ARCHIVES MASSACHUSETTS INSTITUTE OF TECHNOLOGY JUL 3 0 2015 Ultrasonic Inspection Methods LIBRARIES for Defect Detection and Process Control in Roll-to-Roll Flexible Electronics Manufacturing by Nigel Costello Kojimoto B.S. Mechanical Engineering, Massachusetts Institute of Technology (2012) Submitted to the Department of Mechanical Engineering in partial fulfillment of the requirements for the degree of Master of Science in Mechanical Engineering at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY June 2015 @ Massachusetts Institute of Technology 2015. All rights reserved. Author.. Signature redacted - - - - - _ L epartment ot Mechanical Engineering A 4Signature redacted Certified May 8, 2012 .a Principal Research Scientist, Departme 7 Brian W. Anthony f Mechanica ngineering /"is- Accepted by......... 9pervisor Signature redacted David E. Hardt Professor, Department of Mechanical Engineering Graduate Officer 2 Ultrasonic Inspection Methods for Defect Detection and Process Control in Roll-to-Roll Flexible Electronics Manufacturing by Nigel Costello Kojimoto Submitted to the Department of Mechanical Engineering on May 8, 2012, in partial fulfillment of the requirements for the degree of Master of Science in Mechanical Engineering Abstract Roll-to-roll flexible electronics manufacturing technologies require new sensing and measurement capabilities for defect detection and process control. This body of work investigates the use of ultrasound, specifically Lamb and longitudinal waves, as a sensing modality and measurement technique for thin film roll-to-roll manufacturing. A variety of custom wedges were designed and machined from multiple materials to test their suitability at launching ultrasonic Lamb waves along Polyethylene Terephthalate (PET) film. Using hydrogel wedges, the fundamental antisymmetric (Ao) Lamb wave mode was successfully propagated a distance of 2 cm. ABS wedges were found to initiate longitudinal waves, which were detected at a distance of 5 cm. InstronTM based extension-tensile experiments revealed that the longitudinal wave is affected by tension in the PET, exhibited by a general increase in attenuation and a decrease in wave speed during plastic deformation; such variation may have process control applications. Thesis Supervisor: Brian W. Anthony Title: Principal Research Scientist, Department of Mechanical Engineering 3 4 Acknowledgments This work would not have been possible without the contributions of the following individuals. I would first and foremost like to thank my advisor, Brian Anthony. Brian has in incredible volume of knowledge and experience to draw from and was excellent at suggesting new research directions as well as refocusing my efforts whenever I became unsure of how to proceed. I am incredibly luck to have had Brian serve as my advisor. My colleagues in the Computational Instrumentation Lab were an all around pleasure to both work and eat lunch with. I find it rare for lab mates to spend so much time outside of lab together and the fact that we could hang out at lunch everyday without getting sick of each other was very special. I would like to specifically thank Xian Du for working with me on this project and teaching me much about roll-to-roll manufacturing. Shawn Zhang, Ina Kundu, Tylor Hess, and Matthew Gilbertson were always willing to let me bounce my design ideas off of them, John Lee provided much needed EECS expertise, and Jon Fincke helped me better understand wave physics. I would like to thank Gerry Wentworth from the LMP Machine shop for teaching me fabrication techniques and always being willing to help improve my process plans. Shaoting Lin from the Soft Active Materials Lab was incredibly helpful in creating and casting the hydrogel wedges. Last, I would like to thank my family for always supporting me and encouraging me to do my best, and my girlfriend, Sam Ordonez, for always being there for me and keeping me healthy and happy. 5 6 Contents 1.2 19 Ultrasonic Testing Techniques for Thin Film Polymers Destructive Ultrasonic Testing . . . . . . . . . . . 19 1.1.2 Scanning Acoustic Microscopy (SAM) . . . . . . . 19 1.1.3 Ultrasonic Atomic Force Microscopy (UAFM) . . 20 Ultrasound Wave Modes . . . . . . . . . . . . . . . . . . 21 . . . 1.1.1 . 1.1 Surface Acoustic Waves (SAW) . . . . . . . . . . 22 1.2.2 Lamb Waves . . . . . . . . . . . . . . . . . . . . . 23 1.3 Ultrasound Wedges . . . . . . . . . . . . . . . . . . . . . 25 1.4 Governing Equations for Ultrasound . . . . . . . . . . . 27 1.4.1 Lamb Wave Characteristic Equations . . . . . . . 27 1.4.2 Equations for Wedge Based Lamb Wave Excitation 30 1.4.3 Single Element Transducer Equations . . . . . . . 32 . . . . . . 1.2.1 System Design 33 . . . . . . . . . 34 Angle Wedges . . . . . . . . . . . . . . . . . . . . . . . . 34 2.3 Variable Angle Wedges . . . . . . . . . . . . . . . . . . . 35 2.4 Ultrasound Gel Wedges . . . . . . . . . . . . . . . . . . . 37 2.5 Hydrogel Wedges . . . . . . . . . . . . . . . . ., . . . . . 38 Acetone Vapor Bath for 3D printed ABS Smoothing 40 . . . . . . . 2.5.1 . 2.2 ................. . Transducers .... . 2.1 Wedge Mounting Fixtures . . . . . . . . . . . . . . . . . 43 2.7 Benchtop Web Tensioner . . . . . . . . . . . . . . . . . . 44 7 . . 2.6 . 2 17 Introduction . 1 2.8 3 Zebra Printer . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45 Experiments 49 3.1 Plastic Wedge Benchtop Experiments . . . . . . . . . . . . . . . . . . 49 3.1.1 Acrylic Wedges . . . . . . . . . . . . . . . . . . . . . . . . . . 49 3.1.2 ABS Single Angle Wedges . . . . . . . . . . . . . . . . . . . . 50 3.1.3 ABS Variable Angle Wedges . . . . . . . . . . . . . . . . . . . 54 3.1.4 Teflon Wedges . . . . . . . . . . . . . . . . . . . . . . . . . . . 58 3.2 Instron Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . . 59 3.3 Gel Wedge Benchtop Experiments . . . . . . . . . . . . . . . . . . . . 68 3.3.1 Ultrasound Gel Wedge . . . . . . . . . . . . . . . . . . . . . . 68 3.3.2 Edge Exit Hydrogel Wedges . . . . . . . . . . . . . . . . . . . 70 3.3.3 Surface Exit Hydrogel Wedges . . . . . . . . . . . . . . . . . . 74 4 Conclusions 81 4.1 Longitudinal Waves for Process Control . . . . . . . . . . . . . . . . . 81 4.2 Contact Ultrasound Generated Lamb Waves . . . . . . . . . . . . . . 82 A MATLAB Code 83 B LabVIEW Code 89 C Wedge Drawings 91 8 List of Figures 1-1 A schematic of one SAM embodiment. The single transducer pulses and receives, imaging the sample as it is translated by the stage. . 1-2 . . 20 A schematic of one UAFM embodiment. The AFM probe vibrates in accordance with surface features and probe amplitude and frequency are measured by a laser and photo detector. By sending ultrasonic waves into the stage and sample, the probe vibration characteristics become dependent on subsurface features. . . . . . . . . . ... . . . . . 1-3 21 Cross section showing SAW propagating horizontally along the surface of a thin film and underlying substrate. Notice that the wave amplitude decays with depth into the medium. 1-4 . . . . . . . . . . . . . . . . . . 22 Example Lamb waves propagating horizontally through a section of a continuous plate. The symmetric mode (left) and antisymmetric mode (right) are differentiated by their symmetry over the plate midplane, represented by the dashed line. The arrow indicates the direction of w ave travel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1-5 Olympus acrylic wedge (ABWML-7T-90) with miniature screw-in transducer (C548-SM) for propagating SAW through steel. . . . . . . . . . 1-6 23 26 Figure shows the difference between through transmission and pulse echo ultrasound configurations. Two transducers are required for through transmission, one to pulse and one to receive, while only one transducer is used for pulse echo, because the single transducer both pulses and receives any returning echoes. . . . . . . . . . . . . . . . . . . . . . . 9 26 1-7 Dispersion curves showing the 0th and 1st symmetric (So and S1) and antisymmetric (Ao and A 1 ) modes for 76 ym thick PET. . . . . . . . 1-8 29 Group velocity dispersion curves showing the 0th and 1st symmetric (So and S1 ) and antisymmetric (Ao and A 1 ) modes for 76 pm thick PET. 29 1-9 The angle -y at which a wave refracts at the interface between two different materials is governed by Snell's Law. . . . . . . . . . . . . . 31 1-10 Schematic showing wedge based excitation of a Lamb wave. The transducer launches an ultrasonic wave that refracts off the bottom of the wedge. Note that the wave is still longitudinal when inside the wedge. The critical angle of incidence 4 from Eq. 1.10 is also shown. . . . . . 31 1-11 When a single element transducer fires, the near field region contains many wave fronts constructively and destructively interfering with each other. In the far field, the single wave front diverges at angle a. . . . 32 . . . . . . . . . . . . . . . . . . . . . . . . . . 33 2-1 Overall system design. 2-2 Wedges machined from ABS (left) and Teflon (right). Note that the Teflon wedge is larger due to their slower speed of sound and thus the larger near field distance. . . . . . . . . . . . . . . . . . . . . . . . . . 2-3 35 Variable angle wedge machined from Teflon. Wedges are composed of a circular element that is free to rotate inside a larger wedge base. Note that the wedge acoustic wave exit location is angle dependent. .... 2-4 36 Variable angle wedge machined from ABS. Wedges are composed of a carriage that rides on a round base, allowing the wedge acoustic wave exit location to be independent of the angle. Carriages are secured to a laser cut acrylic mount which allows for the accurate selection of angle. 36 2-5 Image showing the bottom of an ultrasound gel based wedge. A 3D printed ABS support structure contains the ultrasound gel while a 12 pim thick PET layer glued to the support base prevents the gel from leaking out. Notice the abundance of bubbles present in the gel. . 10 . . 38 2-6 Mold with UV cured hydrogel wedge. Mold is composed of a 3D printed ABS base and laser cut acrylic sides. Note that although there visible bubbles along the mold surface, the hydrogel itself is essentially bubble free. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 2-7 Hydrogel Wedge in ABS support structure. . . . . . . . . . . . . . . . 40 2-8 Edge exit vs. surface exit wedge geometry. In edge exit, the center of the ultrasound beam exits at the wedge tip, whereas for surface exit, the beam center exits before the wedge tip and the entire beam within 6 dB of the beam center exits through the wedge bottom surface. 2-9 . . 40 Acetone vapor bath used to smooth surface of 3D printed ABS parts. Liquid acetone rests on the bottom of the closed container, releasing vapors. The 3D printed part rests on a stand to prevent it from coming in contact with the liquid acetone. . . . . . . . . . . . . . . . . . . . . 41 2-10 Comparison between wedge molds of similar sizes that were exposed to the acetone vapor bath for 1.5 hours (top) and 2 hours (bottom). Molds are similar but not of identical design, as the edge exit mold (top) and surface exit mold (bottom) differ in wedge tip geometry. Note that the mold with longer exposure has much greater radius of curvature along the edges and especially around holes. . . . . . . . . 42 2-11 Mount designed to investigate the effect of acoustic wedge preload on PET substrate. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 2-12 Mount designed to vary the spacing between acoustic wedges continuou sly. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 2-13 Benchtop tensioner constructed to perform experiments on PET under tension. Tensioner utilizes a ratchet mechanism to hold the PET taught. 44 2-14 Zebra thermal transfer printer used for printing patterns on PET film. 45 2-15 Example hollow square grid pattern printed with the Zebra thermal transfer printer. Hollow squares are roughly 420 pm wide. PET substrate was printed on left to right. . . . . . . . . . . . . . . . . . . . . 11 46 2-16 Example dot grid pattern printed with the Zebra thermal transfer printer imaged at 20x magnification. diam eter. Dots are roughly 336 pm in . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 2-17 CAD concept of fully integrated Zebra printer based roll-to-roll manufacturing and inspection set-up. Note that this concept uses two cameras, one wide view with backlight to locate areas of interest and one with high magnification on a translation stage to investigate areas of interest. Sensing is not limited to the optical domain and ultrasonic sensors could also be utilized in a similar set-up. . . . . . . . . . . . . 3-1 Single angle ABS wedges used in experiment on the benchtop tensioner in a through transmission configuration. 3-2 . . . . . . . . . . . . . . . . 50 ABS single angle wedge average of 100 through transmission signals for wedge spacing of 1.27 cm. 3-3 48 . . . . . . . . . . . . . . . . . . . . . . 51 ABS wedge transmission signal map created by varying the spacing between wedges in increments of 1.27 mm. Greater magnitude indicates more signal activity in that region. Expected locations of various acoustic waves are overlayed onto the map. . . . . . . . . . . . . . . . 3-4 51 Example frequency-time decomposition for single angle ABS wedges spaced 1.27 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. . . . . . . . . . . . . . . . . . . . . . . . 53 3-5 Variable angle ABS wedges used in experiment on the benchtop tensioner. 54 3-6 ABS variable angle wedge average of 100 through transmission signals for wedge angle of 300 with respect to horizontal. Note the large number of individual waves when compared to Fig. 3-2. . . . . . . . . 12 55 3-7 Simulation showing the expected locations of the various possible waves propagated by the ABS variable angle wedges. Note that signal attenuation was not considered and equal magnitudes of the different waves is not a serious result. From left to right, the various signals are the longitudinal wave, the symmetric Lamb wave, the back and forth reflection in the wedge carriage only, the back and forth reflection in wedge base only, the antisymmetric Lamb wave, and the back and forth reflection through the entire wedge. 3-8 . . . . . . . . . . . . . . . . . . . . . . . . 55 Variable angle ABS wedge transmission signal map created by varying the wedge angle in increments of 5' between 30'-90' with respect to horizontal. Greater magnitude indicates more signal activity in that region. Expected locations of various acoustic waves are overlayed onto the m ap. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3-9 57 Example frequency-time decomposition for variable angle ABS wedges at 300 and spaced 2.54 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. . . . . . . . . . . . . . . . . . 57 3-10 Configuration with Teflon wedges that was able to propagate measurable ultrasound waves. . . . . . . . . . . . . . . . . . . . . . . . . . . 58 3-11 Schematic showing the experimental setup on the Instron machine. Transmission signal through the PET travels out of the page . . . . . 60 3-12 Experimental setup on the Instron Machine. Note the use of the spring loaded mount to apply preload on the wedges and the rollers to mimic roll-to-roll boundary conditions. . . . . . . . . . . . . . . . . . . . . . 60 3-13 Instron stretch experiment of 76 pm thick PET monitored by ABS ultrasound wedges spaced 4.45 cm apart.- Figure shows that the maximum of the ultrasonic transmission signal power spectrum decreases reversibly as the PET sample is stretched. PET was stretched at a rate of 0.1 mm /min. . . . . . . . . . . . . . . . . . . . . . . . . . . . 13 61 3-14 Representative force profile for Instron stretch experiment of 76 prm thick PET. Note the elastic region on the left and the plastic region on the right. PET was stretched at a rate of 0.5 mm/min. . . . . . . 