Ultrasonic technologies enable ultra-fine-pitch, low-temperature bonding
02/01/2001
Lorenzo Parrini, ESEC SA, Cham, Switzerland
overview
Ultra-fine-pitch wire bonding requires higher stability of ultrasonic vibrations than achieved by current ultrasonic transducers. BGA packages and low-temperature bonding demand higher bonding frequencies and force the design of stable, high-frequency converters. The demand for higher throughput and improved placement accuracy necessary for ultra-fine-pitch bonding necessitates new requirements in the dynamic mechanical behavior of the horn. Also, the bonding of new materials like copper requires optimal vibration frequencies of the ultrasonic transducers. To address this, new high-frequency ultrasonic transducers and a new mounting approach have been devised.
 |
Right: Close-up of a fine-pitch device.
Below: A wider bond range contributes to reducing manufacturing costs. (Photos courtesy of ESEC)
Ultra-fine-pitch wire bonding and copper bonding require a higher stability and robustness of the ultrasonic vibrations generated by the transducer compared to current ultrasonic horns [1, 2]. Furthermore, BGA applications and the prospect of being able to decrease the temperature of the wire bonding process demand higher bonding frequencies [3-6] that impose additional challenges on the design of stable high-frequency ultrasonic transducers [7].
 |
The design of ultrasonic transducers [8] uses such modern tools as the finite-element method (FEM) [9] and laser interferometry [10, 11]. In the present work, the advantage of these new techniques has been considered in order to design an advanced, high-frequency ultrasonic transducer for wire bonding. The clamping flange of the ultrasonic transducer also addresses a design issue that has plagued previous ultrasonic systems. The flange should be set in longitudinal nodes of the ultrasonic field, where no vertical amplitude occurs, in order to minimize the loss of ultrasonic energy in the fixing system. The loss of ultrasonic energy in the fixing system can never be totally removed [12], but this problem has been solved with the new ultrasonic transducer [13].
FEM modeling
The model of the transducer used for the FEM simulations is presented in Fig. 1. FEM modeling enables the designer to find out which vibration modes can be excited in a given frequency range and to characterize the corresponding displacement fields accurately. Sixteen different modes can be found for 110-140kHz, and every mode is characterized by a particular displacement field and resonant frequency. Not all modes can be excited at the same time. The excitation of a specific mode depends on the exact frequency and on the symmetry of the forcing stress produced by the piezoelectric driving elements. For example, assuming radially symmetric stress generated by the piezo, the right side of Fig. 2 shows the simulated transducer frequency response. The vibration amplitude at the tip of the transducer is calculated, and two maxima are evident. The geometry of these modes is shown on the left side of Fig. 2. These two modes are the pure longitudinal vibration modes of the new transducer occurring between 110 and 140kHz.
 Figure 1. FEM model of the new ultrasonic transducer. |
Because series electrical resonance occurs for each excited mechanical vibration mode of the transducer, it is possible and necessary to control the fidelity and accuracy of the FEM simulations by comparing them to the experimental electrical frequency responses taken at the network analyzer. The main resonance of the transducer is at 124kHz, with a secondary resonance at 137kHz. These correspond to the two modes shown in Fig. 2. The simulated resonance frequencies are a close match, at 122kHz and 136kHz.
New design concepts
Several new approaches have steered the design of the new ultrasonic transducer.
It is made of titanium; this metal and two others commonly used for ultrasonic horns are compared in Table 1. Titanium is a good choice because it is relatively light, has a high mechanical quality factor, high tensile strength, and a high elastic limit. The highest ultrasonic vibration amplitudes can only be attained on titanium horns. Also, the thermal expansion coefficient of Ti is lower than that of steel or Al, which reduces the thermal drift exhibited by wire bonders. This dimensional stability plus the relatively low density of titanium improves the wire bonding speed and placement accuracy.
The new transducer is characterized by a "Unibody" concept [7], meaning that no interface occurs between the capillary clamping and the piezo. In this way, the ultrasonic energy loss is minimized and controlled. It is also solid (instead of having concave sections), which allows higher stored mechanical energy and higher stability of the oscillations. The new horn is symmetric with respect to the horizontal plane, which eliminates parasitic transversal vibration components along the vertical direction. The ultrasonic amplifier of the new horn is a stepped cylinder, permitting the highest ultrasonic amplification [8].
