NIed. & Biol. Eng. & Comput., 1977, 15, 712-715
Technical note Silicone-rubber adhesives as encapsulants for microelectronic implants; effect of high electric fields and of tensile stress 1 Introduction
WE HAVE found silicone-rubber adhesives, such as Midland Silicones 'Silastoseal E', Dow Corning 3140 and Dow Coming 734, to be satisfactory encapsulants for thick-film microcircuits such as the small radio-stimulating receiver shown in Fig. 1. Similar receivers have been used for visual prosthesis (DONALDSON, 1973), cerebellar stimulation for relief of epilepsy (BRINDLEY, FALCONER et al., 1977) and ventral root stimulation for the control of incontinence (unpublished) in man. A batch of six receivers is operating faultlessly after three years submersion in saline in vitro. The maximum electric field present is of the order of 2x 10= V/m.
Fig, 1 Implantable r.f, receiver which gives a satisfactory life when encapsulated in silicone-rubber adhesive. The outside diameter of the printed coil is 10.7 mm. The maximum electric field is about2 • 10 a V/m
We have sought also to encapsulate in a similar manner a thick-film intersection unit (DONALDSON, 1973 ; DONALDSON and SAVER, 1973; IVALL, 1975) in which switching is performed by sealed-junction beam-lead bipolar transistors (Figs. 2a and b). The expense and additional complexity of placing the microcircuit in a hermetic enclosure would be avoided if a successful silicone-rubber encapsulation could be achieved; unfortunately, this has proved difficult and this note is a brief account of the difficulty. 2 Likely sources of trouble
There are two differences between the conditions under which the radio receiver operates, and those under which the intersection units operate, which are likely to be relevant. One is that, in the latter, the electric field reaches, in some regions, a much higher value, of the order of 106 V/m, because of the close spacing of the beams of the transistors; the gap between adjacent
712
beams is in some cases as small as 30Fn (Fig. 3). The other is that, because silicone-rubber adhesives shrink slightly on cure (about 1-5% linear for Dow Coming 3140), when an intersection unit is encapsulated in a rigid protective tray (Fig. 2a) a system of tensile stre3s is set up. The stresses will be distributed in a way which depends on how the encapsulation is carried out, but generally will tend to pull the rubber away from the surface of the circuit components. If this happens, the water vapour present in the rubber, as a result of implantation, can condense in the microscopic voids so formed; eventually, liquid will cause a short circuit at some vital point, or electrolysis of conductors, and the unit will fail.
Fig 2a Implantable thick-film bank of switching circuits, an "intersection unit" which does not always give a satisfactory life when encapsulated in silicone-rubber adhesive. The substrate is 25 mm long, The maximum electric field is about 106 Vim, The protective case is of glassfibre-reinforced d.a,p. Wires are p.t.f.e.-coated platinum
........................................................................... .9...............",................;...............,'................,',,..~ Fig. 2b Theoretical circuit of the intersection unit. n-p-n- transistors and diodes. B T 2483. p - n - p transistors, B T 2604. A l l resistors are 10 k ~ + 2 0 % . The unit comprises 20 2-input A N D gates, one input to all gates being common. In fabricating 20 gates, the aim is to obtain 19 g o o d ones
Medical & Biological Engineering & Computing
November 1977
The high electric fields follow from the voltages at which the intersection unit is required to operate (about 30) and the arrangement of the beams in the limited
Fig. 3 Sealed-junction transistor of the type used. The small scale divisions are I 0 0 ttm apart
range of sealed-junction transistor chips commercially available; there is nothing that can be done about them. The tensile stresses in the rubber, however, should depend on the encapsulation procedure. We have looked at the life obtained from intersection units, operated under saline, which had been encapsulated by a variety of procedures to see what correlation, if any, exists.
