In this paper, the oxygen plasma bonding process for fusion bonded silicon wafers has been characterized by a new approach. The mechanical reliability of bonded microstructures was determined using burst tests and Weibull statistic analyses. The fracture characteristic of the bonded system is considered to depend on the stress distribution, the defect distribution and the fracture surface energy at the bond. Using Weibull theory, it is possible to extract the Weibull modulus m and the mean fracture uniform tensile stress per unit length, σfc, from the measured data. These quantities make it possible to compare the joint defect distribution and the fracture surface energy at the bonded interface for the processing conditions under observation. These experiments also demonstrate that it is possible to distinguish between these quantities under certain conditions. The fracture probability for different annealing temperatures has been evaluated and found to agree with previous results from surface energy measurements. It is shown that the bond fracture probability increases with annealing times in the range of 10-100 h. The saturated bond strength value is considerably enhanced by oxygen plasma activation prior to bonding. In this study, plasma activations at room temperature and 300 °C compare to chemical activations in hot nitric acid annealed at 120 °C and 700 °C, respectively. The tendency to form voids at elevated temperatures, e.g. 300 °C, is increased by the oxygen plasma treatment. If the surface energy is considered to be homogeneous over the bonded interface, the Weibull modulus m is an indirect measure of the defect distribution, low m values indicate a wide spectrum of defect types, whereas a high m value narrows the defect distribution responsible for fracture. The Weibull modulus m is shown to be valuable for evaluation of the bonded interface. It is demonstrated that a more scattered defect distribution emerges for in situ bonded wafers as compared to ex situ, and annealing at 300 °C for 90 h as compared to room-temperature storage. However, the defect distribution becomes increasingly more narrow with storage time. These variations may be due to either changes in microcracks or void configuration or inhomogeneities in the fracture surface energy over the bond interface.
[1]
G. Kissinger,et al.
Void-free silicon-wafer-bond strengthening in the 200–400 °C range
,
1993
.
[2]
U. Gösele,et al.
Semiconductor wafer bonding
,
1998
.
[3]
U. Gösele,et al.
Low temperature wafer direct bonding
,
1994
.
[4]
Lars Stenmark,et al.
Weibull fracture probability for silicon wafer bond evaluation
,
2000
.
[5]
Y. Hama,et al.
Structural stabilization induced by oxygen plasma post-exposure of films deposited from tetraethoxysilane
,
1997
.
[6]
Werner Langheinrich,et al.
Application of oxygen plasma processing to silicon direct bonding
,
1993
.
[7]
K. Hjort,et al.
Surface energy as a function of self-bias voltage in oxygen plasma wafer bonding
,
2000
.
[8]
Manfred Reiche,et al.
Hydrophobic silicon wafer bonding
,
1994
.
[9]
Stefan Bengtsson,et al.
Formation of Silicon Structures by Plasma‐Activated Wafer Bonding
,
2000
.
[10]
K. Hjort,et al.
Oxidation and Induced Damage in Oxygen Plasma In Situ Wafer Bonding
,
2000
.
[11]
M. Reiche,et al.
Time‐Dependent Surface Properties and Wafer Bonding of O 2 ‐ Plasma ‐ Treated Silicon (100) Surfaces
,
2000
.
[12]
B. Roberds,et al.
Chemical Free Room Temperature Wafer To Wafer Direct Bonding
,
1995
.