of 7 nm. SEM images were taken during tensile deformation. Figure 13 shows the stress-strain curve measurement of the Al film sample. The Young's modulus of the specimen is measured to be 71.5 GPa by linear fitting, consistent with bulk materials (68.5-71 GPa). The Young's modulus is also consistent with the 100 nm Al film measured by the MEMS sensors [62]. The flow stress is determined to be~460 MPa, which is low compared to~700 MPa [62]. This may be related to the thickness and density of the film used.

Stress-Strain Curve
An Al thin film of 510 nm in thickness was subjected to tensile deformation on the device in the FIB/SEM system with an actuation step size of 7 nm. SEM images were taken during tensile deformation. Figure 13 shows the stress-strain curve measurement of the Al film sample. The Young's modulus of the specimen is measured to be 71.5 GPa by linear fitting, consistent with bulk materials (68.5-71 GPa). The Young's modulus is also consistent with the 100 nm Al film measured by the MEMS sensors [62]. The flow stress is determined to be ~460 MPa, which is low compared to ~700 MPa [62]. This may be related to the thickness and density of the film used.

Conclusions
A small sized MEMS device was designed and fabricated to study quantitatively, the mechanical properties of thin film materials in situ during TEM/SEM observation. The device was tested in SEM. Piezoresistive sensors were integrated onto the device to measure quantitatively the mechanical properties of materials. This device allows the study of microstructure-property correlations in small materials and understanding of deformation mechanisms at nanometric and atomic scales. The main outcome of this work may be summarized as follows: (1) Sensors A and B have displacement sensitivities of 37.4 μV/nm and 4.8 μV/nm.