Method-0: The microcircuit, bearing conventionallymounted (face-down) transistors, is placed in the bottom of a diallyl-pthalate (d.a.p.) tray and silicone rubber is poured over until the tray is full. Bubbles are removed by vacuum eentrifugation (DONALDSON and SAYE~, 1975). In the crucial region of high electric field between the beams of the transistors (Fig. 3), contraction of the rubber upon cure can occur neither in directions parallel to the substrate, nor perpendicular to it. Method-l: As for method-0, but the transistors are mounted inverted (face-up). Beams are gold-wire bonded down to conductor pads on the substrate. In regions of high field, contraction can now occur perpendicular to the plane of the substrate, but not parallel to it. Method-2: The microcircuit, bearing conventionally mounted transistors, is encapsulated while supported in a temporary p.t.f.e, tray. After vacuum centrifugation and cure, the microcircuit is removed, allowing contraction to occur in the rubber in directions parallel to the substrate, but not in regions of high field, perpendicular to it. The assembly is then glued permanently into a d.a.p, tray which it closely fits. As a result of the thin glue line, curing of the glue produces negligible further stress. Method-3: As for method-2, but the transistors are mounted inverted. In regions of high electric field, contraction of rubber is now possible both parallel and perpendicular to the substrate. These four encapsulation procedures go some way toward allowing stress-relieving to occur in the encapsulating rubber in 0, 1, 2 and 3 orthogonal directions, respectively.
3 Encapsulation procedures
These have been discussed elsewhere 1976). Recapitulating briefly, they are:
i lo0 _,,~r~.q~.,~.e.a._. . . . . . . . ~ _.~.~..~.4
Z
.
_
No. o f test
! : .** .....
1200
-Z 9 , ,
IC
(DONALDSON,
--
*
~
_--
9
--
A-
.
.
.
--
A-
.
.
.
--
* * *
.
.
.
_--
~ .~
-
&
--
--
.
9
----
.
9
-- --
.
.
-- __
"
120
NO. o f hours
period~
0
1 Nurnberof
~lr 'new'3140
2 shrink
9 734
|
-
3
directions
A 'o|d'3140
Fig. 4 Working lifetimes of submerged intersection units, encapsulated in one of three materials and by one of four procedures. Tested at 30 V, 1:10markspaceratio
Medical & Biological Engineering & Computing
4 Test procedure
Early in this work it became clear that, in microcircuits destined to fail when submerged, what was damaging was not so much the fact of submersion as of submersion coupled with the fields consequent upon electrical use. For example, two from a batch of four circuits encapsulated by method-0 were submerged and electrically tested continuously; both failed within 10 days. The other two were submerged and left for four months. At the end of that time they were tested and found to work correctly, Testing was continued and these circuits also failed within a further 10 days. It seemed therefore that microcircuit life should be measured in terms of submerged operating time. For lack of test equipment we were unable to test all circuits continuously; we tested them instead in batches of six to each of two testers. The testers ran all the time, so each microcircuit operated for one sixth of the time, 12 hours on, 60 hours off. Operating life was taken as the number of hours running submerged before the first fault appeared somewhere in the microcircuit. 4.1 W h a t is a reasonable life on test? A visual prosthesis has to deliver stimuli of up to about 30 V amplitude with mark-space 1:10. We aim at a life of 5 years or more. Assuming a patient uses his prosthesis for 2 ' 4 h out of the 24, and if the prosthesis delivers pulse trains having gaps of four train-lengths between successive trains, then the total time for which the prosthesis will be delivering stimuli will be 0.1 year, about 1000 h. In the event, microcircuits which survived 1200h submerged operation (switching 3 0 V pulses November 1977
713
with mark-space ratio of 1:10) were deemed to have shown a satisfactory life; 1200 h is convenient as representing 100 12 h operating periods, extending over a period upwards of seven months.
the probability that substrates having numbers of transistors other than nine will run successfully for 1200h (Fig. 5a). This predicts that, for example, a simple 1.0
4.2 Experimental material 35 substrates were used in these experiments Although each is intended ultimately to support 19 gates, in order to conserve stocks, transistors were in fact only connected into four gates, so that each substrate bore nine transistors (Figs. 2a and b). Most were encapsulated in Dew Coming 3140; a few were in 734. Some effects of choice of encapsulant used are discussed below. Most were submerged in saline at room temperature; a few were in saline maintained at 37~ Our results do not show any difference in lifetime depending on which temperature was used. 5 Results Results are shown in Fig. 4. A relation between lifetime and number of shrink directions is evident. For example, if one considers geometric mean lifetime, taking the life of circuits which exceed 1200 h as equal to 1200 h, we have:
number of shrink directions: geometric mean lifetime, h
0
1
2
3
45
188
213
343
suggesting that relieving tensile stress in the encapsulant in as many directions as possible is beneficial. However, if one asks the question what fraction of circuits achieve the required lifetime, as a function of the number of shrink directions, we find: number of shrink directions: 2 3 proportion of circuits achieving required life: 4/11 3/10 suggesting that stress relief is better at preventing short lives than at promoting long ones. The results are complicated by the fact that, while this work was being carried out, Dew Corning 3140 became temporarily unavailable, and it was necessary to resort to old material which had exceeded its shelf life. Re-examining the results for circuits with two and three shrink directions, broken down by encapsulatirlg material, gives rise to the results in Table 1 : These results suggest (albeit tentatively in view of the small number involved): (a) 'New' 3140 is a better encapsulant than 'old' 3140 or 734. (b) Using new 3140, it is helpful to relieve stress in as many directions as possible. (c) A circuit in new 3140, maximally stress relieved, has a probability of 0.67 of achieving the required life.
