Microassembly and micropackaging of implantable systems
The successful realization of an implantable sensor system for medical applications requires an overview on how individual electronic, electromechanical or mechanical components can be assembled and later packaged in order to allow a long failure-free operation inside the patient. This chapter provides insights into common microassembly methods and electronic packaging methods including hermetic packaging and non-hermetic polymeric encapsulation. The latter particularly requires the implant component technology and assembly methods to be selected very thoroughly in order to reliably protect the electronics from body fluids. Hence, microassembly methods and micropackaging concepts have to match in order to fabricate implants that withstand the harsh bodily environment for years or even decades.
Monitoring body parameters using wireless electronic implants started in the late 1950s. Heart rate, blood pressure, temperature and brain activity of various animal models were recorded. The simplicity of the circuits restricted assembly methods to soldering of discrete components (batteries, resistors, capacitors, coils, diodes, transistors) and the packaging consisted of casting the circuits in a polymeric encapsulant that functioned as a moisture barrier (Mackay, 1993). At the same time (1958), the first fully implantable cardiac pacemaker was developed and used (Zhou and Greenbaum, 2010). The large market for cardiac pacing was accompanied by the development of technology that worked reliably in electrolytic environments. The polymer packaging of implanted devices was supplanted by hermetic packages, protecting delicate circuitry against the corrosive effect of humidity. This was partially due to the lessons learnt in a field which developed in parallel, that of electronics for military aircraft and space flight applications, where very high-reliability is mandatory and electronic failure modes were extensively researched (Thomas, 1976). Today, pacemakers are highly developed electronic devices with multiple transmitting, sensing and stimulation features, fabricated by applying various assembly technologies, including microwelding, wire bonding and flip-chip bonding (Gossler, 2007). Among others, these technologies are state-of-the-art for fabricating miniature or micro implants for sensing and transmitting of various body parameters. In this chapter, common microassembly methods are presented, followed by an introduction into various aspects and methods of micropackaging.
Substrates can be fabricated from silicon, glass, ceramic as well as rigid and flexible polymers. The conductive elements of the substrates can be patterned with thick-film metallization (some μm to some 10 μm), or thin-film (some 10 nm to some 100 nm). Some substrate technologies even allow for integrated active and passive electronic components.
Bare integrated circuits (dies), housed integrated circuits (ICs) and surface-mount devices (SMD), such as capacitors, resistors, diodes, inductors, batteries, etc., are assembled for building the actual electronic circuitry.
A variety of different wire technologies is used for interconnection of components, for building antennas and coils, electrode contacts, etc. Some wires, for example bond wires, are bare metal while others are insulated.
In order to be able to integrate these components that are fabricated in a broad variety of technologies, a set of established assembly methods is available. The following section provides an overview of common methods suitable for assembly of implantable micro components.
The assembly of a variety of components to an implantable microsystem often requires the combination of multiple assembly methods. Depending on the component technology, one can usually choose from a small number of different methods, for example an SMD component might be soldered or conductively glued to a substrate; a wire might be soldered or welded, an IC might be wire bonded, flip-chip bonded or microflexed, etc. However, the freedom of choice is often restricted by the fact that some methods might not be combined with others easily or at all or that one method or another might not be compatible with the envisioned micropackaging concept. For example, under some circumstances, conductive gluing might not result in long-term stable contacts when encapsulated in a polymer but might be very reliable inside a hermetic package. Also, the different assembly methods might only be combinable in a certain order, especially when aggressive atmosphere (low or high pressures, plasma) or high temperatures that might damage individual components of the assembly are involved, for example, delicate sensors. Another criterion for selecting a method for microassembly of components is the minimum feature size or maximum integration density of individual electrical contacts. In the following sections, methods of microassembly are presented and, where applicable, examples of implantable microsystems that were fabricated by these methods are given as well as the state-of-the-art of the highest integration density achievable.
Adhesive bonding employs the use of a chemical agent to join two or more bond partners by establishing a mechanical link between their surfaces. The underlying bonding process can be surface adsorption (intermolecular or interatomic attraction forces between surface and adhesive), chemical bonding (van-der-Waals forces, valence bonds), diffusion (interdiffusion of adhesive into the structure of polymeric bond partners), electrostatic attraction and mechanical interlocking (Zhou, 2008).
Since successful adhesive bonding requires the bond surfaces to be in an appropriate condition, pre-treatments like abrasion, grit or shot blasting, laser ablation, etching, anodizing, plasma exposure, corona discharge, sonification and the use of primers or adhesion promoters are required. The most suitable has to be selected for a given adhesive in order to establish a bond of desired properties and to permit the adhesive to properly wet the surface of the bond partners (Zhou, 2008).
Adhesives are commonly applied to the target structure by manual or automated dispensing, by screen-printing, or by pin-transfer (Zhou, 2008). Most adhesives used today are polymers (omitting inorganic cements), which cure under heat, moisture, pressure or radiation (typically ultraviolet light). Besides the mechanical fastening, additional functionality can be achieved, such as electrical insulation (natural property of adhesives), electrical conduction (isotropic or anisotropic), heat conduction or light conduction (optical guide wire) by addition of a suitable filler material.
According to the literature (see following paragraph), the types of adhesive most commonly used in miniature implant assembly are cyanoacrylates (‘superglue’), epoxies and silicone. Each of these represents an entire family of chemicals. If the adhesive is used outside an implant’s hermetic package, an important criterion is its water absorption (leading to swelling), which is closely related to the ion concentration in the adhesive. Low ion concentration and low water uptake is usually required to ensure best stability over time. Information on the ionic content is given on the adhesive datasheet of some suppliers.
As examples for adhesive bonding in assembly of micro implants, Cameron et al. (1997) describe the use of cyanoacrylates for mechanical fastening of chip to carrier; silver-filled epoxy was used for establishing an electrical contact between bond wires and copper coil wires as well as for connecting iridium and tantalum wires to pads of ICs in the fabrication of BION® microstimulators. Anisotropically conductive adhesive was used in the manufacture of an implantable intraocular sensor system (Ha et al., 2010). Rubehn et al. (2011) used optically transparent epoxy for joining a polymeric microfabricated optical guiding structure assembled to a shaft-type flexible microelectrode with an optical connector for fibre optics, allowing the realization of multimodal optogenetic neural probes. Silicone adhesive is widely used in implant making as both an electrical insulator and an encapsulant (Donaldson, 1994; Rushton et al., 2002).
Plasma bonding is a technology often applied in the fabrication of micro-fluidic devices. The two bond partners, typically at least one of which is a polymer, are chemically joined by bringing them into close contact, after exposure to a plasma. Either one or both of the bond partners have to undergo plasma treatment before bonding.
A common application for plasma bonding is permanently covering a polydimethylsiloxan (PDMS) base structure that contains microflu-idic channels or similar structures with a PDMS foil functioning as a lid. Therefore, one or both surfaces to be bonded are exposed for some seconds to a low-power plasma (e.g. oxygen as reaction gas) causing the surfaces to be chemically activated. Solvents such as methanol, ethanol or isopropyl alcohol can be applied to the base structure onto which the lid is placed. The solvent causes the lid to float on the base structure, allowing some time to align the lid relative to the base. Eventually, when the solvent is evaporated, the lid establishes a close contact with the base and irreversibly bonds to it (Jo et al., 2000). In some cases, the application of pressure to ensure a tight contact between the bond partners is also required.
Numerous different (and also dissimilar) materials can be bonded by plasma bonding. Among other material combinations, the authors successfully applied this method in the fabrication of implantable microdevices for joining PDMS to PDMS (Schuettler et al., 2011b), PDMS to silicon, PDMS to glass (Guenther et al., 2008) and PDMS to polyimide (Schuettler et al., 2001). Unlike adhesive bonding, plasma bonding does not require an adhesive and is therefore safe to use when smearing of excess adhesive has to be avoided, for example in proximity to sensors or electrode contacts.
