Available online at www.sciencedirect.com
Journal of Non-Crystalline Solids 354 (2008) 1407–1410 www.elsevier.com/locate/jnoncrysol
Lithium aluminosilicate glass-ceramic containing Na2O for low-temperature anodic bonding in microelectronic mechanical systems Yourong Huang *, Zhu Cui, Xiping Gao, Changjiu Li, Zhenan Gu Quartz and Special Glass Institution, Quartz Glass Department, China Building Materials Academy, Guanzhuang, Beijing 100024, PR China Available online 26 November 2007
Abstract In this study, one kind of lithium aluminosilicate-b-quartz glass-ceramic containing Na2O has been chosen, which can be used for lowtemperature bonding with a higher bonding rate for the manufacture of microelectronic mechanical systems. The microstructure of the glass-ceramic is analyzed by transmission electron microscopy, high resolution transmission electron microscopy, selected-area diffraction, energy dispersive spectrometry and X-ray diffraction. The influence of the heat treatment process of the glass on the properties and the microstructure of the glass-ceramics is discussed. The microstructure of the glass-ceramic has the continuous glassy region with higher alkali metal ion content, which benefits the low-temperature anodic bonding. 2007 Elsevier B.V. All rights reserved. PACS: 61.72; 82.45; 87.59.L Keywords: Glass-ceramics; Conductivity; Microstructure; Microcrystallinity; Thermal properties
1. Introduction Silicon-to-glass anodic bonding plays a very important role in the fabrication of stacked wafer structures for microelectronic mechanical systems (MEMS) such as microsensors, microactuators, micro flow devices and so on [1]. In order to prevent the degradation of integrated circuits and metal leads, and minimize the bonding-induced stress, bonding is better to be performed below 180 C [2]. However, Pyrex glass is usually used as substrate glass mainly because of its thermal expansion coefficient, 32.5 · 10 7/ C, suitable to the wafer. But bonding with the Pyrex glass has to be performed at higher temperature, usually at 300 C, otherwise the time required for complete bonding will be so long to be unacceptable in the commercial. So it is necessary to find a new glass that not only has the ther-
mal expansion coefficient suitable to the wafer, but also can make the bonding perform below 180 C. In this study, one kind of lithium aluminosilicate-bquartz glass-ceramic contained Na2O has been chosen, which can be used for low-temperature bonding at 180 C with two times bonding rate more than that of Pyrex glass. The microstructure of the glass-ceramic was tested by transmission electron microscopy (TME), high resolution transmission electron microscopy (HRTEM), selected-area diffraction (SAD), energy dispersive spectrometry (EDS) and X-ray diffraction (XRD). The influence of the heat treatment process of the glass on the properties and microstructure of the glass-ceramic was analyzed. 2. Experimental procedure 2.1. Materials preparation
*
Corresponding author. Tel.: +86 010 51167369; fax: +86 010 65794978. E-mail address:
[email protected] (Y. Huang). 0022-3093/$ - see front matter 2007 Elsevier B.V. All rights reserved. doi:10.1016/j.jnoncrysol.2007.02.093
The parent glass was prepared in the system Li2O– Al2O3–SiO2 contained Na2O from 0% to 4% in weight
1408
Y. Huang et al. / Journal of Non-Crystalline Solids 354 (2008) 1407–1410
Table 1 Program of heat treatment No.
1 2 3 a
Table 2 a and qv of glass-ceramics
Nucleation
Crystallization
Ta (C)
t (h)
T (C)
t (h)
630 630 630
16 16 16
730 730 730
12 14 16
No.
aa 300 C (·10 7/ C)
qvb 298 C (·1011 X cm)
Parent glass 1 2 3
53.9 39.0 31.1 26.8
4.73 3.1 2.6 2.4
a
The control accuracy of temperature is ±2 C.
