Grundlegende Untersuchungen der Mechanismen des Wedge-Wedge Bondens

Obwohl die Ultraschall-Drahtbondtechnik seit den 1960ern verbreitet angewendet wird, sind die zugrundeliegenden Mechanismen noch nicht vollständig verstanden. Das eingereichte Projekt widmet sich diesen Mechanismen sowie deren Einflüsse auf das US-Bonden. Insbesondere die Reibungs- und die US-Materialerweichungs-Phase bilden die Projektschwerpunkte. Diese Phasen spielen eine große Rolle zum Verständnis der Oxidentfernung und der Bildung der Mikroverschweißungen.Die Oxidschichten auf den Oberflächen von Draht und Substrat behindern die Verbindungsbildung. Beide o.g. Phasen beeinflussen die Oxidentfernung im Hinblick auf Ablösung und Transport aus der Verbindungszone. Aufgrund geringer Dicke der natürlichen Oxidschicht kann die Reinigung nur sehr schwer in Echtzeit während des Verbindungsprozesses beobachtet werden. Daher werden hier Mikropartikel und intransparente Schichten zur Emulation der natürlichen Oxide verwendet um sowohl schrittweise und Echtzeit-Beobachtungen durchzuführen und dann die Bewegungspfade der Oxide abzuleiten.Die Berührpunkte der Oberflächen sind die Startpunkte der Mikroverschweißungen. Der Einfluss der US-Materialerweichung auf diese Bereiche ist noch ungeklärt. In diesem Projekt wird daher das Substrat gezielt strukturiert, um so die Verformung der Struktur gezielt zu analysieren. Aufbauend wird ein Modell der Materialerweichung abgeleitet. Darüber hinaus werden die Mikroverschweißungen zeitdiskret charakterisiert und deren Wachstum und Bondfestigkeit beobachtet.Die Oxid-Metall-Kontaktbereiche, die Metall-Metall-Kontaktbereiche und die Mikroverschweißungen verändern sich über den Bondprozesses. Das fehlende Wissen über die dynamische Veränderung der Grenzflächen ermöglicht keine genaue Aussagen zur Bindungsgeschwindigkeit und -festigkeit. Um eine Messung der Kräfte beim Bonden in Echtzeit zu ermöglichen, wird ein hochauflösendes Kraftmesssensorarray erforscht, hergestellt und untersucht. Dieses wird direkt unter der Bondstelle eingebettet und somit integraler Teil des Bondversuches. Die hohe Messempfindlichkeit in tangentialer und normaler Kraftrichtung ermöglicht eine genaue Charakterisierung des Prozesses. Mit den Informationen werden insbesondere die Bildung lokaler Mikroverschweißungen und deren Aufbrechen untersucht.Im gesamten Projekt wird die Versuchsplanung mittels Design-of-Experiments genutzt und der Einfluss der Prozessparameter abgebildet. Schließlich wird ein empirisches Modell zur Vorhersage der Vorgänge an den Grenzflächen der zeitabhängigen Metalloxidverteilung, Rauheitsänderungen sowie der lokalen und globalen Bondqualität erstellt. Basierend auf dieser Analyse sollen neue Topographien sowohl auf dem Substrat als auch auf dem Draht hergestellt und ausgewertet werden, um eine Verbesserung der Bindungsgeschwindigkeit und -qualität zu erreichen. Alle Erkenntnisse geben einen tiefen Einblick in die Bindungsmechanismen und schließen so Lücken in diesem Bereich des Prozessverständnisses.