0"5 A Probability Of survival
0.1
10
20
30
40
Number of t r a n s i s t o r s
Fig. 5 Probability of achieving a given working life with different numbers of transistors, given that the probability of achieving that life with nine transistors is (1t) O"67 and (13) O. 56 implantable animal microstimulator, comprising three transistors and made according to this technology, will run for 1200 h with probability 0-87, a reliability which many physiologists would accept. On the other hand, the probability of getting a substrate with 19 gates (39 transistors) to run for 1200 h submerged is only 0'17, and the probability of an entire visual prostheses, containing perhaps 16 such 19-gate substrates, lasting 1200 h is negligible. Since the success of two circuits out of three is scant evidence on which to assess a technology, one may, with a little more confidence, take the figure of 5 out of 9, for substrates having either two or three shrink directions encapsulated in new 3140. This gives Fig. 5b. Evidently the prospects for visual prosthesis are now more dismal than ever, but the probability of success for the simple 3-transistor stimulator is still better than 0" 8. 7. Mo:tes of failure Dew Corning 3140 and 734, although at first transparent, arc inclined to turn cloudy after prolonged
6 Discussion
Accepting for the moment that (c) above is justified, and assuming that the probability of failure of a substrate depends on the number of transistors on it and that all transistors are equally likely to fail, one can easily plot
Table 1. Proportion of circuits achieving required life No. of shrink directions Material 734 'Old" 3140 "New' 3140
714
2 0/1 1/4 3/6
3 1/5 0/2 2/3
total 1/6 1/6 5/9
Fig. 6 Appearance of transistor in a failed gate. Note the growth bridging the gap between collector and base beams. Picture taken through encapsulating rubber
Medical & Biological Engineering & Computing
November 1977
immersion in saline, so that when something goes wrong with a circuit encapsulated in one of them, it is not always possible to see what has happened. We have nevertheless managed to distinguish three modes of failure, as follows: (i) Breakdown in the high-field region between transistor beams, accompanied by dendritic growth. Figs. 6 and 7 show working faces of transistors where such
Fig. 7 Another failed transistor, again photographed through encapsulating rubber, A much wider growth has occurred here from co/Jector beam to emitter metal//sat/on
breakdowns have occurred. The point of failure is clear; a chocolate-brown growth, probably gold chloride, has formed between one beam and its neighbour. (ii) Breakdown in regions of moderate (10 s V/m) field between Pt-Au conductors on the substrate (Fig. 8). Such failures are puzzling, since we know from other work that thin-film (gold on tantalum on glass) test gaps, insulated with silicone-rubber adhesive and submerged, operate for long periods at fields as high as 2 . 5 x 106 V/m. The occasional failure of thickfilm Pt-Au gaps on ceramic at l0 s V / m remains to be investigated. (iii) Faults unaccompanied by any apparent physical deterioration in the circuit or its encapsulation; the circuit behaves as if one or other of the transistors had a collector-base breakdown. They occasionally disappear after a day or two, sometimes temporarily, sometimes permanently. More often, a fault appears permanently and does not disappear if the circuit is removed from saline and dried out; we assume that contamination of the transistor has occurred, perhaps by sodium ions, as a result of a defect in the siliconnitride coating. 8 Conclusions
Silicone-rubber adhesives are reliable encapsulants for microelectronic implants at field strengths of the order of 10" V/m. Thick-film/nitrided-beam-lead/silicone-rubber-adhesive technology provides a fairly simple way of making very small implants and, at voltages of 3 or 4 V, such implants are probably reliable. However, as one raises the voltage to 30 or 40 V, unreliability begins to
Medical & Biological Engineering & Computing
appear as insulation failure in regions of field strength 105-106 V/re. Reliability is nevertheless probably still adequate for very simple implants, but insufficient for complex ones. For complex, high-voltage implants, microcircuits should be hermetically enclosed to keep water vapour away from regions of high electric field. Outside the hermetic package, the gaps between conductors entering the package--a region filled with silicone
Fig. 8 Breakdown at the substrate/rubbef interface, A mass of dark material has migrated into what should be an insulating gap. Photographed through encapsulating rubber
rubber--should be chosen for a maximum field strength in the range 10~-10 5 V/m. Where hermetic encapsulation is not used, but a rigid outer case is employed to provide mechanical protection, some thought should be given to the relief of tensile stress in the encapsulating rubber. P. E. K. DONALDSON E. SAYER
M R C Neurological Prostheses Unit I Windsor Walk, London SE5 8BB, England.
