Reliability and Durability Enhancement of Large-area Mullite-cordierite Composite Substrates for Semiconductor Probe Card

Da-Eun Hyun Korea Testing Laboratory Jwa-Bin Jeon Kwangwoon University Yeon-Ji Choi Korea Testing Laboratory Yeon-Sook Lee Korea Testing Laboratory Yong-Nam Kim Korea Testing Laboratory Minkyung Kim Kwangwoon University Seunghoon Ko Kwangwoon University Sang-Mo Koo Kwangwoon University Weon Ho Shin Kwangwoon University Chulhwan Park Kwangwoon University Dong-Won Lee Kwangwoon University Jong-Min Oh (  jmOH@kw.ac.kr ) Kwangwoon University


Introduction
The probe card is a component that directly tests the operation of the IC chip integrated on the wafer. In recent years, the probe card has become important because not only size of the pads has decreased, but also spacing between the pads has become narrow with the miniaturization of electronic devices [1,2].
The ceramic substrate of a probe card has several required characteristics. Firstly, the probe card must have a low thermal expansion coe cients (TECs) similar to that of silicon for the wafer-level burn in test. The probe card cannot be used in to wafer-level burn-in test due to the unacceptable dimension errors due to the mismatch in the thermal expansion of the probe card and the silicon wafer [3]. Secondly, it is crucial to control the sintering shrinkage of the ceramic substrate of the probe card with many probe pins. The positions of the processed holes can deviate from the designed positions due to the shrinkage of the ceramic substrate during the sintering process. Therefore, it is important to consider ceramics with low TEC as candidate materials for semiconductor probe card and investigate the various factors affecting the sintering of ceramics.
Cordierite (2MgO·2Al 2 O 3 ·5SiO 2 ) is considered a promising material for the probe card substrate owing to its attractive properties such as high thermal shock resistance, high resistivity, high refractoriness, chemical stability, and low TEC. Consequently, cordierite has been widely used in various applications, such as electrical insulator, refractories, lters, membranes, substrates for integrated circuit boards, and microwaves [4][5][6][7]. However, it is di cult to obtain cordierite ceramics with a dense structure as cordierites have a narrow sintering temperature and are di cult to sinter using the solid-state process [8][9][10].
Many studies have investigated different methods to improve the densi cation of cordierite ceramics by adding sintering aids. Banjuraizah et al. [11] found that the addition of MgO to cordierite glass improves the densi cation, but excessive addition of MgO increased the TEC. Chen [12,13] also increased the density of cordierite-based glass ceramics by adding the sintering aids like CaO and ZnO. The excess use of these sintering aids, however, caused an increase in the TEC and dielectric loss. Therefore, it is critical to select suitable materials to fabricate a dense ceramic substrate with a low TEC for use in probe cards.
Mullite (3Al 2 O 3 ·2SiO 2 ) is not only well-known as a material that improves the mechanical properties of cordierite [6, 14,15], but is also used in electronic substrates, microelectronic packaging, and as a component in reinforced composites owing to its good electrical resistance, thermal shock resistance, excellent thermal and chemical stability at high temperatures, and good dielectric and mechanical properties [15][16][17][18]. Albhilil et al. [9] reported that the relatively low mechanical strength of cordierite can be improved by polymorphism transformation, although the TEC slightly increases by the addition of mullite. Subramanian et al. [19] proposed that chip detachment and device failure can be prevented by matching the TEC of the probe card substrate to that of Si. They also reported that a mullite-cordierite composite with 35:65 vol% has a TEC that matches with Si.
In this study, we investigated the composition ratio of the composite ceramics and the factors that in uence the sintering to fabricate a large-area mullite-cordierite composite substrate with excellent physical and mechanical properties. To investigate the effect of the particle size distribution on the sintered body, the microstructures of the composite granules before and after attrition milling were observed. The change in the sheet resistance and the exural strength of the large-area mullite-cordierite composite substrate during various environmental tests were measured to evaluate its reliability and durability.

Materials and processing
Commercially available powders of mullite (DAIHAN REFRACTORIES MATERIAL, Korea) and cordierite (Eastking Industrial Limited, China) were used as starting powders and labeled M and C, respectively.