The capillary clamping system of the new transducer is symmetric with respect to the distribution of masses and the distribution of stresses in the ultrasonic field. The new clamping system is drawn in the middle section of Fig. 3. The fixing screw is tightened into a special nut, and the mass of the nut compensates the mass of the head of the screw. At both sides of the horn tip, the same boundary conditions are valid for the ultrasonic stress field, and the propagation of the ultrasonic waves is therefore totally symmetric.
The experimental frequency response of the new transducer is plotted in Fig. 3 as a function of the frequency before and after cutting the new capillary clamping system. Due to the additional mass, the capillary clamping reduces the main mode frequency from 133kHz to 128kHz, but the main resonance remains very clean. If a conventional capillary clamping had been cut in the horn, whereby the thread to tighten the screw is cut directly into the transducer body, the main resonance at the bottom of Fig. 3 would have been split. Measurements with the laser interferometer have demonstrated that this splitting is due to the excitation of a parasitic transversal mode along the direction of the axis of the screw. Such modes can make the vibration of the capillary tip very unstable and are excited because of the asymmetry of the boundary conditions for the ultrasonic stress field at the horn tip.
 |
The mounting flange of the transducer has a new geometry derived by FEM simulations. The geometry allows the horn to be fixed not only at longitudinal nodes, but also in so-called radial nodes of the ultrasonic field. This results in a total decoupling of the horn and the welding device. When a conventional flange becomes constrained, for example, by a screw tightened to the force assembly, the ultrasonic field in the flange and along the transducer is perturbed. The transducer is therefore sensitive with respect to the clamping. In contrast, the new mounting flange in free state exhibits a zero displacement region, because the large width of the membrane allows the radial waves to build up a node (radial node). When the flange becomes constrained at this radial node, the ultrasonic field in the flange and along the transducer remains totally unperturbed. The transducer is insensitive with respect to the clamping, and no ultrasonic coupling is occurring between the transducer and the welding device.
 |
Testing
Table 2 summarizes the improvements that have been achieved with the new transducer compared to a conventional one, including the improvements in touch-down robustness. The high accelerations of the transducer during bonding create background mechanical oscillations that are excited along the transducer, and which disturb the touch-down detection process of the wire bonder. These touch-down disturbances can be compensated for by higher settings of the touch-down level for ball and wedge bonding, but this raises the minimum attainable impact, thus reducing the fine-pitch capabilities of a wire bonder. Touch-down disturbances can also be decreased by increasing the thickness of the mounting flange. This is normally undesirable, but the new mounting design decouples the transducer from the wire bonder, allowing the thicker flange.
Figure 3. Experimental frequency response of the new ultrasonic transducer (without fixing flange). The impedance is plotted as a function of the frequency between 50 and 150kHz before (above) and after (below) cutting the capillary clamping system. The new symmetric capillary clamping system is drawn in the middle.
The higher frequency of the transducer is beneficial for low-temperature and copper-bonding applications, where the lower temperature and the stiffer material properties can be compensated for by the higher energy contribution from the horn. The higher frequency also causes the better ball roundness obtained with the new horn, since at higher frequency, a lower vibration amplitude is required for the same ball bonds. The higher maximum amplitude of the new transducer is necessary to bond thicker wires.
 |
Tables 3 and 4 show process data from an evaluation of a wire bonder by a user. The data were collected on the same wire bonder after process optimization using a conventional transducer and later the new transducer. The bonded product was always the same — a BGA bonded at 160°C with the 70µm fine-pitch process.
Table 3 reports the touch-down disturbances, which were measured on the wire bonder during the evaluation, the touch-down levels, which were set on the wire bonder, and the measured impact (ball and wedge). The suggested parameters for the 50µm fine-pitch process are reported as a reference. The new transducer reduces the touch-down disturbances on the wire bonder at both the ball and the wedge sides. This has allowed us to set up lower touch-down levels on the machine and to achieve lower impact values, which are required to produce small ball and wedge bonds.
 |
The influence of the lower impact on the process results is better illustrated in Table 4, which lists the measured wedge pull-strength, ball diameter, and ball height, all with corresponding CPK, for the above-mentioned process with both transducers. The target parameters for this process according to the user are also reported as a reference. With a conventional transducer, it is impossible for the wire bonder to achieve the required specifications for wedge pull-strengths and ball diameter and heights. If equipped with the new transducer, the wire bonder can meet the specifications.
 |
Figure 4. Fine-pitch 50µm process on a BGA (ball bonds), bonded at 120°C with the new transducer.
