SOME ASPECTS OF MULTILAYER CERAMIC CHIP CAPACITORS FOR HYBRID CIRCUITS

It is shown that both European and American standardization committees specify the dimensions in the same grid and that there is no standardization of the thickness in relation to the capacitance. Of the ceramic dielectrics, special attention is paid to the instability of capacitance and losses of type II materials due to temperature and voltage treatment. The conductivity of the inner electrodes determines the losses at MHz in type dielectrics. It is shown that the economically interesting partial replacement of Pd by Ag can cause an increase in the losses: other substitute metals seem more promising. The last aspect discussed is the end terminations of the capacitors. In the ease of silver/palladium end terminations the Ag/Pd ratio should be smaller than two.


INTRODUCTION
Multflayer ceramic chip capacitors (MCCCs, Figure 1) are an invaluable aid to efficient hybridization. This fact has been accepted by an increasing number of hybrid circuit manufacturers and, as a result, MCCCs are now to be found in a wide range of military and civil equipments. However, feedback from end.users does indicate that an explanation of certain facets of MCCC development technology would give the user a better insight into the apparently unpredictable behaviour of MCCCs. The aim of this paper is to help the user to appreciate some design considerations before formalizing his hybrid circuit.
Again, emerging from feedback, the following properties are of importance and will be discusssed:

DIMENSIONS
Initially, MCCCs were only used in highly professional fields; out of the hesitant growth of MCCCs a standardization of dimensions has been born. Both the EIA and IEC (the American and European standardization committees) now specify their preferred sizes in the same grid ( Figure 2). When the size of a capacitor chip is quoted, the first dimension is length, i.e. from end terminal to end terminal. Dimensions are given in 10 -2 inches, e.g. 0805: the length is 8 x 10 -2 inches and the width 5 x 10 -2 inches (2 mm and 1,25 mm respectively). For the thickness, only a maximum (1,9 mm) and a minimum (0,5 mm) value is specified.
It is noteworthy that to date there is no standardization of thickness specified in relation to capacitance. An end-user with a specific objective must therefore bear this point in mind when ordering control the properties of ceramic barium titanate are the application of additives to the material to prevent grain growth during firing. Another method is the partial substitution of barium and titanium oxide to shift the Curie temperature to the vicinity of room temperature to make use of the peak in the capacitance versus temperature curve ( Figure 3). The Curie temperature is that temperature above which the ceramic has a cubic structure; below the Curie point the material has a tetragonal structure in which the c/a ratio is 1,01. This means that one of the crystallographic axes is 1% longer than the other two. This coincides with the splitting up of the crystals in domains of the same crystallographic orientation, as shown in Figure 4. only are the crystals very fine, but also the peak in e(T) is suppressed by additives. The aging of the capacitance can be less than 1% per time decade (logarithmic scale).
The (re)orientation of the domain structure is influenced by the crystal size and the defect structure of the material. Figure 5 shows the capacitance of an X7R dielectric (specification: C -+ 15% between -55C and +125C) as a function of time. It can be seen that the aging rate is only 0,55%/time decade. Figure 6 shows the capacitance of a Z5U dielectric +22 (specification: C =_56-/o between +10C and +85C) as a function of time. This is a material in which the  characteristic is achieved by the partial replacement of BaO and TiO2 to shift the Curie temperature to the vicinity of room temperature. It can be seen that the aging rate of this Z5U dielectric is about 6%/time decade. Experiments were carried out to show the influence of heat and voltage treatment on the aging of capacity. Ten capacitors were each given a temperature treatment at 72, 500, 1000 and 2000 hours after the burn-in of the end terminals. This temperature treatment was at 250C for five minutes to simulate soldering conditions. The conclusion was that with a temperature treatment above the Curie temperature, the capacitance is completely rejuvenated regardless of the duration of shelf life. This temperature treatment (i.e. soldering) has the same effect on the dissipation factor (D.F.). Ten other capacitors were subjected to approximately +350 V d.c. for one minute at the same time points. This treatment also has a rejuvenating effect on the capacitance, but not as complete as the heat treatment. It should be noted that either a.c. or d.c. voltages have a rejuvenating effect on both capacitance and dissipation factor.