Their median diameters (D 50 ) of the mullite and cordierite powders were 3.96 µm and 4.23 µm, respectively. The starting powders were then subjected to attrition milling for 24 h to investigate the effect of the particle size and size distribution on the formation of the granules. After that, the slurry was added to the slurry as the sintering aid, and a small amount of MgO powder was also used as the sintering aid. Then, the slurry was ball-milled with ZrO 2 balls for 3 h to form a uniform dispersion and spray dried. At the end of spray drying process, the dried granules were sieved and then uniaxially pressed into cylindrical pellets (36 mm in diameter and 6 mm in thickness) at 80 MPa and sintered at a temperature ranging between 1300℃ and 1450℃ for 4 h. The physical and mechanical properties of the resultant mullite-cordierite composites were analyzed to determine the optimum mixing ratio and the sintering temperature. The processing work ow for mullite-cordierite composites is shown in Fig. 1.

Fabrication of large-area ceramic composite substrate
To manufacture a large-area ceramic composite substrate for the semiconductor probe card, the granules of the mullite-cordierite composite with the optimal composition ratio were uniaxially pressed at 80 MPa into ceramic substrates with a diameter of 320 mm and sintered at 1350℃ for 4 h. Subsequently, the substrates were subjected to surface grinding to reduce the shrinkage and warpage of the ceramic substrate during the sintering process. Considering the frequency and amplitude of the ultrasonic generator, the size of the processing pin and the processing pressure, ultrasonic drilling was performed and 30000 micro-holes were processed through it.

Characterization procedure
The average particle size of the starting powders was measured by a particle size analyzer (PSA, Bluewave, MICROTRAC, Japan). The microstructure of the mullite-cordierite composites was observed by eld-emission scanning electron microscopy (FE-SEM, MIRA3 XMU, TESCAN, Czech Republic). The bulk density and the porosity were determined by the static weighing method according to ASTM C20 [20]. The TECs of the mullite-cordierite composites were measured with a thermomechanical analyzer (TMA, Q400, TA Instrument, USA) in a temperature range of 25-500℃ using a 5℃/min heating rate.
To evaluate the durability and the reliability of ceramic substrates made of mullite-cordierite composites, the specimens were prepared with dimensions of 80 mm 80 mm by processing the fabricated ceramic substrates. The environmental tests such as high temperature storage test, damp heat test, and thermal shock test were conducted to determine the changes in the properties of the specimens under the environmental conditions that cause failure. The high temperature storage (HTS) test was measured in an oven at 200℃ for 100 h. Damp heat (DH) test was conducted at 85℃ and 85% relative humidity for 100 h in a chamber (WKE 64/70, WEISS, USA). The thermal shock (TS) test was carried out for 100 cycles using a thermal shock chamber (TSA-41L, ESPEC, Japan). One cycle consists of heating the substrate up to 85℃ with a hold time of 30 min, followed by cooling down to -40℃ with a hold time of 30 min.
After the environmental tests, the sheet resistivity and exural strength of the ceramic substrate were measured. The sheet resistivity was measured using a high resistance meter (Hiresta-UX MCP-HT 800, MITSUBISHI CHEMICAL ANALYTECH, Japan) at 1000 V. Flexural strength was assessed by a three-point exural test using the universal testing machine (UTM, INSTRON, NVLAP, USA). The rectangular bars were prepared with dimensions of 3 mm 4 mm 36 mm, and the crosshead speed was set to 0.5 mm/min. The three-point exural strength ( ) was calculated according to the following formula : where F is the load, L is the length of the span, b is the width of the specimen, and d is the specimen thickness. The exural strength data was computed based on an average of ten measurements of the specimens.

Formation of mullite-cordierite composites granules by spray drying process
The sintering process is used to produce mullite-cordierite composites with high mechanical strengths, low TECs, and dense structures. In the sintering process, several variables, such as the particle size distribution and morphology of the powders, sintering temperature, and sintering aids, affect the properties of the sintered body. The particle size distribution of the powders greatly in uences the reaction rate as well as the sintering behavior. For example, in coarse powders having a low initial density due to non-uniform size and high agglomeration, the particles are separated by a large distance, resulting in a low densi cation rate. Hence, it is necessary to control the particle size distribution of the starting powders.
