Investigation of fluorine induced probe marker discoloration

In semiconductor integrated circuits processing, an insulating layer was usually deposited on wafer surface to protect the chip after devices and metal bonds formed. The protection layer on the metal bonds should be selectively removed by photolithography and reactive plasma etching before wire bonding. CF₄/O₂ plasma reactive etching usually formed fluorine contaminated layer on the surface of aluminum pads. Wet treatment to remove fluorine contaminations on aluminum bond pads was used conventionally. However, in memory integrated circuits, a polyimide layer is coated on the protection layer to be a radiation shielding. The addition of polyimide layer as protection coating on the insulating layer prohibited the wet cleaning. Then discoloration of probe marker was found on the Al bond pads after circuit probing. To verify the reliability issue of probe marker discoloration and investigate the possible discoloration origins, a systematically harsh environmental testing and materials characterizations were carried out. All of the samples with and without probe marker discolorations pass the reliability testing and final test. Even so, it is believed that probe marker discolorations would possibility degrade the bonding yield. Therefore it is necessary to figure out the possible origins for discolorations and propose a modified process to improve it. Scanning electron microscopy (SEM) and energy dispersive X-ray spectroscopy (EDS) investigations indicate that a few fluorine content exist in the discoloration marker region while no detectable fluorine signal is observed in the normal marker region. These results confirmed that the marker discoloration is resulted from the fluorine contaminations. Micro Raman scattering spectra from normal and discoloration areas demonstrated that a signal located at about 850 cm^-1 from normal marker area is vanished in discoloration marker area while a signal located at about 2250 cm^-1 appeare- . It is believed that at least one complex compound is formed on the discoloration area. Although we try to identify the chemical formula of this complex compound from Raman database. However, we can´t find any compound that matched with our observation. From the variation of Raman spectra from different marker area, this complex compound could possibly be amorphous aluminum and aluminum oxide corroded by fluorine. This can explain why the 850 cm^-1 Raman signal disappear for discoloration marker area and the bonding yield was a little degraded due to the poor adhesion. However, it is believed that most of the corrosion layers would be removed during wire bonding action. To improve the issue, a modified process based on two different passivation/polyimide etching process was proposed. Results of the composition analyses of the bond pads performed by using an energy dispersive X-ray spectrometer show that fluorine has been reduced when compared to the original process. From the results obtained so far, it seems that extra wet strip can remove the polymer contained slight fluorine after metal etching can resolve probe marker discoloration issue. It is concluded that two photo and etching steps for passivation and polyimide had been proven to be a highly reliable pad opening process used in IC fabrication.