Soldering of Ceramics with Metal

V. A. Presnov, M. P. Yakubenya

Introduction

Metal-ceramic constructions obtained by soldering ceramic parts to metal parts are finding ever wider application in the electrovacuum and capacitor-manufacturing industries. With the proper selection of metal and ceramic parts and their good soldering, metal-ceramic constructions have the following advantages:

1) resistance to the action of high temperatures and thermal shocks;
2) high mechanical strength;
3) high stability under the action of electrical voltages, both high-voltage and high-frequency;
4) high dielectric properties.

The concept of a “soldered joint of ceramics with metal” goes beyond the ordinary concept of a “soldered joint,” which is applied, for example, to the soldering of two metals with some solder.

The process of soldering metal to ceramics is a complex process, often multistage, associated with the necessity of preliminary metallization of the ceramics (for some methods). The physical phenomena that occur during soldering have still been studied very poorly.

The present article sets forth certain data obtained in the study of the processes of soldering steatite ceramic VK-92 to metal.

Methods of Soldering Metal to Ceramics

The coating of ceramics with glaze and colored metals for decorative purposes has been known to mankind since the most ancient times. However, the methods of coating ceramics used for decorative purposes could not satisfy the requirements for the technical coating of ceramics with metal. Vacuum-tight soldered joints of ceramics with metal were obtained only 15 years ago. During this time several different methods have been proposed for obtaining a hermetic and mechanically sufficiently strong soldered joint of ceramics with metal.

a) Soldered joints based on metal oxides

If a metal oxide, for example, $\mathrm{Cu}_2\mathrm{O}$, is applied to ceramics and fired in the appropriate manner, diffusion of the oxide into the ceramics occurs; at the same time the oxide of the metal interacts with the corresponding components of the ceramics, forming a physicochemical compound.

By subjecting the oxide-treated ceramic surface to the action of a reducing atmosphere, a thin layer of metal is obtained, firmly held on the ceramic surface. A ceramic surface treated in this way can be soldered to a metal part by the usual methods for soldering metal surfaces.

b) Solders Based on Pressed Powders

It is possible to obtain a strong bond between metal and ceramic by pressing metal powder together with a raw (unfired) ceramic body.

After the metal powder has been pressed into the ceramic part, the latter is fired in a specially selected gaseous medium that prevents complete oxidation of the metal portions of the part. When this method is used, special attention must be paid to the proper selection of materials.

The metal and the ceramic must have identical thermal-expansion characteristics over a wide temperature range. Poor selection of the components leads to the formation of cracks arising as a result of thermal stresses that occur during cooling of the part after firing. The method of joining ceramics to metal by pressing in metal powders can be successfully applied in cases where subsequent mechanical machining of the ceramic part is required.

c) Solders Based on Titanium Hydride and Active Metals

Ceramics can be soldered directly to metal by using titanium hydride or an active metal (zirconium, titanium, etc.). For soldering, the ceramic is coated with the hydride of an active metal and assembled with the metal part together with a hard solder. The assembly thus prepared is heated in vacuum or in an atmosphere of a pure, dry inert gas to the melting temperature of the solder. At the same time, the hydride decomposes according to the reaction: \(\mathrm{ZrH_2 \to Zr + H_2}\), \(\mathrm{Ag + Zr \to ZrAg}\), with formation of an alloy of the active metal with the solder. The active metal acts on the ceramic and metal surfaces in such a way that the liquid solder simultaneously wets both the ceramic surface and the metal surface, soldering them together.

The presence of hydrogen (released during decomposition of the hydride) is not required in this process. The active metal itself, pre-alloyed with silver, gives the latter the property of wetting both the ceramic surface and the metal surface when the silver is melted in high vacuum.

Recommended compositions are: 5–15% silver and 95–85% zirconium. The zirconium–silver system has the property of providing good wetting and soldering on zirconium and alumina ceramics. Soldering is carried out in vacuum or in a well-purified dry inert medium.

