Two groups of faults have occurred in our totally implanted pacemakers. One of these has been due to shrinkage of the encapsulating medium (in this case, epoxy resin) during curing. The time taken to fully cure this resin is approximately nine to 12 months, and this is one of the reasons why some pacemaker faults develop around that time. The second group of faults has been due to the leakage of cell electrolyte reaching the electronic components. Ingress of body fluid accelerates the flow of this electrolyte along the connecting wires and onto the circuit. Fortunately, the body fluid is, by the time it has passed through only a thin membrane of resin, completely deionized and in itself is nonconducting. The rate of ingress of this body fluid varies with the resin used from as little as 1 mm of penetration in 12 months with the hard resins to complete saturation of a 2-cm cube in 12 months with the softer resins.FIGURE 1 shows the stress patterns around a glass-encapsulated diode. As a rough guide, the closer together the bands are, the greater the stress. Differential forces of over 200 pounds per square inch have been measured by semiconductor strain gauges. Under this sort of stress, mechanical damage can occur to the glass encapsulation of the diode. To eliminate this fault, the diode was encapsulated in silicone rubber prior to its assembly. This has the effect of smoothing out the stress pattern (FIGURE 2).The electrolyte, a 40% aqueous solution of sodium hydroxide, leaks from the cells, travels along the connecting wires or any interface, and eventually reaches the electronic components. This leakage was found to take place rapidly along the tinned-copper wire used for the battery connections, since the sodium hydroxide reacted with the tin and formed a tube around the copper wire down which the electrolyte could flow.Solder joints also present a problem, because of the affinity of the sodium hydroxide for the tin contained in the solder; thus, once the electrolyte reaches a joint, erosion takes place quickly. FIGURE 3 illustrates the effect of this erosion on the connecting wire to a mercury cell. The joint circled in the illustration is badly corroded and was initially a large area of bright solder, about twothirds of which has been corroded away.To eliminate all these faults, our first step was to place a discrete assembly in an hermetically sealed can. On the basis of available reject figures, this step alone would give a tenfold improvement in reliability. We then decided to produce the circuit in thin-film form, which, from the reliability figures available, would give a further fivefold improvement, In our current thin-film techniques, the basic component is the glass substrate with its associated resistive, capacitive, and interconnecting patterns. This substrate is manufactured automatically, enabling consistently reliable components to be economically produced. The components are deposited under closely controlled vacuum conditions. The material used for the substrate is borosilicate glass, the t...