Soldering commonly serves two purposes at the same time: mechanical fastening as well as electrical joining. Soldering involves the joining of two or more metallic surfaces by a molten filler, leading to metallurgical bonds. By convention, if the filler melts below 450 °C, the process is called soldering (or soft-soldering); above this temperature, the process is called brazing (Humpston and Jacobson, 2004).
The filler material is commonly called solder and usually consists of a metal alloy, which traditionally comprises lead to keep the melting point low. Lead, however, was banned from use in solder for most applications by the introduction of the European Directive 2002/95/EC ‘RoHS’ (Reduction of Hazardous Substances) in 2004. Currently, one of the few exceptions named in this directive is the use in medical devices, although according to Directive 2011/65/EC, this exception expires by July 2014. AIMDs are further excluded, however, it is generally expected that this exception is also limited in time.
A good overview of solder materials and guidelines for selecting appropriate solder material is given by Martin (1999). Besides the mechanical stability (e.g. tensile strength), the thermal and electrical conductivity and other alloy-specific properties, the melting point plays an important role, and a solder material might be selected so that the melting temperature does not interfere with the preceding and remaining assembly steps. Common solder alloys melt between 180 °C and 250 °C, however, low-melting point solders based on bismuth or gallium are available that melt below 150 °C (Humpston and Jacobson, 2004). Although low-melting points might ease the soldering process, it should also be noted that after soldering, the solder should not be exposed to temperatures higher than 50 K below the melting point to avoid weakening the joints. This temperature might be reached during subsequent assembly or packaging steps or in the sterilization process.
Modern solders substitute lead by elements such as bismuth or indium to adjust the melting point of the alloy to the desired level. It is important not to mix these solders with lead-based solder, since this might result in an alloy with a very low and difficult to predict melting point and which thus compromises the solder joint reliability and reproducibility.
To allow molten solder to flow and to wet the surfaces to be joined, an additional agent is used, called flux. Flux is usually activated at soldering temperature and removes oxides from the surfaces to be wetted by the solder. Three types of fluxes are widely used: rosin mildly activated (RMA) fluxes, water-soluble fluxes and no-clean fluxes (Blackwell, 2000). Fluxes have to be removed after soldering in order to avoid corrosion and accelerated interme-tallic growth that may eventually lead to a bond failure (Licari and Enlow, 1998). Therefore, either water or solvent-based flux removers are to be used. The no-clean fluxes are designed not to be removed after soldering by giving them a very low activity and a chemistry that encloses all active agents after soldering thus passivating them (Blackwell, 2000). They still should be removed in implanted devices, however. It is of fundamental importance that implanted devices have very clean surfaces, especially before they are encapsulated in polymers, as explained later in Section 4.4.5.
• Soldering by using solder paste consisting of solder beadlets in a flux matrix, dispensed to the substrate. The component to be soldered is placed on the dispensed paste, whose adhesiveness keeps it in place. After all solder joints of a substrate are prepared like this, the entire substrate is heated above the melting temperature of the solder beadlets, causing the actual solder joints to establish. Heating is locally applied by the use of laser light or more generally by hot air stream or by using an oven or a hot plate.
• Soldering using a bath of molten solder. The components to be soldered are mechanically joined to the substrate by a thermally stable adhesive. Then, the substrate is moved upside-down over an edge of molten solder. The solder bath wets the metal surface, solidifies and establishes the joint.
When using solder in combination with thick-film or thin-film metallization, one must be aware of the tendency of the metallization film to diffuse into the solder. Soldering to gold pads, even of layer thicknesses around 10 μm and thicker, should be avoided. Platinum and platinum/gold is much more resistant to diffusion and should be used instead (Holmes and Loasby, 1976).
One has to decide, depending on the application, which of the techniques – solder alloys and fluxes – is most suitable. An example for solder joints made from a tin/lead/silver alloy (Sn62Pb36Ag2) in clinically used implants is the Sacral Anterior Root Stimulator, which has been successfully used since the late 1970s for treating urinary bladder and bowel incontinence in spinal cord injured patients (see Section 4.4.5). The same implant utilizes brazing for joining platinum-iridium (Pt80Ir20) wires to platinum electrode contacts using gold as filler material (Donaldson, 1987). The use of solder in implantable microsystems such as an implantable retina stimulator is shown in Fig. 4.1. A similar technological approach was taken to fabricate a nerve cuff electrode with multiplexer module attached to it (Fig. 4.3).
Fig 4.1 Implantable retina stimulator (type ELMA 1) of the German Retina Implant Project (sponsored by the German Federal Ministry of Education and Research), photography by AG1: Opthalmology and Technology. Left side: Coil and surface-mount components soldered to a flexible polyimide substrate. Right side: A stimulator chip is bonded to the substrate using the microflex technique. The entire system was coated in Parylene C and the receiver electronics and the stimulator chip were additionally casted in silicone rubber. Source: Reproduced with permission from IOP Publishing: Figure 12c in Stieglitz, 2009.
A challenge of microsystems technology is the connection to the macro world, for example to robust medical-grade cables. There are number of cable technologies available on the market with proven high-reliability and, if possible, one should take advantage of using them. Besides soldering, which has the drawback of connecting dissimilar metals in a potentially wet environment, welding of medical-grade cables is a good option. Different to soldering, welding causes both components to melt, forming a metallic transition phase. Microwelding does not usually require a filler material. The five most common microwelding techniques are:
1. Resistive spot welding. Welding of two metal parts by clamping both between two opposing electrode contacts while introducing an electrical current i (Fig. 4.2, left). The electrical resistance of the weld partners causes the current to develop heat, resulting in melting of the material (Zhou, 2008).
Fig 4.2 Joining two weld partners A and B by resistance welding. The black area indicates the location of the molten and solidified material. Left: Spot welding using two opposing electrodes, introducing a vertical current i. Right: Parallel gap welding permitting welds on insulating substrate material.
2. Resistiveparallel gap welding. Welding of two metal parts by pushing one part down onto the other using two parallel electrodes (Fig. 4.2, right). By setting the current amplitude and pulse slope, the welding current is directed into the interface layer between the two metals, melting both surfaces and establishing a weld (Johnson and Knutson, 1976; Schuettler et al., 2008; Harman, 2010).
3. Laser welding. Focused laser radiation, usually in the infrared spectrum, causes local heating of both weld partners causing them to melt. In laser microwelding, the beam is focused to a diameter of down to 10 μm, permitting very small welds (Zhou, 2008). Common laser welders are, for example, Nd:YAG lasers that operate at 1064 nm wavelength, emitting pulses of 100 μs to 100 ms width, radiating power in the range of 10–100 W. A disadvantage of laser welding (e.g. compared to resistance welding) is the absence of a clamping force to ensure a good contact between both weld partners (Schuettler et al., 2008; Zhou, 2008).
4. Electrical arc welding. For joining two very fine wires, the cut ends are placed face to face while the wires are electrically energized using opposing voltages, resulting in the formation of an electrical arc bridging the gap between the cut ends. The wires are moved towards each other forming a smooth melting zone (Zhou, 2008).
5. Electron beam welding. Electrons are accelerated to about 2/3 of the speed of light using an electrical field. The resulting electron beam is focused by a magnetic lens onto the target, to a spot of a few 10 μm in diameter, causing metal conductors to melt (Zhou, 2008). This method requires the welding target to be placed in a vacuum chamber and is used only rarely for interconnections but rather applied to joining metal pieces, for example, welding a lid to a case, forming a hermetic housing.
Welding works best when the two weld partners are identical materials. However, this is often impossible since different functional elements need to have different physical properties. Therefore, one has to carefully select the combination of materials and the method of welding. A broad overview of suitable combinations of implantable metals is given by Zhou (2008). Depending on the corrosion affinity of the metals to be joined, the welding has to take place in inert atmosphere, for example pure Argon shielding gas.
Figure 4.3 shows an example of parallel gap welding: for the fabrication of a multiplexer module mounted directly to a neural cuff electrode, platinum-iridium (Pt80Ir20) wires were welded to 30 μm platinum thick-film metallization printed on alumina. The screen-printed alumina acts as an adapter to mediate between a medical-grade cable and a flexible thin-film substrate, which was connected to the screen print by using microflex technology, as described later (Schuettler et al., 2000).