63 3-15 Instron stretch experiment of 76 pm thick PET monitored by ABS ultrasound wedges spaced 0.64 (top), 4.45 (middle), and 5.08 cm (bottom) apart. Figures show the peaks of the power spectrum as a function of stretched length. PET was stretched at a rate of 0.5 mm/min. Note that different amplifier gains were used for each wedge separation, preventing the comparison of power spectrum peak values of different wedge separations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 3-16 Instron stretch experiment of 76 jim thick PET monitored by ABS ultrasound wedges spaced 0.64 (top), 4.45 (middle), and 5.08 cm (bottom) apart. Figures show the peaks of the power spectrum as a function of stretch force. PET was stretched at a rate of 0.5 mm/min. Note that different amplifier gains were used for each wedge separation, preventing the comparison of power spectrum peak values of different wedge separations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 3-17 Instron stretch experiment of 76 jpm thick PET monitored by ABS ultrasound wedges spaced 0.64 (top), 4.45 (middle), and 5.08 cm (bottom) apart. Figures show the time of flight arrival of the three largest peaks in the ultrasonic transmission signal as the PET sample is stretched. The peak locations are found using a peak fitting algorithm that is not as effective at lower signal to noise ratios, accounting for the large errors at higher strains and wedge separation distances. PET was stretched at a rate of 0.5 mm/min. . . . . . . . . . . . . . . . . . . . . . . . . . 67 3-18 Example through transmission signal for ultrasound gel wedges spaced 0.64 cm apart. Notice how all of the various waves seem to blend together much more than the through transmission examples of other wedge types. This is likely due to interference caused by bubbles in th e gel. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 69 3-19 Experimental setup with edge exit hydrogel wedges. Wedges rest on PET that sits on an acoustically uncoupled wooden desk. . . . . . . . 70 3-20 Example through transmission signal for edge exit hydrogel wedges spaced 1.27 cm apart . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 3-21 Edge exit hydrogel wedge signal map created by slowly incrementing the spacing between wedges. Expected locations of various acoustic waves are overlayed onto the map. The signal line with the most gradual slope that does not have an expected overlay is due to a wave propagating though mounting hardware and is present when the wedges are not in contact with the web. . . . . . . . . . . . . . . . . . . . . . . . 72 3-22 Example frequency-time decomposition for edge exit hydrogel wedges spaced 1.27 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. . . . . . . . . . . . . . . . . . . . . . . . 73 3-23 Experimental setup with surface exit hydrogel wedges. Wedges rest on PET that sits on an acoustically uncoupled wooden desk. Note that the tip of the hydrogel wedge protrudes from the ABS support structure, differentiating the surface exit wedges from the edge exit wedges in Fig. 3-19. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 74 3-24 Example through transmission signal for surface exit hydrogel wedges spaced 1.27 cm apart . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 3-25 Comparison of through transmission signal in PET on the table and PET in the benchtop tensioner. Both signals were obtained using the same surface exit hydrogel wedges with a separation of 1.27 cm. . . . 76 3-26 Surface exit hydrogel wedge signal map created by slowly incrementing the spacing between wedges. Expected locations of various acoustic waves are overlayed onto the map. . . . . . . . . . . . . . . . . . . . . 77 3-27 Example frequency-time decomposition for surface exit hydrogel wedges spaced 1.27 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. . . . . . . . . . . . . . . . . . . . . . . . 15 78 B-i Front panel of LabVIEW code providing communication between the computer and digital oscilloscope. Code allows various oscilloscope acquisition settings to be selected and processes and saves collected data. 89 B-2 Labview code providing communication between the computer and digital oscilloscope. Code allows various oscilloscope acquisition settings to be selected and processes and saves collected data. . . . . . . . . . 90 . . . . . . . . 91 C-2 Drawings of ABS variable angle wedge base. . . . . . . . . . . . . . 92 C-3 Drawings of Variable angle ABS wedge carriage. . . . . . . . . . . . 93 C-4 Drawings of Single angle Teflon wedge. . . . . . . . . . . . . . . . . 94 C-5 Drawings of Mold base for edge exit hydrogel wedge. . . . . . . . . 95 C-6 Drawings of Mold base for surface exit hydrogel wedge. . . . . . . . . 96 . . . . C-1 Drawings of ABS single angle wedge. . . . . . . . . . 16 Chapter 1 Introduction Currently, there is an industry push to scale up roll-to-roll flexible electronics manufacturing [27]. One approach involves "printing" circuits on flexible polymer substrates [27]. Roll-to-roll manufacturing brings with it benefits that include large area, high throughput and low cost, with diverse applications that include flexible displays, solar panels, radio-frequency identification tags (RFIDs), medical devices, bio-integrated sensors, microfluidic devices, and computing platforms [12]. However, before this technology can be fully realized at scale with high quality, there are a number of challenges that must be solved. One of these challenges is to develop methods for defect detection and process control. This can be challenging not only because printed features can be on the micro or even nano scale over meter scale distances, but also because motion and vibrations in the non-rigid substrate, if of some scale or larger, interfere with sensor coupling. Some optical methods have been applied to the control of such processes [12] but many inspection modes have yet to be investigated or understood. This work explores the viability, challenges, and implementation issues of using ultrasound as a sensing mode for the defect detection and process control of roll-to-roll flexible electronics manufacturing. Reflection and transmission signals can be analyzed in order to estimate physical properties of the medium through which it propagated. One primary goal was to use contact ultra- sound to initiate and propagate ultrasonic Lamb waves in polyethylene terephthalate (PET), a commonly used polymer substrate, and to evaluate the sensitivity of Lamb 17 waves to defects and changes in process parameters. Ultrasound based measurements could then be combined with optical based measurements into a multimodal sensing platform in order to observe and control geometry and material variations. Prior work has shown that the excitation of Lamb waves is possible in PET through laser based ultrasound [11, 13] and in other materials with contact wedge ultrasound [2, 34, 35]. The author is not aware of any documented work to excite Lamb waves in PET using transducers in contact with the material, hereafter referred to as contact ultrasound. Contact ultrasound was chosen and pursed in this research with the hypothesis that it had the potential to provide advantages over laser based ultrasound. One of these advantages is propagation directionality, because the majority of acoustic energy in contact ultrasound propagates in a specific direction, whereas in laser based ultrasound, the energy propagates radially from the laser spot. Another advantage is that for contact ultrasound the initial introduction of energy into the system is spread over the transducer face and not concentrated into a spot with the potential to ablate the PET surface at high energies. Contact ultrasound is a less expensive technology than laser based ultrasound to implement and laser based ultrasound also requires either the selection of a specific laser frequency for which PET is not transparent or the deposition of an additional layer that can absorb the laser frequency. Finally, contact ultrasound allows for more control over the excitation frequencies. These advantages are hypothesized to allow contact ultrasound to provide stronger, more desirable signals with greater propagation distances than laser based ultrasound. Note that in a manufacturing environment the strategy for making and maintaining contact will have to be addressed. The remainder of this chapter is an overview and background, including prior art and theory, in ultrasonic testing methods relevant or adaptable to thin film polymers. Chapter 2 describes key hardware components that were acquired or manufactured for this research. Chapter 3 describes the experiments performed in this study and discusses the results and findings. Finally, Chapter 4 summarizes and comments on the key findings of this work. 18 1.1 Ultrasonic Testing Techniques for Thin Film Polymers The following sections include prior art and research in ultrasonic testing methods for, or adaptable to, thin film polymers. 1.1.1 Destructive Ultrasonic Testing Ultrasound has been used to excite a sample and induce delamination. Haidara et al. [16] used ultrasonic excitation to destructively test the adhesive performance of vapor deposited aluminum on PET. The sample, which was submerged and attached to the tip of a piezoelectric transducer actuated acoustic amplifier, was excited until the aluminum layer completely separated from the PET base. Although interesting, a nondestructive method for testing adhesion or other film properties is preferred if used for continuous process control. 1.1.2 Scanning Acoustic Microscopy (SAM) Scanning Acoustic Microscopy (SAM), a schematic of which is shown in Fig. 1-1, involves the use of a high frequency ultrasonic transducer to measure relative sample thickness or elastic modulus locally. By displacing the sample in known increments, the relative measurements can be assembled to form an image. SAM can achieve relatively high resolution, on the order of microns [22]. However, SAM suffers from limited signal penetration due to the increased attenuation experienced by high frequency ultrasound [28] and the point by point scanning method limits its suitability for large area measurement. Balogun et al. [3] utilized a laser based SAM technique to image sub-surface features in a silicon wafer sample. Lisy et al. [22] imaged internal damage in oriented polypropylene due to high-speed projectile impact with SAM. Cros et al. [10] used SAM to resolve 2-4 lim diameter ceramic particles in PET and the adhesion between PET and an aluminum coating. 19 Transducer Stage Figure 1-1: A schematic of one SAM embodiment. The single transducer pulses and receives, imaging the sample as it is translated by the stage. 1.1.3 Ultrasonic Atomic Force Microscopy (UAFM) Ultrasonic Atomic Force Microscopy (UAFM), a schematic of which is shown in Fig. 12, incorporates ultrasound into atomic force microscopy (AFM). A piezoelectric transducer is used to vibrate the cantilevered probe tip at its resonant frequency. By sending ultrasonic waves into the stage and sample, the probe vibration characteristics become dependent on subsurface features because the resonant frequency changes due to interaction forces and local sample stiffness when the probe contacts the sample. The changing resonant frequency is then mapped as the probe scans the sample to form an image. UAFM is capable of achieving very high resolutions, on the order of tens of nanometers [38], but suffers from similar signal penetration and scanning speed problems as SAM, limiting its applicability to large area inspection and process control. Passeri et al. [31] utilized UAFM for the mechanical characterization of thin films and for the detection of subsurface defects, including delamination, voids, cracks, and dislocations. Hu et al. [17] imaged subsurface gold structures under a polystyrene layer with UAFM. Kwak et al. [20] used UAFM to image SiO 2 patterns under a polymethyl methacylate (PMMA) layer. Shekhawat and Dravid [38] used a variation of UAFM they called scanning near-filed ultrasound holography (SNFUH) to successfully image subsurface 15 nm diameter gold particles and perform subsurface 20 Laser Probe Sample Stage Figure 1-2: A schematic of one UAFM embodiment. The AFM probe vibrates in accordance with surface features and probe amplitude and frequency are measured by a laser and photo detector. By sending ultrasonic waves into the stage and sample, the probe vibration characteristics become dependent on subsurface features. metrology in microelectronic structures. McGuigan et al. [23] used UAFM to measure debonding near cracks in the brittle glass layer of a nanocomposite with PET substrate. Gaimei et al. [14] used UAFM to image subsurface defects in SiO, coated PET and biaxially oriented polypropylene (BOPP). Nalladega et al. [26] applied UAFM towards the characterization of defects in flexible circuits. 1.2 Ultrasound Wave Modes The simplest form of ultrasonic testing involves the excitation and measurement of longitudinal waves. However, additional wave modes exist that also have uses in measurement and sensing. The following sections include prior art and qualitative background for two of these additional modes, surface acoustic waves and Lamb waves, including their applicability to the inspection of thin film polymers. 21 Surface Acoustic Waves (SAW) 1.2.1 Surface acoustic waves (SAW) are waves that travel along the surface of an elastic medium whose amplitude decays with depth into the medium [2, 36], as shown in Fig. 1-3. They consist of the superposition of a longitudinal wave and a vertical shear wave. SAW waves are dispersive, meaning that the velocity at which they travel varies with wave frequency, causing a single wide bandwidth temporal impulse to disperse into separate frequency components as it propagates. SAW based measurements utilize the dependence of SAW dispersive characteristics on the sample elastic properties to estimate those properties. Wave Amplitude Depth Figure 1-3: Cross section showing SAW propagating horizontally along the surface of a thin film and underlying substrate. Notice that the wave amplitude decays with depth into the medium. When used on thin films, these SAW based systems are primarily applied towards measurements of thin films on significantly thicker substrates [9, 24, 36] such as depositions on silicon wafers. This way, the majority of the SAW energy propagates along the thin film with little energy propagating into the substrate. The dispersive nature of SAW allows for more information to be determined faster, thus making SAW based systems more suitable for process control and large area inspection than SAM or UAFM methods. However, SAW are physical phenomenon that require the substrate or sample to be thicker than the depth to which the energy penetrates, which, depending on the excitation frequencies used, will not always the case for 22 flexible electronics whose substrates are already thin film polymers. Cote et al. [9] used laser generated surface acoustic waves (SAW) to measure the elastic modulus of a thin polymer layer coating on a stainless steel substrate. Mileham et al. [24] developed a SAW based system that employed interdigital transducers (IDT) to measure the mass loss due to water outgassing during the curing of thin polymer films, as well as the glass transition temperature of acoustically thin films and film resonance for acoustically thick films. Films with thicknesses that were significantly smaller than the SAW wavelenth were categorized as acoustically thin whereas films with significantly larger thicknesses were categorized as acoustically thick. Schoedel [36] described a laser excited SAW based system capable of measuring thickness, hardness, elastic modulus, and Poisson's ratio in thin films. 