The technique of soldering ceramics to metal with active alloys is very attractive, since soldering takes place directly (in a single-stage process). This method will undoubtedly find broad practical application after it has been improved to permit operation in furnaces with gaseous media that ensure reduced evaporation of the metal and increased productivity.

d) Soldering Using Metallization of Ceramics with Powders of Refractory Metals

The method of metallizing ceramics with powders of refractory metals (for example, molybdenum) was first developed by the German firm Telefunken [1]. The technology of ceramic metallization subsequently underwent a whole series of modifications and is now widely used with the use of

by using various powders of refractory metals: tungsten, molybdenum, rhenium, with additions of other metals such as iron, nickel, cobalt, copper, etc.

Obtaining a ceramic-to-metal joint by this method is carried out in three stages.

1. Coating the ceramic with a paste made from molybdenum or tungsten powders, followed by their sintering in a specially selected gas atmosphere consisting of hydrogen and nitrogen with the addition of a small amount of air at a temperature of 1300–1350°C.

2. Coating the molybdenum or tungsten with nickel. Usually this is done by a similar method, i.e., first the layer of molybdenum sintered onto the ceramic is coated with a paste of nickel powder, and then it is sintered in a hydrogen atmosphere at a temperature of 1000°C.

3. Soldering the metallized ceramic part to the metal part by means of silver or eutectic solder.

The described method of soldering a metal part to a ceramic one makes it possible to obtain vacuum-tight joints that ensure preservation of a high vacuum for a long time during repeated heating and cooling of the joint.

The author of the method, Pulfrich, emphasized the importance of the chemical composition of the ceramic for obtaining a strong joint with metal. It was considered that the good adhesion of zirconia-containing ceramics to metal is explained by a chemical reaction between zirconium oxide and the intermediate metal. However, Pulfrich did not indicate to what oxide the intermediate metal must be oxidized, and did not detect a transition layer between the metal and the ceramic.

Later, Novodvorskii [2] showed that, taking Pulfrich’s method as a basis, metal can be soldered not only to zirconia-containing ceramics.

Another modification of Pulfrich’s method was proposed by Nolte [3] and consists in adding manganese to the molybdenum.

As for the mechanism of soldering, up to the present time several hypotheses have been advanced. In addition to Pulfrich’s hypothesis concerning a chemical reaction between zirconium oxide and the intermediate metal, another hypothesis was proposed, according to which the metal diffuses into the ceramic and the bond is created by a special arrangement or rearrangement of crystals in the transition layer [4]. However, this hypothesis does not explain the fact that there is no strong bond between the ceramic and the metal when molybdenum is fired in a strictly reducing atmosphere.

The following hypothesis [4] consists in assuming solid-phase reactions due to the loss of atomic bonds under thermal excitation and the mutual diffusion of various atoms in the contacting parts of the component. In our opinion, the latter hypothesis is of interest. There are very few experimental data relating to the study of the nature of ceramic-to-metal soldering, since until now researchers have been interested in technological questions.

The hypotheses listed proceed from the fact that a strong bond between ceramic and metal is accompanied by a transition layer whose composition and structure differ. The absence of experimental data proving the presence of a transition layer has led some investigators to the thought that the bond of metal with ceramic is effected by forces of surface adhesion and mechanical interlocking. A bond of this kind is effected without the formation of a transition layer of altered composition and structure.

Thus, for the study of the soldering of ceramics with metal, the question of the presence or absence of a transition region in the joint is of primary interest. In the event of a positive answer to this question

it is necessary to ascertain the paths of formation and the structure of the transition region. The authors set themselves the task of answering these questions.

As the objects of investigation, joints of metal with steatite vacuum-tight ceramics, obtained by the method of metallization with refractory metals (molybdenum), were selected. The indicated ceramics are used for the manufacture of envelopes of metal-ceramic radio tubes. The choice of the ceramics and of the metallization method was dictated above all by production requirements.