Fig 4.3 Concept of hybrid integration of different component technologies and assembly methods for fabrication of a microimplant. Top: Photograph of system before silicone rubber encapsulation. Bottom: Sketch of technologies and methods involved. 1: Polyimide flex substrate. 2: Electroplated electrode site. 3: Thermosonically placed gold bumps. 4: Silicon chip. 5: Soldered SMD component. 6: Screen printed alumina substrate. 7: Gap welded Ptlr wire.
Wire bonding is a standard interconnection technique used for electrically connecting microchips to the terminals of a chip package or directly to a substrate (Harman, 2010). Wire bonding technology can either be categorized by the bonding method (ball–wedge or wedge–wedge) or the actual mechanism that creates the metallic interconnection between wire and substrate (thermo-compression, ultrasonic or thermosonic). The ball–wedge method is illustrated in Fig. 4.4. The bonding machine uses a capillary through which the bond wire is threaded. An electrical spark (1) is used to form the ‘ball’ (2) at the end of the bond wire. The ball is pressed to the bond pad (3) under the application of heat (thermo-compression), ultrasound energy (ultrasonic) or both (thermosonic), establishing the first bond. The capillary is moved to the target pad (4–6) where the wire is bonded by the same forces as in step 3, using the rim of the capillary to form the ‘wedge’. The capillary is removed while the wire is clamped, causing the wire to tear (7).
Wedge–wedge bonding utilizes slightly different bonding equipment than ball–wedge bonding. The most obvious difference is the bonding capillary, which resembles a very fine pointed stamp with an integrated wire feeding channel. The bonding process is sketched in Fig. 4.5. The bond reaches through the channel in the bonding tool (1). The latter is used for forcing the wire onto the bond pad. Then, ultrasonic and/or thermal energy is applied, forming the first ‘wedge’ (2). Subsequently, the bonding tool is moved to the target pad (3) and the second ‘wedge’ is formed (4). The wire tears by moving the bonding tool upwards while the bond wire is clamped (5).
Each bonding method as well as each bond mechanism is accompanied by a range of advantages and drawbacks. While thermosonic bonding is most commonly used (the latest method developed), the decision for either ball–wedge or Wedge–wedge bonding has to be made depending on the application. Table 4.1 lists some selected properties of the two methods.
|Technique||Ball–wedge bonding.||Wedge–wedge bonding.|
|Strength||Highest processing speed in automated bonding.||Allows higher integration densities.|
|Finest assembly||Smallest bonding pitch: 70 μm using 25 μm diameter wire.||Smallest bonding pitch: 25 μm using 18–20 μm diameter wire.|
|Chip–chip assembly||Bond capillary touches bond pad during second bond, making this method unsuitable for chip-tochip bonding (risk of pad damage). However, there are workarounds like the use of stand-off-stitches, basically requiring a bond-stud to place on the second pad to which the wire is ‘wedged’.||Allows chip-to-chip bonding (bond tool does not touch the bond pads).|
|Ribbon bonding||No bonding of ribbon-shaped wires possible.||Can be used for bonding comparably wide ribbon-shaped wires, suitable for high-power and high-frequency interconnects.|
Source: Beck (2007); Harmann (2011).
The predominant material for bond wire is gold alloy (> 90% Au), doped with additives such as beryllium to optimize for properties such as possible loop height, elongation at break, temperature strength, breaking load or tensile strength. Other wire technologies are established for technological niches like high-current applications, assemblies restricted to low processing temperatures, or bonds with enhanced mechanical strength: aluminium wire, copper wire, insulated bond wire, palladium bond wires, aluminium coated gold wires, etc. Typical wire diameters are 25.4 and 17.8 μm (1 and 0.7 mil) (Heraeus, 2009).
Once a chip is wire bonded to a substrate, the delicate wires have to be protected to prevent adjacent wires touching each other or being sheared off during handling. This protection is usually achieved by applying an opaque, low-viscosity polymer resin (glob-top) that cures after dispensing to a hard sphere, completely covering chip, bond wires and contact pads. If required, the flow of the low-viscosity glob-top across the substrate can be limited by first dispensing a dam around the bonded chip, curing the dam and filling up the volume surrounded by the dam. Glob-top resin cures under ultraviolet light exposure or heat.
Wire bonding is used in almost all microelectronic devices and, hence, widely applied in the fabrication of pacemakers (Gossler, 2007), cochlear implants (according to pictures by Advanced Bionics: (Advanced Bionics, 2011)), BION® microstimulators (Loeb et al., 2001) and many other implanted devices. However, wire bonding competes with flip-chip bonding, which has the advantage of permitting the use of the entire chip area for interconnects, as explained later.
Microflex bonding is a variation of wire bonding that permits array-type high-density interconnection of rigid substrates (silicon dies or screen-printed ceramics) and flexible substrates (polyimide, parylene, etc.). Microflex is applied using a thermosonic wire bonder and was successfully carried out using 25.4 and 17.8 μm diameter bond wire, permitting an integration density of one contact per 100 × 100 μm2 or one contact per 90 × 90 μm2, respectively (Meyer et al., 2001; Kisban et al., 2007; Schuettler et al., 2008; Stieglitz et al., 2009).
The process flow for microflex bonding is sketched in Fig. 4.6. (1) The bond wire protruding from the bonding capillary is flamed to a ball by an electric spark. (2) The capillary is aligned to the bond pad of a (rigid) substrate or silicon chip. (3) The first stud is placed using thermosonic bonding. (4)The wire is clamped, so that the lifting of the capillary causes it to tear. (5)The wire is flamed to a ball. (6) The flex substrate is positioned over the first stud. (7) The second stud is placed, riveting the flex substrate to the rigid substrate. (8) The capillary is removed and the clamped wire tears.
The original method (Meyer et al., 2001) does not include the placement of the first stud. However, this stud serves two functions: (1) it improves mechanical bond strength compared to direct riveting of the flex substrate to the bond pad and (2) it permits an encapsulant to reach the area between flex and rigid substrates by introducing a small gap. If no hermetic packaging is possible, the presence of the encapsulant prevents humidity condensing and initiating corrosion (see Section 4.4).
Microflex was used in the development of the German Retina Implant (Fig. 4.1), in the assembly of a nerve cuff multiplexer (Fig. 4.3) and for interconnection between flexible ribbon cables and silicon-based shaft-type neural probes (Hetke et al., 2003; Kisban et al., 2007). A modification of the microflex-technique permits the connection of two flexible substrates (Schuettler et al., 2000) as shown in Fig. 4.7, instead of the more common connection between a flexible and a rigid substrate.
A very common method for array-type interconnection of chips to substrates (rigid and flexible) is flip-chip bonding. Here, the chip is flipped over, having its contact pads facing towards the substrate. The electrical contact between chip and substrate is established in one of several different ways, for example by solder bumps (either on the chip or on the substrate), conductive polymer flip-chip, anisotropic conductive flip-chip, wire flip-chip or metallurgy flip-chip (Lau, 1996).
An example of a flip-chip bonding process is sketched in Fig. 4.8. Solder bumps are deposited on the contact pads of the IC chip (1), which is aligned upside-down to the contacts of the substrate beneath using a flip-chip bonding machine. After placing the chip on the substrate, ensuring that the solder bumps are in good contact with the substrate pads, substrate and chip are heated up causing the solder to melt and equally wet both contacts (2). After cooling down, the chip is electrically and mechanically connected to the substrate. If demanded by the application, a polymeric underfill encapsulant resin is applied, which creeps between chip and substrate, filling the gaps between the contacts and is then cured to a solid state, usually by heating (3).
Fig 4.8 (1–3) Example of flip-chip bonding process using solder bumps as interconnects and underfill encapsulant. 1: Flipped chip. 2: Substrate (e.g. printed circuit board). 3: Solder bumps. 4: Molten and resolidified solder bumps establish electrical contacts. 5: Polymeric underfiller insulates contacts against each other and against environment and mechanically stabilizes assembly.