INNER ELECTRODES
The firing temperature of the ceramic dielectrics, which is between 1200 C and 1400 C, and the oxidizing circumstances, compel us to use a noble metal with a high melting point. Another factor determining the choice of electrode material is the kind of dielectric used. In ceramics in which no bismuth-containing compound is used palladium with its melting point of 1550C is attractive.
The electrical properties are of course affected by the electrodes. The continuity of the electrodes has an influence on the spread of capacitance and the conductivity determines the series losses. The losses due to series resistance are given by" tan coRsC (2) in which: tan i dissipation factor; co 2zrf (f frequency (Hz); Rs series resistance (2) C capacitance (F).
The electrode resistance is determined by grinding away one of the end terminations until the capacitor is short-circuited. It is then end terminated again. The resistance is measured by the four-point contact method as shown in Figure 7. The square resistance is calculated as follows: V I

RD=-x -xNs (3) I in which"
Rr square resistance in f',  The capacitance and losses of six experimental lots of NP0 capacitors (500-1000 pF) were measured on a Danbridge 1MHz capacitance deviation bridge CB1, and the Rr of the electrodes determined. Figure 8 shows the dissipation factor plotted against the square resistance; the straight line shows the relative proportionality. The main contribution to losses at this frequency is the resistance of the inner electrodes (the ceramic used is a secondary contribution).
Internationally accepted specifications state that at MHz the dissipation factor must be lower than 10 x 10 -4 at a maximum capacitance of 1000 pF.
With the rather arbitrary choice of a Rn < 100 m2, this demand is amply fulfilled.
The price of the electrode material greatly influences the product price, particularly with high capacitances. Recent developments speak of a partial replacement of Pd by Ag. 2 From the point of view of firing, there is no objection as can be seen in the phase diagram, Figure 9, for instance in materials with a firing temperature of 1250 C, 50% of the Pd could be replaced by Ag. However, from the point of view of conductivity, there might be objections. As can be seen in Figure 10 The silver-palladium phase diagram.
lOO Ag aim at the replacement of precious metal electrodes by base metals, e.g. nickel or cobalt. There could be no objections to the use of these base metals on the ground of series resistance as can be seen from the If the capacitor is to be glued onto the circuit, the question is how the bonding strength of the connection is influenced by the surface state of the end termination (which may be oxidized or sulphurized after shelf life). This raises a second question: whether or not the series resistance is increased and, if so, by how much.
In the functioning phase, the quality of package sealing is of importance. In a partially sealed package (i.e. one still subject to atmospheric conditions), there will be the possibility of silver migration. Hermetically sealed packages raise a degassing problem of both components and mounting material, e.g. remnants of flux in the solder bond, solving media or softener in the glued contacts and absorbed moisture.
Summarizing, one can say that in the development of end terminations, one has not only to deal with the adhesive force between the metal layers and the  In the second experiment, a method was chosen which was a closer adaptation for multilayer configurations. In this experiment, the voltage was 50 V d.c., the electrode distance 4 mm, and a drop of deionized water was carefully placed on the chip, avoiding contact with the end terminals.
The first series of experiments proved that a 5% replacement of Ag by Pd was sufficient to prevent migration. However, the second series, with condensation, proved that the Ag/Pd ratio should be <2. In the second test it was also found that silver migration was not stopped by protecting the end terminations with an electrolytic Ni layer.

STORAGE INFLUENCE
During storage, oxidization and sulphurization of the end terminals can occur. As can be seen in Figure 11, the resistance against sulphur increases with the Pd content, although at an Ag/Pd ratio of 2 the optimum is not reached. Using an Ag/Pd ratio of 2, we there- mounted using the techniques mentioned; a good shelf life is also guaranteed.