Results of the investigations

Microscopic investigations and measurements of the microhardness of the ceramics convince us that the ceramics adjacent to the metal change their properties in the course of metallization (Figs. 1–3). The color of the ceramics adjacent to the metal changes (Fig. 1), its microhardness increases (Fig. 2), as does its mechanical strength. The latter conclusion may be reached by examining Fig. 3, which shows a photograph of an internal joint of ceramics with metal. Indeed, calculation of the thermal stresses in the joint shows [5] that the stresses are maximal at the boundary between the ceramics and the metal. If, however, rupture of the joint occurs at a distance of up to 100 mic-

Fig. 1. 1 — ceramics, 3 — metal.

ron from this boundary, it is natural to assume that the increase in the strength of the boundary layer of the ceramic occurred as a result of metallization of the ceramic. (In Fig. 3, the crack in the ceramic is clearly visible.)

Fig. 2.
1—non-metallized ceramic, 3—metallized ceramic.

Vertical axis: \(H,\ \mathrm{kg/mm^2}\).
Horizontal axis: distance from the edge of the specimen in microns.

As specially conducted experiments [6] showed, metallization also changes the dielectric properties of the ceramic. Numerical values of the measurement of dielectric losses are given in Table 1. It should be noted here that, during metallization of the ceramic, there is only a slight change in the numerical value of \(\operatorname{tg}\delta\) under cor-

Fig. 3.
1—ceramic, 2—silver, 3—molybdenum, 4—crack in the ceramic.

preservation of the general character of the dependence of \(\operatorname{tg}\delta\) on temperature and frequency (Figs. 4 and 5).

Table 1

The experimental data presented already indicate that, when steatite ceramics are metallized with powders of refractory metals, a change occurs in the properties of the ceramic layer adjoining the metal. Consequently, the experimental investigations indicate the forma-

Fig. 4. Dependence of \(\operatorname{tg}\delta\) on temperature

\(d = 0.34\) mm

\(\lambda = 300\) m

...of the transition region in the ceramic during its metallization. In this case, a quantitative increase in the substance of the transition region leads to a considerable increase in \(\operatorname{tg}\delta\) [6].

The transition region may be formed either as a result of simple diffusion of metal atoms into the ceramic, or as a result of more complex processes. Diffusion of molybdenum into the ceramic, however, does not ensure a strong bond between the parts being brazed, since when molybdenum is fired into the ceramic in a strictly reducing atmosphere, i.e., under conditions most favorable for diffusion, no strong bond between molybdenum and the ceramic is obtained.

For obtaining a strong braze, it is necessary, as shown by practical data and special experiments [13], to fire molybdenum into the ceramic in a weakly oxidizing atmosphere, in which oxidation of the molybdenum is possible.

Experimental data obtained in the study of brazing glass to metal show that, in order to obtain a good-quality braze, the metal must be oxidized to a definite type of oxide (for glasses with acidic properties, the best brazes are obtained when the metal is oxidized to oxides of lower valences).

Fig. 5. Dependence of \(\operatorname{tg}\delta\) on frequency

\(d = 30\ \text{m}\)
\(\lambda = 0.34\ \text{mm}\)

It was of interest to determine the influence of the type of metal oxide on the quality of the braze between ceramic and metal. To resolve this question, ceramic parts were metallized with pastes based on lower and higher oxides of molybdenum. The specimens were metallized in a weakly oxidizing atmosphere. If in the first case a strong braze was obtained, then in the second case it was not possible to obtain a strong braze (Table 2). Thus, it was established that, in order to obtain a strong braze, oxidation of molybdenum at the boundary with the ceramic to the basic oxide is necessary. This conclusion is also confirmed by X-ray structural-analysis data.

Table 2

Mechanical tensile strength of the braze

To prove that a strong braze of metal with ceramic is obtained as a result of an acid–base interaction of molybdenum oxides with the ceramic components, it is necessary to detect in the braze the product of this interaction, i.e., to establish the presence of a new substance in the transition...

region of the joint. We succeeded in solving this problem by carrying out an X-ray structural analysis of the specimens. Since we were working with polycrystalline substances, it was natural to adopt the powder method of analysis.

In all, during the course of the work about 500 X-ray photographs were taken, and the exposures in taking individual X-ray photographs reached 10–12 hours.

The interpretation of the X-ray photographs of the specimens was carried out by comparison with X-ray photographs of phases that could be expected in the composition of the specimens under investigation [14].