The integration density of flip-chip interconnects strongly depends on the technology applied (see Table 4.2). Flip-chip is used in state-of-the-art implantable pulse generators (Gosser, 2007) as well as in implantable microsystems such as Utah-array based wireless implants (Kim et al., 2009).
The methods of microassembly mentioned above all establish mechanical and/or electrical joints that are permanent and cannot be disconnected without damaging the assembly. However, reversible joints might be favourable at least when connectors are needed, for example when connecting an electrode lead to the implant electronics as in cardiac pacemakers. Besides this, establishing an electrical contact using a spring can have various advantages: (1) the spring provides superior reliability over a welded or soldered contact (Donaldson, 1984); (2) the spring permits the otherwise very difficult assembly of a system (Loeb et al., 2001) (see Fig. 4.9); (3) in contrast to a rigid assembly, a spring can compensate thermal stress during manufacturing steps, for example during curing of adhesives at elevated temperatures or during steam sterilization (Kane et al., 2011). As spring contacts, Loeb used gold-plated Elgiloy® (CoCrNi), while Donaldson reported on excellent performance of platinized gold wires.
Aarts used an array of micromachined gold springs in the shape of cantilevers for establishing electrical contacts between shaft-type silicon micronee-dles and a silicon base (Aarts et al., 2008). The concept of this method is sketched in Fig. 4.10. It was developed for assembling three-dimensional needle-type neural electrode arrays.
Fig 4.10 Concept of spring-loading a silicon shaft electrode to a silicon base. The shaft is inserted into a cavity in the base where a cantilever-type gold spring establishes an electrical contact. In a subsequent step, the contact area is electrically insulated by a polymeric encapsulant. 1: Silicon shaft. 2: Electrode site. 3: Metal track. 4: Silicon base. 5: Gold cantilever spring.
While addressing these subjects, the packaging must not interfere with implant functionality, for example the sensor elements and the communication with an external control unit or with other implants. In this section, we focus on aspects related to packaging for implant protection against moisture, rather than on biocompatibility, which is subject of a different chapter of this book. As mentioned in Section 4.3, the selection of the most suitable packaging method should be strongly influenced by the methods applied for assembly of the microsystem to be packaged (and vice versa).
Protection of an implanted system against moisture and mechanical stress is likely to be the most challenging aspect of packaging, especially when a high degree of miniaturization is required. Among the failure modes for electronics, humidity induced corrosion is usually one of the dominant factors. Depending on their intended use, implanted systems, for example cochlear implants, operate in a wet environment for an entire human life. Moisture, once condensed to liquid water, can affect electronic circuits in multiple ways:
• Galvanic corrosion of two dissimilar metal parts joined together, for example gold wire bonds to aluminium pads (Osenbach, 1993).
• Dissolution of inorganic passivation layers of semiconductors, for example silicon oxides and silicon nitrides (Osenbach, 1993).
• Evolution of gas by electrolysis (Donaldson et al.,2011), followed by pH shift and perhaps mechanical damage by volume displaced by gas.
• Corrosion of thin-film metal tracks of semiconductors, for example aluminium (Thomas, 1976).
• High resistance shorting path for current in high-impedance circuits, for example quartz oscillators (Ko and Spear, 1983).
• Swelling-induced mechanical stress of polymeric components, such as underfillers used in flip-chip bonding (Lau, 1996).
• Low-resistance shorting of tracks by conductive saline (Donaldson, 1976).
All these effects must be avoided in high-reliability implants. In general, this can be achieved through two complementary strategies: (1) preventing moisture from the body reaching the electronic circuitry by sealing it inside a hermetic can, or (2) carefully selecting electronic and mechanical components that are not affected by moisture, designing the circuit avoiding high impedances and casting it in a suitable polymeric encapsulant.
Presuming the package is perfectly hermetic, one has to ensure that all packaged components and surfaces are dry. Polymers are prone to store vapour originating from the humid environmental atmosphere. Therefore, wire insulations, electronic component packages, etc., have to be dried before sealing. Furthermore, a water film of the thickness of many monolayers is usually adsorbed at all dry surfaces within one second of exposure to ambient air (Thomas, 1976), including metal, ceramic and glass, and needs to be removed. In the literature, multiple methods on how to dry hermetic packages before sealing can be found that target drying various materials. These methods involve process times between 5 minutes and 100 hours, employing elevated temperatures (110–398 °C), vacuum (− 1.33 × 10− 5 mbar to atmospheric pressure) and purges of dry nitrogen (20 ×) or helium (1 ×). Table 4.3 provides an overview of some drying methods. The drying procedure needs to be optimized and evaluated for each individual implant design.
The maximum permissible level of humidity inside a package is controversially discussed. It is often stated that the onset of corrosion or degradation of passivation layers requires the presence of liquid water. Liquid water is commonly defined as at least three atomic layers of water molecules. Hence, this is the first possible limit one could use: The humidity it takes to completely condense to a three-monolayer film of water covering all surfaces inside the package. Since the real surface inside a package is often difficult to calculate (considering porosity, surface roughness, etc.) and the case that water condenses equally across all surfaces, independent of local temperature, hydrophobic or hydrophilic surface properties, etc., is rather unlikely, this measure is difficult to apply. Another, more practical limit is the concentration c (usually expressed in parts per million, ppm) of water molecules in the package atmosphere. Based on this measure two limits are described in the literature: as long as c ≤ 6000 ppm inside a package with an internal pressure of one atmosphere, vapour cannot condense at positive temperatures; and the dew point is below 0 °C. Condensation at negative temperatures leads to the immediate formation of ice crystals. In contrast to liquid water, ice being a solid is considered chemically non-reactive and hence has no corrosive effect. Since condensation can occur earlier based on surface contaminants (e.g. salts), a safety margin is introduced: As a result, the concentration of c = 5000 ppm is often used as a conservative limit for humidity inside a package (Greenhouse, 2000) and is a requirement for class K military and space-graded circuits (Licari and Enlow, 1998). Although corrosive processes can start in the presence of three monolayers of water, it was found that c = 17 000 ppm of water is required to sustain reactions of etching aluminium tracks, leading to circuit failure (Thomas, 1976).
Among the mechanical and electrical properties, biocompatibility and chemical stability, the permeability to gases is a dominating criterion for the selection of materials suitable for fabricating implantable packages. Figure 4.11 shows the time it takes for the interior of a package to reach 50% of the exterior humidity. Depending on the thickness of the package wall, different wall materials perform very differently. While all polymer families (silicones, epoxies and fluorocarbons) provide barrier properties insufficient for long-term implantation, glasses and especially metals can be considered hermetic. The barrier properties of ceramics are commonly presumed to be somewhere between metals and glasses; the gas permeability of semiconductors (e.g. silicon) is similar to that of metal conductors. Although Fig. 4.11 is based on calculations and is valid only for a package with an internal volume of 2 × 103 mm3, it gives a good impression of the barrier properties of potential packaging material families. All of the material families shown in Fig. 4.11 have members that are well accepted by the body and are considered biocompatible.
Fig 4.11 Permeability of potential packaging material classes Source: Reproduced with permission from IEEE (Figure 1 in Traeger, 1977).
When deciding on using non-hermetic packaging, one can follow two strategies: (1) use a polymer encapsulant that provides good barrier properties against moisture and apply the encapsulant to a thickness that allows safe operation of the implant for the intended implantation time or (2) use a polymer that provides excellent adhesion to all surfaces to be coated and has a low Young’s modulus, for example an elastomer, in order to minimize mechanical stress that might lead to crack development.
Obviously, the first strategy is only an option for short-term implants. Commonly, such implants have been built using epoxy (Kenney et al, 2000) and Parylene C (Ramachandran et al., 2007) as encapsulant. Depending on the encapsulant, the complexity of the potted circuits and the overall electronic concepts (maximum voltage, DC or AC, etc.), such implants can operate over a time of several months up to some years.