1.2.2 Lamb Waves Lamb waves are similar to surface acoustic waves except they propagate throughout the entire thickness of thin elastic mediums [2]. The waves travel parallel to the plate surfaces while individual particle motion is primarily perpendicular to the direction of wave travel. Lamb waves propagate in two distinct ways; symmetric, where the particle motion is symmetric about the plate midplane, and antisymmetric, where the motion is antisymmetric about the plate midplane. Example Lamb wave mode profiles are shown in Fig. 1-4. -------------------- Ir- Symmetric Lamb Wave Antisymmetric Lamb Wave Figure 1-4: Example Lamb waves propagating horizontally through a section of a continuous plate. The symmetric mode (left) and antisymmetric mode (right) are differentiated by their symmetry over the plate midplane, represented by the dashed line. The arrow indicates the direction of wave travel. In an elastic medium, an infinite number of symmetric and antisymmetric prop- 23 agation modes, denoted (Sn) and (A,), respectively, are theoretically possible. Like SAW, Lamb waves are dispersive, facilitating the measurement of various sample properties including elasticity and thickness [2]. A series of Lamb waves obtained by scanning a wave source and receiver in one direction, translating them perpendicular to the direction of wave travel and parallel to the plate, can even be used to reconstruct an elastic property or thickness based image of the scanned area by solving an inverse problem [32]. Lamb wave based inspection systems are used on a variety of sample materials such as aluminum, steel, and composites [34, 42] with advantages that include the ability to quickly inspect large areas with little attenuation and good sensitivity to defects including cracks and delaminations [42]. This makes them a potentially good fit for large area inspection of thin film polymer based flexible electronics and a better fit than SAM and UAFM. Some potential drawbacks to using Lamb waves include decreased resolution in comparison to SAM and UAFM methods. Abdelrahman et al. [1] used a pair of lead zirconate titanate (PZT) transducers to generate Lamb waves in an aluminum plate. One transducer was bonded to the plate's top surface while the other was bonded symmetrically on the bottom surface. By operating the transducers in or out of phase, symmetric or antisymmetric waves, respectively, were selectively excited. Rogers [34] used variable angle acrylic wedges to propagate Lamb waves and accurately measure elastic constants of aluminum, steel, and glass. Yeum et al. [42] used Lamb waves to find delamination defects in a composite plate. By monitoring Lamb waves with a piezoelectric transducer network, delaminations were located using the property that the fundamental antisymmetric mode (Ao) slows down when it passes through an area of delamination while the fundamental symmetric mode (So) remains the same. Rose and Ditri [35] demonstrated Lamb wave based adhesive bond inspection in aluminum plates and its potential for significantly decreased scan times over conventional bulk wave normal incidence techniques. Pei et al. [32] utilized a dry contact method of A0 mode Lamb wave excitation and measurement to accurately measure thickness in a steel pipe and image a defective aluminum plate. 24 There has also been research related to the propagation of Lamb waves in PET. Okabe et al. [30] measured reflected wave intensity of water submerged PET to measure elastic constants. They observed dip points in the magnitude of reflection at certain incident angles due to leaky Lamb wave excitation. Nakaso and Yasujima [25] used a point focus beam transducer to measure Lamb wave acoustic emissions in ceramic coated PET film. Desmet et al. [11] utilized laser based excitation and measurement of Lamb waves in PET and explored the effects of film stress on the So and AO mode dispersive relations. So and AO modes were measured 0.8 cm from the laser source. Futatsugi et al. [13] used laser generated Lamb waves to simulate acoustic emissions caused by cracks in the SiO_ coating of PET film. A network of four transducers then used the arrival of the So mode to locate the source origin. The So mode was measured up to 2.5 cm from the source. Thus, Lamb wave measurement with contact transducers has been achieved in PET, but Lamb wave excitation in PET with contact ultrasound has not. 1.3 Ultrasound Wedges Ultrasound wedges are used to change the direction of a propagating wave through refraction. A transducer is placed perpendicular to one of the wedge faces, while the other wedge face lies flat against the object to be inspected. The transducer launches a longitudinal wave into the wedge that then refracts at the wedge-object interface, changing the direction of travel. This process is described more quantitatively in Sec. 1.4.2. An example wedge and transducer unit is shown in Fig. 1-5. To be effective, a coupling fluid, such as ultrasound gel, is commonly added to solid interfaces, such as the transducer-wedge and wedge-object interfaces, to increase acoustic coupling between the two surfaces. The coupling fluid fills the tiny air gaps in the interface, allowing acoustic waves to more easily pass through. Ultrasound wedges can be used in two different ways, the pulse receive method, also known as through transmission, and the pulse echo method, as shown in Fig. 16. In through transmission, a transducer emits an ultrasonic pulse which propagates 25 Figure 1-5: Olympus acrylic wedge (ABWML-7T-90) with miniature screw-in transducer (C548-SM) for propagating SAW through steel. through the wedge and into the sample. The wave continues to propagate through the sample until it reaches a new wedge. The wave can then enter the wedge and be received by a new transducer. In pulse echo, a single transducer emits a pulse which propagates through the wedge and into the sample. The wave continues to propagate through the sample until it reaches a discontinuity that causes it to reflect back. This "echo" is then picked up by the same wedge and received by the same transducer. Pulsing and Receiving Transducer Pulsing Transducer Receiving Transducer Wedge Wedges Smple Sample Pulse Echo Through Transmission Figure 1-6: Figure shows the difference between through transmission and pulse echo ultrasound configurations. Two transducers are required for through transmission, one to pulse and one to receive, while only one transducer is used for pulse echo, because the single transducer both pulses and receives any returning echoes. Tucker and Bender [41] investigated using wedged ultrasound for the inspection of wood-plastic composites. Son and Lee [39] looked at alternative wedge materials for 26 SAW based outdoor touch panels. Some researchers [5, 18, 41] have explored using a liquid for the wedge material. Wedges can also be used to excite Lamb waves [2, 34]. However, at this time, the author is not aware of any work involving wedge based Lamb wave excitation in PET or other polymer films. 1.4 Governing Equations for Ultrasound The following sections contain the equations and theory that govern wedge based Lamb wave excitation and drove the design of various parts and experiments in this study. 1.4.1 Lamb Wave Characteristic Equations The elastic wave equation describes how waves propagate in an elastic isotropic homogeneous medium (1.1) pii = f + (A + 21 )V(V - u) - uV x (V x u) where p is the material density, u is the displacement vector, f is the driving force, and V is a vector of partial derivative operators. A and p are the Lam6 parameters which parameterize the elastic moduli for isotropic homogeneous media. For derivation of the elastic wave equation, refer to Bedford and Drumheller [4]. Using Helmholtz decomposition, the displacement vector u can be expressed as u = V<D + V x T (1.2) where <D is the scalar potential and IF is the vector potential. Assuming a thin plate of thickness d in the z direction, infinitely extent in the x and y directions, and wave propagation in the x direction, the following sinusoidal equations can be used with the Helmholtz decomposition in Eq. 1.2 to solve Eq. 1.1 <P = F(z)e(wt-kx) 27 (1.3) and (1.4) 41 = G(z)ei(wt-kx). Here, F and G represent unknown functions of z, w is the angular frequency, t is time, and k is the wave number. By applying a boundary condition of zero stress at the plate surfaces to the resulting equations, the following two characteristic equations can be derived: tan(od/2) tan(ad/2) 4ak 2 (02 - k 2 ) 2 tan(#d/2) (2 - k2)2 tan(ad/2) 4ack2 ( and representing the symmetric and anti-symmetric modes of Lamb wave propagation, respectively. Here, a2 CT - k2 , 2 = k 2 , cl is the longitudinal wave velocity, ct - t is the shear (transverse) wave velocity and cp is the Lamb wave phase velocity. Note that the wavenumber k is related to w and c, by k = w/cp. For the complete derivation of the Lamb wave characteristic equations, refer to Bedford and Drumheller [4]. Numerical methods can then be used to find the roots of Eq. 1.5 and 1.6 as frequency and phase velocity pairs. The continuous curve of roots, W vs k, define the dispersive characteristics of a particular mode. This work focuses on the fundamental symmetric (So) and antisymmetric (Ao) modes because they can be excited at all frequencies. fc, below which they by f, = ct/2d [13]. For Higher order modes exhibit a cutoff frequency, will not propagate. The cutoff frequency can be estimated PET with ct = 1000 m/s and d = 76 um, f, = 6.6 MHz, which is a higher excitation frequency than those considered in this work. Once the phase velocities for the various symmetric and antisymmetric modes have been calculated, group velocity, Cg, is given by the slope of the dispersion curve: c = W (1.7) Group velocity is the velocity at which the envelope of a group of waves propagates through space [2]. An example of dispersion curves is given in Fig. 1-7, which shows 28 Dispersion Curves for 76 urn thick PET 4000 F ---- - 3500 --- 3000 - - - AO3 L - _Al - E 2500 S0 2000 - 0 (D Cl) c. 1500 - - - -- - - - - - 1000 / 500 'I o' 0 10 5 15 Frequency (MHz) 20 25 25 30 Figure 1-7: Dispersion curves showing the Oth and 1st symmetric (SO and S1) and antisymmetric (A0 and A 1 ) modes for 76 pm thick PET. Group Velocity Curves for 76 urn thick PET 2000 1800 So AO 1600 A~ - 1400 1200 P 0 1000 ,---------- 800 ~ -I. 600 400 200 0 0 5 10 15 Frequency (MHz) 20 25 25 30 Figure 1-8: Group velocity dispersion curves showing the Oth and 1st symmetric (So and SI) and antisymmetric (A0 and A 1 ) modes for 76 pum thick PET. 29 the 0th and 1st symmetric and antisymmetric modes for 76 pim thick PET, with cl = 2310 m/s and ct = 1000 m/s [13]. Note that this curve is in the form of cp vs frequency f = w/27r but contains the same information as w vs k. An example of group velocity dispersion curves is given in Fig. 1-8. The MATLAB code used to generate these graphs can be found in Appendix A. The longitudinal speed of sound in a solid can be written in terms of the material's Young's modulus E, Poisson's ratio v, and density p [33] such that E(1- v) p(1+ v)(1 - 2v) For typical values of Poisson's ratio where 0 < v < 0.5, longitudinal speed of sound cl is strictly increasing. 1.4.2 Equations for Wedge Based Lamb Wave Excitation As discussed in Sec. 1.3, ultrasound wedges can be used to change the direction of wave propagation in accordance with Snell's law [19], Sin(4) Sin(y) ci c,' (1.9) where / is the angle of incidence, 7 is the angle of refraction, ci is the speed of sound in the incident material, and c, is the speed of sound in the refracted material. Both angles are measured with respect to the line normal to the material interface, as shown in Fig. 1-9. To use the angle wedge method to excite Lamb waves, the wedge must preferentially direct energy along the sample layer. This requires the angle of refraction -/ to be 90', which transforms Snell's law into the following form, c = sin(O) (1.10) where cp is the phase velocity inside the sample, cl is the longitudinal wave velocity 30 cj Interface cr Figure 1-9: The angle y at which a wave refracts at the interface between two different materials is governed by Snell's Law. in the wedge, and # is now the critical angle of incidence required to excite a Lamb wave. This process is illustrated in Fig. 1-10. Wedge amb Wave Transducer A A Substrate Figure 1-10: Schematic showing wedge based excitation of a Lamb wave. The transducer launches an ultrasonic wave that refracts off the bottom of the wedge. Note that the wave is still longitudinal when inside the wedge. The critical angle of incidence 4 from Eq. 1.10 is also shown. Note that for cl > cp, there theoretically exists no critical angle # for which a Lamb wave can be generated. In practice, Eq. 1.10 is not a hard rule, although excited 30 % symmetric and antisymmetric modes in general do have phase velocities around of that predicted by Eq. 1.10 [34]. Potential reasons for a range of phase velocities excited by a single wedge angle include the wave interactions being more complex than modeled or variations in local material properties. The excitation frequency is a stricter determinant of which frequency and phase velocity pair on the dispersion 31 curves is excited [34]. Thus, a single wedge can therefore be used to excite multiple Lamb wave modes depending on the excitation frequency. 1.4.3 Single Element Transducer Equations When a single element transducer fires, the ultrasonic beam created consists of two different regions: the near field and the far field [28]. The near field of an ultrasonic beam fluctuates due to constructive and destructive interference between the many wave fronts created by the transducer surface. These waves merge into one in the far field region, as illustrated in Fig. 1-11. The distance N from the transducer element's face to the boundary between the near and far field regions is given by N= D2 f [28], 4c, (1.11) where D is the diameter of the transducer element, f is the transducer frequency, and cl is the longitudinal speed of sound in the material through which the wave propagates. Ultrasonic beams in the far field diverge over time [28], as illustrated in Fig. 1-11. The angle a at which the beam diverges can be calculated for a specific loss from the beam center magnitude. The -6 dB beam spread angle can be calculated by a = 2Sin- (0-514c) fD (1.12) [28]. Transducer ~----~------ a N Figure 1-11: When a single element transducer fires, the near field region contains many wave fronts constructively and destructively interfering with each other. In the far field, the single wave front diverges at angle a. 32 Chapter 2 System Design This chapter summarizes the various key hardware components that were acquired or manufactured for this study and the reasons for their selection. The overall system design can be found in Fig. 2-1. m puser PicoScope JSR 5444B Ultrasonics Oscilloscope e d on Roll-to-Roll Systerm bdl l Triggers Microcontroller Transducers Wedges c f i PET Web Figure 2-1: Overall system design. a Cdigital osilsoe(iocPR30)0 omuiainbewe h cmue n Ultrasound wedges are placed in contact with the PET web on a roll-to-roll system. A pulser/receiver (JSR Ultrasonics DPR300) excites the pulsing transducer and processes the signal from the receiving transducer before sending the signal to a digital oscilloscope (PicoScope 5444B). Communication between the computer and oscilloscope is achieved using a LabVIEW application, which can be found in Appendix B, allowing data to be saved and analyzed. Triggering was performed using a microcontroller (mbed LPC1768) in order to synchronize the oscilloscope and pulser. The following sections contain further details on individual system components. 33 2.1 Transducers Two 1 MHz miniature screw-in Centrascan transducers (C548-SM) were purchased from Olympus. The transducers have a 10 mm element diameter a -6 dB bandwidth of roughly 0.6-1.5 MHz. These identical transducers, one to pulse and one to receive, were used in a through transmission configuration. The transducers were chosen to have a relatively small size to prevent the near field from becoming too large and a relatively low frequency to maximize signal through the PET substrate. This is because attenuation typically increases with higher frequencies [28], and comes at the cost of larger wavelength. The Centrascan transducers were chosen because they have a larger bandwidth compared to other transducer types and a wide range of frequency excitation was desired. All experiments were performed with these 1 MHz transducers. 