The new transducer permits the wire bonder to attain the required impact values for the 50µm fine-pitch process (see Table 3). Figure 4 shows the ball bonds of a 120°C, 50µm process on a BGA. The same wire bonder could not achieve a 70µm process when equipped with a conventional transducer, as shown in Table 4.
Conclusion
A new, high-frequency ultrasonic transducer has been conceived, designed, prototyped, and tested. The transducer is made of titanium and vibrates at the resonance frequency of 125kHz.
 |
It is fixed on the wire bonder with a flange whose special geometry was calculated by means of FEM simulations and laser interferometry. This flange allows a total decoupling of the transducer and the wire bonder. Process tests showed that the mounting of the new transducer on a wire bonder improved the fine-pitch and copper-bonding capabilities of the machine. The introduction of the new transducer and its higher vibration frequency allows the operating temperature of the wire bonder to be reduced for BGA and low-temperature applications.
Acknowledgments
We thank The Swiss Confederation for its financial support.
References
- G. Harman, Wire Bonding in Microelectronics, McGraw-Hill, New York, p. 11, 1997.
- N. Onda, V. Jaecklin, S. Arsalane, "High Frequency Wire Bonding for Ultra Fine Pitch Applications," Semicon/Test, Assembly & Packaging Proceedings, Singapore, p. 159, 1998.
- H. Ramsey, C. Alfaro, "High-frequency Enhancement for Ambient Temperature Ball Bonding," Semiconductor International, p. 93, August 1997.
- V. Jaecklin, "Room Temperature Ball Bonding Using High Ultrasonic Frequencies," Semicon/Test, Assembly & Packaging Proceedings, Singapore, p. 208, 1995.
- J. Tsujino, H. Yoshihara, K. Kamimoto, Y. Osada, "Welding Characteristics and Temperature Rise of High Frequency and Complex Vibration Ultrasonic Wire Bonding," Ultrasonics International 1997 Proceedings, Delft, Holland, p. 59, 1997.
- B. Langenecker, "Effect of Ultrasound on Deformation Characteristics of Metals," IEEE Transactions on Sonics and Ultrasonics, Vol. Su-13, No. 1, p. 1, 1966.
- A. Safabakhsh, Kulicke & Soffa Inv., "Unibody Ultrasonic Transducer," US Patent, 5,469,011, 1995.
- R. Millner, Ultraschalltechnik, Physik-Verlag, Weinheim, p. 380, 1987.
- P. Derks, "The Design of Ultrasonic Resonators with Wide Output Cross-sections," PhD Thesis, Eindhoven Institute of Technology, Eindhoven, Holland, 1984.
- J. Tsujino, H. Yoshihara, K. Kamimoto, Y. Osada, "High-frequency Longitudinal-transverse Complex Vibration Systems for Ultrasonic Wire Bonding," 1997 IEEE Ultrasonics Symposium Proceedings, p. 849, 1997.
- M. Bougoin, J. Luc, "New Piezoelectric Ceramic for High Power Transducers," Ultrasonics International 1989 Proceedings, Elsevier, Amsterdam, The Netherlands, p. 584, 1989.
- P. Whelan, M. Whelan, Verity Instruments Inc., "Ultrasonic Transducer and Mount," US Patent, 5,364,005, 1994.
- L. Parrini, ESEC SA, "Ultrasonic Transducer with a Flange for Mounting on an Ultrasonic Welding Device, in Particular on a Wire Bonder," European Patent, EP 0 934 794 A1, August 1999.
Lorenzo Parrini received his diploma in nuclear engineering in 1991 from Politecnico di Milano. He completed his doctorate in 1995 at the Federal Institute of Technology (EPFL) in Lausanne, Switzerland, focusing on mechanical spectroscopy in metal matrix composite materials. He worked at the European Center for Nuclear Research (CERN) in Geneva studying nuclear particle accelerators, and at the Federal Institute of Technology (ETHZ) in Zurich, Switzerland. Parrini joined ESEC in 1997, where he heads ultrasound development. He has authored 27 scientific publications. ESEC SA, Hinterbergstrasse 32, CH-6330, Cham, Switzerland; ph 41/41-749-5934, fax 41/41-749-5222, e-mail [email protected].