The second strategy permits the fabrication of long-term implantable devices and is applied to the fabrication of for example bladder controllers by Finetech Medical Ltd, Welwyn Garden City, UK (Fig. 4.12). Only components that function in the presence of humidity are used, such as chip resistors, chip capacitors, metal-can capacitors, glass-sealed diodes, polyurethane enamelled copper wire (receiver coil), and are soldered to screen-printed alumina substrate (PtAu on Al2O3). After thorough cleaning and drying, the circuit is moulded in room temperature vulcanization (RTV) silicone rubber adhesive. Although silicone adhesive is a very weak moisture barrier (see Fig. 4.11) it was identified in the 1970s as most suitable encapsulant because (1) it provides good adhesion to all encapsulated surfaces even after long-term soaking in water (Donaldson, 1994), (2) it has a very low Young’s modulus (Donaldson, 2007) and (3) it provides sufficient barrier properties for ions and salts (Donaldson, 1981, 1991; Donaldson et al., 2011).
Fig 4.12 Photo of Sacral Anterior Root Stimulator by Finetech Medical, Hertfordshire, UK, used for bladder and bowel evacuation by electrical stimulation in spinal cord injury. 1: One of the three receiver/stimulator units. 2: Male part of Craggs Connector. 3: Female part of Craggs Connector. 4: Ptlr Cooper Cable. 5: Stimulation electrode array.
Adhesion is the key property of encapsulants (Donaldson, 1996; Ardebili and Pecht, 2009). Although water vapour quickly penetrates the encapsulant driven by forces of diffusion, it cannot condense to liquid water as long as the encapsulant adheres to the surfaces of the implant circuit without any voids or cavities. The sole presence of vapour does not affect the circuit’s function (the circuit designer has to take care of that). The uptake of vapour causes a mild swelling of the silicone rubber. In contrast to silicone elastomers, very rigid encapsulation materials develop cracks as a reaction to swelling. The cracks are quickly filled with liquid leading to circuit failure (Kenney et al, 2000). This effect is avoided by using soft silicone rubber, which complies with the swelling (and also complies with shrinking during crosslinking). The fact that silicone is a barrier to ions such as Na+ and Cl− causes osmosis to act as a force opposing the vapour diffusion (Donaldson, 1991). As long as the implant surfaces were thoroughly freed from ionic residuals before silicone moulding, the osmotic gradient draws water molecules from the silicone (Mackay, 1993). The semipermeable property of silicone rubber also prevents to a great extent chemical compounds of the solder used for assembling the circuit being released into the body (Donaldson et al., 2011).
It is crucial to apply the polymeric encapsulant void-free, in particular the surface of the implant circuit has to be covered perfectly. In case of enclosing bubbles, the encapsulation can fail in different ways, depending on the location of the void. In the case where the void is located between two conductor lines that are electrified, the water vapour diffuses into the void where it condenses, eventually shorting the two conductors. If the voltage is high enough, the condensed water is dissociated and the resulting gas expands potentially bursting the encapsulant. Another failure mode is sketched in Fig. 4.13. Here, one of the voids is located at the surface of the implant, while another void is in the encapsulant bulk (Fig. 4.13a). Once the implant is immersed in electrolyte (body fluid), water molecules travel through the encapsulant and condense in the voids (Fig. 4.13b). Depending on the surface (soluble or two dissimilar metals forming a galvanic element) the condensed water can have a corroding effect, causing ions to be dissolved in the water (Fig. 4.13c). The presence of ions locally affects the osmotic gradient and attracts more water (Fig. 4.13d). The osmotic pressure inside the void increases with increasing ion concentration, eventually causing the encapsulant to detach from the implant surface (Fig. 4.13e) or even to completely flake off (Fig. 4.13f), causing the implant to fail. To prevent this kind of failure one has to ensure to bubble formation is avoided during the encapsulation process, for example by applying centrifugation (Ko and Spear, 1983), by casting in a vacuum (Lovely et al., 1986) or both (Donaldson and Sayer, 1975).
When selecting a material for coating electronic circuitry, one has to bear in mind that the volume resistance of polymers can be reduced by as much as eleven orders of magnitude by ionic impurities in low parts per million in the presence of water vapour (Licari, 2003). Based on a list by Donaldson (2007), the main demands on a polymeric implant encapsulant are:
• The Young’s modulus should be low in order to permit shrinkage (during curing) and swelling (during soaking) without developing cracks. Further, a soft encapsulant moderates better between soft tissue and hard electronics.
• It should have a high electrical volume resistivity and a high breakdown voltage, providing some freedom for high integration density in the circuit design. These properties must not degrade substantially over time, once implanted.
In the past, four classes of polymers were widely used: epoxy (Lovely et al., 1986), silicone rubber (Donaldson, 1994), Parylene (Loeb et al., 1977) and polyurethane (Edell, 2004; Accellent Inc., 2011). All these classes have representatives that are USP class VI listed for use in implants, which means that these materials comply with the biocompatibility test standards issued by the United States Pharmacopedia (USP). Epoxy was used extensively up to the 1970s as a pacemaker encapsulant and was eventually identified as the potential cause for many early devices malfunctions. The encapsulant defects were most clearly disclosed where the epoxy cracked or delaminated from wires, allowing direct current to flow and short batteries or capacitor discharge units (Davis and Siddons, 1965; Fisher et al., 1976; Donaldson, 1978, 1996). The development of cracks can be attributed to swelling of the brittle, non-compliant polymer (Young’s modulus 200 000–600 000 psi) during soaking (Kenney et al., 2000). Polyurethane and polyurethane elastomers (Young’s modulus 290–3600 psi) were developed as a material similar to silicone rubber but with improved insulation properties and mechanical toughness. They were and still are used as an encapsulant and as insulator of pacemaker leads. However, the mechanical stability of polyurethane was found to degrade in the biological environment developing stress cracks. These cracks were promoted by the oxidant OHCl, produced, for example, by neutrophilic granulocytes (Edell, 2004). Furthermore, polyurethane elastomers degrade by a process called metal ion oxidation, a reaction triggered by metal ions dissolved from the encapsulated metal conductor (especially in the presence of silver), which led to insulation failure in pacemaker leads (Love, 2006). Nevertheless, polyurethane, as well as epoxy, is still widely used as encapsulant in headers for pacemakers and neuromodulators, embedding the metal contact parts that form the female connector for the electrode leads (Accellent Inc., 2011). Parlyene C is a USP class VI indexed representative of the Parylene family, which is deposited at room temperature from the gas phase, and, therefore, has the ability to fill very fine cracks and pores. Table 4.4 gives an impression of its ability to reach inside small gaps. However, the Young’s modulus of Parylene C is rather high (400 000 psi) and it is difficult to establish good chemical adhesion to surfaces. Some materials require pre-treatments, such as plasma or primer coatings, while other materials do not (Hassler et al, 2010).This makes it difficult to obtain a reliable encapsulation of an implant constructed from multiple materials. In addition, the mechanical properties of Parlyene C were observed to degrade over time in moist environment, allowing cracks to develop (Edell, 2004) and the encapsulant to eventually fail. Nevertheless, Parylene C might still be an adequate choice for encapsulating microsystems for short-term or mid-term implantation, as it was successfully done with a one-channel retinal prosthesis implanted for a period of three months in a cat (Schanze et al., 2007), although it it should be noted that this very simple system was additionally coated in silicone rubber. Members of the silicone rubber family meet all demands of a perfect encapsulant as listed above, unfortunately, no one member meets them all. Most importantly, silicone rubber is reported to be the most biostable of the aforementioned encapsulant families (Edell, 2004) and provides hydrothermally stable joints to many implant circuit surfaces, while having a suitable Young’s modulus of below 1000 psi. Many silicone formulations are USP class VI indexed.
|Gap (μm)||Depth (mm)|
Source: Ramachandran et al. (2007).
Non-hermetic packaging requires the proof that all packaged components are inert to an extent to which potential elutes cannot pass the encapsulant or only to a non-critical extent to the host tissue. This proof has been delivered for silicone rubber by more than 2500 patients who received the Finetech Medical bladder controller (Finetech Medical, 2011). New devices designed by Finetech Medical applying this technology are entering the market: an implantable drop foot stimulator (STIMuSTEP® (Finetech Medical, 2008)) and an implantable grasp prosthesis (STIMuGRIP® (Spensley, 2007)). Besides the practical experiences, recent investigations showed very good barrier properties for rubber with respect to metal corrosion products (Donaldson et al., 2011).