2.2 Angle Wedges Commercially, many angle wedges are made from acrylic which has a longitudinal speed of sound ci = 2, 750 m/s. This makes sense for creating Lamb waves in materials like steel or aluminum, which can have phase velocities of c, 5,500 m/s, but for materials like PET which require phase velocities closer to cP ~ 1, 800 m/s, as shown in Fig. 1-7, wedges with slower longitudinal speeds are required. the different materials considered for the angle wedges. Material ABS Polyethylene (LDPE) Styrene Butadiene (SBR) Teflon RTV Silicone cl (m/s) 2,230 1,950 1,920 1,390 670-1,160 Table 2.1 shows Ultimately, wedges were Loss (dB/cm) 11.1 A 5 MHz 2.4 A 5 MHz 24.3 A 5 MHz 3.9 A 5 MHz 3.7 A 1 MHz Table 2.1: Materials considered for angle wedges [6, 7, 8]. manufactured from ABS and Teflon due to their ease of machining and availability. Wedges were machined using a mill and are shown in Fig. 2-2. The general wedge 34 shape was modeled after the Olympus SAW wedge (ABWML-7T-90) shown in Fig. 15. Note that 0 = 90 - #, where # is from Sec. 1.4.2. Figure 2-2: Wedges machined from ABS (left) and Teflon (right). Note that the Teflon wedge is larger due to their slower speed of sound and thus the larger near field distance. Wedges were designed such that near field effects would be contained within the wedge. This caused the Teflon wedges, with N'refon = 29.0 mm, to be larger than the ABS wedges, with NABS = 18.0 mm, due to its lower longitudinal speed of sound, in accordance with Eq. 1.11. The center of the propagated wave was designed to go through the wedge tip under the assumption that this would maximize the transmitted acoustic energy. An alternative design is presented in Sec. 2.5 and experiments later showed that this assumption does not hold, as discussed in Sec. 3.3.3. Drawings of both ABS and Teflon single angle wedges can be found in Appendix C. Although the Teflon wedges were able to conduct acoustic waves, all wedges made from Teflon were unable to transmit those acoustic waves along the PET film and ultimately did not work. For more discussion, please see Sec. 3.1.4. 2.3 Variable Angle Wedges Variable angle wedges were also designed and fabricated to test the effect of wedge angle on Lamb wave signal strength and to confirm the critical angle for Lamb wave 35 propagation. Variable angle wedges also allow for the selection of different Lamb modes as demonstrated by Rogers [34]. Wedges were machined from Teflon as shown in Fig. 2-3, and from ABS as shown in Fig. 2-4. Figure 2-3: Variable angle wedge machined from Teflon. Wedges are composed of a circular element that is free to rotate inside a larger wedge base. Note that the wedge acoustic wave exit location is angle dependent. otate ineto Variable 0 Carriage Fron Vie Figure 2-4: Variable angle wedge machined from ABS. Wedges are composed of a carriage that rides on a round base, allowing the wedge acoustic wave exit location to be independent of the angle. Carriages are secured to a laser cut acrylic mount which allows for the accurate selection of angle. Note that the fundamental wedge design varies between the two causing different properties, the most significant of which is the wedge acoustic wave exit location. The location that the acoustic wave exits the Teflon wedges is dependent on the wedge angle, while the ABS wedges use a remote center of motion mechanism to allow the 36 wedge acoustic wave exit location to be constant. This gives the Teflon wedges an undesirable property in which the same wedge spacing for different wedge angles will require a wave to propagate over different distances in the sample. It also limits the minimum possible propagation distance in the sample for large 0. Similar to the single angle wedges in Sec 2.2, the variable angle wedges were designed such that the near field effects would be completely contained within the wedges, with NTeon = 29.0 mm and NABS = 18.0 mm. However, the near field distance was not completely contained within the rotating components of the wedges for ease of manufacturing and to prevent the wedges from becoming too large. It is our hypothesis that having the near field interrupted by the brief discontinuity in the material would not have a significant effect on signal shape, although this was never tested. Drawings of the ABS variable angle wedge can be found in Appendix C. 2.4 Ultrasound Gel Wedges Attempts to generate Lamb waves using the ABS and Teflon wedges proved to be ineffective and the primary reason for this was determined to be a wedge material longitudinal speed of sound that was still too high, preventing the excitation of slow enough phase velocities in accordance with Eq. 1.10. Liquid wedge materials were pursued for their generally lower longitudinal speeds of sound when compared with solid materials. Initially considered liquids included water, with a speed of 1480 m/s, and ultrasound gel, with a speed of 1580 m/s. Ultrasound gel was selected on the basis that it was more viscous than water and thus easier to contain and prevent leaks. To create the ultrasound gel wedges, a 3D printed ABS hollow support structure was fabricated, and a 12 pm thick PET film was glued to the bottom. The support structure was then filled with Parker Laboratories Aquasonic Clear ultrasound gel. An Image of the completed ultrasound gel wedge can be found in Fig. 2-5. The PET film bottom prevents the ultrasound gel from escaping the wedge and allows the ultrasound signal to pass through with limited attenuation. PET was chosen as 37 the bottom film material to match well with the target PET sample. These simple manufacturing methods unfortunately did not prevent the occurrence of air bubbles in the ultrasound gel. Air bubbles are very undesirable as they cause ultrasound waves to reflect and can significantly interfere with a propagated signal. The bubbles may be removable with the use of a vacuum chamber but unfortunately access to one was not available at the time. Figure 2-5: Image showing the bottom of an ultrasound gel based wedge. A 3D printed ABS support structure contains the ultrasound gel while a 12 Pm thick PET layer glued to the support base prevents the gel from leaking out. Notice the abundance of bubbles present in the gel. 2.5 Hydrogel Wedges To eliminate the problem of bubbles interfering with the ultrasound signal and the need for a solid thin film wedge bottom, an experimental hydrogel from MIT's Soft Active Materials Laboratory was selected to be the wedge material [21, 40, 43]. With a longitudinal speed of sound around 1540 m/s, the hydrogel has similar acoustic properties to that of water and ultrasound gel, but is also able to remain solid at room temperature. To create the gel, a liquid pre-gel solution, consisting of chemical 38 components including alginate and acrylamide, is mixed together and cured using ultraviolet light to form a Polyacrylamide-alginate (PAAm-alginate) hydrogel. For additional details on the gel forming process, please refer to Lin et al. [21]. The liquid pre-gel solution allows the gel to be cast in molds bubble free. For this process to be successful, however, the molds themselves must be partially transparent. The molds, shown in Fig. 2-6, are made of a 3D printed ABS base and laser cut acrylic sides. The ABS base was smoothed using an acetone vapor bath smoothing process, described in Sec. 2.5.1, to prevent the 3D printed ridge pattern from appearing on the cast hydrogel wedges. Once cured, the hydrogel wedges were placed in a 3D printed support structure that also held the transducers in place, as seen in Fig 2-7. Figure 2-6: Mold with UV cured hydrogel wedge. Mold is composed of a 3D printed ABS base and laser cut acrylic sides. Note that although there visible bubbles along the mold surface, the hydrogel itself is essentially bubble free. Two geometries of hydrogel wedges were considered. One where the ultrasound beam exits at the wedge tip edge, and one where the ultrasound beam exits such that the entire beam within 6 dB of the beam maximum exits through the wedge bottom surface, determined with a beam spread of a = 70 from Eq. 1.12. These two geometries are illustrated in Fig. 2-8 and both fit into the same ABS support structure. Drawings of both the edge exit and surface exit hydrogel wedge molds can be found in Appendix C. 39 Figure 2-7: Hydrogel Wedge in ABS support structure. Transducer Wedge Transducer Wedge -6 dB -6 dB Surface Exit Edge Exit Figure 2-8: Edge exit vs. surface exit wedge geometry. In edge exit, the center of the ultrasound beam exits at the wedge tip, whereas for surface exit, the beam center exits before the wedge tip and the entire beam within 6 dB of the beam center exits through the wedge bottom surface. 2.5.1 Acetone Vapor Bath for 3D printed ABS Smoothing Parts of the Hydrogel wedge molds were printed in ABS using a Stratasys uPrint SE fused deposition modeling (FDM) 3D printer. Even when printed at the highest resolution, the limitations of the FDM process leave tiny ridges along the part surface that are undesirable for wedge molds and would leave an uneven wedge coupling surface. An acetone vapor bath was used to smooth the ABS surface. Acetone vapors are able to slowly melt ABS, allowing the surface ridges to flow and settle on the surface, smoothing it without significantly compromising dimensionality. 40 Some acetone vapor bath methods heat the acetone to increase the production of vapors, speeding up the smoothing process and allowing the part to be smoothed on the order of minutes. The smoothing process can be accomplished without the added heat as acetone will still produce vapors at room temperature, but the process takes longer to complete, on the order of hours. The process used for smoothing the molds can be found in Fig. 2-9. Acetonee Closed Figure 2-9: Acetone vapor bath used to smooth surface of 3D printed ABS parts. Liquid acetone rests on the bottom of the closed container, releasing vapors. The 3D printed part rests on a stand to prevent it from coming in contact with the liquid acetone. A small amount of acetone is poured into a closed container. The 3D printed part is then placed on a stand to keep it out of contact from the liquid acetone. Sealing the container traps vapors from the acetone, slowly smoothing the part. A smoothing time of 1.5 hours seemed to work well for the molds, whereas 2 hours caused the molds to lose some dimensionality and round part edges, as shown in Fig. 2-10. Also of note is that there seems to be a gradient of vapor concentration in this method, causing the bottom of the part to smooth faster than the top. This is especially a problem for smoothing relatively tall parts. Lining the sides of the container with a liquid absorbing medium such as paper towels can help to distribute the vapors more 41 evenly throughout the container, although such a technique was not necessary for this project given the relatively short parts. Once the parts have been appropriately smoothed, they are removed from the container and must set to allow remaining traces of acetone to evaporate, "drying" the surface. Touching the parts while the surfaces are still soft can leave fingerprints and other undesirable marks. 00000 010 0000 Figure 2-10: Comparison between wedge molds of similar sizes that were exposed to the acetone vapor bath for 1.5 hours (top) and 2 hours (bottom). Molds are similar but not of identical design, as the edge exit mold (top) and surface exit mold (bottom) differ in wedge tip geometry. Note that the mold with longer exposure has much greater radius of curvature along the edges and especially around holes. 42 2.6 Wedge Mounting Fixtures Two types of mounting hardware were designed and constructed for the ultrasound wedges. The first, shown in Fig. 2-11, includes springs that preload the wedges into the PET substrate with constant force. This mechanism allows the preload applied to the wedges to be more consistent across trials. Additionally, the thumbscrews enable the amount of applied preload to be adjusted. This feature allows the effect of wedge preload on the generated and propagated ultrasound signal to be investigated. Figure 2-11: Mount designed to investigate the effect of acoustic wedge preload on PET substrate. The second mount, shown in Fig. 2-12, was designed to vary the distance between wedges in fine increments. This allows the effect of wedge spacing on the propagated ultrasound signal to be investigated. In particular, this reveals the effect of dispersion, as discussed in Sec. 3.3. The mount uses a 1/4" - 20 threaded rod to allow for 0.05" (1.27 mm) of movement for each full turn of the knob. 43 continuFigure 2-12: Mount designed to vary the spacing between acoustic wedges ously. 2.7 Benchtop Web Tensioner a rollA benchtop web tension setup, shown in Fig. 2-13, was constructed to simulate A to-roll system and allow for simple tensioning of the PET web in lab experiments. difficult to ratchet mechanism is used to pull the PET taught. The actual tension is the web finely control with the mechanism, and the tensioner is simply used to pull to support taught until the roller constraints begin to slip. This is enough tension the wedges' weight without significant sagging in the web. PET Film Fo:erp moves with ratchet mechanism Roller tFixed zeE g Pw Ratchet Mechanism Close-Up Ratchet Mechanism under Figure 2-13: Benchtop tensioner constructed to perform experiments on PET taught. tension. Tensioner utilizes a ratchet mechanism to hold the PET 44 Zebra Printer 2.8 A Zebra thermal transfer printer (11OXi4), shown in Fig. 2-14, was purchased to print patterns on PET film. The printer, which is primarily used for printing barcodes, uses a thermal printhead to melt "ink" off of a ribbon and fuse it onto a substrate. Various Zebra ribbon formulations were tested on PET, including wax (5319), wax/resin mix (5586), and resin (5095) of which the resin ribbon showed the best performance in terms of smallest printable features and adhesion. The stated resolution of the printhead is 24 dots/mm, but the physical limitations of the thermal transfer process prevents the printing of single pixel sized dots. The smallest printable features were found to be 2-4 times larger than the resolution. Figure 2-14: Zebra thermal transfer printer used for printing patterns on PET film. An example printed hollow square grid pattern is shown in Fig. 2-15. The hollow squares are roughly 420 pm wide or about 10x larger than the stated resolution. Any smaller than that and the printer is unable to prevent the hollow portion of the pattern from being filled in. Also note the printing defects that arise near the pattern edge. Fig. 2-16 shows an example dot grid pattern imaged at 20x magnification. The dots are roughly 336 jim in diameter. The large magnification factor reveals the printhead interaction with the ribbon. The black portions that comprise most of the dot area have corners that indicate the individual pixels of the printhead. Circles are 45 Figure 2-15: Example hollow square grid pattern printed with the Zebra thermal transfer printer. Hollow squares are roughly 420 pum wide. PET substrate was printed on left to right. Figure 2-16: Example dot grid pattern printed with the Zebra thermal transfer printer imaged at 20x magnification. Dots are roughly 336 Mm in diameter. 