The analysis of the X-ray photographs was greatly complicated by the complex composition of the interacting components, as a result of which many lines were present on the X-ray photographs; moreover, the lines characterizing one substance were superposed on the lines of other substances. In addition, little of the new substance could form in the transition region, and therefore the lines characterizing it on the X-ray photographs could be weak and not present in their full number.

To solve this difficult experimental problem we approached the ceramic–metal boundary from two sides: from the metal side and from the ceramic side, by grinding the specimens.

An X-ray study of oblique polished sections of specimens of the ceramic-to-metal joint showed that, in the X-ray photographs of the transition region, along with lines belonging to reflections from known crystalline phases, new lines are observed—namely: 19.2, 20.1, 22.5, 25.1, 25.8, 26.5, 28.3, 29.0, 30.1 (Table 3). These lines cannot be explained by reflection of X-rays from the crystalline structures of the ceramic, molybdenum, and its oxides. In addition, the relative intensity of some of the lines belonging to the ceramic decreased. The latter effect is especially clearly expressed in X-ray photographs taken with chromium radiation [8].

If an increase in the relative intensity of one or another line may indicate that a line of another substance has been superposed on the line in question, then a decrease in relative intensity is associated with a rearrangement of the atoms in the crystalline lattice under investigation. This, together with the discovery of new lines on X-ray photographs of oblique polished sections of the joint, indicates that an interaction occurred between the oxides of molybdenum and the components of the ceramic, as a result of which a substance was formed with a crystalline lattice different from the crystalline lattices of the starting substances.

Having found the presence of a new substance in the transition region of the ceramic, it is natural to pose the following question: which constituents of the ceramic and of the metallic layer interact with the formation of a new reaction product?

Proceeding from the general hypothesis of acid–base interaction, it could be expected that the ceramic components with acidic properties enter into interaction with the basic oxides of molybdenum. The acidic properties in the ceramic studied are imparted chiefly by SiO₂ and B₂O₃, with SiO₂ being found in the ceramic in the form of a free oxide.

Therefore we carried out an X-ray structural analysis of special briquettes made from a mixture of MoO₂ and molybdenum with silica and boracite.

The study of the briquettes proved the possibility of the formation, under the conditions considered, of molybdenum silicates and borates (Tables 4 and 5).

The process of formation of molybdenum silicates and borates proceeds stepwise, through a series of intermediate active states [7]. Some of the lines obtained on the X-ray photographs in the study of oblique polished sections of the ceramic-to-metal joint coincide with the lines characteristic of molybdenum silicates and borates [7]. Thus, the X-ray studies confirm the idea that, on the ceramic side, into interaction with the bas—

Table 3

...with the basic oxides of molybdenum there enter components having acidic properties, such as silica and boric anhydride. At the same time, the possibility is not excluded that the basic oxides of molybdenum interact with other components of the ceramic.

The supposition, put forward by one of us as early as 1952, that the processes of soldering glass to metal and ceramics to metal have common features—namely, that in both the first and the second cases a strong bond between dissimilar materials is formed as a result of acid–base interaction of oxides—is confirmed experimentally.

This supposition is probably of a general character. It covers not only the case of soldering metal to ceramics possessing acidic properties, but also the case of soldering to ceramics possessing basic properties. For the latter, the metal fired into the ceramic must be oxidized to an acidic oxide. Pinkus [19], in investigating the soldering of metal to alumina ceramic, obtained results confirming this supposition.

Table 4

Another confirmation of this assumption is provided by the results obtained by Vyatkin for magnesite ceramics [10].

The principal result of our work is the substantiation of the assumption that a strong bond between glass and ceramics and a metal is ensured as a result of the chemical interaction of the components constituting the glass and ceramics with the metal oxides; in this process a transition layer is formed. The transition layer contains the products of the indicated chemical interaction. In contrast to the soldering of glass to metal, in the soldering of ceramics to metal solid-phase reactions play an essential role. The process of solid-phase interaction is a diffusion process. If the materials being sintered consist of a single substance (metal grains on the surface of the ceramic), then, when the temperature is raised and the mobility of the particles increases, the possibility arises of particle exchange, leading to the sintering of individual grains into a continuous metallic layer. The bond between the metallic layer and the ceramic is likewise effected as a result of diffusion processes, but they are more complex in comparison with the sintering of homogeneous metallic grains.