The particle size distributions of the starting powders before and after attrition milling are shown in Fig.  2. As shown in Fig. 2(a), for the mullite powder, the peck of the particle size distribution curve (which has a unimodal distribution) moved toward the smaller particle size, and the D 50 value decreased from 3.96 to 1.75 µm after attrition milling. As shown in Fig. 2(b), the particle size distribution of the cordierite powder shifted from a bimodal distribution to a unimodal distribution, and the D 50 value decreased from 4.23 to 3.23 µm after attrition milling. The SEM images in Fig. 2 indicate that the initial powders with various particle sizes became homogeneous powders with uniform and small particle sizes following the milling process. It was expected that the starting powders with sizes over 10 µm would be crushed during the milling process. Figure 3 shows the microstructure of the granules formed by spray drying the mixed powders of mullite and cordierite before and after attrition milling. It shows the dependence of the particle size distribution of the mullite and cordierite powders on the formation of granules. Granules formed using the initial starting powders before attrition milling were found to have different morphologies (Fig. 3(a)). In contrast, the granules formed using starting powders with homogeneous size distribution obtained through the attrition milling process were found to be spherical ( Fig. 3(b)). Moreover, the D 50 value of the unmilled mixed powders (147.3 µm) was larger than that of the attrition milled mixed powders (88.31 µm). The span de ned as (D 90 -D 10 ) / D 50 , indicates the width of the particle size distribution. The unmilled mixed powders had a larger span. The span before the attrition milling was 1.22 and span after the attrition milling was 0.71. These results show that the particle size and the size distribution of the initial powder affects the morphology of granular powders obtained by spray drying. In a previous work, it was shown that the granules obtained from ner initial powders and powders with a narrow size distribution formed smoother and denser structures, when compared to the large initial powders. Furthermore, the spherical granules of the powders have good owability, which is an important factor in the fabrication of dense and mechanically strong ceramics [21]. Therefore, the spherical granules obtained by spray drying the ne powders after attrition milling are suitable for fabricating the mullite-cordierite composite pellets and substrates.
3.2 Characterization of the sintered mullite-cordierite composite pellets produced by varying the cordierite content It is important to determine the optimal composition ratio and sintering temperature to improve the mechanical and physical properties of the sintered mullite-cordierite composite pellets before fabricating a large-area composite ceramic substrate. First, the effect of the composition ratio on the sintered mullitecordierite composite pellets was characterized. Figure 4 shows the SEM micrographs of the fractured surfaces of the mullite-cordierite composite pellets with various cordierite weight ratios (0, 10, 30, 50, 70, and 100 wt%) sintered at 1300℃. A porous structure was observed in the sample with 100 wt% cordierite, while the sample with 100 wt% mullite exhibited a more compact structure. For samples of pure cordierite and mullite, the density (cordierite = 2.55 g/cm 3 and mullite = 3.02 g/cm 3 , respectively) was measured to be similar to the average theoretical density (cordierite = 2.53 g/cm 3 and mullite = 3.17 g/cm 3 , respectively) [22], which is attributed to the structure of the samples, as shown in Fig. 4(a) and 4(f). As the mullite content in the composite pellets increased, the density noticeably increased as shown in Fig.  4(c)-(e), compared to Fig. 4(b). Thus, increasing the content of mullite not only accelerates densi cation during the sintering process, [23] but also increases the overall density of the samples as the density of mullite is higher than that of cordierite. Furthermore, a previous study reported that in batches with > 90 wt% cordierite, the cracks formed by cristobalite transition from β to α weakened the structure of the sample [22]. Figure 5 shows the effect of the cordierite content on the TEC and the exural strength of the sintered composite pellets. Generally, to improve the thermal shock resistance, the ceramic substrate for probe cards requires a low TEC similar to Si. As shown in Fig. 5(a), the TEC of the samples decreased with increasing cordierite content. The sample containing 70 wt% cordierite was observed to have the lowest TEC (2.88 ppm/℃) among