46 difficult to print at this scale due to the limited number of pixels available. The gray border on the dots is likely residual resin ink from the ribbon that was not thermally fused to the PET. The consistent accumulation of the resin ink to the right of the dots indicates directionality involved in the printing process. The large accumulations were likely the last parts of the resin ink to separate from the ribbon, indicating that the PET was printed on left to right. The printer communicates to the computer through the Zebra Programming Language (ZPL) II. ZPL II has very limited capabilities, and all patterns must be specified by either the position and size of rudimentary shapes or as binary images compressed into ASCII Hex format. Matlab scripts were written to compose ZPL code capable of producing the desired patterns. The primary motivation for purchasing the printer was to print circuit mimicking patterns on long strips of PET for simultaneous optical inspection experiments. Another motivation was to use it as an already fabricated roll-to-roll printing machine that could be used to print patterns and control the web in an integrated set up. Additional sensors could also be installed to experiment with real time process control and defect detection. A CAD concept of such a set-up can be found in Fig. 217. Unfortunately, there was not enough time to pursue this fully integrated set-up, although it would be interesting for future research. Some thermal transfer printers have already seen some use for printing circuits. T M , a line of conductive thermal transfer A company called iimak offers Metallograph ribbons. The ribbons have conductive particles mixed into the ribbon ink, which are then sintered together during the thermal transfer process. Thus, the printer could also be used in future research to print large scale circuit samples for experiments. 47 Wide View Camera High Magnification Camera on Linear Stage ..... .......... .. .......... .... ..... ...... .. N Rollers Figure 2-17: CAD concept of fully integrated Zebra printer based roll-to-roll manufacturing and inspection set-up. Note that this concept uses two cameras, one wide view with backlight to locate areas of interest and one with high magnification on a translation stage to investigate areas of interest. Sensing is not limited to the optical domain and ultrasonic sensors could also be utilized in a similar set-up. 48 Chapter 3 Experiments This chapter summarizes experiments conducted for this research. All experiments utilized 76 pm thick PET and 1 MHz transducers. 3.1 Plastic Wedge Benchtop Experiments This section summarizes benchtop experiments conducted with Acrylic, ABS and Teflon wedges. Unless otherwise specified, all experiments were conducted in a through transmission method, where a transducer and wedge pair propagate a signal through the PET sample that is measured by a second wedge and transducer pair. 3.1.1 Acrylic Wedges Olympus acrylic wedges (ABWML-7T-90) were used to test if the PET film could carry an ultrasound signal of observable amplitude. Although designed for propagating SAW through steel, the acrylic wedges were able to transmit a weak longitudinal signal when spaced 1.27 cm apart. This was a quick but promising result that suggested that wedges, if better matched for PET and made from materials with slower speeds of sound, may be able to excite the desired Lamb waves. 49 3.1.2 ABS Single Angle Wedges Custom machined ABS single angle wedges were the first custom wedges evaluated on PET. The ABS wedges were able to generate a strong longitudinal signal through PET up to distances of about 5.08 cm. The pulse echo method was also attempted with an ABS wedge, but there were too many reflections inside of the wedge itself making it impossible to measure what signal, if any, was propagating into the PET and reflected back from the PET edge. Fig. 3-1 shows the transmission setup. Transmission signals were captured from 20-100 ps after the initial excitation pulse at a sampling rate of 125 MHz. Ultrasound gel was used as the couplant in the transducer-wedge and wedge-PET interfaces. An example transmission signal is shown in Fig. 3-2 for a wedge separation of 1.27 cm. The continuous spacing wedge mount was used to vary the spacing between wedges in increments of 1.27 mm from 0.26-3.17 cm. For each value of wedge spacing, 100 separate transmission signals were captured and averaged together to improve the signal to noise ratio. Figure 3-1: Single angle ABS wedges used in experiment on the benchtop tensioner in a through transmission configuration. To better visualize the different types of waves and compare signals acquired at different wedge separation distances, the transmission data was processed and various wedge spacings were stacked together to create a transmission signal map, shown in Fig. 3-3. Signal processing consists of taking the magnitude of the Hilbert transform of each signal, a method of envelope detection [37] that helps to visualize the location 50 Transmission Signal for ABS Wedge Separation of 1.27 cm 1.5 Longitudinal 1 0.5 1- 0 Hardware Disturbance 0 c -0.5 Longitudinal Reflection - Iip - -1 - -1.5 25 I I Ii 30 35 40 j I 50 45 Time (us) 60 55 65 70 Figure 3-2: ABS single angle wedge average of 100 through transmission signals for wedge spacing of 1.27 cm. Processed Transmission Signal Map for ABS Wedges 0 3 -1 2.5 -2 -3 c 2 1.5 i0) C .r. -5 -6 0.5 20 -7 30 40 50 60 Time (us) 70 60 90 Figure 3-3: ABS wedge transmission signal map created by varying the spacing between wedges in increments of 1.27 mm. Greater magnitude indicates more signal activity in that region. Expected locations of various acoustic waves are overlayed onto the map. 51 of the oscillating wave, and then taking the natural log of the resulting signal to better visualize lower energy signals. Last, the map magnitudes are linearly shifted such that the greatest magnitude is zero. Brighter areas on the map indicate a stronger wave presence. Expected locations of various acoustic waves in the PET are overlayed onto the map. These expectations are calculated using stated material properties and geometry of the ABS wedges and PET sample. The longitudinal reflection represents the wave that either propagates through the length of the transmitting wedge three times or propagates through the length of the receiving wedge three times due to a reflection at the wedge tip and transducer-wedge interface and travels through the PET as a longitudinal wave. Comparing the overlayed expected waves with the transmission signal map, it is clear that the strongest signal in the PET corresponds with the longitudinal wave. Also, looking at the slope of the energy in the middle of the map, the second strongest signal is most likely the reflected longitudinal wave. The thin vertical line around 50 1ps is actually a very short high frequency wave caused by the data acquisition hardware that also appears when the wedges are not in contact with the web. Noticeably absent is any indication of a symmetric or antisymmetric Lamb wave. The overylay also makes it apparent how difficult it would be to differentiate a symmetric Lamb wave from the longitudinal wave at these length scales. Looking at the transmission signal map makes it easier to then go back to individual through transmission signals and identify the different wave components. Thus, in Fig. 3-2, the wave around 29 ps is the longitudinal wave and the wave around 55 [is is the reflected longitudinal wave. The wave around 50 [is is the wave due to the hardware disturbance. Frequency based analysis can also be performed on the transmission data by creating a frequency-time decomposition map to visualize wave dispersion at different excited frequencies. These maps decompose a through transmission signal into individual frequency components and show the relative strength of these components throughout the signal. This allows the speed of waves at specific frequencies to be determined, leading to the measurement of sample dispersive characteristics. 52 Frequency-time decomposition was performed on each through transmission signal by convolving it with a test wave consisting of a single frequency five period sinusoid subject to a normal distribution envelope. The magnitude of the Hilbert transform is used to indicate the envelope of the convolved signal, for which greater magnitudes indicate a greater presence of the initial test frequency. This process is then repeated with different test frequencies and the resulting signals are stacked together to form a frequency vs. time map. A cutoff magnitude is then selected, above which all magnitudes are considered the same value, to better visualize smaller magnitude features. Last, all magnitudes are linearly shifted down such that the greatest magnitude is zero and linearly shifted in time such that only the time the waves spend propagating through PET is considered, removing the time spent propagating through the wedges. Theoretical group velocity curves are then modified to find time of flight using wedge spacing and overlayed onto the frequency-time decomposition for comparison. Frequency-Time Decomposition for ABS Wedge Separation of 1.27 cm i / 0 Longitudinal Wave -0.005 -Symmetric Lamb S- 6 -- -0.01 Antisymmetric Lamb -0.015 5 -0.02(0 T- -0.025 -0.03 3 M L- -0.035 2o 2 -0.04 2 0 -- 0.045 1 0 5 10 15 25 20 Time in PET (us) 30 35 40 Figure 3-4: Example frequency-time decomposition for single angle ABS wedges spaced 1.27 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. An example frequency-time decomposition can be found in Fig. 3-4 for a wedge separation of 1.27 cm. The presence of the longitudinal wave is very strong, and it 53 is difficult to determine whether the symmetric Lamb wave is also present. There is also no indication of an antisymmetric Lamb wave. These results agree with those from the transmission signal map. 3.1.3 ABS Variable Angle Wedges Custom machined variable angle wedges were used to investigate the effect of wedge angle on the propagated signal and to confirm the optimal angle for Lamb wave generation. The experimental setup is shown in Fig. 3-5. Wedges were separated at a fixed distance of 2.54 cm with data acquisition settings the same as in Sec. 3.1.2. The variable angle wedges allowed the wedge angle to be varied between 30'-90' with respect to horizontal in 5' increments. The same wedge angle was used for both the transmitting and receiving wedges in all experiments. Ultrasound gel was used as the couplant in the transducer-wedge, carriage-base, and wedge-PET interfaces. An example through transmission signal is shown in Fig. 3-6. Figure 3-5: Variable angle ABS wedges used in experiment on the benchtop tensioner. The results from various propagation angles look deceivingly promising with many more waves propagated than in experiments with the single angle wedge. However, the variable angle wedges are now composed of a separate carriage and base, creating the potential for an additional reflection at the carriage-base interface. Where reflections in the single angle wedges could only take one path, back and forth through the 54 Transmission Signal for Variable Angle ABS Wedge at 30* Longitudinal 1.5 Reflection in Base 1 0.5 F 0) 0 Cu -0.5 Reflection in Full Wedge -1 Reflection in Carriage -1.5 -220 40 30 50 60 Time (us) 70 80 90 100 Figure 3-6: ABS variable angle wedge average of 100 through transmission signals for wedge angle of 300 with respect to horizontal. Note the large number of individual waves when compared to Fig. 3-2. Variable Angle ABS Wedge Wave Arrival Simulation 6 - Longitudinal Wave --- Symmetric Lamb Antisymmetric Lamb Longitudinal Reflections 4 -2 ___ V V Animen Lamb V U) 2 -4 -6L 20 I I I I 30 40 50 60 70 80 90 100 Time (us) Figure 3-7: Simulation showing the expected locations of the various possible waves propagated by the ABS variable angle wedges. Note that signal attenuation was not considered and equal magnitudes of the different waves is not a serious result. From left to right, the various signals are the longitudinal wave, the symmetric Lamb wave, the back and forth reflection in the wedge carriage only, the back and forth reflection in wedge base only, the antisymmetric Lamb wave, and the back and forth reflection through the entire wedge. 55 entire wedge, reflections in the variable angle wedge have three potential paths that can be measured by the receiving transducer. These reflections can occur in either the transmitting or receiving wedge and include reflecting back and forth in the entire wedge, only in the carriage, and only in the base. Fig. 3-7 shows a simulation of the location of the various expected waveforms, including the longitudinal wave, Lamb waves, and reflected longitudinal waves. The simulation is created using the stated material properties of the ABS wedges and PET. Comparing the simulation in Fig. 37 to the data in Fig. 3-6, it becomes clear that all measured waves can be attributed to longitudinal waves or longitudinal reflections. This was a common pattern for all of the tested angles, although the individual wave magnitudes varied. A transmission signal map, shown in Fig. 3-8, can be created for the variable angle wedge, but with varying wedge angles instead of wedge spacing. All waves in the map now appear as vertical lines instead of diagonal lines because the wave propagation distance is independent of the wedge angles. With good agreement between the expected and measured wave locations, the transmission signal map further illustrates the absence of measured Lamb waves. Frequency based analysis also suggests that little dispersion is going on. An example frequency-time decomposition can be found in Fig. 3-4 for wedges at an angle of 300. The presence of the longitudinal wave is once again very strong. There is very little energy along the symmetric Lamb group velocity curve, although there are some high energy areas that intersect with the antisymmetric curve. However, because these regions are very vertical and show little dispersion, they are most likely due to longitudinal reflections as opposed to antisymmetric Lamb waves. These results agree with those from the transmission signal map. These results suggest that despite having a lower longitudinal speed of sound than acrylic, ABS is still not a practical material for propagating Lamb waves in PET. However, in retrospect, given that later results showed limited Lamb wave propagation distances, 2.54 cm may also be too wide a separation for realistically detecting Lamb waves. It would be worth redesigning the variable angle wedges such that they only cover angles between 30'-60' and could be placed closer together. Additionally, 56 Processed Transmission Signal Map for Variable Angle ABS Wedges 0 90 -0.5 -1 80 -1.5 70 -2 F -2.5 c60 -3 -3.5 C 50 -4 t* -4.5 6. 40 -5 -5.5 Time (us) Figure 3-8: Variable angle ABS wedge transmission signal map created by varying the wedge angle in increments of 50 between 30'-90' with respect to horizontal. Greater magnitude indicates more signal activity in that region. Expected locations of various acoustic waves are overlayed onto the map. Frequency-Time Decomposition for Variable Angle ABS Wedge at 300 / 7 R - - 6 -- - -0.02 Symmetric Lamb Antisymmetric Lamb -0.0412 / - Longitudinal Wave I -0.06 5 -0.086 4 -0.1 \1 lb U_ 93 2 1 0 10 20 -0.12 -0.14 I 40 30 u. -0.16 a. -0.18 50 60 Time in PET (us) Figure 3-9: Example frequency-time decomposition for variable angle ABS wedges at 300 and spaced 2.54 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. 57 a coupling medium that is better matched to ABS could reduce reflections in the carriage-base interface, allowing more energy to go into other transmission modes. 3.1.4 Teflon Wedges Custom machined Teflon wedges were evaluated for Lamb wave excitation in PET. Teflon was expected to perform better than ABS due to its slower speed of sound but this was not the case. Surprisingly, both single angle and variable angle Teflon wedges were not able to propagate any ultrasonic waves through PET, not even longitudinal waves for small wedge separation distances. The wedges were however able to propagate the longitudinal wave through themselves and even through both wedges sandwiching the PET film, a configuration shown in Fig. 3-10. It seems that as soon as the longitudinal signal has to refract in PET, it attenuates. This was a surprise because the Teflon wedges were able to generate a signal through a steel sample and Teflon wedges have previously been applied to inspect wood-plastic composites [41]. One potential reason could be a mismatch in acoustic impedance between Teflon and PET but the root cause was not investigated further. Transducer Wedge PET Film Figure 3-10: Configuration with Teflon wedges that was able to propagate measurable ultrasound waves. 