References
DONALDSON, P. E. K. (1973) An experimental visual prosthesis. Prec. 1EEL 120, 281-298. BRINDLEY, G. S., FALCONER, M. A., FENTON, G. W., FENWICK, P. B. C., POLKEY, C. E. and RUSHTON, D. N., (1977) Chronic cerebellar stimulation in the treatment of generalised epilepsy: a case report. E.E.G. Clin. Neurophysiol. 41, 539-540. DONALDSON, P. E. K. and SAYER, E. (1973) Toward a high-definition visual prosthesis. IERE. Conference on Hybrid Microelectronics, Canterbury, 261-263. ]VALE, Z. E. (1975) Artificial vision progresses. Wireless World 81, 156-157. DONALDSON, P. E. K. (1976) The encapsulation of microelectronic devices for long-term surgical implantation. IEE Trans. BME-23, 281-285. DONALDSON, P. E. K. and SAYER,E. (1975) A vacuum centrifuge for void-free potting of implantable hybrid microcircuits in silicone. Med. & Biol. Eng. 13, 595596.
November 1977
715
Technical note Silicone-rubber adhesives as encapsulants for microelectronic implants; effect of high electric fields and of tensile stress 1 Introduction
WE HAVE found silicone-rubber adhesives, such as Midland Silicones 'Silastoseal E', Dow Corning 3140 and Dow Coming 734, to be satisfactory encapsulants for thick-film microcircuits such as the small radio-stimulating receiver shown in Fig. 1. Similar receivers have been used for visual prosthesis (DONALDSON, 1973), cerebellar stimulation for relief of epilepsy (BRINDLEY, FALCONER et al., 1977) and ventral root stimulation for the control of incontinence (unpublished) in man. A batch of six receivers is operating faultlessly after three years submersion in saline in vitro. The maximum electric field present is of the order of 2x 10= V/m.
Fig, 1 Implantable r.f, receiver which gives a satisfactory life when encapsulated in silicone-rubber adhesive. The outside diameter of the printed coil is 10.7 mm. The maximum electric field is about2 • 10 a V/m
We have sought also to encapsulate in a similar manner a thick-film intersection unit (DONALDSON, 1973 ; DONALDSON and SAVER, 1973; IVALL, 1975) in which switching is performed by sealed-junction beam-lead bipolar transistors (Figs. 2a and b). The expense and additional complexity of placing the microcircuit in a hermetic enclosure would be avoided if a successful silicone-rubber encapsulation could be achieved; unfortunately, this has proved difficult and this note is a brief account of the difficulty. 2 Likely sources of trouble
There are two differences between the conditions under which the radio receiver operates, and those under which the intersection units operate, which are likely to be relevant. One is that, in the latter, the electric field reaches, in some regions, a much higher value, of the order of 106 V/m, because of the close spacing of the beams of the transistors; the gap between adjacent
712
beams is in some cases as small as 30Fn (Fig. 3). The other is that, because silicone-rubber adhesives shrink slightly on cure (about 1-5% linear for Dow Coming 3140), when an intersection unit is encapsulated in a rigid protective tray (Fig. 2a) a system of tensile stre3s is set up. The stresses will be distributed in a way which depends on how the encapsulation is carried out, but generally will tend to pull the rubber away from the surface of the circuit components. If this happens, the water vapour present in the rubber, as a result of implantation, can condense in the microscopic voids so formed; eventually, liquid will cause a short circuit at some vital point, or electrolysis of conductors, and the unit will fail.