At a certain complexity of the electronic circuits, it becomes difficult, or even impossible to prove that none of the polymer-packaged circuit elements fails within the projected implantation time. Instead, it is faster, easier and cheaper to seal the electronics in a gas- and water-tight hermetic package and prove only that the package is properly sealed. Within a dry hermetic package one can use components, voltages and materials that might not be suitable when only coated by a polymer. Furthermore, a hermetic package can protect the circuit.
The most common hermetic package material is metal, usually titanium or MP35N alloy (Zhou, 2008), because these metals are very resistant to corrosion, provide sufficient mechanical strength and are relatively easy to process at acceptable costs. Such metal cans consist of two shells that are joined together by laser welding, after inserting the battery and electronics. Welding takes place in an inert atmosphere, commonly consisting of a mixture of argon and helium. Argon acts as a protective (anticorrosive) atmosphere during the laser process, while helium is used as tracer gas for subsequent hermeticity tests. The basic architecture of a traditional electronic implant is shown in Fig. 4.14.
Fig 4.14 Cross section of basic architecture of a traditional electronic implant. 1: Substrate (printed circuit board). 2: Surface mount component (glued or soldered). 3: Wire bonded bare chip. 4: Weld post. 5: Battery. 6: Hermetic can. 7: Communication coil. 8: Encapsulant. 9: Electrical feedthrough. 10: Wire (to connector, electrode, sensor, etc.).
Obviously, a metallic housing attenuates an alternating magnetic field by eddy currents, hence the coupling between the coils of an implant and an external transmitter coil is very poor and most of the transmitted power is dissipated as heat (implants must not warm up above 39 °C, according to ISO 45502–1). Implants that are powered by inductive coupling either directly or via inductive charging of a battery need to have their receiver coil placed either outside the metal-can as in the Freehand System (Smith et al., 1998) or most cochlear implants (Clark, 2003), or have to employ a hermetic housing material that is non-conductive, such as ceramics (some cochlear implants (Clark, 2003)), ActiGait System by NeuroDan/Otto Bock (Otto Bock, 2007), BION® microstimulator (Zhou and Greenbaum, 2010) or glass, as used in the earlier generation of BIONs® (Loeb et al., 2001; Schulman et al.,2008).
Ceramics (Al2O3 or ZrO2) provide barrier properties similar to that of metal. Besides the advantage of being insulators, ceramics are much more brittle than metals and might not meet the international standards, for example hammer impact tests according to ISO 45502-2-3 as proposed for cochlear and auditory brainstem implants. Glass was and still often is used as feedthrough insulating material. However, the mechanical robustness is inferior to that of ceramics, which appears to be replacing glass slowly in this field.
The electrical connection between hermetically packaged electronics and electrically operated sensors and actuators requires the use of feedthroughs. Traditionally, these feedthroughs are made of metallic pins that protrude through an electrically insulating bead of glass, glass ceramic or ceramic, which sits in a ring-shaped metal bulkhead. The bulkhead is welded or brazed to the metal housing. This technology is relatively demanding on space and is commercially applied to implants with up to 16 feedthroughs. The seal between contact pin, glass bead and metal bulkhead is either established chemically (so-called reactive seal) or by mechanical press-fit. In the latter, the seal is established by heating pin (e.g. platinum), glass ceramic bead and titanium bulk head above the glass melting temperature, allowing the glass to melt, and cooling it down again. The dissimilar thermal expansion coefficients of metals and glass result in a mechanical press-fit, tight enough to prevent substantial gas leakage. The reactive seal is based on, for example, borosilicate glass reacting with the native oxide layer of the pin, which might be made from tantalum (Zhou and Greenbaum, 2010). The scheme of one-channel feedthroughs is sketched in Fig. 4.15.
Fig 4.15 Cross section of electrical feedthroughs. Left: Compressive or reactive seal. Centre: Active brazing seal. Right: Non-active brazing seal. 1: Metal pin. 2: Feedthrough insulator. 3: Bulk head. 4: Brazing filler material. 5: Pre-metallization for non-active fillers.
Today, instead of using the predominantly amorphous material glass as insulation material, ceramics based on crystalline or partly crystalline metal oxide, usually based on zirconium (Zr) or aluminium (Al) are becoming more popular because they provide superior reliability over time. Ceramic materials such as sapphire (crystalline pure Al2O3) and ruby (chromium doped crystalline Al2O3) are commonly used for commercial implants. Ceramic to metal seals are usually established by brazing. A filler material is melted (at ≥ 450 °C), establishing a bond between metal and ceramic. The filler can either be of an active or non-active nature. Active filler materials use additives that are activated at brazing temperature and chemically react with the ceramics. Non-active fillers require the ceramic to be metallized prior to brazing, for example by sputtering or chemical/physical vapour phase deposition (Zhou and Greenbaum, 2010).
As an alternative, the joint between metal and ceramic can be achieved by diffusion bonding. This is a joining process wherein the principal mechanism is interdiffusion of atoms across the interface. To establish a bond, the bonding partners are pressed together at elevated temperatures (usually 50–70% of melting temperature in the Kelvin scale) for a relatively long time. The stronger the compressive forces the shorter the bonding time required (Zhou, 2008). To alumina, metals such as tungsten, platinum, molybdenum, stainless steel and niobium have been successfully diffusion bonded. Niobium can also be diffusion bonded to zirconia. A comprehensive overview on ceramic to metal sealing techniques is given by Jiang in Zhou and Greenbaum (2010).
To join the metal bulkhead to the metal case of the implant, various metal-to-metal joining methods can be applied, such as fusion welding, ion beam welding, resistance welding or laser welding. The latter is the most popular method.
The aforementioned feedthrough technologies are suitable for devices that use a very limited number of electrical channels, for example up to 16. The new technologies developed involve the use of screen-printing in combination with either low-temperature co-fired ceramics (LTCC) or high-temperature co-fired ceramics (HTCC) or traditional alumina substrate based screen-printing. While the latter can be used to provide feedthroughs running horizontally on the substrate (e.g. 360 feedthroughs through a 15 mm diameter capsule (Schuettler et al., 2010)), co-firing of ceramics allows the fabrication of vertical feedthroughs at densities of 20–55 feedthroughs per mm2 (Guenther et al., 2011; Ordonez et al., 2011). Besides providing a superior feedthrough density, these technologies allow the use of flip-chip assembly of ICs inside the package and flip-chip or thermo-compression bonding for assembly of flexible electrode arrays outside the package (Greenberg et al., 2011), as sketched in Fig. 4.16. Since these packages have very small internal volumes, life-time prediction based on Helium fine leak tests can provide reasonable data when the package contains a desiccant.
Fig 4.16 Concept of miniature hermetic package utilizing vertical feedthroughs through a ceramic substrate sandwich. The gold contacts that join the flex substrate to the package have to be insulated in a polymeric encapsulant. 1: Integrated Circuit. 2: Metal or ceramic lid. 3: Ceramic substrate. 4: Metal feedthroughs. 5: Flip-chip contacts. 6: Flex substrate (leads to electrode array). 7: Gold interconnects.
Besides electrical feedthroughs, optical feedthroughs are required for applications in the field of optogenetics. In general, these can be realized by glass windows brazed to or press-fit into a metal bulkhead, as known from hermetic LED and photodiode packages.
Most implanted electronic devices that are intended to stay in the body for long or unlimited time require appropriate packaging to prevent water-induced failure. The minimum requirement for a hermetic package is the protection of the active electronics of the implant. When the electronics consist of one single chip only (and perhaps a few discrete components), the chip area that contains the critical circuitry has to be covered by a lid. In this case, the silicon substrate forms the base of the hermetic package. This type of miniature package is often referred to as chip-scale packaging, which by definition demands the packaging not being larger than 1.2 times the area of the chip (Ardebili and Pecht, 2009).