percent from reagent grade chemicals. The batch was fully mixed and melted in a corundum crucible at 1580 C for 6 h in a MoSi2 electric furnace. The melt was poured into a stainless steel die to form a piece and annealed in a muffle. According to differential thermal analysis (DTA) technology, the program of heat treatment for crystallization was made out as Table 1. 2.2. Property measurement The thermal expansion coefficient a was measured by using a horizontal dilatometer (Netz 402PC). The volume electrical resistivity qv was tested by a super-high resistivity meter (HZ2513) with a measurement bridge at 100 kHz. 2.3. Microstructure analysis and identification of the main crystalline phases TEM (JEM-2000FX) and HRTEM (JEM-2010) were used to investigate the microstructures of the glass-ceramic. Specimens for TEM analysis were prepared by argon-ionthinned in the conventional manner. Those for HTEM were made out by powder method with copper mesh. The element distribution and phase character in microstructure were tested by EDS (model: Voyager3105) and SAD, respectively. Because of the strong rays, the use of SAD and EDS should be very careful to prevent the destruction in microstructure. The main crystalline phase of the glassceramic was identified with powder XRD (Dam-IIIA) technology. 2.4. The anodic bonding test The anodic bonding test was performed at 180 C with an applied voltage of 500 V for 30 min. The sample’s diameter was 76 mm with 0.8 mm in thickness. As comparison, the Pyrex glass was also used for bonding test in this study. 3. Results and discussion 3.1. The effect of crystallization treatment on thermal expansion coefficient a and electrical resistivity qv The results shown in Table 2 indicated that the thermal expansion coefficient and electrical resistivity of the glassceramic were strongly influenced by the crystallization time
b
The measurement error of a is less than ±3%. The measurement error of qv is less than ±1%.
in the heat treatment process. The thermal expansion coefficients decreased with the increase of crystallization time, from 53.9 · 10 7/ C to 26.8 · 10 7/ C. When crystallization time was 14 h, the thermal expansion coefficient of the glass-ceramic, Sample 2, was 31.1 · 10 7/ C, matched to the wafer. The electrical resistivity also decreased with increase of crystallization time, from 4.73 · 1011 X cm to 2.4 · 1011 X cm, almost a half of the parent glass. It is generally known that the thermal expansion coefficient of Li2O–Al2O3–SiO2 glass-ceramic would decrease with crystallization process because of that crystal phase with lower thermal expansion coefficient would be formed in parent glass. On the other hand, the electrical resistivity would increase with Li+ ions entering into the crystal structure, decreasing the mobility of it. But in this study, the electrical resistivity decreased with the crystallization process. It could be attributed to the special microstructure formed in the parent glass contained Na2O during the heat treatment. 3.2. The microstructure analyze Fig. 1 shows that a great number of micro grains, with 50–80 nm in diameter, was formed in a continuous matrix phase and the density of grains was increased with the lengthening of crystallization time, obviously. HRTEM test (Fig. 2) shows clearly that the grains in irregular shapes were embedded in the continuous phase, and the interface between the grains and the continuous phase with 3–5 nm in width was a kind of unsharp transition layer, which indicated that ions exchange has happened. The SAD pattern (Fig. 3(a)) of the grains was some individual dots, implied that the grain had crystal characters. And the SAD pattern (Fig. 3(c)) of the amorphous was a disperse light spot, suggested it was an uncrystallized glass phase. In this study, it failed to identify the crystal category by the regulation of parallelogram because of the unsystematic in the SAD pattern. XRD technology was used to test the kind of crystal for the block glass. It shows (Fig. 4) that the crystal in the glass after heat treatment process was a b-quartz solid solution, Li2O–Al2O3–2SiO2. A lot of b-quartz solid solution with lower thermal expansion coefficient was formed in the glass during crystallization process, which was the reason for the decrease of thermal expansion coefficient of the glass-ceramics with the increase of crystallization time.
Y. Huang et al. / Journal of Non-Crystalline Solids 354 (2008) 1407–1410
1409
Fig. 3. SAD pattern and EDS pattern of Sample 3: (a) SAD pattern for the micro grain; (b) EDS pattern for the micro grain; (c) SAD pattern for the amorphous area and (d) EDS pattern for the amorphous area.
