Reliability of piezoelectric microelectomecanical systems in humid conditions

Since its discovery, silicon-based microelectromechanical systems (MEMS) have had a significant impact on the technological development of sensors and actuators. Using piezoelectric materials deposited as thin-films on a substrate has enabled the mass production of miniaturized devices with high sensitivity and low power consumption. Piezoelectric MEMS (piezoMEMS), therefore, holds the promise of revolutionizing the sensor and actuator technology and is anticipated to dominate the future MEMS market. Ferroelectric thin-films have high dielectric and piezoelectric compliances, which makes them attractive for piezoMEMS. Lead zirconate titanate (PZT) is an essential ferroelectric material for numerous thin-film piezoMEMS applications. Including inkjet-printers, ultrasound transducers, energy harvesters, autofocus lenses, gas-detectors, and ferroelectric random-access memories. A key technological and scientific challenge for sustaining the development of thin-film piezoMEMS is improving their reliability and lifetime in realistic and harsh operating conditions. Elevated temperatures, high humidity levels, mechanical shocks, and stresses are all lifetime-limiting stressors that devices must endure in real-life, not encountered in a lab-scale environment. For the widespread adaption of thin-film piezoMEMS in sensor and actuator technology, it is necessary to mitigate the impact of such stressors. Understanding which physical and chemical degradation mechanisms imposed by the operating ambient affects device performance is, therefore, essential on both a microscopic and macroscopic scale. The ambient humidity has proved to be one stress-factor that severely degrades the lifetime and reliability of piezoMEMS. However, humidity-related effects are complex and diverse, and different degradation-mechanisms can dominate, depending on the type of piezoMEMS-device. The goal of this thesis is to establish the critical stress-factors governing humidity-related degradation of thin-film, PZT-based piezoMEMS structures, and devices. Both test-structures and factual piezoMEMS devices were studied to bridge the gap between applied and fundamental research on reliability. Four manuscripts the main findings of this work. A thin-film piezoelectric micro-mirror developed at SINTEF MiNaLab was chosen as a model thin-film piezoMEMS device for studying the effects of humidity from a device perspective. LaNiO3/Pt/Ti/SiO2 deposited on Si-substrates was chosen as a template functional stack for depositing 1 or 2 μm PZT thin-films by either pulsed laser deposition (PLD) or chemical solution deposition (CSD). Two PZT-compositions were studied; undoped PbZr0.52Ti0.48O3 and Ba-doped "hard" Pb1-xBaxZr0.40Ti0.60O3. The electrodes, deposited by magnetron-sputtering, was varied depending on the type of study being conducted. Based on initial test-results from temperature-humidity-bias tests of micro-mirrors, simple test-structures, including pads, cantilevers, and diaphragms, were designed and fabricated to study the various aspects related to humidity-induced degradation in detail. Thin-film deposition and structuring were done in the cleanroom facilities at SINTEF MiNaLab and Penn State University's. Characterization of the reliability and lifetime of devices and test-structures were done in close collaboration with the Trolier-McKinstry group at Penn State University and SINTEF Industry. Optical, scanning electron microscopy, atomic force microscopy and X-ray diffraction were used for microstructural characterization. Various test set-ups, dedicated for studying the reliability and lifetime of piezoMEMS devices and test-structures, were developed to address the effects of humidity. The template set-up was specifically designed to be retrofittable to the different characterization-equipment available at the relevant lab-facilities. This set-up had a Raspberry Pi based platform for controlling the ambient humidity and temperatures, and consisted of a compact, 3D-printable environmental-chamber with viewports for studying the samples in-situ during the experiments. A high-temperature polyimide testcircuit with an integrated substrate heater was designed, making it possible to conduct electromechanical, ferroelectric, dielectric, leakage, and time-dependent measurements, simultaneously on multiple samples under controlled operating conditions. As a first test case, temperature-humidity-bias tests were conducted on micro-mirrors with 1 μm Pb0.9Ba0.1Zr0.40Ti0.60O3 deposited by PLD. 250 nm Au and a 10 nm Ti with 10 wt. % W adhesionlayer to PZT was used as the top electrode system. Unipolar actuation of 20 VPP offset around 10 VDC, corresponding to the device's maximum intended operating voltage, were carried out at substratetemperatures from 25 to 175C. The ambient humidity was kept constant at 22 g/m3, corresponding to 95 % relative humidity (RH) at 25 oC. Initial characterization of the pristine devices showed characteristics, consistent with general ferroelectric theory. During the time-dependent measurements, however, device failure from water-induced degradation preceded that of timedependent dielectric breakdown at all temperatures. In general, such degradation was manifested as local evaporation of the stack-material, forming craters along the edges and on the surface of the used electrodes. Increasing the temperature decreased the number of craters and increased the median time-to-failure, which correlated well with the Brunauer-Emmet Teller theory for water-adsorption on