Fig 2a Implantable thick-film bank of switching circuits, an "intersection unit" which does not always give a satisfactory life when encapsulated in silicone-rubber adhesive. The substrate is 25 mm long, The maximum electric field is about 106 Vim, The protective case is of glassfibre-reinforced d.a,p. Wires are p.t.f.e.-coated platinum
........................................................................... .9...............",................;...............,'................,',,..~ Fig. 2b Theoretical circuit of the intersection unit. n-p-n- transistors and diodes. B T 2483. p - n - p transistors, B T 2604. A l l resistors are 10 k ~ + 2 0 % . The unit comprises 20 2-input A N D gates, one input to all gates being common. In fabricating 20 gates, the aim is to obtain 19 g o o d ones
Medical & Biological Engineering & Computing
November 1977
The high electric fields follow from the voltages at which the intersection unit is required to operate (about 30) and the arrangement of the beams in the limited
Fig. 3 Sealed-junction transistor of the type used. The small scale divisions are I 0 0 ttm apart
range of sealed-junction transistor chips commercially available; there is nothing that can be done about them. The tensile stresses in the rubber, however, should depend on the encapsulation procedure. We have looked at the life obtained from intersection units, operated under saline, which had been encapsulated by a variety of procedures to see what correlation, if any, exists.
Method-0: The microcircuit, bearing conventionallymounted (face-down) transistors, is placed in the bottom of a diallyl-pthalate (d.a.p.) tray and silicone rubber is poured over until the tray is full. Bubbles are removed by vacuum eentrifugation (DONALDSON and SAYE~, 1975). In the crucial region of high electric field between the beams of the transistors (Fig. 3), contraction of the rubber upon cure can occur neither in directions parallel to the substrate, nor perpendicular to it. Method-l: As for method-0, but the transistors are mounted inverted (face-up). Beams are gold-wire bonded down to conductor pads on the substrate. In regions of high field, contraction can now occur perpendicular to the plane of the substrate, but not parallel to it. Method-2: The microcircuit, bearing conventionally mounted transistors, is encapsulated while supported in a temporary p.t.f.e, tray. After vacuum centrifugation and cure, the microcircuit is removed, allowing contraction to occur in the rubber in directions parallel to the substrate, but not in regions of high field, perpendicular to it. The assembly is then glued permanently into a d.a.p, tray which it closely fits. As a result of the thin glue line, curing of the glue produces negligible further stress. Method-3: As for method-2, but the transistors are mounted inverted. In regions of high electric field, contraction of rubber is now possible both parallel and perpendicular to the substrate. These four encapsulation procedures go some way toward allowing stress-relieving to occur in the encapsulating rubber in 0, 1, 2 and 3 orthogonal directions, respectively.
3 Encapsulation procedures
These have been discussed elsewhere 1976). Recapitulating briefly, they are:
i lo0 _,,~r~.q~.,~.e.a._. . . . . . . . ~ _.~.~..~.4
Z
.
_
No. o f test
! : .** .....
1200
-Z 9 , ,
IC
(DONALDSON,
--
*
~
_--
9
--
A-
.
.
.
--
A-
.
.
.
--
* * *
.
.
.
_--
~ .~
-
&
--
--
.
9
----
.
9
-- --
.
.