Fig. 1. TEM micrograph of the LAS glass-ceramic after crystallization process (space bar 200 nm): (a) Sample 1 for crystallization time 12 h and (b) Sample 3 for crystallization time 16 h. Fig. 4. XRD pattern of Sample 3.
Table 3 Element distributiona for the grains and the amorphous area in Sample 3 by EDS
Fig. 2. HRTEM micrograph of Sample 3 (space bar 50 nm).
EDS (Fig. 3(b) and (d)) was used to test the element distribution for the grains and the amorphous area. The results (Table 3) indicated that the chemical compositions of them were quite different, especially in sodium distribution. The sodium content of amorphous glass phase was 1.7 times as much as that of the grain phase. This result could also be attributed to crystallization of parent glass. Because it was more difficult for sodium ions to get into the structure of b-quartz solid solution for its larger ion radius, so they had a strong tendency towards to enrich in the continuous amorphous glass phase during the crystallization of the parent glass to form sodium-rich glass phase, which caused the decrease of the electrical resistivity in the glass-ceramic with the increase of crystallization time.
Element (wt%)
Glass phase
Grain phase
Si–K Na–K Mg–K Al–K As–K Zr–K Li–K
64.88 5.07 0.41 27.23 0.60 0.31 1.49
77.67 2.99 0.00 16.77 0.00 0.71 1.86
Total
100
100
a
The measurement error of EDS is less than ±3%.
3.3. The bonding test result It is well known that the essential for the anodic bonding was the transference of alkali metal ion in glass. So the mobility of alkali metal ion plays an important part in the anodic bonding. In general, glass with lower electrical resistivity would incline to a larger mobility of alkali metal ions, which could conduce to the low-temperature anodic bonding with a higher bonding rate. As it is shown in Fig. 5, Sample 2 with the lower electrical resistivity, 2.6 · 1011 X cm, was used for the low-temperature anodic bonding at 180 C, which had about two times bonding rate larger than that using the Pyrex glass that had a higher
1410
Y. Huang et al. / Journal of Non-Crystalline Solids 354 (2008) 1407–1410
low-temperature bonding of 180 C with higher bonding rate.
Fig. 5. Bonding surface photos of substrate glass after anodic bonding at 180 C with an applied voltage of 500 V for 30 min. (The dark gray areas were bonded parts.): (a) Glass-ceramic Sample 3 and (b) Pyrex glass.
electrical resistivity of 1 · 1015 X cm. The result was consistent with the analysis mentioned above. 4. Conclusions By the microstructure change produced in parent glass during heat treatment process, the lithium aluminosilicate-b-quartz glass-ceramic contained Na2O has a less electrical resistivity and a lower thermal expansion coefficient suitable to the wafer. The glass-ceramic could be used for
(1) Silicon-to-glass anodic bonding could be performed at 180 C using lithium aluminosilicate-b-quartz contained Na2O glass-ceramic with about two times bonding rate higher than that using the Pyrex glass; (2) the glass-ceramic obtained by heat treatment consisted of the continuous glassy region with higher alkali metal ion content, and a great number of bquartz solid solution grains with lower thermal expansion coefficient; (3) the glass-ceramic with alkali metal ion-rich continuous glassy region had a less electrical resistivity and then a higher alkaline mobility, which benefited the low-temperature anodic bonding. The thermal expansion property of the glass-ceramic could be adjusted by controlling the density of the b-quartz solid solution, which contributed to the match of the thermal expansion between the glass and the wafer.
Acknowledgement The authors are grateful to Mr Lu Miao for supplying the tests of anodic bonding. References [1] P.W. Barth, Sensor Actuator A21–A23 (1990) 919. [2] S. Shoji, H. Kikuchi, H. Torigoe, Sensor Actuator A 64 (1998) 95.