The steatite ceramic “VK—92,” which we investigated, as already noted, contains free silica, SiO₂, in its composition. Boracite is also introduced into the ceramic body. The oxides of silicon and boron, both in the solid and in the liquid state, have a quite definite distribution of valence chemical bonds between atoms. The basic structural unit of both silicate glasses and crystalline silica is the tetrahedron. Crystalline silica consists entirely of tetrahedra regularly arranged in space. Silicon is located at the nodes of the spatial lattice, and oxygen in the bridges. The character of the bonding forces is predominantly covalent, although there is also a certain proportion of an ionic component.

The possibility of diffusion of silicon atoms with complete rupture of all bonds, i.e., with complete departure from the tetrahedron, is extremely difficult; diffusion, however, is nevertheless possible, and in fact takes place under conditions of practically complete preservation of all valence bonds. A process of this kind of diffusion can be imagined if, following R. L. Müller, one allows for the possibility of switching (transformation) of valence bonds between atoms [11]^1). The switching of valence bonds is possible only under thermal excitation. The diffusion process, therefore, can be observed practically only at elevated temperatures. Along with the simplest switchings of neighboring valence bonds, rearrangements of atoms are possible, occurring owing to analogous incomplete ruptures. As a result, the activation energy for diffusion will be considerably lowered.

Admitting the possibility of diffusion of particles by the method of partial rupture and transformation of valence bonds, the process of a strong joining of ceramics to metal may be represented as follows.

The layer of metal deposited on the ceramic is partially oxidized on the side facing the ceramic. During oxidation of the metal, forces act that draw oxygen atoms into interaction with the crystal lattice of the metal. As a result of the interaction, a chemical compound of oxygen with the metal is formed. The crystal lattice of the metal in its surface layer thereby undergoes a certain change. The chemically reorganized cells are not separated by a sharp boundary from the more internal cells, and a crystalline bond is preserved between the former and the latter.

^1) The idea of a similar kind of diffusion was expressed to the authors by R. L. Müller in a private conversation.

Molybdenum + boracite

Table 5

$MoO_2$ + boracite

Between the metal oxide and certain components of the ceramic (silica, boric anhydride), interaction is possible, with the formation of the corresponding reaction products in the transition layer. Since both silica and boric anhydride are acidic oxides, the reaction of their interaction will proceed the more readily, the more basic the properties of the metal oxides. In the transition layer, silicates and borates of the metals will thereby be formed.

The interaction process may be imagined as a diffusion process, proceeding with partial rupture of valence bonds. The activation energy of such a diffusion process will be considerably less than the energy required for the complete tearing out of a given particle and its transfer to another place.

The process of formation of the reaction product in the transition layer of the ceramic may be represented visually as proceeding according to the scheme shown in Fig. 6.

The bond of the metal with the ceramic is effected by means of intermediate oxygen atoms according to the scheme shown in Fig. 7. Precisely because the bond is effected by means of an intermediate element—oxygen—partial oxidation of the metal layer facing the ceramic is necessary. The binding element may be not only oxygen atoms, but also other atoms, for example carbon, nitrogen, boron, capable of giving a paired-electron bond with metals.

\[ \begin{array}{ccccc} 2\,Me_{n}^{I}O_{m} &&&& 2\,Me_{p}^{II}O_{k}\\ \downarrow && \searrow\!\!\!\!\!\!\nearrow && \downarrow\\ Me_{n}^{I} &&&& Me_{p}^{II}\\ \hline Me_{n}^{I}Me_{p}^{II}O_{2m} &&&& Me_{n}^{I}Me_{p}^{II}O_{2k} \end{array} \]

Fig. 6.

\[ \begin{array}{c} \text{binding}\\ \text{element}\\[2mm] O - Me_{r} - O\\ \phantom{O -}\,|\\ -\,Me_{p}\ \left.\begin{array}{c}\\[-1mm] \\ \end{array}\right\}\ \text{oxide}\\ \phantom{-\,Me_{p}}|\\ O \end{array} \]

Fig. 7.