-- __
"
120
NO. o f hours
period~
0
1 Nurnberof
~lr 'new'3140
2 shrink
9 734
|
-
3
directions
A 'o|d'3140
Fig. 4 Working lifetimes of submerged intersection units, encapsulated in one of three materials and by one of four procedures. Tested at 30 V, 1:10markspaceratio
Medical & Biological Engineering & Computing
4 Test procedure
Early in this work it became clear that, in microcircuits destined to fail when submerged, what was damaging was not so much the fact of submersion as of submersion coupled with the fields consequent upon electrical use. For example, two from a batch of four circuits encapsulated by method-0 were submerged and electrically tested continuously; both failed within 10 days. The other two were submerged and left for four months. At the end of that time they were tested and found to work correctly, Testing was continued and these circuits also failed within a further 10 days. It seemed therefore that microcircuit life should be measured in terms of submerged operating time. For lack of test equipment we were unable to test all circuits continuously; we tested them instead in batches of six to each of two testers. The testers ran all the time, so each microcircuit operated for one sixth of the time, 12 hours on, 60 hours off. Operating life was taken as the number of hours running submerged before the first fault appeared somewhere in the microcircuit. 4.1 W h a t is a reasonable life on test? A visual prosthesis has to deliver stimuli of up to about 30 V amplitude with mark-space 1:10. We aim at a life of 5 years or more. Assuming a patient uses his prosthesis for 2 ' 4 h out of the 24, and if the prosthesis delivers pulse trains having gaps of four train-lengths between successive trains, then the total time for which the prosthesis will be delivering stimuli will be 0.1 year, about 1000 h. In the event, microcircuits which survived 1200h submerged operation (switching 3 0 V pulses November 1977
713
with mark-space ratio of 1:10) were deemed to have shown a satisfactory life; 1200 h is convenient as representing 100 12 h operating periods, extending over a period upwards of seven months.
the probability that substrates having numbers of transistors other than nine will run successfully for 1200h (Fig. 5a). This predicts that, for example, a simple 1.0
4.2 Experimental material 35 substrates were used in these experiments Although each is intended ultimately to support 19 gates, in order to conserve stocks, transistors were in fact only connected into four gates, so that each substrate bore nine transistors (Figs. 2a and b). Most were encapsulated in Dew Coming 3140; a few were in 734. Some effects of choice of encapsulant used are discussed below. Most were submerged in saline at room temperature; a few were in saline maintained at 37~ Our results do not show any difference in lifetime depending on which temperature was used. 5 Results Results are shown in Fig. 4. A relation between lifetime and number of shrink directions is evident. For example, if one considers geometric mean lifetime, taking the life of circuits which exceed 1200 h as equal to 1200 h, we have:
number of shrink directions: geometric mean lifetime, h
0
1
2
3
45
188
213
343
suggesting that relieving tensile stress in the encapsulant in as many directions as possible is beneficial. However, if one asks the question what fraction of circuits achieve the required lifetime, as a function of the number of shrink directions, we find: number of shrink directions: 2 3 proportion of circuits achieving required life: 4/11 3/10 suggesting that stress relief is better at preventing short lives than at promoting long ones. The results are complicated by the fact that, while this work was being carried out, Dew Corning 3140 became temporarily unavailable, and it was necessary to resort to old material which had exceeded its shelf life. Re-examining the results for circuits with two and three shrink directions, broken down by encapsulatirlg material, gives rise to the results in Table 1 : These results suggest (albeit tentatively in view of the small number involved): (a) 'New' 3140 is a better encapsulant than 'old' 3140 or 734. (b) Using new 3140, it is helpful to relieve stress in as many directions as possible. (c) A circuit in new 3140, maximally stress relieved, has a probability of 0.67 of achieving the required life.
0"5 A Probability Of survival
0.1
10
20
30
40
Number of t r a n s i s t o r s
Fig. 5 Probability of achieving a given working life with different numbers of transistors, given that the probability of achieving that life with nine transistors is (1t) O"67 and (13) O. 56 implantable animal microstimulator, comprising three transistors and made according to this technology, will run for 1200 h with probability 0-87, a reliability which many physiologists would accept. On the other hand, the probability of getting a substrate with 19 gates (39 transistors) to run for 1200 h submerged is only 0'17, and the probability of an entire visual prostheses, containing perhaps 16 such 19-gate substrates, lasting 1200 h is negligible. Since the success of two circuits out of three is scant evidence on which to assess a technology, one may, with a little more confidence, take the figure of 5 out of 9, for substrates having either two or three shrink directions encapsulated in new 3140. This gives Fig. 5b. Evidently the prospects for visual prosthesis are now more dismal than ever, but the probability of success for the simple 3-transistor stimulator is still better than 0" 8. 7. Mo:tes of failure Dew Corning 3140 and 734, although at first transparent, arc inclined to turn cloudy after prolonged
