2021 Volume 62 Issue 10 Pages 1583-1588
Low-melting glass with an optimal composition should be developed for application to Cu electrodes in multilayered ceramic capacitors (MLCCs). This study evaluated the glass transition temperature, plating solution resistance, and wettability of glass melts on Cu substrates for low-melting glass with a conventional composition and compositions of BaO, TiO2, ZnO, and V2O5 added to 46SiO2/27B2O3/27Na2O (mol%). Moreover, we produced a Cu electrode paste for an MLCC using the designed glasses and evaluated the characteristics in terms of terminal electrode sintering. The plating solution resistance was improved by adding TiO2 to 46SiO2/27B2O3/22Na2O/5BaO (mol%) glass. Furthermore, adding V2O5 improved the glass melt fluidity while maintaining the plating solution resistance, thereby improving the Cu base wettability. Compared with conventional glass, the 45.0SiO2/17.6B2O3/16.6Na2O/4.9BaO/2.1V2O5/8.8TiO2/4.9ZnO (mol%) composition formed glasses with low transition temperatures, good Cu wettabilities, and superior plating solution resistances. The fired film of the Cu electrode paste using this new glass was inferior in sintering as compared with the conventional glass and did not become densified. As a result of measuring the amount of carbon in the fired film, it was found that the new glass has a larger amount of residual carbon than the conventional glass. In addition, a reaction layer with ceramic is confirmed in the new glass, and there is concern about ceramic embrittlement due to crystal precipitation and cracks. For practical use of new glass as an electrode paste material, removal of residual carbon and suppression of ceramic reaction are considered to be future issues.
Low-melting glass is widely used in electric and electronics applications, including ceramic bonding and binders for metal powder sintering.1,2) The composition of low-melting glass used as an electronic component material is known to be SiO2–B2O3–PbO-based or AEO–B2O3–ZnO-based (AE: alkaline earth metal oxide). From the viewpoint of reducing environmental load and chemical stability, the development of new low melting point glass is desired.3)
One example of the use of low-melting glass is multilayer ceramic capacitors (MLCCs).4–6) Generally, MLCCs are composed of a ceramic dielectric, internal electrodes, and external electrodes, which are typically made of BaTiO3, Ni, and Cu, respectively.7) The typical problems with MLCCs are delamination and short-circuiting,8–10) which may be caused by factors such as faulty Cu adhesion between the external electrodes and BaTiO3 of the ceramic dielectric or the formation of blisters at the Cu–BaTiO3 interface and inside the Cu.11,12) In other words, the ingress of water or plating solution through gaps formed in the Cu-ceramic interface or in the Cu electrode film can cause various issues.7) Additionally, similar problems occur when, due to the inferior plating solution resistance of the binder glass itself, glass components leach into the plating solution, thereby reducing the adhesion of the external electrodes and ceramic dielectric. It is therefore advisable to use a glass composition with a high acid resistance and good wettability on the electrodes and ceramic; in addition, there must be no gaps in the Cu-ceramic interface or in the Cu electrode film. However, few studies have investigated the plating solution resistance of glass, and no reports are available on the wettability of oxide melts and metal powders, specifically of molten glass on solid Cu.
The purpose of this study was to develop low-melting glass with a new, optimum composition for electronic device applications. Conventionally, low melting point glass used for MLCC copper terminal electrodes includes barium borosilicate glass and zinc borosilicate glass, which have relatively good wettability with copper and easily form a dense film, but are weak in acid resistance.13) There was a drawback. In order to improve the acid resistance, the authors thought that it was necessary to be SiO2-rich as a component forming the glass former. However, since the SiO2 former has a rigid structure, it tends to raise the softening point of the glass and deteriorate the wettability to copper and ceramics. A composition range that can be vitrified even if it is rich in SiO2 is wide, and a composition that aims to lower the melting point was selected based on the SiO2–B2O3–Na2O composition, and additional components were added for the purpose of improving acid resistance and wettability.14) For glass physical characteristics, we measured the glass transition temperature, plating solution resistance, and wettability on Cu of SiO2–B2O3–Na2O glasses. We also evaluated the performance for actual use as an external electrode Cu paste for MLCCs.