all the samples (except the pure cordierite pellet) and exhibited excellent thermal shock resistance. However, compared to the TEC of Si (3.2-3.9 ppm/℃) [24], it can be assumed that all composite pellets with 30-70 wt% cordierite have good thermal stability. As shown in Fig. 5(b), the result is evidently observed that the exural strength of the samples containing up to 30 wt% cordierite is higher than that of the samples containing over 50 wt% cordierite. This is because mullite has higher mechanical strength than cordierite. Thus, the exural strength of samples with a large amount of mullite is higher. In addition, the gure shows the correlation between density and exural strength. The exural strength increases with increasing density as shown in Fig. 4 and Fig. 5(b). Therefore, it can be concluded that the composition ratio of 70 wt% mullite and 30 wt% cordierite is optimum for forming mullite-cordierite composites due to its low TEC and high exural strength. Figure 6 shows the TEC and the exural strength of sintered mullite-cordierite composite pellets of 70:30 wt% for different sintering temperatures ranging from 1300℃ to 1450℃. As shown in Fig. 6(a), the TEC of the samples sintered at 1300℃ and 1350℃ was 3.45 ppm/℃ and 3.42 ppm/℃, respectively, which are similar to the TEC of Si. A signi cant increase in TEC can be observed for the samples sintered at 1400℃ and 1450℃. This is probably due to the thermal stress induced by the mismatch of the TEC between the glassy phase and the crystalline phases of mullite [25]. This result explains why the sample sintered at 1350℃ is better able to withstand thermal stresses, compared to others samples prepared at different sintering temperatures. The physical and the mechanical properties, such as the density, porosity, exural strength, and TEC, of the mullite-cordierite composites (70 wt% mullite-30 wt% cordierite) for different sintering temperatures are listed in Table 1. The density at 1350℃ is higher than that at 1300℃. As the sintering temperature increases, some cordierite changes into glassy phase, which lls into the porosity of the mullite-cordierite composites [4,22]. Over 1400℃, however, all the cordierite melts and the sample consists of only the glassy phase, and the mullite content leads to decrease in density, leading to the large size of pores and the formation of macro cracks [25]. There is a high correlation between the density and the exural strength, with higher density leading to higher exural strength.  Table 1. The mullite-cordierite composite pellet sintered at 1350℃ exhibits the most compact structure, which improves its mechanical properties. It can be con rmed that the exural strength is the highest at 1350℃ (264 MPa). The composite pellets sintered at other temperatures except 1350℃, by contrast, show low exural strength due to their relatively low density. Table 1 Bulk density, porosity, exural strength, and thermal expansion coe cient (TEC) of the mullite-cordierite composites at different sintering temperatures.

Characterization of mullite-cordierite composite pellets sintered at different temperature
3.4. The reliability and durability of the large-area mullitecordierite composite substrate Based on the above results, a large-area mullite-cordierite composite substrate was fabricated by sintering at 1350℃ using granules having a composition ratio of 70 wt% mullite-30 wt% cordierite. The large-area ceramic composite substrate is 320 mm in diameter. The thickness of the substrate is approximately 5.7 µm, and the relative standard deviation of the thickness is 0.07%, which means that it has a highly uniform thickness as shown in Fig. 7(a). To form the ne pitch of the probe card, 30000 micro-holes were processed in the large-area substrate through ultrasonic drilling. The size and position deviation of the holes were measured to investigate the shrink rate of the large-area substrate that occurred during the sintering process, as shown in Fig. 7(b) and 7(c). The SEM images in Fig. 7(b) show the cross sections of the ceramic substrate with the processed micro-holes. The measured average sizes of the top, middle and bottom section of the processed holes were 416.3 µm, 400.1 µm and 387.8 µm, respectively, having a deviation of < 5%. Figure 7(c) shows enlarged SEM images of the four sections marked #1 to #4 in Fig. 7(a). The distance between the holes was measured in four places in each of the SEM images, and the standard deviation of the distance was found to be < 3%. These results reveal that the processed holes of the large-area ceramic substrate are formed evenly and that the shrinking of the substrate hardly occurred during the sintering process.