58 3.2 Instron Experiments An InstronTM machine (Instron 5869) was used to conduct tensile experiments on PET with the goal of evaluating how ultrasound could be used to measure tension locally for process control in a roll-to-roll system. Fig. 3-11 shows a schematic of the. experimental setup while Fig. 3-12 shows the actual experimental setup with the various hardware components. The Instron grips the PET film in its jaws and pulls to apply tension to the web in a controlled fashion while the sample is also monitored by ultrasound wedges. The Instron allows for the measurment of upper jaw displacement and applied axial force. Ultrasound gel is used to couple the wedges to the PET film. The PET samples used for these experiments were 10.8 cm wide and had 21.6 cm lengths that were exposed between the Instron jaws. The PET sample is also simultaneously monitored optically for a separate experiment working toward multimodal sensing, which accounts for the red light in Fig. 3-12. The spring loaded mount is used to maintain a set preload on the wedges while PVC idler rollers help to simulate roll-to-roll boundary conditions on the system. The original intent of the Instron experiments was to explore what effects tension in the PET web has on ultrasonic Lamb waves and to confirm the results from Desmet et al. [11] utilizing contact ultrasound. Unfortunately, at the time of Instron availability, the best working wedges were the single angle ABS wedges which are only able to excite longitudinal waves. However, Instron experiments showed that the longitudinal wave is also affected by tension in the PET web and may have potential for generating useful measurements in a manufacturing setting. One effect web tension has on the longitudinal wave is signal attenuation. The results of a stretch and relax experiment, presented in Fig. 3-13, show that the peak of a signal's power spectrum decreases reversibly as the PET sample is stretched. With a linear fit R 2 value of 0.93, this relationship can roughly be approximated as linear for the range of stretch forces considered. The PET was stretched elastically at a rate of 0.1 mm/min from negligible tension to 30 N as measured by the Instron and then slowly released. Note that although related, force measured by the Instron is not an 59 IStretcl h PET Wedge Thumbscrew 0.t 2 Fixed Spring Fixed Figure 3-11: Schematic showing the experimental setup on the Instron machine. Transmission signal through the PET travels out of the page. Figure 3-12: Experimental setup on the Instron Machine. Note the use of the spring loaded mount to apply preload on the wedges and the rollers to mimic roll-to-roll boundary conditions. 60 exact measure of tension in the PET web due to the angle of the PET with respect to the Instron jaws, the change in that angle as the PET is stretched, and friction within the rollers. It is however a close enough approximation to begin to see some trends. An ultrasound transmission signal was taken roughly every 1 N of additional tension. A fast Fourier transform (FFT) is applied to the signal to obtain its power spectrum, of which the maximum is used as an indicator of signal attenuation. Originally, the maximum amplitude of the transmission signal was used to look at attenuation, but the FFT power spectrum method proved to be less affected by signal noise. Later experiments also averaged 100 consecutive captures of the transmitted signal to reduce signal noise. Instron PET Stretch and Relax Experiment 0 o 1.4 - 0 1.35- ~ - 1.3 1.25 (L0 - 2 o Stretching Relaxing 2 Linear Fit (R=0.93) - 1.45 0 1.2 - E 2 c 1.15- O S 1.1J0 1.05 0 5 10 15 Stretch Force (N) 20 25 30 Figure 3-13: Instron stretch experiment of 76 pm thick PET monitored by ABS ultrasound wedges spaced 4.45 cm apart. Figure shows that the maximum of the ultrasonic transmission signal power spectrum decreases reversibly as the PET sample is stretched. PET was stretched at a rate of 0.1 mm/min. The underlying cause for the longitudinal signal to attenuate with tension is unclear. One possibility is that the tension applied to the polymer causes the polymer strands to become more oriented and preloaded in a way that makes them less flexible and able to move in other directions. However, more experiments will be necessary to confirm this hypothesis. Regardless of the underlying cause, this attenuation relationship could be applied towards measuring local tension in the web for process 61 control. Contact ultrasound often involves a non-zero force keeping the transducers and wedges in contact with the sample. Following work done by Gilbertson [15], who found that the elastic and thus acoustic properties of soft tissue are affected by the ultrasound probe contact force applied by the sonographer, significantly affecting elastic modulus measurements. It was thus hypothesized that the wedge contact force may have an effect on the propagated signal. To investigate this potential effect, experiments were conducted where the wedge preload into the PET was varied using the spring loaded mount described in Sec. 2.6. Instron experiments were then conducted at various preloads for wedges spaced 0.64, 4.45, and 5.08 cm apart. Wedge preload is calculated by measuring spring compression with calipers and using the spring constant to calculate force. Wedge separation is measured from the wedge beam exit location which for the ABS wedges is at the tip. The PET sample is replaced after each tensile test for every new wedge preload and separation pair. PET samples were stretched such that the Instron jaws were displaced at a rate of 5 mm/min from 0 to 30 mm and ultrasound transmission data was taken every 0.5 mm of stretch. The rate of stretch is significant because the PET does noticeably relax during the experiment. Note that like tension, the displacement of the Instron jaws is not an exact measure of the stretch of the PET due to the angle of the PET with respect to the Instron jaws, and does not even have a linear relationship due to the change in that angle as the PET is stretched. It is however also a close enough approximation to begin to see some trends. Force data was recorded in Instron experiments and Fig. 3-14 shows a typical force profile. Note that the actual forces applied to the web are much larger than those in Fig. 3-13, and on the order of 800 N for 30 mm of stretch, reaching well into the plastic region of PET. Transmitted ultrasound signals were processed for their power spectrum maximum, shown with respect to stretch length in Fig. 3-15 and stretch force in Fig. 3-16. The variability of the power spectrum magnitudes with different preloads suggests the importance of maintaining a repeatable preload in an inspection system. The plots 62 PET Instron Stretch Experiment Force Profile - 900 800700- - 600 2 500 ~400CD300200100- 00 5 10 15 Stretch Length (mm) 20 25 30 Figure 3-14: Representative force profile for Instron stretch experiment of 76 pm thick PET. Note the elastic region on the left and the plastic region on the right. PET was stretched at a rate of 0.5 mm/min. also show the general decrease in power spectrum peak magnitude with web tension and stretch force. This trend becomes less clear for the wedge separation of 5.08 cm, especially with respect to stretch force. This suggests an upper limit on practical wedge separation for process control. Interestingly, data taken with wedge separation distances of 0.64 and 5.08 cm show a general decrease in power spectrum peak with an increase in preload whereas data taken with a wedge separation distance of 4.45 cm shows the opposite trend. One potential cause of this discrepancy are variations in the ultrasound gel and coupling interface between the wedges and the PET. The ultrasound gel was difficult to apply in a consistent manner and variations in the ultrasound gel interface were observed to have a significant effect on the propagated signal magnitude. It is possible that the gel coupling interface more significantly contributes to the signal magnitude than preload, causing a scaling factor of signal magnitudes in these experiments to be essentially random. Additional experiments are needed to clarify the exact effect preload has on the transmitted signal magnitude. One potential way of improving the repeatability of the coupling interface is to use water as a couplant instead of gel. Gel was chosen for these experiments because the PET sample was oriented vertically and there were concerns that water, being 63 PET Instron Stretch Experiment 0.64 cm Wedge Spacing 12 -___5.13 N Preload - - - 9.82 N Preload 10 8 E 6 CL 4 eO 0 a_ 2 ------------------------n i 10 5 0 25 20 15 Stretch Length (mm) 30 PET Instron Stretch Experiment 4.45 cm Wedge Spacing 14 a) E CL 12 -- 10 ----- 0- 8 - 3.66 N Preload 4.83 N Preload 8.94 N Preload -\ 0. 0) 4 0 0~ C 10 5 0 25 20 15 Stretch Length (mm) 30 PET Instron Stretch Experiment 5.08 cm Wedge Spacing 5- -- 6.01 N Preload 11.87 N Preload CU 0) a- E U 1N 3:0. 0 a 5-~ . N I 0 5 10 15 Stretch Length (mm) - -- 5 20 25 30 Figure 3-15: Instron stretch experiment of 76 pm thick PET monitored by ABS ultrasound wedges spaced 0.64 (top), 4.45 (middle), and 5.08 cm (bottom) apart. Figures show the peaks of the power spectrum as a function of stretched length. PET was stretched at a rate of 0.5 mm/min. Note that different amplifier gains were used for each wedge separation, preventing the comparison of power spectrum peak values of different wedge separations. 64 PET Instron Stretch Experiment 0.64 cm Wedge Spacing 12 10 0 E 2 6 82a 5.13 N Preload 9.82 N Preload --- 300 200 100 0 500 400 Stretch Force (N) 700 600 800 90C PET Instron Stretch Experiment 4.45 cm Wedge Spacing 14 - 12 .10 N Preload N Preload 8.94 N Preload ___3.66 - -4.83 cc 4) -N C, 0 0- 0 0 PET Ca 300 200 100 500 400 Stretch Force (N) Instron Stretch Experiment 600 5.08 cm Wedge 700 800 901 Spacing I--.. I E a) 5 -- CD, 0. 3: 0 d- 6.01 N Preload 11.87 N PreloadN 0 100 200 300 500 400 Stretch Force (N) 600 700 800 900 Figure 3-16: Instron stretch experiment of 76 ttm thick PET monitored by ABS ultrasound wedges spaced 0.64 (top), 4.45 (middle), and 5.08 cm (bottom) apart. Figures show the peaks of the power spectrum as a function of stretch force. PET was stretched at a rate of 0.5 mm/min. Note that different amplifier gains were used for each wedge separation, preventing the comparison of power spectrum peak values of different wedge separations. 65 much less viscous than the gel, would flow from the coupling interface during the experiment. However, if the experimental hardware was altered to allow the PET to be oriented near horizontally, a constant cycling flow of water could be used to reduce variability in the coupling interface. In addition to looking at the power spectrum peak magnitudes, the time of flight arrival of the three largest peaks of the Instron experiment transmission data was also analyzed for changes in propagating wave speed or frequency. A peak fitting algorithm [29] was used to estimate the time of flight arrival of each peak. This data is presented in Fig. 3-17. Note that the peak fitting algorithm is not as effective at lower signal to noise ratios, causing errors for larger values of strain and wedge separation. The time of flight arrival data shows that preload has little to no effect the propagating wave speed or frequency. Additionally, the lines created by the peak arrival times remain parallel to each other throughout the experiment. This means that the frequency of the propagated longitudinal wave remains constant while the PET is being stretched. The wave speed, however, does vary with tension. Notice that the peak locations appear to have two distinct slopes, with a linear region between about 0-10 mm of stretch and another linear region between about 10-30 mm of stretch. The 10 mm transition area could correspond with the transition from elastic to plastic strain, shown in Fig. 3-14. The positive slope of the peak locations during high values of strain is also interesting because it indicates that the longitudinal speed of sound is decreasing as the PET is plastically deformed. Since speed of sound can be calculated solely based on the material's Young's modulus, Poisson's Ratio, and density, as shown in Eq. 1.8, then the plastic deformation must be causing a decrease in Young's modulus, a decrease in Poisson's ratio, an increase in density, or some combination of the three. Desmet et al. [11] explains that stretching a polymer foil causes the polymer chains to align in a preferred orientation relative to the stretching direction, leading to an increase in the elastic stiffness along the axis of alignment. It is however difficult to determine whether the elastic stiffness perpendicular to the stretching direction, in the direction of ultrasound wave propagation, increases for 66 PET Instron Stretch Experiment 0.64 cm Wedge Spacing 29 28.5 -- 28 - 0 o 1st peak 5.13 N Preload 2nd peak 5.13 N Preload o 3rd peak 5.13 N Preload 1st peak 9.82 N Preload -__2nd peak 9.82 N Preload 27.5 27 265- 3rd peak 9.82 N Preload 26 25.5 25 0 5 10 15 20 25 35 30 40 45 50 Stretch Length (mm) PET Instron Stretch Experiment 4.45 cm Wedge Spacing 46 + 45 0 ++ ++ + 41 0 5 10 ++ 42+ + + + 15 ++ + +++ 4 - 401 ++ + 44- + + 4++ + + + 0 1st peak 3.66 N Preload 0 0 2nd peak 4.83 N Preload 3rd peak 4.83 N Preload 1st peak 8.94 N Preload 2nd peak 8.94 N Preload 3rd peak 8.94 N Preload 2nd peak 3.66 N Preload 3rd peak 3.66 N Preload 1st peak 4.83 N Preload + ++ + + + + 20 11 25 30 35 40 45 50 Stretch Length (mm) PET Instron Stretch Experiment 5.08 cm Wedge Spacing 48 - 47- 0 ___ 1st peak 6.01 N Preload 2nd peak 6.01 N Preload 3rd peak 6.01 N Preload 1st peak 11.87 N Preload ___ 2nd peak 11.87 N Preload 3rd peak 11.87 N Preload o 46- 0 45 44- 4ll3 0 5 10 15 20 25 30 Stretch Length (mm) 35 40 45 50 Figure 3-17: Instron stretch experiment of 76 pm thick PET monitored by ABS ultrasound wedges spaced 0.64 (top), 4.45 (middle), and 5.08 cm (bottom) apart. Figures show the time of flight arrival of the three largest peaks in the ultrasonic transmission signal as the PET sample is stretched. The peak locations are found using a peak fitting algorithm that is not as effective at lower signal to noise ratios, accounting for the large errors at higher strains and wedge separation distances. PET was stretched at a rate of 0.5 mm/min. 67 similar reasons, remains unchanged, or decreases due to the polymer being oriented in a perpendicular direction. It is common for Poisson's ratio to increase post yield, potentially accounting for the change in speed of sound. Alternatively, if Young's modulus does not decrease enough and Poisson's ratio does not increase enough, then the change in speed of sound must also be caused by an increase in density. Assuming conservation of mass, an increase in density either indicates that the PET width or thickness is decreasing enough to compensate for the increase in length. This could be due to tension induced denser packing of polymer chains. If the PET width changes negligibly, additional research could allow this trend to be used to monitor changes in PET thickness due to plastic deformation. 3.3 Gel Wedge Benchtop Experiments Once ABS and Teflon were determined to be ill suited towards Lamb wave excitation, gel materials were considered to serve as the wedge core due to their low longitudinal speeds of sound and easy handling. Both ultrasound gel and an experimental hydrogel were utilized to create gel based wedges and conduct through transmission experiments. 