For this study, we used samples with various oxides added to a ternary eutectic SiO2–B2O3–Na2O base glass composition and an actual glass sample for comparison. Table 1 lists the glass compositions. We weighed SiO2 (Sigma-Aldrich, 99.9% pure), Na2CO3 (Sigma-Aldrich, ≥99.5% pure), CaCO3 (Sigma-Aldrich, ≥99.0% pure), BaCO3 (Sigma-Aldrich, ≥99.0% pure), Al(OH)3 (Wako, ≥95.0% pure), TiO2 (Sigma-Aldrich, ≥99.0% pure), ZnO (Sigma-Aldrich, ≥99.0% pure), V2O5 (Sigma-Aldrich, ≥99.5% pure), and H3BO3 (Sigma-Aldrich, ≥99.5% pure) to the specified compositions and mixed them thoroughly using an aluminum mortar. The glass samples were produced by placing 20 g of the mixed powder in a platinum crucible, melting and retaining at 1200°C for 15 min under an air atmosphere, then pouring onto a Cu plate.
The glass transition temperature (Tg), a thermodynamic property of glass, was measured using a differential scanning calorimeter (DSC3300SA, Bruker). As samples for analysis, we used powder obtained by pulverizing the bulk glass in an agate mortar. Of this powder, 50 mg and an alpha-alumina reference sample were placed in separate platinum dishes, left to stand on a specified heat sink, and measured. The measuring conditions were controlled as follows: temperature elevation rate 20°C/min, measuring temperature range RT-1100°C, air atmosphere, flow rate 50 mL/min.
We evaluated the plating solution resistance based on the change in weight of the bulk glasses upon immersion in an Sn plating solution. The bulk glass was produced by placing the glass powder in a Pt crucible, melting in air at 1200°C, and pouring into a carbon mold. After annealing at 400–450°C, it was then cooled in a furnace. The obtained bulk glass was molded into approximately 10 mm × 10 mm × 10 mm cubes, and the surfaces were polished using waterproof abrasive paper. After calculating the surface area of the bulk glass using a micrometer, it was immersed in 100 mL of an Sn plating solution at pH = 3.4, and the weight variation was measured every 30 min for 180 min.
The contact angles of the glass melts on a Cu substrate were measured in a horizontal tube furnace with SiC heating elements, as shown in Fig. 1. The droplet shape of the glass melt allows for direct video capture from the viewing hole in the side of the tube furnace. The actual temperature during measurements was recorded by placing a Pt/Pt–13Rh thermocouple next to the sample in the furnace. We used N2–2%H2 gas for the experiment. A digital camera stabilized on a tripod was used for video recording, and the contact angle was measured by analysis of the captured images. The experimental procedure was as follows: the glass pellet was placed in the center of the Cu substrate, the Cu substrate was placed horizontally in the center of the horizontal furnace tube, and N2–2%H2 gas was introduced into the oven at 200 mL/min. The furnace was heated to 800°C and maintained at 800°C for 1 h. Images of the drop-shaped glass were acquired at 10-min intervals for 60 min.
Schematic illustration of gas-tight horizontal SiC furnace.
Table 2 lists the components of the Cu terminal electrode paste (hereafter referred to as “Cu paste”). Glass samples pulverized using a ball grinder into particles with a diameter of 3 µm were used as the glass frit. By dosing, mixing, and kneading each sample with three roll mills, we produced a Cu paste with evenly distributed components.
The prepared Cu paste was dipped on the end of a prepared MLCC bare chip with dimensions of 20 mm × 12 mm × 12 mm to form a 70–90-µm-thin layer after firing. The sample was maintained at 150°C for 10 min to produce a dried layer. Next, we used a mesh belt sintering furnace (47-MT-6841-20AMC-36, Koyo Thermo Systems) to fire the dried layer according to the temperature profile shown in Fig. 2 to obtain a fired Cu electrode. We observed the surface and cross-sections of the fired samples using a scanning electron microscope (TM4000, Hitachi) to evaluate the film structure and density.
Temperature and atmospheric conditions of tunnel furnace used for paste firing.
The dried Cu paste film was placed in a mesh belt sintering furnace, and the samples were removed from the furnace upon reaching 600, 700, or 800°C. The residual carbon in the Cu electrode film was measured using a carbon/sulfur analyzer (EMIA-320V, Horiba). A thermogravimetry/mass spectroscopy device (TG-MS; MS9610, Bruker AXS) was used to heat the dried Cu paste film in an He gas atmosphere for 10°C/min up to 900°C and to measure the generation behavior of mass number 44 (CO2).