Various methods are established for hermetically attaching the lid, such as anodic bonding, fusion bonding, eutectic bonding, glass frit bonding and thermo-compression bonding. Harpster (2005) reported on miniature silicon-based implants using glass lids made from Pyrex 7440 bonded to the silicon chip. These lids were attached by anodic (also called electrostatic) bonding, utilizing elevated temperatures (300–400 °C) and an electrical voltage applied between lid and chip base of 800–1500 V. The resulting electrical field generates strong attraction forces bringing both bond partners in intimate contact, permitting the creation of chemical bonds (Harpster et al., 2005). Eutectic bonding involves a filler (e.g. gold) to form a eutectic bond between the base and the lid. Saeidi demonstrated successful eutectic bonding for an implantable stimulator chip using a silicon lid eutectically bonded to the chip, forming a hermetic chip-scale package (Saeidi et al., 2010; Schuettler et al., 2011c).
An alternative approach to lidding is to coat the active electronics area with non-polymeric barrier layers. This concept was successfully applied to miniature implants by using metals, for example electroplated gold (Najafi, 2003), vapour phase deposited silicon carbide (Hsu et al., 2007), diamond (Xiao et al., 2006) or intermittent layers of silicon nitride and silicone oxide (Haemmerle et al., 2002; Seidl et al., 2009).
A major problem associated with the deposition of non-polymeric barrier layers and also with some of the lidding techniques is the high process temperature required. In particular, when hybrid assemblies involving solder joints and low-melting point polymeric adhesives have to be lidded, most chip-scale hermetic packaging technologies cannot be applied.
The hermeticity of traditional implant packages as used, for example, in cardiac pacemakers is determined by using a two step procedure: firstly, a gross leak test is performed, indicating imperfect seals with leak rates down to R ≥ 10− 5 atm.cm3s− 1. If no gross leak is found, helium fine leak tests are performed. The military standard MIL-STD-883 (method 1014) suggests two different methods. Both refer to exposing a hermetically sealed package to helium at a certain combination of pressure and exposure time (so-called helium bombing) and forcing helium through potential leaking channels into the package. If helium is introduced in the package, it will leak out during successive fine leak measurements. Applying a mathematical model to the measured helium leak rate permits an estimate of the leakage rate of water vapour through the package wall. Both bombing methods differ only in the selection of helium exposure parameters dependent on the package volume. Although MIL-STD-883 suggests rejection criteria for measured leak rates, these criteria are usually regarded as too lenient for long-term implanted devices. Furthermore, there is some criticism on the bombing method itself: valve effects on the micro-scale might interfere with the ability of the pressurized helium to enter the package or with its subsequent release from the package. Instead, sealing the package in a pure or diluted helium atmosphere increases the resolution of the leakages rate measurements and is therefore widely used in commercial implant fabrication. This method is called helium back filling.
When using packages with very small volumes, fine leak tests are of limited use because the sensitivity of today’s leakage test equipment is insufficient of. A common detection limit is a RHe = 1 × 10− 12 atm cm3s− 1. If a package is backfilled with 100% helium at atmospheric pressure p. a true helium leakage rate of LHe = RHe/p = 1 × 10− 12 cm3s− 1 can be calculated. This helium leakage rate is converted to a water vapour leakage rate LHe = 4.71 × 10− 13 cm3s− 1. The time t (in s) it takes for an amount of water vapour 1 atm to collect inside the package can be calculated using Equation [4.1], where c = 5000 ppm is the maximum permissible vapour concentration (see Section 4.4.3) and atm is the partial water vapour pressure inside the body (Greenhouse, 2000; Zhou and Greenbaum, 2010). Obviously, the time is directly proportional to the volume V of the package. For a miniature package of V = 10 mm3 = 0.01 cc, the maximum time predictable using helium leakage tests is t = 1.816 × 109 s = 57 years and 7 months.
For packages of much smaller volume (e.g. V ≤ 1 mm3), helium leakage testing cannot predict the potential life-time within the time scale useful for chronic implants, even if the package is perfectly sealed, since the sensitivity of leakage testers is not good enough.
There are two alternative methods proposed by the MIL standards for detecting leaks in packages: detection of leakage using a radioactive tracer gas (blend of krypton 85 and air), and optically measuring the deformation of the package in response to a pressurized atmosphere. Unfortunately, none of these methods is applicable to detect very small leakage rates in micropackages (Vanhoestenberghe and Donaldson, 2011). As an alternative, vapour leakage might be measured directly using humidity sensors integrated in the packages. The hermeticity can be investigated in accelerated life-time tests, submersing the package in saline solution at elevated temperature and recording the interior humidity over time (Harpster et al., 2005). However, one has to be careful not to confuse a fine leak with out-gassing of packaged components (Schuettler et al., 2011a), see Section 4.4.2. When using humidity sensors as leakage indicators to evaluate a packaging concept, desiccants (or water getters) must not be used since they keep the humidity inside the package at an absolute minimum until saturation. Then, suddenly and after a potentially long time, the humidity rises quickly, making it difficult to draw conclusions about the effectiveness of the drying and packaging method under development.
In case of water vapour leaking into the package at a low rate or outgassing of components inside the package, the use of a desiccant (or water getter) is recommended, keeping the atmosphere inside the package dry. Desiccants can bind water molecules either chemically or physically. Chemical binding is usually a non-reversible process based on metal oxides reacting with oxygen. The process is exothermal and therefore generates heat. The most prominent chemical desiccant is calcium oxide (CaO). For implanted devices, the development of heat is to be avoided; furthermore, system integration is difficult with a non-reversible desiccant that collects humidity from the environment before sealing. Another group of getters is more attractive: those whose adsorption process is reversible, based on physical binding. Three major technological classes of physical desiccants exist. (1) Silica gel: a highly porous silicon dioxide offering a large surface for water molecules to adsorb. Silica gel works most efficiently at room temperature at humidity levels greater than 30% relative humidity (RH), where 100 g gel adsorbs 15 g H2O. At 60% RH, 100 g silica gel is able to adsorb 35 g H2O. (2) Molecular sieves: a framework of pores and open cavities, whose sizes are designed to adsorb molecules of certain polarity and size. Various base materials are used, such as glasses, aluminosilicate minerals, zeolites, clays, charcoals and active carbons. Molecular sieves adsorb moisture over a wide temperature range and are most efficient at a humidity of 8% RH and greater, where they adsorb about 22–25 g H2O per 100 g desiccant, independent of the RH level. (3) Activated clay: layers of silicate that are attracted to each other by electrostatic forces. Water molecules can adsorb between these layers. This naturally occurring desiccant is most efficient at very high levels of RH. In the range below 30% RH its performance is between that of silica gel and molecular sieves. At 5% RH, 100 g clay adsorbs about 5 g H2O. With rising humidity level, the adsorption capacity increases almost linearly to about 20 g H2O at 60% RH (Crossno, 2011). Figure 4.17 shows the performance of the three physically acting getter materials as a function of ambient humidity.
Fig 4.17 Performance of three different physically adsorbing desiccants at 25 °C as a function of ambient relative humidity. Source: Adapted from Desiccare, 2010.
Although the performance of physically binding desiccants is best at medium or high humidity levels, good results were obtained with molecular sieves in miniature implant packages (Loeb et al., 2001; Schuettler et al., 2011a). These desiccants have the ability to extend the life-time of an implant dramatically. Despite the problem described in the section ‘Limits of the helium fine leak test’ above, the use of desiccants inside the package permits a reasonable lifetime estimation: the time until the desiccant with known absorption capacity is situated with water is calculated and is added to the time it takes to reach critical humidity levels inside the package. As reported by Cameron for the case of the BION® microstimulator, the use of 1 mm3 desiccant material extends the time it takes to reach 100% RH from an initially predicted 200 days (without desiccant) to 9000 years (Cameron, 1997).
The first part of this chapter provides a broad overview on state-of-the-art microassembly methods, including adhesive and plasma bonding, solder, resistive and laser microwelding, wire and flip-chip bonding, and spring-loading. Wherever applicable, examples of implanted devices are named that utilize the particular assembly method in their fabrication process. The second part of this chapter introduces the reader to the basics of implant packaging, giving advantages and drawbacks associated with non-hermetic polymeric encapsulation of electronic circuitry as well as those related to hermetic implant packaging.