In order to evaluate the reliability and the durability of the large-area ceramic composite substrate, the substrate was cut into specimens with dimensions of 80 mm 80 mm and subjected to environmental tests, such as the HTS test, DH test, and TS test. Figure 8 shows the changes in the sheet resistance of the specimens during the environmental tests. In the HTS test, the specimens were placed in an oven at 200℃ for more than 100 h. There is no signi cant difference between the changes in the sheet resistance of the specimens before and after the HTS test. It is clear that the ceramic substrate is stable even when exposed to high temperatures for a long time. In the DH test, the specimens were exposed to a relative humidity of 85% at 85℃. The rate of change in resistance was nearly constant. It can be seen that the absorption of moisture hardly occurred due to the high temperature and the high humidity. Compared to the other environmental tests, a larger change was observed in the sheet resistance of the specimen when it was subjected to TS test of over 100 cycles in a temperature range from − 40℃ to 85℃. The change in sheet resistance was caused by the impact of the thermal shock. However, the rate of change in the resistance was not large, which means that ceramic substrate was not signi cantly affected by sudden temperature changes. In Kiattisaksophon et al. [4], a mullite-cordierite composite of 30:70 wt% was subjected to the TS test by heating the sample to 1100℃ and cooling to room temperature. No damage was observed on the surface of the composite body until the 8th cycle, but the spot of micro-cracks was observed in the 9th cycle. Compared to previous studies [4,9,14,24,26], the TS test was performed in a harsher environment at low temperature, although the applied temperature range is slightly different. In all the environmental test results, the fabricated specimens exhibited a low resistivity change < 10%, and no micro-cracks on the surface of the sample were observed even after 100 cycles. These results prove that the ceramic substrate has excellent reliability and durability.
The in uence of the environmental tests on the exural strength of the ceramic substrate is shown in Table 2. Specimens with dimensions of 3 mm 4 mm 36 mm were prepared for the measurement of exural strength, and the result was computed as the average of ten measurements of the specimens. On comparing the exural strength of the specimens before and after each environmental test, the exural strength of all the specimens was observed to decrease slightly after the environmental tests. It can be assumed that the thermal stress can generate an impurity, which affects the physical properties [26]. In the research of Cheng et al. [27], it was reported that the exural strength of the specimen with the best thermal property was about 90 MPa and the loss rate of exural strength was 13.12%, after 30 thermal shock cycles. Compared to the composites reported in previous studies [27,28], the fabricated large-area mullite-cordierite composite substrate has excellent mechanical strength. This is because it has a high exural strength of 200 MPa or more and the rate of change in the exural strength is less than 10%, after 100 cycles of all environmental tests. Consequently, the large-area substrates fabricated using spherical 70 wt% mullite-30 wt% cordierite granules with homogeneous particle size distribution were not considerably affected after environmental tests to accelerate fatigue failure. These results prove the high reliability and durability of the fabricated composite.

Conclusions
The sintered mullite-cordierite composite pellets were produced from spherical composite granules, which were prepared by spray drying the ne mixed powders obtained from attrition milling. The composition ratio and the sintering temperature were varied to obtain the composite pellet with excellent physical and mechanical properties. In particular, the mullite-cordierite composite pellet containing 70 wt%-mullite and 30 wt%-cordierite sintered at 1350℃ showed excellent performance. The fabricated composite has a bulk density of 3.07 g/cm 3 , a porosity of 0.19%, a TEC of 3.42 ppm/℃, and exural strength of 264 MPa. Although the composite pellet with 30 wt% cordierite sintered at 1300℃ was a higher TEC than the composite pellet with 70 wt% cordierite, 70 wt% mullite-30 wt% cordierite was chosen as the best composition ratio because its TEC is within the allowable range for the probe card substrate, and its exural strength is higher than that of samples containing other amounts of cordierite. Based on this result, a large-area ceramic composite substrate with a diameter of 320 mm and a thickness of 5.7 µm was successfully fabricated. Moreover, the rate of change in sheet resistance and the exural strength was con rmed to be within 10% in the fabricated ceramic substrate after exposure to different environmental tests, such as HTS test, DH test, and TS test, indicating good reliability and durability.
Thus, the mullite-cordierite composite can be considered as a suitable substrate material for the probe card. Figure 1 Flow diagram of the experiment procedure.      The ceramic substrate fabricated by using mullite-cordierite composite containing 70 wt% mullite-30 wt% cordierite sintered at 1350℃.