3.3.1 Ultrasound Gel Wedge The ultrasound gel wedges were used in experiments with the benchtop tensioner. Transmission signals were captured from 30-110 pis after the initial excitation pulse at a sampling rate of 125 MHz. Water was used as the couplant between the PET wedge base and the PET sample while the transducers were placed in contact with the ultrasound gel wedge core. Water was used because the adhesion between the PET base and the 3D printed ABS wedge support structure was poor and there were concerns that an ultrasound gel couplant would cause the PET base to stick and peel off. Four different wedge spacings were tested, including 0.64, 1.27, 1.91, and 2.54 cm. For each value of wedge spacing, 100 separate transmission signals were captured and averaged together to reduce the effect of random noise. 68 An example through transmission signal can be found in Fig. 3-18 for a wedge spacing of 0.64 cm. Transmission Signal for Ultrasound Gel Wedge Separation of 0.64 cm 3 2 1 76 0 CM, -1 -2 I I I I I 30 40 50 60 70 80 90 100 110 Time (us) Figure 3-18: Example through transmission signal for ultrasound gel wedges spaced 0.64 cm apart. Notice how all of the various waves seem to blend together much more than the through transmission examples of other wedge types. This is likely due to interference caused by bubbles in the gel. Specific waves in the ultrasound gel wedge through transmission data were very difficult to pinpoint when compared to through transmission signals from other wedge types. Once the receiving transducer begins to measure a signal, there is a lot of ringing and the signal never really settles down. Waves from various parts of the signal all seem to blend into each other. This is likely due to the numerous bubbles inside the ultrasound gel that reflect and interfere with the propagating signal. The transmitted signal in this form is simply too complex to be useful. However, as a wedge material, the ultrasound gel was able to propagate strong signals and a bubble-free wedge of a similar design could perform well. 69 3.3.2 Edge Exit Hydrogel Wedges To solve the bubble interference problem encountered with the ultrasound gel wedges, solid wedges made from a UV cured hydrogel were fabricated. One advantage of the hydrogel wedges is that they remain moist enough that they do not need a couplant to transmit signal. Experiments were not performed on the benchtop tensioner as in other cases and were instead performed on an unstressed sample of PET resting on a wooden desk. This configuration is shown in Fig. 3-19. The PET was not acoustically coupled to the desk. This represents a different boundary condition than in prior experiments, the consequences of which are discussed further in Sec. 3.3.3. Transmission signals were captured from 35-115 ps after the initial excitation pulse at a sampling rate of 125 MHz. An example transmission signal is shown in Fig. 3-20 for a wedge separation of 1.27 cm. The continuous spacing wedge mount was used to vary the spacing between wedges in increments of 1.27 mm from 0.26-2.54 cm. For each value of wedge spacing, 100 separate transmission signals were captured and averaged together to reduce the effect of random noise. Edge Exit Hydrogel Wedge Wedge flush with support structure face PTFilm on Wod Figure 3-19: Experimental setup with edge exit hydrogel wedges. Wedges rest on PET that sits on an acoustically uncoupled wooden desk. 70 Transmission Signal for Edge Exit Hydrogel Wedge Separation of 1.27 cm I V 1 1 5 Longitudinal 4- Symmetric Lamb? 3 2 -o Antisymmetric Lamb 0 Hardware to -Mounting Longitudinal Reflection -3 -440 50 60 80 70 90 100 110 Time (us) Figure 3-20: Example through transmission signal for edge exit hydrogel wedges spaced 1.27 cm apart. The through transmission signal achieved with the hydrogel wedges is fairly clean and easy to locate wave fronts and tails. To better visualize the different types of waves, the transmission data was processed and various wedge spacings were stacked together to create a transmission signal map, shown in Fig. 3-21. Signal processing consists of the same method described in Sec. 3.1.2. Expected locations of various acoustic waves in the PET are also overlayed onto the map. The overlayed expected waves match fairly well with the measured signals. The longitudinal, antisymmetric Lamb, and longitudinal reflected waves are all clearly represented in the measured data. Unfortunately, given the proximity of the ex- pected symmetric Lamb wave to the dominating longitudinal wave for small wedge separations, it is difficult to tell whether the symmetric Lamb wave appears in the data or not. Note that there is an additional diagonal line in the data that shows up in the middle of the map. This line is due to signal that propagates through the wedge mount and is apparent even when the wedges are not in contact with the PET. 71 2.4 2 E Processed Transmission Signal Map for Edge Exit Hydrogel Wedges .. Longitudinal Wave - - - -2 Symmetric Lamb Antisymmetric Lamb Longitudinal Reflection S1.8 *1.6 1.4 Q 12 -1 -2 -3C C - Q 0 -4 0D U) C CD 0) -5 "a M- 0 o 0.6 -7 0.4 40 50 60 80 70 Time (us) 90 100 110 incrementing the spacing between wedges. Expected Figure 3-21: Edge exit hydrogel wedge signal map created by slowly with the most gradual slope that does not have locations of various acoustic waves are overlayed onto the map. The signal line and is present when the wedges are not in contact an expected overlay is due to a wave propagating though mounting hardware with the web. C Understanding the transmission signal map allows for the identification of individual signals in the transmission data. Thus, in Fig. 3-20, the wave around 47 ps is the longitudinal wave, the wave around 60 ps is the antisymmetric Lamb wave, and the wave around 88 ps is the reflected longitudinal wave. The wave around 80 ps is the wave due to signal going through the wedge mounting hardware. Also, it is possible that there is a small symmetric Lamb wave around 50 ps although the features there could very well just be the tail of the longitudinal wave. Similar to Sec. 3.1.2, frequency-time decomposition maps can be created to visualize wave dispersion at different excited frequencies. An example frequency-time decomposition can be found in Fig. 3-22 for a wedge separation of 1.27 cm. The presence of the antisymmetric Lamb wave is apparent for frequencies in the vicinity of 1 MHz and the high energy portion of the map follows and shows agreement with the theoretical group velocity curve. The symmetric Lamb wave, if present, is difficult to distinguish from the longitudinal wave for the frequencies excited. These results agree with those from the transmission signal map. Frequency-Time Decomposition for Edge Exit Hydrogel Wedge Separation of 1.27 cm 7.0 / I /---Antisymmetric 6 Wave Symmetric Lamb Lamb -Longitudinal -- c. 5 -0.1 CO Cr U- -0.15 LL 3 2 -0.2 0 5 10 15 25 20 Time in PET (us) 30 35 40 Figure 3-22: Example frequency-time decomposition for edge exit hydrogel wedges spaced 1.27 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. 73 3.3.3 Surface Exit Hydrogel Wedges The surface exit hydrogel wedges consist of a slightly different geometry than the edge exit wedges, essentially allowing the entire acoustic wave within 6 dB of the center magnitude to propagate through the wedge bottom surface. Through transmission experiments conducted with the surface exit hydrogel wedges utilized similar experimental parameters as the edge exit hydrogel wedges in Sec. 3.3.2. The primary difference was that the wedge geometry prevented them from having as small wedge separation distances. For the surface exit wedges, wedge separation was varied between 0.64-3.18 cm in increments of 1.27 mm. Note that wedge separation here is measured from the beam center exit location which does not correspond to tip to tip distance for these surface exit wedges. An example transmission signal is shown in Fig. 3-24 for a wedge separation of 1.27 cm. Surface Exit Hydrogel Wedge Wedge protrudes from support structure face PFilm on Wood Figure 3-23: Experimental setup with surface exit hydrogel wedges. Wedges rest on PET that sits on an acoustically uncoupled wooden desk. Note that the tip of the hydrogel wedge protrudes from the ABS support structure, differentiating the surface exit wedges from the edge exit wedges in Fig. 3-19. The through transmission signal obtained with the surface exit wedges is similarly desirable to those obtained with the edge exit wedges, although there are some no- 74 Transmission Signal for Surface Exit Hydrogel Wedge Separation of 1.27 cm 5 I I Longitudinal 4- Symmetric Lamb? 3_ 2- Antisymmetric 1 Lamb - C0 -2 - Longitudinal Reflection -3.4-5 40 50 60 80 70 90 100 110 Time (us) Figure 3-24: Example through transmission signal for surface exit hydrogel wedges spaced 1.27 cm apart. table differences. Primarily, the magnitude of the longitudinal reflected wave is much smaller. This makes sense because more of the beam energy is able to propagate through the wedge bottom surface instead of getting reflected off of the front face. It also appears that most of this acoustic energy went straight into the longitudinal wave as it has more peaks at its maximum magnitude. Last, the signal that was propagating through the mounting hardware is no longer apparent. This is also likely due to the fact that less acoustic energy is reflected in these surface wedges. The similarity between the edge exit and surface exit wedges can be exploited to characterize the waves in Fig. 3-24. The wave around 47 ps is the longitudinal wave, the wave around 56 ps is the antisymmetric Lamb wave, and the wave around 88 ps is the reflected longitudinal wave. Also, once again, it is possible that there is a small symmetric Lamb wave around 50 ps although the features there could simply be the tail of the longitudinal wave. As mentioned in Sec. 3.3.2, the presence of the PET on the wooden desk represents 75 a different boundary condition from prior experiments. It can still be analogous to a roll-to-roll system if acoustic wedges contacted the PET where it rested on an idler roller. A comparison between portions of the transmission signals obtained with PET resting on the table and PET in the benchtop tensioner is presented in Fig. 3-25 for surface exit hydrogel wedges with a spacing of 1.27 cm. The longitudinal waves in both conditions are very comparable, with the tensioner wave starting a little later yet ending comparably, thus having slightly less energy. The antisymmetric Lamb waves, however, are separated by a few ps, with the tensioner wave arriving later and with less magnitude. Resting PET on the table thus proved to be the superior configuration for propagating antisymmetric Lamb waves. If this result is due to the tension in the web, the decrease in magnitude agrees with trends found in Sec. 3.2. The decrease in antisymmetric wave speed, however, is unexpected and contradicts findings by Desmet et al. [11] where antisymmetric wave speed increased with tension. Further research will be needed to clarify the reasons for the contradicting results. Transmission Signals for Surface Exit Hydrogel Wedge Separation of 1.27 cm 5 4 -PET on Table PET in Tensioner 3 C -2 -3 -4 -5 45 50 55 Time (us) 60 65 Figure 3-25: Comparison of through transmission signal in PET on the table and PET in the benchtop tensioner. Both signals were obtained using the same surface exit hydrogel wedges with a separation of 1.27 cm. 76 Processed Transmission Signal Map for Surface Exit Hydrogel Wedges Longitudinal Wave Symmetric Lamb 3 A 1 Antisymmetric Lamb Longitudinal Reflection 2.5 .U -3 C,, 2 C a) a) 2 -4 CL C -5 i 1.5 0 1 -7 40 50 60 70 80 90 100 110 Time (us) Expected Figure 3-26: Surface exit hydrogel wedge signal map created by slowly incrementing the spacing between wedges. locations of various acoustic waves are overlayed onto the map. A transmission signal map, shown in Fig. 3-26, can be created just as it was in Sec. 3.3.2. The transmission signal map is also very similar to that made with the edge exit wedges except the line created by the mounting hardware signal is no longer there and the antisymmetric Lamb wave is distinguishable for longer wedge separation distances. These factors suggest that the surface exit hydrogel wedge geometry is superior to that of the edge exit hydrogel wedge. Frequency-time decompositions can also be generated as they were in Sec. 3.3.2, an example of which can be found in Fig. 3-27 for a wedge separation of 1.27 cm. Similar to the edge exit hydrogel wedges, the presence of the antisymmetric Lamb wave is apparent for frequencies in the vicinity of 1 MHz and shows reasonable agreement with the theoretical group velocity curve. The symmetric Lamb wave, if present, is difficult to distinguish from the longitudinal wave for the frequencies excited. These results agree with those from the transmission signal map. In the future, transmission data from transducers with wider bandwidth or the superposition of multiple transducers with different frequencies could be used to measure group velocity for a wider range of frequencies and reconstruct a larger portion of the dispersion curves. Frequency-Time Decomposition for Surface Exit Hydrogel Wedge Separation of 1.27 cm 7/ 6 6 Longitudinal Wave Symmetric Lamb Antisymmetric Lamb -0.05:E -0.1 4 -0.151 U 3 U-0.2 2 -0.25 011 0* 5 10 15 25 20 30 35 40 Time in PET (us) Figure 3-27: Example frequency-time decomposition for surface exit hydrogel wedges spaced 1.27 cm apart. Theoretical wave group velocity dispersion curves are overlayed on top. 78 Although it is notable that the hydrogel wedges were able to propagate a clear antisymmetric Lamb wave, the magnitude is still dwarfed by the longitudinal wave. This means that the majority of the acoustic energy manifests as the longitudinal wave. This is a problem because the smaller the magnitude of the Lamb wave, the shorter the distance it can travel before it attenuates and is no longer distinguishable from noise. In our experiments, the furthest wedge separation for which the antisymmetric wave remains distinguishable is about 2 cm. Future work investigating improvements in the current Lamb wave propagation scheme to direct more of the acoustic energy into the Lamb waves instead of the longitudinal wave could improve achievable propagation distances. 79 80 Chapter 4 Conclusions In this body of work, ultrasonic longitudinal and antisymmetric Lamb waves were successfully propagated along PET. For a detailed discussion of longitudinal wave For a detailed discussion of Lamb wave related findings please refer to Sec. 4.1. related findings please refer to Sec. 4.2. 4.1 Longitudinal Waves for Process Control Longitudinal waves were successfully transmitted through 76 pm thick PET using contact wedge ultrasound with a variety of wedge materials. ABS wedges were shown to be able to transmit longitudinal waves up to 5.08 cm. Instron based extensiontensile experiments revealed that the longitudinal wave is affected by tension in the PET, exhibited by a general increase in attenuation and a decrease in wave speed during plastic deformation. The transmitted wave magnitude was also found to be sensitive to wedge preload and the wedge-PET coupling interface. Any serious attempt at integrating wedge based contact ultrasound into a roll-to-roll system will have to find a way to maintain repeatability of those parameters. Additionally, the use of a coupling fluid on or in the vicinity of flexible electronics would have to be carefully engineered so as not to damage the electronics themselves. 