 · Home· Table of Contents · Manufacturing | Nondestructive evaluation of disbonding area in chip scale packages using resonance ultrasound spectroscopyMasahiko Hirao, Keiji Sato, Hirotsugu OgiGraduate School of Engineering Science, Osaka University, Machikaneyama 1-3, Toyonaka, Osaka 560-8531, Japan Takekazu Miya Sonix K.K., Hyakunin-cho 2-6-7, Sinjuku-ku, Tokyo 169-0073, Japan. Contact |
Keywords: resonance ultrasound spectroscopy (RUS), chip scale packaging (CSP), resonance frequency, disbanding, thermal cycling test.
Microelectronics is rapidly evolving toward high-density, compact, and cheap structures. This movement includes developing the chip scale package (CSP) assemblies, which are miniature semiconductor packages adopting lower pin counts than the exist- ing ones. To complete the advanced performance, there is a severe demand for the reliability [1]. During manufacturing, the dicing and bonding processes bring in thermal stresses, which may cause delamination between elements. Delamination hinders the heat flow and, together with the moisture evaporation, may lead to corrosion, cracking, and disconnection. Current inspection practice relies on the immersion-type C-scan-ultrasonic imaging technique along with other methods including the electrical monitoring for disconnections. This ultrasonic method uses the high-frequency focused transducers and can construct the fine two-dimensional images by digitizing the reflection intensities from the boundaries. There are, however, some drawbacks that the inspection is rather time consuming and also the CSPs have to be placed in a water tank. On-line inspection over all the produced devices then has not been realized to date.
The flip-chip CSP assemblies consist of bare silicon monocrystal, ceramic base, and thin polymer layer between them (underfill); roughly, a triple-layered rectangular- parallelepiped composite (Fig.1). Conventional C-scan imaging demonstrated that the sample CSPs were originally flawless. To introduce the flaws, we heated up a sample to 523K (or 573K) and then cooled it with chilled water. After repeating several cycles,partial separation appeared at the chip-to-underfill interface. Figure 2 shows the C- scan image of 40% delamination; the upper bright portion indicates strong reflection intensity from the disbonded interface. Such an image was obtained using a focal transducer at 110 MHz and scanning steps of 0.025mm, taking about 90 sec.
Fig 1: Geometry of tested CSP assembly(a=13mm b=11mm, c=0.45mm, d=0.9mm). | 
Fig 2: A C-scan ultrasonic image of delaminated boundary between chip and underfill.
|
This study presents a dry ultrasonic inspection method for inspecting the CSP assemblies: the resonance-ultrasound-spectroscopy (RUS) method [2,3]. A CSP sample was put in a horizontal position on three pins, two of which contained piezo- electric transducers (pinducers) (see Fig.3). We measured the mechanical resonance spectra by driving one pinducer with continuous sinusoidal wave (20Vp-p) and detect- ing the oscillation amplitude through the other pinducer. Frequency scans resulted in the spectra as shown in Fig.4, where comparison is made between the flawless and flawed samples with 40%, 60%, and 85% of fractional disbanding area. Only the CSP weight (~0.4g) and the three-point contact served to accomplish the ultrasonic-vibration transmission and reception. Neither coupling agent nor an external load was necessary.
Fig 3: Measuring setup for resonance ultrasound spectroscopy.
| 
Fig 4: Resonance spectra from flawless and flawed CSP samples.
|
The spectrum contained a number of resonance peaks, from which we deter- mined the resonance frequencies by fitting Lorentzian function and finding the center frequencies. The present system finishes measuring the spectrum typically in 5 sec when stepping 1000 frequencies. The resonance peak height varies depending on the supporting positions, but the resonance frequencies are independent of them. Minimum constant load (dead weight) also contributes to realize free vibration and the stable measurement of mechanical resonance spectrum.
We observe in Fig.4 that the damage decreases the resonance frequencies. The thermal cycling test introduced delamination at the chip-underfill interface and lowered the overall rigidity of the structure. Figure 5 plots the relation between the three resonance frequencies and the delamination. The resonance-frequency shift shows a linear dependence on the disbonding area. Scattering of the resonance frequency over 30 flawless samples was only within 2%, being much less than the damage effect. The resonance peak height also diminishes with the damage. This too could be a damage indicator, but the trend is not straightforward.
Fig 5: Shifts of three resonance frequencies and damage fraction. I-bar shows the data scattering for flawless samples.
|
The RUS measurement with point-contacting transducers has a potential of providing a robust, quick, and reliable inspection means for the CSP assemblies. Future work should include enhancing the sensitivity to tiny delamination and studying the mode identification. The most critical is the chip-to-base joint (via bumps) along the chip's edge, whose defects are difficult to find with conventional inspection methods. Theoretical and experimental investigation on ultrasonic resonance could determine the key resonance mode, which exhibits the maximum sensitivity to the damages at this important part.
The CSP samples with materials data were supplied by Semiconductor Company, Matsushita Electrics Corporation.
| © AINDT , created by NDT.net | |Home| |Top| |