The main objective of the project has been achieved. The oxide removal process has been completely revealed in detail. It consists of four steps including the occurrence of cracks, the detachment of oxides, the milling of these oxides into small particles and the transportation of these particles to the peripheral region. Four mechanisms drive the transportation of the oxide particles: metal penetration, oxide flow, pushing and metal splash. Metal penetration and oxide flow play the most significant role in the transportation process. During the transportation process, some oxide particles agglomerate together into larger particles. Microweld formation and breakage were studied via molecular dynamics simulations. They showed that microwelds can be formed and broken in an extremely short time. The formation and breakage of microwelds take place continuously during the bonding process. The local positions where old microwelds were broken still offer the opportunity for the formation of new microwelds, even with the same asperities. The horizontal movement of the wire significantly changes the surface topographies. Under the same conditions, a high stiffness of the materials has a negative effect on the microweld growth. The surface topography, especially the tall asperities, significantly affects the change of microwelds. A larger approaching distance or a larger deformation of asperities significantly helps the microweld growth. A large vibration amplitude makes the microweld changes faster while it does not necessarily increase the shear stress and the microweld area. A sensor array based on the dice-and-fill method was accomplished containing 12 sensors in total. Due to their size an interposer chip had to be used to electrically contact each sensor of the array individually. During the development of the manufacturing process various challenges have been overcome. First, suitable parameters for mechanically structuring the brittle PZT with precision dicing were found. Chipping on the ceramic surface can be reduced by using a rotational speed of 30,000 rpm and a dicing blade with the least grain size. By setting the feed rate to 0.5 mm/s no cracks are initiated during dicing. For filling the diced kerfs, different materials were tested. Polyimide turned out to be unsuitable due to layer tension as a consequence of shrinkage during the curing process. With epoxy based fillers Vitralit® UD 2018 and Vitralit® 2020 the highest bonding strength can be obtained with the former showing a smaller coupling effect among sensor columns Thermocompression bonding of gold and tin proved to be a suitable technology for contacting the sensor chip to the interposer chip. An average shear strength of more than 20 MPa is achieved with a parameter set using 230 °C, 200 s and 7,8 MPa, respectively. An upper electrode Au showed a much higher bondability compared to Al. As the thickness increases, the bonding strength but also the coupling effect among the columns increase. The local tangential forces have been measured in-situ by the manufactured sensor array and the underlying mechanisms were derived. The two central columns undergo the largest tangential force, the 4 corner columns provide the smallest force signals, and the 6 peripheral columns undergo the largest increase of contact area. In the first approx. 5 ms, a fast increase (sticking) – plateau (sliding friction dominating) – fast increase (increasing significance of sticking contact behavior) of the force was detected on all sensor columns. After this stage, a decrease of force was observed on the central columns due to reduced local normal force. As more microwelds were formed, the force amplitude became constant. For peripheral and corner columns, a slow increasing force amplitude appeared due to the expansion of contact area and the increase of the local normal force. Compared to central columns, the signal at peripheral and corner columns indicated a more sliding friction. An energy flow model was further quantified according to the detected relative motions. The majority of the US energy flows to the vibration induced friction at the wire/substrate and the wire/tool interfaces, and the vibration induced microwelds formation, deformation and breakage. Even though the rest items are significant to the process, only a little amount of energy is delivered to them. A good coupling of the normal force and the US power could guide more energy to the wire/substrate interface for microwelds formation. If the two factors are not well coupled, even though more energy can be provided to the wire/substrate interface, the breakage of microwelds will take up a big part. Different surface structures have been tested for wire bonding and it is found that the deposited strips and straight ditches facilitate oxide removal and can enlarge process window. For the deposited strip structure, the edges of strips helped break the oxide scale. The oxide transportation occurred earlier and more oxides can be removed, compared to bonding on smooth surfaces. Thus, a relatively high bonding strength can be achieved. As for bonding on ditch structures, aluminum and aluminum oxide were continuously cut from the wire, accumulated in the ditches, pressed and finally squeezed to the outside of the ditches to form long chips. Accompanying with the growing of such chips, metal splashes were generated which further enhanced the oxide removal process. Under the same driving current, the bonding strength on straight ditches was up to 4 times higher than that on smooth surfaces.

Projektbezogene Publikationen (Auswahl)

• Revealing of ultrasonic wire bonding mechanisms via metal-glass bonding. Materials Science and Engineering: B, 236, pp.189-196
  Long, Yangyang; Dencker, Folke; Isaak, Andreas; Li, Chun; Schneider, Friedrich; Hermsdorf, Jörg; Wurz, Marc; Twiefel, Jens & Wallaschek, Jörg
• Self-cleaning mechanisms in ultrasonic bonding of Al wire. Journal of Materials Processing Technology, 258, pp.58-66
  Long, Yangyang; Dencker, Folke; Isaak, Andreas; Hermsdorf, Jörg; Wurz, Marc & Twiefel, Jens
• Quantification of the energy flows during ultrasonic wire bonding under different process parameters. International Journal of Precision Engineering and Manufacturing- Green Technology, 6(3), pp.449-463
  Long, Yangyang; Schneider, Friedrich; Li, Chun; Hermsdorf, Jörg; Twiefel, Jens & Wallaschek, Jörg
• Contact mechanics and friction processes in ultrasonic wire bonding-Basic theories and experimental investigations. Journal of Sound and Vibration, 468, p.115021
  Long, Yangyang; Twiefel, Jens & Wallaschek, Jörg
• Investigations on the mechanism of microweld changes during ultrasonic wire bonding by molecular dynamics simulation. Materials & Design, 192, p.108718
  Long, Yangyang; He, Bo; Cui, Weizhe; Ji, Yuhang; Zhuang, Xiaoying & Twiefel, Jens
• Impact of surface texture on ultrasonic wire bonding process. Journal of Materials Research and Technology, 20, pp. 1828-1838
  Long, Yangyang; Arndt, Matthias; Dencker, Folke; Wurz, Marc; Twiefel, Jens & Wallaschek, Jörg