The structural features of the reacting substances, and the radii of action of the spheres of the chemical forces of interaction, play an important role in the process of forming a strong bond between ceramic and metal.

The mechanism of formation of a strong bond between ceramic and metal described in the present work has been experimentally confirmed in a number of investigations carried out on joints obtained by the method of metallizing ceramics with refractory metals \([7, 12, 13]\), and on joints obtained by another method \([10]\).

It should be noted, however, that the described mechanism of formation of a strong bond between dissimilar materials is by no means the only possible one, universal and applicable to all types of joints.

The joining of dissimilar materials in other cases is also possible as a result of the diffusional dissolution of one substance in another when they are heated, with the formation of solid solutions. A sufficiently strong joining of materials is also possible as a result of the action of surface-interaction forces.

The experimental data accumulated up to the present time are still insufficient to answer the question of the mechanism of formation of a strong bond in all known methods of soldering ceramics to metal.

Conclusions

On the basis of the investigations carried out, the following principal conclusions may be drawn:

1. A vacuum-tight joint between ceramics and metal is achieved as a result of acid–base interaction of the ceramic components with molybdenum oxides. The possible influence on the strength of the joint of the action of surface-adhesion forces and mechanical interlocking is of secondary importance.

2. In order to obtain a strong joint, the decisive factor is therefore the controlled oxidation of molybdenum to a definite oxide at the boundary with the ceramic. If the ceramic has acidic properties, then the surface of the molybdenum adjacent to the ceramic must be oxidized to a basic oxide; and, conversely, if the ceramic has basic properties, then the surface of the molybdenum adjacent to the ceramic must be oxidized to an acidic oxide. Probably, the greater the difference in the acidity–basicity of the properties of the interacting components, the stronger the joint obtained.

3. The source of oxygen for the oxidation of molybdenum is the gaseous medium. The preferential oxidation of molybdenum on the side adjacent to the ceramic is evidently promoted by oxygen adsorbed by the ceramic surface.

4. The principal phenomena occurring when ceramics are soldered to metal can be explained from the standpoint of diffusion theory. The diffusion mechanism of the joint is confirmed experimentally by X-ray studies.

5. As a result of the chemical interaction of the ceramic components with molybdenum oxides, a transition layer (transition region) is formed between the ceramic and the molybdenum, consisting of reaction products. The formation of the transition layer proceeds stepwise, through a series of intermediate active states.

6. The transition layer may contain molybdenum silicates and borates, as well as products of the interaction of molybdenum oxides with the more complex components of the ceramic.

Literature

1. H. Pulfrich—Ceramic–to–Metal Seals, U. S. Pat. 2163407, June 20, 1939.
2. Yu. B. Novodvorskii. Abstracts of reports of the interregional Siberian conference on questions of electrical insulation. Tomsk, SFTI, 1949.
3. H. I. Nolte and R. Spurch—Television Engineering, 1, November, 1950.
4. Bernside—Radio Corporation of America Review, march, 1954.
5. V. A. Presnov and E. N. Alekseeva. Transactions of SFTI, no. 36.
6. M. P. Yakubenya. Ibid.
7. V. A. Presnov, M. P. Yakubenya. Transactions of SFTI, no. 36.
8. M. P. Yakubenya. Ibid.
9. A. Pincus—Ceramic Age, march, 1954.
10. A. V. Dyatkin, V. A. Presnov. Transactions of SFTI, no. 36.
11. R. L. Muller. ZhPKh, 28, no. 7.
12. V. N. Zhdanova, M. P. Yakubenya. Transactions of SFTI, no. 36.
13. V. A. Presnov, M. P. Yakubenya, E. N. Alekseeva. Ibid.
14. V. I. Arkharov. Transactions of the Institute of Metal Physics of the Ural Branch of the Academy of Sciences of the USSR, issue 11, 31–43, 1950.

Siberian
Physico-Mechanical Institute.

Errata