Table 3 lists the Tg values of the studied glasses. The external Cu electrodes of MLCCs are commonly fired at approximately 800°C and must soften and melt below this temperature. The glass compositions in this study were all vitrified with Tg values corresponding to glass with a low melting point of approximately 500°C. Replacing the alkali metal oxide Na2O in the base glass with alkaline-earth metal oxide BaO resulted in densification of the glass structure and increased stability,15) e.g., chemical durability and non-crystallization, but resulted in a higher Tg due to the increased additives. Even upon replacing 9 mol% of B2O3 in 5Ba glass with TiO2, an amphoteric oxide, no increase in Tg was observed. Tg decreased by replacing 5 mol% of B2O3 in 5Ba glass with ZnO, another amphoteric oxide. The addition of ZnO at 5 mol% or higher tended to lower the reduction resistance.16) Furthermore, we found that Tg decreased by adding the amphoteric oxide V2O5 to 5Ba–9Ti–5Zn glass.
The plating solution resistances of the studied glasses are shown in Fig. 3. An increase in plating solution resistance compared with that of conventional glass was confirmed for all studied glass compositions. Glass component leaching was suppressed by replacing Na2O in the base composition with BaO, an alkaline-earth metal oxide because Ba2+ can link the non-bridging oxygens produced by silicon anions in the 3D network, and because in contrast to forming a –Si–O–Ba–O–Si– structure, Na+ disrupts the network. The addition of even a small amount of ZnO is known to have the same effect as that of BaO.17) Moreover, all compositions where 9 mol% of B2O3 was replaced with TiO2 had superior plating solution resistances with hardly any leaching. TiO2 is an amphoteric oxide with a redox equilibrium, where the Ti3+/Ti4+ equilibrium shifts to the low-valence side with increasing temperature and decreasing PO2 in the atmosphere.18) This is because TiO2 in the melted glass incorporates into the glass structure as Ti3+, making it more difficult to produce non-bridging oxygen. Additionally, the ionic radii of Ti ions are larger than those of Si ions, and since Ti ions bind weakly with O, it is believed that the plating solution resistance is improved without increasing Tg.
Time dependence of weight loss by plating solution for glasses.
The contact angle measurements of the glass melts on Cu substrates at 800°C are shown in Fig. 4. The wettability of melted glass with a Tg of 500°C or higher was poor, as shown in Table 3. In contrast, the contact angles with the Cu substrate were 90°C or less for the conventional glass with a Tg of 500°C or less, the base glass, and the 5Ba–9Ti–5Zn + 5 mass%V2O5 glass. V2O5 has been reported to reduce the surface tension of melted glass and improve its fluidity,19) and it is believed that, as a result, the wettability of the 5Ba–9Ti–5Zn + 5 mass%V2O5 glass was improved.
Time dependence of apparent contact angle of glasses on Cu substrate at 800°C.
Regarding the firing evaluation of Cu paste, the denseness of the film is a property that should be prioritized. In addition, it is desirable that there is no problem such as glass float or blister that hinders the function as an electrode. The SEM images of the fired surface and cross-sections of the Cu paste samples produced using conventional glass (paste A), base glass (paste B), and 5Ba–9Ti–5Zn + 5 mass%V2O5 glass (paste C) are shown in Fig. 5. Whereas paste A formed a relatively dense film, hardly any densification occurred for paste B. Paste C showed partial densification, but many voids were present, and most of the glass was unevenly distributed over the surface. It was clear from these results that the sinterability of the Cu film varies substantially depending on the glass composition. As shown in Table 2, the Cu powder, glass frit diameter, and additional binder were the same for the Cu pastes produced in this study. Furthermore, since no substantial difference was observed in terms of the wettability on Cu of the glass pellets for each of the three pastes, the large differences in the density of the Cu electrode films are believed to arise from the suppression of glass fluidity by residual carbon.
SEM image of the firing surface and cross section of a sample with a different glass composition fired at 780°C in a nitrogen atmosphere with an oxygen concentration of 5 ppm or less.