gold-surfaces. Consequently, humidity-degradation was significantly more aggressive at roomtemperature than at elevated temperatures. Despite this, humidity-induced device-failure was encountered also at elevated, even as the ambient approached 15 % RH. In comparison, no devicefailure within the experimental timeframe was detected at 25oC and 35 % RH. Based on the findings, it was proposed that humidity-induced degradation stems from the electrolysis of water. To study the coupling between water-electrolysis and degradation, a comparative study of Pt and Au-based circular top electrode-pads with a diameter of 400 μm on top of 2 μm undoped PbZr0.52Ti0.48O3 thin-films deposited by CSD, was conducted. It was shown that humidity-related degradation couples directly to the stack's electrochemical activity towards the electrolysis of water. As a result of the evolved hydrogen and oxygen gas at the top and bottom electrodes, electrochemical compression lead to cracking and delamination of the piezoelectric layer and top electrodes. Consequently, electrothermal breakdown-events occurring through the newly formed cracks, resulted in time-dependent breakdown, significantly sooner in humid compared to dry conditions. Exemplified for 40 VDC, the median time-to-failure for Au was here three orders of magnitude of time larger than for Pt, at room-temperature in 95 % RH. The results were not consistent with a degradation of the electrical properties of the electrode/PZT interface properties, as commonly encountered in dry conditions. Furthermore, the post-failure craters were considerably larger in humid than in dry conditions, suggesting that additional contributions to leakage, presumably from protonic currents, are involved during breakdown. Based on the findings, a degradation-model for thin-film piezoMEMS in humid conditions degradation was proposed. By studying how surface-currents and the electrode size affected humidity-induced degradation, it was shown that the presence of surface-water is essential for degradation. When the operating ambient exceeds about 50 % relative humidity, protonic surface currents from the anode to the cathode dominate the measured leakage and drives degradation. Since irreversible degradation, including microcracking and electrode-delamination, cannot be removed by drying up the surface, the pristine state of humidity-degraded devices can only be partly recovered by flushing with dry N2 or post-degradation annealing. Increasing the radius of the electrode from 50 to 1000 μm decreased the median time-to-failure by two orders of magnitude, presumably due to the increase in the exposed electrochemically active areas. Accordingly, the median time-to-failure displayed an electrode areadependence rather than a circumference-dependence. Lastly, the use of atomic-layer deposited Al2O3 humidity barrier-layers for improving device reliability, was studied. The dynamic behavior and lifetime of bare and encapsulated micro-mirrors with the initial stack of 1 μm Pb0.9Ba0.1Zr0.40Ti0.60O3 deposited by PLD, were studied. A two orders of magnitude of time increase in the median time-to-failure, in 95 % relative humidity was demonstrated, yet at the cost of decreased piezoelectric and ferroelectric properties, presumably related to the additional encapsulation-step. The encapsulated devices failed by the same mechanisms as for the bare devices due to breaches in the barrier-layer at localized points of the device. The breaches were found to stem from a combination of defects in the barrier-layer, cracks in the PZT-film induced from wire-bonding or sputtered particles from processing, and electrode-delamination. Electrothermal breakdown-events on the membrane decreased the reliability due to distortions in the dynamical behavior of the micro-mirrors during operations, and the electrode-routings were identified as critical flaws. From the above findings, some mitigation-strategies for reducing the impact of humidity, are proposed: (i) Decreasing the overall electrochemical activity of the stack will reduce the amount of water-electrolysis, and thereby the degradation-rate. (ii) Since, within the framework of water-electrolysis, protons generated on the anodeside move towards the cathode-side of the stack, the bottom electrode should be chosen as the positrode. This, to reduce the area of the electrochemically active materials being exposed to the ambient during operation, thereby reducing the proton-transport through the piezoelectric layer. (iii) The electrochemical activity can be reduced by impeding the proton mobility in PZT and along exposed surfaces, e.g. by reducing the physisorption of water-molecules on the exposed surfaces. (iv) Reducing the number of electrode-defects, such as grain-grooves, sputtered particles, cracks or pinholes will reduce the diffusion of protons and hydrogen through the electrode. This can be done by e.g. increasing the electrode thickness when possible. (v) Since the electrolysis of water is essentially a DC-phenomena involving the transport of protons in the form of hydronium or hydroxide-ions, the electrochemical activity will further be reduced by using high-frequency (>100 Hz) bipolar actuation, when such operation is relevant (e.g. for FRAM-applications). (vi) It is imperative to encapsulate piezoMEMS-devices. Therefore, improved electromechanical durability, improved coverage over complex structured surfaces, and reduced water-vapor transmission rate of the humidity barrier-layer will reduce the impact of water. Multi-layered coatings deposited by atomic layer deposition show promising results in this regard