6 Discussion
Accepting for the moment that (c) above is justified, and assuming that the probability of failure of a substrate depends on the number of transistors on it and that all transistors are equally likely to fail, one can easily plot
Table 1. Proportion of circuits achieving required life No. of shrink directions Material 734 'Old" 3140 "New' 3140
714
2 0/1 1/4 3/6
3 1/5 0/2 2/3
total 1/6 1/6 5/9
Fig. 6 Appearance of transistor in a failed gate. Note the growth bridging the gap between collector and base beams. Picture taken through encapsulating rubber
Medical & Biological Engineering & Computing
November 1977
immersion in saline, so that when something goes wrong with a circuit encapsulated in one of them, it is not always possible to see what has happened. We have nevertheless managed to distinguish three modes of failure, as follows: (i) Breakdown in the high-field region between transistor beams, accompanied by dendritic growth. Figs. 6 and 7 show working faces of transistors where such
Fig. 7 Another failed transistor, again photographed through encapsulating rubber, A much wider growth has occurred here from co/Jector beam to emitter metal//sat/on
breakdowns have occurred. The point of failure is clear; a chocolate-brown growth, probably gold chloride, has formed between one beam and its neighbour. (ii) Breakdown in regions of moderate (10 s V/m) field between Pt-Au conductors on the substrate (Fig. 8). Such failures are puzzling, since we know from other work that thin-film (gold on tantalum on glass) test gaps, insulated with silicone-rubber adhesive and submerged, operate for long periods at fields as high as 2 . 5 x 106 V/m. The occasional failure of thickfilm Pt-Au gaps on ceramic at l0 s V / m remains to be investigated. (iii) Faults unaccompanied by any apparent physical deterioration in the circuit or its encapsulation; the circuit behaves as if one or other of the transistors had a collector-base breakdown. They occasionally disappear after a day or two, sometimes temporarily, sometimes permanently. More often, a fault appears permanently and does not disappear if the circuit is removed from saline and dried out; we assume that contamination of the transistor has occurred, perhaps by sodium ions, as a result of a defect in the siliconnitride coating. 8 Conclusions
Silicone-rubber adhesives are reliable encapsulants for microelectronic implants at field strengths of the order of 10" V/m. Thick-film/nitrided-beam-lead/silicone-rubber-adhesive technology provides a fairly simple way of making very small implants and, at voltages of 3 or 4 V, such implants are probably reliable. However, as one raises the voltage to 30 or 40 V, unreliability begins to
Medical & Biological Engineering & Computing
appear as insulation failure in regions of field strength 105-106 V/re. Reliability is nevertheless probably still adequate for very simple implants, but insufficient for complex ones. For complex, high-voltage implants, microcircuits should be hermetically enclosed to keep water vapour away from regions of high electric field. Outside the hermetic package, the gaps between conductors entering the package--a region filled with silicone
Fig. 8 Breakdown at the substrate/rubbef interface, A mass of dark material has migrated into what should be an insulating gap. Photographed through encapsulating rubber
rubber--should be chosen for a maximum field strength in the range 10~-10 5 V/m. Where hermetic encapsulation is not used, but a rigid outer case is employed to provide mechanical protection, some thought should be given to the relief of tensile stress in the encapsulating rubber. P. E. K. DONALDSON E. SAYER
M R C Neurological Prostheses Unit I Windsor Walk, London SE5 8BB, England.
References
DONALDSON, P. E. K. (1973) An experimental visual prosthesis. Prec. 1EEL 120, 281-298. BRINDLEY, G. S., FALCONER, M. A., FENTON, G. W., FENWICK, P. B. C., POLKEY, C. E. and RUSHTON, D. N., (1977) Chronic cerebellar stimulation in the treatment of generalised epilepsy: a case report. E.E.G. Clin. Neurophysiol. 41, 539-540. DONALDSON, P. E. K. and SAYER, E. (1973) Toward a high-definition visual prosthesis. IERE. Conference on Hybrid Microelectronics, Canterbury, 261-263. ]VALE, Z. E. (1975) Artificial vision progresses. Wireless World 81, 156-157. DONALDSON, P. E. K. (1976) The encapsulation of microelectronic devices for long-term surgical implantation. IEE Trans. BME-23, 281-285. DONALDSON, P. E. K. and SAYER,E. (1975) A vacuum centrifuge for void-free potting of implantable hybrid microcircuits in silicone. Med. & Biol. Eng. 13, 595596.
November 1977
715