81 4.2 Contact Ultrasound Generated Lamb Waves Lamb waves were very difficult to generate in 76 ym thick PET using a contact method. Lamb wave propagation was attempted with a variety of different wedge materials, including acrylic, ABS, Teflon, ultrasound gel, and an experimental hydrogel, many of which are not typically used for wedge contact ultrasound. Of the wedges tested, the surface exit hydrogel wedge had the best results. However, in all experiments the symmetric mode was unable to be distinguished from the longitudinal wave, if present at all. The antisymmetric mode Ao was apparent but was dwarfed in magnitude by the longitudinal wave and attenuated quickly, becoming indistinguishable from noise for wedge spacings greater than 2 cm. This performance is inferior to Futatsugi et al. [13] who measured laser excited symmetric Lamb modes So 2.5 cm from the source and Desmet et al. [11] who measured laser excited So and Ao modes 0.8 cm from the source. These results unfortunately do not support the initial premise that contact ultrasound could provide stronger, more desirable signals with greater propagation distances than laser based ultrasound. These findings also call into question whether the claim that Lamb waves can propagate farther than other acoustic waves is applicable to waves in PET, although the difference in observed propagation distances could be attributed to differences in initial magnitude. Future methods allowing acoustic energy to be better directed into Lamb waves instead of longitudinal waves could improve achievable Lamb wave propagation distances. Laser based ultrasound may be more appropriate for PET Lamb wave inspection, although the small propagation distances still pose a problem for large area inspection. Laser based ultrasound would also remove the need for a consistent coupling fluid application procedure. The impulse nature of the exciting laser could allow for a greater bandwidth of frequencies to be excited in the PET than with a contact transducer. However, to be used successfully for the inspection of flexible electronics with polymer substrates, either a specific laser frequency for which PET is not transparent or a material that can absorb the laser frequency and be deposited in an additional layer without affecting the electronics must first be found. 82 Appendix A MATLAB Code Symetric Lamb Mode Characteristic Equation function [sym] = symmetricMode(f,c,Cl,Ct,d) % Symmetric Lamb Equation % % % % % f : c : Cl: Ct: d : [Hz] frequency [m/s] Lamb wave phase velocity [m/si Longitudinal wave velocity [m/si Shear wave velocity [m] Plate thickness w = 2*pi*f; % circular frequency % wave number k = w/c; alpha = ((w/Cl)^2-k^2)^0.5; beta = ((w/Ct)^2-k^2)^0.5; sym (k^2-beta^2)^2*cos(alpha*d/2)*sin(beta*d/2)+... 4*k^2*alpha*beta*sin(alpha*d/2)*cos(beta*d/2); = sym = real(sym)+imag(sym); end 83 Antisymmetric Lamb Mode Characteristic Equation function [asym] = antiSymmetricMode(f,c,Cl,Ct,d) % Symmetric Lamb Equation % % % % % f : c : Cl: Ct: d : [Hz] frequency [m/si Lamb wave phase velocity [m/si Longitudinal wave velocity [m/si Shear wave velocity [m] Plate thickness w = 2*pi*f; % circular frequency k = w/c; % wave number alpha = ((w/Cl)^2-k^2)^0.5; beta = ((w/Ct)^2-k^2)^0.5; asym = (k^2-beta^2)^2*sin(alpha*d/2)*cos(beta*d/2)+... 4*k^2*alpha*beta*cos(alpha*d/2)*sin(beta*d/2); asym = real(asym)+imag(asym); end 84 Lamb Wave Group Velocity Calculation function [cgVec] =vGroup(fVec,cpVec) %Calculates group velocity for wave based on phase velocity % % fVec cpVec - % cgVec - [Hz] vector of considered frequencies [m/s] vector of phase velocities (same lenght as fVec) [m/si vector of group velocities at frequencies in fVec w = 2*pi*fVec; % circular frequency k = w./cpVec; % wave number %dMat is matrix for taking centered difference for first derivative dMat = diag(ones(1,length(fVec)-1),1)+... diag(-ones(1,length(fVec)-1),-1); dMat(1,1) = -1; dMat (end, end)=1; dw = dMat*w'; dk = dMat*k'; % cg = dw/dk cgVec = dw./dk; end 85 Lamb Wave Characteristic Equation Solver U/ Material Parameters clear d = 76.2e-6; %[m] PET thickness (0.003") %d = 177.8e-6; %[m] PET thickness (0.007") C1 = 2310; %[m/s] Ct = 1000; %[m/s] %% Phase Velocity cpsOGuess = 1800; % [m/si initial guess - symmetric phase velocity cpa0Guess = 200; % [m/s] initial guess - antisymm phase velocity options = optimset('fzero'); fVec = [0.01:0.01:0.1, 0.2:0.1:30]*1e6; cpsOVec = zeros(size(fVec)); cpaOVec = zeros(size(fVec)); %Calculate S_0 and A_0 for i = 1:length(fVec); % [Hz] cps = fzero(@(c) symmetricMode(fVec(i),c,Cl,Ct,d),... cpsoGuess,options); cpsoVec(i)=cps; cpsOGuess = cps; %[cpa,fval,exitflag,output] cpa = fzero(@(c) antiSymmetricMode(fVec(i),c,Cl,Ct,d),... cpa0Guess,options); cpaOVec(i) = cpa; cpaOGuess = cpa; end % Calculate S_1 slfstart = round(1.7e5*d*10)/10; %finds slfstart to 2 sig digits slfVec = (slfstart:0.1:(fVec(end)/1e6))*1e6; cpslGuess = 3000; cpslVec = zeros(size(slfVec)); for i = 1:length(slfVec); % [Hz] cps = fzero(@(c) symmetricMode(slfVec(i),c,Cl,Ct,d),... cpslGuess,options); cpslVec(i)=cps; cpslGuess = cps; end 86 % Calculate A_1 aifstart = round(1e5*d*10)/10; %finds alfstart to 2 sig digits alfVec = (alfstart:0.1:(fVec(end)/1e6))*1e6; cpalGuess = 3000; cpalVec = zeros(size(alfVec)); for i = 1:length(alfVec); % [Hz] cpa = fzero(@(c) antiSymmetricMode(alfVec(i),c,Cl,Ct,d),... cpalGuess,options); cpalVec(i) = cpa; cpalGuess = cpa; end figure plot(fVec/1e6,cps0Vec,'-b',fVec/1e6,cpaOVec,'-g',... slfVec/1e6,cpslVec,'--b',alfVec/1e6,cpalVec,'--g') xlabel('Frequency (MHz)') ylabel('Phase Velocity (m/s)') title('Dispersion Curves for 76 um thick PET') legend('S_0','A.0','SJ1','A-1') %axis([0 100 0 3000]) XU Group Velocity cgsOVec cgaOVec cgslVec cgalVec = = = = vGroup(fVec,cpsOVec); vGroup(fVec,cpaOVec); vGroup(slfVec,cpslVec); vGroup(alfVec,cpalVec); figure plot(fVec/1e6,cgsOVec,'-b',fVec/1e6,cgaOVec,'-g',... slfVec/1e6,cgslVec,'--b',alfVec/1e6,cgalVec,'--g') xlabel('Frequency (MHz)') ylabel('Group Velocity (m/s)') title('Group Velocity Curves for 76 um thick PET') legend('S_0','A_0','S_1','AJ1') 87 88 Appendix B LabVIEW Code ChavIdA PSsaAvies US Powered ChRageA ugs PotTriggersamples Open usigSanofNu~rn 20- S- COAnd. Ar u SerW Number Offset Foew hono Dss -loorem swrwks TOtaI Stocks TggrDiedioe AutoTrigeronie (nns) Rie Save Locatuop Tngei ac Trigge Threshold (16-bit counts) Trigger Delay (samples) Figure B-1: Front panel of LabVIEW code providing communication between the computer and digital oscilloscope. Code allows various oscilloscope acquisition settings to be selected and processes and saves collected data. 89 Totalt Serial Number Bloksp if there isan error Khastatus AO SttI Four Channel Device j PS5X4XB devices =j File Save Location USB Powered -0 3 I .................. . I *PS5X4XB devices USB Powered oscilloscope. Figure B-2: Labview code providing communication between the computer and digital oscilloscope acquisition settings to be selected and processes and saves collected data. Code allows various Appendix C Wedge Drawings Note all units in mm. (D 30.000 0 0 0 0 19.1 0 0.8 3.2 3.2 15.9 25.4 1.6 60.0005. 4.8 1.1 30.5- Figure C-1: Drawings of ABS single angle wedge. 91 4.8 9.5 Q) Ij '0 Ci0 '0 25.4 8.1 Figure C-2: Drawings of variable angle ABS wedge base. 92 C) Q) CC 4.8 4.8 6.4 15.9 19.1 9.5 3. - (NN _, HF ) C -'T c-'i 25.4 N Figure C-3: Drawings of variable angle ABS wedge carriage. 93 30.000 0 1 0 19.1 2.7 3.2 15.9 6.4 1.6 38.1 4.8 50.000 14.6 - 14.3 43.4 Figure C-4: Drawings of single angle Teflon wedge. 94 10.1 21.2 9.5 6.4 0' 14.7 19.1 32.4 Figure C-5: Drawings of mold base for edge exit hydrogel wedge. 95 10.1 5 21.2 AA 0 CD C14 'C 14.7 j3 19.1 2.4 Figure C-6: Drawings of mold base for surface exit hydrogel wedge. 96 Bibliography [1] A. Abdelrahman, U. Amjad, D. Jha, K. S. Tarar, and W. Grill. Zero order mode selective excitation and highly resolved observations of lamb waves. Health Monitoring of Structural and Biological Systems 2011, 7984:798413, 2011. WOS:000294550700031. [2] American Society for Nondestructive Testing. Ultrasonic Testing. In Patrick 0. Moore, editor, Nondestructive testing handbook, volume 7, pages 100-106. [Columbus, Ohio] : American Society for Nondestructive Testing , [1998-2012], 1998. [3] Oluwaseyi Balogun, Garrett D. Cole, Robert Huber, Diane Chinn, Todd W. Murray, and James B. Spicer. High-spatial-resolution sub-surface imaging using a laser-based acoustic microscopy technique. Ieee Transactions on Ultrasonics Ferroelectrics and Frequency Control, 58(1):226-233, January 2011. WOS:000286386100022. [4] A. Bedford and D. S. Drumheller. Introduction to Elastic Wave Propagation. Wiley, Chichester England ; New York, 1 edition edition, August 1994. [5] R. Briers, 0. Leroy, and G. Shkerdin. A liquid wedge as generating technique for lamb and rayleigh waves. Journal of the Acoustical Society of America, 102(4):2117-2124, October 1997. WOS:A1997YE03400024. [6] Onda Corporation. Acoustic Properties of Plastics. http: //www. ondacorp. com/ images/Plastics.pdf, 2003. Accessed:2015-05-21. [7] Onda Corporation. Acoustic Properties of Rubbers. http://www.ondacorp. com/images/Rubbers. pdf, 2003. Accessed:2015-05-21. [8] Onda Corporation. Acoustic Properties of Solids. http: //www. ondacorp. com/ images/Solids.pdf, 2003. Accessed:2015-05-21. [9] R. Cote, T. Van der Donck, J.-P. Celis, and C. Glorieux. Surface acoustic wave characterization of a thin, rough polymer film. Thin Solid Films, 517(8):26972701, February 2009. WOS:000263927300024. [10] B. Cros, M. F. Vallat, and F. Augereau. Characterization of aluminium coated poly(ethylene terephtalate) films by acoustic microscopy. Journal of Materials Science, 32(10):2655-2660, May 1997. WOS:A1997XB29700019. 97 [11] C. Desmet, U. Kawald, A. Mourad, W. Lauriks, and J. Thoen. The behavior of lamb waves in stressed polymer foils. Journal of the Acoustical Society of America, 100(3):1509-1513, September 1996. WOS:A1996VK27300036. [12] Xian Du, Brian W. Anthony, and Nigel C. Kojimoto. Grid-based matching for full-field large-area deformation measurement. Optics and Lasers in Engineering, 66:307-319, March 2015. [13] T. Futatsugi, S. Ogawa, M. Takemoto, M. A. Yanaka, and Y. Tsukahara. Integrity evaluation of SiOx film on polyethylene terapthalate by AE characterization and laser microscopy. Ndt & E International, 29(5):307-316, October 1996. WOS:A1996WD75600008. [14] Zhang Gaimei, Chen Qiang, He Cunfu, Zhu Huiqin, and Li Yuling. Research on the characterization and the properties of SiOx coating on plastic film. In 0. Y. Yun, X. Min, and Y. Li, editors, Printing and Packaging Study, volume 174, pages 486-489. Trans Tech Publications Ltd, Stafa-Zurich, 2011. WOS:000290892000114. [15] Matthew Wright Gilbertson. Electromechanicalsystems to enhance the usability and diagnosticcapabilitiesof ultrasound imaging. Thesis, Massachusetts Institute of Technology, 2014. [16] H. Haidara, Y. Papirer, Mf Vallat, and J. Schultz. Relationships between structural-properties of vapor-deposited metallic-films on to polymer and their relevant adhesive performance. Journal of Materials Science, 28(12):3243-3246, June 1993. WOS:A1993LJ64500018. [17] Shuiqing Hu, Chanmin Su, and Walter Arnold. Imaging of subsurface structures using atomic force acoustic microscopy at GHz frequencies. Journal of Applied Physics, 109(8):084324, April 2011. WOS:000290047000175. [18] X. Jia. Modal analysis of lamb wave generation in elastic plates by liquid wedge transducers. Journal of the Acoustical Society of America, 101(2):834842, February 1997. WOS:A1997WH32300018. [19] P.G. Kenny. Ultrasonic Inspection. In Nondestructive Evaluation and Quality Control, volume 17 of ASM Handbook. ASM International, 2015. [20] Dong Ryul Kwak, Sun Hee Kim, 1k Keun Park, Judith A. Todd, and Chiaki Miyasaka. Visualization of interior structures with nanoscale resolution using ultrasonic-atomic force microscopy. Nanosensors, Biosensors, and Info-Tech Sensors and Systems 2013, 8691:869117, 2013. WOS:000323279600025. [21] Shaoting Lin, Yihao Zhou, and Xuanhe Zhao. Designing extremely resilient and tough hydrogels via delayed dissipation. Extreme Mechanics Letters, 1:70-75, December 2014. 98 [22] F. Lisy, A. Hiltner, E. Baer, 31 Katz, and A. Meunier. Application of scanning acoustic microscopy to polymeric materials. Journalof Applied Polymer Science, 52(2):329-352, April 1994. WOS:A1994NA31700017. [23] A. P. McGuigan, B. D. Huey, G. a. D. Briggs, 0. V. Kolosov, Y. Tsukahara, and M. Yanaka. Measurement of debonding in cracked nanocomposite films by ultrasonic force microscopy. Applied Physics Letters, 80(7):1180-1182, February 2002. WOS:000173896400024. [24] R. D. N. Mileham, J. D. Sternhagen, and D. W. Galipeau. Surface acoustic wave thermogravimetric measurements of thin polymer films. Ieee Transactions on Ultrasonics Ferroelectrics and Frequency Control, 45(5):1273-1280, September 1998. WOS:000076488000022. [25] N. Nakaso and H. Yasujima. Observation of acoustic-emission signals from a ceramic coating layer on polyethylene terephthalate film. Japanese Journal of Applied Physics Part 1-Regular Papers Short Notes & Review Papers, 32(5B):25332535, May 1993. WOS:A1993LF97400091. [26] Vijayaraghava Nalladega, Shamachary Sathish, and Amarjit S. Brar. Characterization of defects in flexible circuits with ultrasonic atomic force miMicroelectronics Reliability, 48(10):1683-1688, October 2008. croscopy. WOS:000259844900011. [27] Arokia Nathan, Arman Ahnood, Matthew T. Cole, Sungsik Lee, Yuji Suzuki, Pritesh Hiralal, Francesco Bonaccorso, Tawfique Hasan, Luis Garcia-Gancedo, Andriy Dyadyusha, Samiul Haque, Piers Andrew, Stephan Hofmann, James Moultrie, Daping Chu, Andrew J. Flewitt, Andrea C. Ferrari, Michael J. Kelly, John Robertson, Gehan A. J. Amaratunga, and William I. Milne. Flexible Electronics: The Next Ubiquitous Platform. Proceedings of the Ieee, 100:1486-1517, May 2012. WOS:000309838000036. Ultrasonic Transducer Technical Notes. [28] Olympus NDT. olympus-ims .com, 2011. Accessed:2015-05-20. https://www. http://terpconnect.umd.edu/-toh/ peakfit.m. [29] Thomas C. O'Haver. spectrum/Interact ivePeakFitter . htm, April 2015. Accessed:2015-05-15. Determination of [30] Y. Okabe, N. Takeda, M. Yanaka, and Y. Tsukahara. the orthotropic elastic constants of thin PET film by an ultrasonic microspectrometer. Ieee Transactions on Ultrasonics Ferroelectrics and Frequency Control, 46(5):1269-1275, September 1999. WOS:000083121800021. [31] Daniele Passeri, Andrea Bettucci, and Marco Rossi. Acoustics and atomic force microscopy for the mechanical characterization of thin films. Analytical and Bioanalytical Chemistry, 396(8):2769-2783, April 2010. WOS:000275946100005. 99 [32] J. Pei, F. L. Degertekin, B. V. Honein, and B. T. KhuriYakub. Erosion/corrosion monitoring with dry contact ultrasonic lamb wave transducers, volume 2945. Spie - Int Soc Optical Engineering, Bellingham, 1996. WOS:A1996BG82A00045. [33] Michael de Podesta. Understanding the Propertiesof Matter. CRC Press, April 2002. [34] Wp Rogers. Elastic property measurement using rayleigh-lamb waves. Research in Nondestructive Evaluation, 6(4):185-208, 1995. WOS:A1995RE46300001. [35] Jl Rose and Jj Ditri. Pulse-echo and through transmission lamb wave techniques for adhesive bond inspection. British Journalof Non-Destructive Testing, 34(12):591-594, December 1992. WOS:A1992KB14600004. [36] K. Schoedel. Ultrasonic thin-film testing. Advanced Materials & Processes, 162(4):35-37, April 2004. WOS:000220709200007. [37] V. Y. Senyurek. Detection of cuts and impact damage at the aircraft wing slat by using Lamb wave method. Measurement, 67:10-23, May 2015. WOS:000351696900002. [38] G. S. Shekhawat and V. P. Dravid. Nanoscale imaging of buried structures via scanning near-field ultrasound holography. Science, 310(5745):89-92, October 2005. WOS:000232477000043. [39] Kyu Tak Son and Chin C. Lee. Design and reliability of acoustic wedge transducer assemblies for outdoor touch panels. Ieee Transactions on Components Packaging and Manufacturing Technology, 1(8):1178-1185, August 2011. WOS:000293752800008. [40] Jeong-Yun Sun, Xuanhe Zhao, Widusha R. K. Illeperuma, Ovijit Chaudhuri, Kyu Hwan Oh, David J. Mooney, Joost J. Vlassak, and Zhigang Suo. Highly stretchable and tough hydrogels. Nature, 489(7414):133-136, September 2012. [41] B. J. Tucker and D. A. Bender. Continuous ultrasonic inspection of extruded wood-plastic composites. Forest Products Journal, 53(6):27-32, June 2003. WOS:,000183679800007. [42] Chul Min Yeum, Hoon Sohn, and Jeong Beom Ihn. Delamination detection in a composite plate using a dual piezoelectric transducer network. Health Monitoring of Structural and Biological Systems 2011, 7984:798406, 2011. WOS:000294550700004. [43] Xuanhe Zhao. Multi-scale multi-mechanism design of tough hydrogels: building dissipation into stretchy networks. Soft Matter, 10(5):672-687, January 2014. 100