Figure 6 shows the SEM images of the fired surfaces and cross-sections of the samples that were pre-processed by maintaining at 450°C for 15 min in a humidified N2 atmosphere, and thus had lower carbon contents after firing at the temperature profile shown in Fig. 2. Some voids were present in the film of paste A, but a relatively dense film was formed. Regarding the increase in voids, we believe that the pre-processing conditions led to solid-phase sintering of the Cu powder into an aggregated state. In paste B, the Cu film densified; however, an uneven distribution of glass over the surface and separation of the ceramic-electrode interface due to electrode swelling were observed. Uneven glass distribution and electrode swelling are issues that occur during the densification of Cu films without the sufficient removal of carbon. Similar to paste A, voids were also observed in paste C; however, a relatively dense film was formed with no uneven glass distribution on the surface. Based on these results, we believe that pretreatment has reduced carbon, resulting in improved fluidity of glass, improved wettability with Cu powder, and film densification. For paste B and paste C, a reaction layer was confirmed at the ceramic-electrode interface. It is considered that the reactivity with the ceramic increased due to the improvement of the fluidity of the glass. The formation of the reaction layer is not preferable because it causes cracks and a decrease in electric capacity, and future improvement is required.
SEM image of the firing surface and cross section of a sample with a different glass composition that was decarbonized at 450°C for 15 hours in advance and then fired at 780°C in a wet nitrogen atmosphere.
Figure 7 shows the carbon residue in dried films using pastes A, B, and C removed from the firing furnace upon reaching 600, 700, and 800°C. Carbon residues originate from organic components such as resin and dispersants in the paste materials, as well as corrosion inhibitors on the surface of the Cu powder that failed to burn or decompose when fired in a non-oxidizing atmosphere. They remain in the film and thus carbonize. For paste A, a radical decrease in carbon occurred between 600 and 800°C, with essentially no carbon at 800°C. In paste B, little reduction occurred, and 137 ppm of carbon remained at 800°C. A radical reduction of carbon was observed for paste C between 600 and 700°C and to a lesser degree between 700 and 800°C. These results are consistent with the densification trends of the fired films and demonstrate that the reduction of carbon residue varies depending on the glass composition.
Residual carbon content of each fired paste sample at 600°C, 700°C, and 800°C under nitrogen atmosphere in a tunnel furnace.
Figure 8 shows the CO2 generation behavior of the dried films of pastes A, B, and C measured using TG-MS. The peaks up to 500°C represent the generation of gas (binder burn out) associated with the burning and decomposition of organic components. Above 500°C, gas generation ceased for some time, then began again above 600°C. The CO2 gas generation above 600°C is believed to be from scattered carbon residue that obtains oxygen from the surroundings. Gas generation peaks were seen for pastes A and C at approximately 650°C, but for paste B, a broad peak was present at a higher temperature of approximately 750°C. Of the three glass compositions, only paste B glass does not contain ZnO.
Mass peak of m/z = 44 (CO2) of each paste measured by quadrupole mass spectrometry.
In conditions where carbon is present, CO2 is generated through reduction reactions as follows.20)
| \begin{equation} \text{ZnO$_{\text{(s)}}$} + \text{C$_{\text{(s)}}$}\to \text{Zn$_{\text{(g)}}$} + \text{CO$_{\text{(g)}}$} \end{equation} | (1) |
| \begin{equation} \text{ZnO$_{\text{(s)}}$} + \text{CO$_{\text{(g)}}$}\to \text{Zn$_{\text{(g)}}$} + \text{CO$_{\text{2(g)}}$} \end{equation} | (2) |
With the aim of developing a new low-melting-point glass with an optimum composition for electronic devices, we conducted a study starting from a composition of SiO2–B2O3–Na2O and obtained the following findings. First, all glass compositions, including conventional glass, were vitrified low-melting glass with Tg of approximately 500°C. Second, the plating solution resistance was improved by adding TiO2 to 46SiO2/27B2O3/22Na2O/5BaO (mol%). Additionally, by adding V2O5, the glass melt fluidity was improved while maintaining the plating solution resistance, thereby improving the wettability on the Cu substrate. Further, compared with conventional glass, the 45.0SiO2/17.6B2O3/16.6Na2O/4.9BaO/2.1V2O5/8.8TiO2/4.9ZnO (mol%) composition has a lower Tg, better wettability with Cu, and superior plating solution resistance. Finally, A fired Cu terminal electrode paste using the optimized glass showed less Cu densification than conventional glass, glass floats were observed in the fired film surface, and more carbon residue was found in the fired film than in conventional glass. In addition, a reaction layer with ceramic is confirmed in the new glass, and there is concern about ceramic embrittlement due to crystal precipitation and cracks. For practical use of new glass as an electrode paste material, removal of residual carbon and suppression of ceramic reaction are considered to be future issues.
The authors would like to thank all members of the Shoei Chemical R&D department who were involved in the preparation of the experimental samples.