The methods required for microfabricating implantable sensor systems have been established for several years or even decades. However, these methods are still not widely applied to actually fabricate implantable microsystems for clinical use. Instead, microfabricating methods are applied to allow the miniaturization of parts or subsystems of implants based on conventional technology. An exemption is the BION® microstimulator, which can be considered a classical microsystem. Currently, the BION® microstimulator is undergoing several clinical trials (Krames et al., 2009). The story of the BION®, which is documented in many scientific publications, is motivating for miniature implant developers and at the same time provides a pool of knowledge about the potential of certain technologies as well as identifying pitfalls that can be avoided in future developments (Kane et al., 2011). The history of cardiac pacemakers, the development of the Brindley bladder controller. cochlear implants and the small BIONs® all provide valuable implant design lessons. These should all be reviewed in order to avoid repeating mistakes. While technology progresses and integration densities become higher by refining microassembly methods, failure of implants still occurs. In many cases, the special requirements for implant (micro) packaging are not taken seriously enough, leading to early failure of the electronics. The assembly and packaging methods have to form part of the design and manufacturing process in order to build a reliable implanted sensor system.
The future development of microassembly methods leading to higher integration densities is paced by the consumer microelectronic industry, for example by developing tools for ultra-fine wire bonding, very high-density flip-chip bonding and methods to stack multiple ICs to multi-chip to modules, using low-loss interconnections. It is expected that gluing will become more widely used, especially anisotropic adhesives for interconnection in flip-chip assemblies. While these developments permit the actual devices to further shrink in size even with increasing complexity, non-hermetic polymeric encapsulation will become less practical and will be further replaced by more space-demanding hermetic packaging technologies (see below). Another development taking place in microassembly today, is the ban of toxic substances that were used in past decades to ensure system reliability, for example lead in solder materials or glass passivation layers. While lead (besides others) is banned today by the European RoHS Directive, more substances might be identified as hazardous in the future and, accordingly, the assembly methods will have to be adapted to the new materials developed. An additional growing demand on implanted devices is compliance with magnetic resonance imaging (MRI). This drives the development of non-ferrous alloys, permitting the patient to be MRI scanned without expecting adverse effects (such as displacement forces or heating). With increasing magnetic field strength installed in the clinical MRI tomograph, it becomes increasingly challenging to develop MRI-safe implanted devices.
The past 15–20 years of international development efforts to produce retinal prosthesis shows that most research consortia as well as companies eventually decided against a non-hermetic, polymeric, low profile encapsulation, not least because of the difficulties associated with the prediction of failure modes and device life-time of the integrated complex circuits. Proving hermeticity of a metal or ceramic housing is much easier, faster and, hence, cheaper. Recently, highly integrated hermetic feedthroughs have become available that allow the interconnection to hundreds or even thousands of stimulation electrode sites placed against the retina. Recent developments in the field of photovoltaic panels, LCDs and organic LED displays were accompanied with the development of new desiccant technology that can be applied to parts of the hermetic package by screen-printing, spin-coating or sputtering, or are applied as a self-adhesive patch (Saes Getters, 2011).
In the past, the use of biocompatible liquid crystal polymer (LCP) as a quasi-hermetic encapsulant for electronic circuits has been investigated. LCP has very low moisture absorption (< 0.04%), low gas penetration rate and can be fusion bonded to itself. Lee demonstrated that LCP can be used for fabricating a polymeric package around a circuit with horizontal feedthroughs to fabricate a low-cost retina implant (Lee et al., 2009). Sukumaran et al. (2011) reported on remarkably low gas permeation in an array of vertical electrical feedthroughs through an LCP substrate, which might lead to a packaging concept similar to that shown in Fig 4.17 but cheaper and easier to fabricate. However, it remains to be proven that 100 μm thin walls of LCP provide sufficient moisture protection. So far, results from in vitro tests based on leakage current measurements in an accelerated aging environment are promising (Lee et al., 2011). The forthcoming years will show whether LCP develops into a real alternative to metal or ceramic packages and if so, it will certainly involve the use of powerful desiccants.
For optimization of system reliability one has to investigate the individual failure modes for the component, assembly or housing technology used. A comprehensive introduction as well as an impressive collection of in-depth details is given in the textbook by Martin (1999), although it does not address specific problems associated with implanted devices. The (correct) use of a polymeric encapsulant was investigated during the 1970s, 1980s and 1990s by P.E.K. Donaldson, whose collection of papers is not only very informative but also a great read (some of them are listed in Section 4.8). The development of the BION® microstimulator was documented by a variety of authors. G.E. Loeb stands out in his descriptions of individual technological development phases over the years (assembly and packaging). His papers are highly recommended. A broader introduction in microassembly methods was edited by Zhou (2008), having one chapter dedicated to microassembly in medical components and devices. This book also addresses micropackaging methods. A good general textbook on hermetic packaging was put together by Greenhouse (2000), explaining all relevant mechanisms, testing methods and calculations for life-time prediction, while Najafi published a variety of papers on micro-scale hermetic packaging technologies for implants. A good overview over his work can be found in Najafi (2003).
Aarts, A.A.A., Neves, H.P., Ulbert, I., Wittner, L., Grand, L., Fontes, M.B.A., Herwik, S., Kisban, S., Paul, O., Ruther, P., Puers, R.P., Van Hoof, C., A 3D slim-base probe array for in vivo recorded neuron activity. Proc IEEE EMBC. 2008:5798–5801.
Accellent Inc, Moulded headers – Accellent Available from. 2001. http://www.accellent.com/solutions/market/neuromodulation/implantable-pulse-generators/device-components/molded-headers.php [[Accessed 12 December 2011].].
Advanced Bionics LLC, Advanced Bionics Europe – HiRes90K Implantat Available from. 2011. http://www.bionicear-europe.com/de/90K_implant/90k.html [[Accessed 12 December 2011]].
Aemtec GmbH, AEMtec – Flip-Chip. 2011. Available from. http://www.aemtec.de/en/services/technologies/flip-chip.html [[Accessed 12 December 2011]].
Cameron, T., Loeb, G.E., Raymond, A., Peck, R.P., Schulman, J.H., Strojnik, P., Troyk, P.R. Micromodular implants to provide electrical stimulation of paralyzed muscles and limbs. IEEE Trans Biomed Eng. 1997; 44(9):781–790.
Crossno, R., Desiccants: A powerful weapon against moisture and humidity for electrical engineering and electronics Available from. Technical information by Sued-Chemie. 2011. http://www.sud-chemie.com/scmcms/web/binary.jsp?nodeId=3781&binaryId=3221 [[Accessed 12 December 2011].].
Desiccare Inc. Bulk desiccants, Technical product information. Available from http://www.desiccare.com/pdf/bulkdata.pdf, 2010. [[Accessed 12 December 2012]].
Donaldson, P.E.K. Hydrothermal stability of joints, using a silicone rubber adhesive, for a range of adherents of interest to makers of surgically-implanted microelectronic devices. Int J Adhes Adhes. 1994; 14(2):103–107.
Finetech Medical Ltd, The STIMuSTEP® system, company product leaflet Available from. 2008. http://www.finetech-medical.co.uk/Portals/0/PL003i00C%20STIMuSTEP%20Leaftlet.pdf [[Accessed 12 December 2011].].
Finetech Medical Ltd, About Us – Finetech Medical. 2011. Available from. http://www.finetech-medical.co.uk/AboutUs/tabid/72/language/en-US/Default.aspx [[Accessed 12 December 2011]].
Greenberg, R., Talbot, N., Ok, J., Neysmith, J., Zhou, D.M. Biocompatible bonding method and electronics package suitable for implantation. US Patent Application Publication; 2011. [No. US 2011/0118808].
Guenther, T., Dodds, C.W.D., Lovell, N.H., Suaning, G.J., Chip-scale hermetic feedthroughs for implantable bionics 30 August–3 September. Proceedings to the Annual Conference of IEEE EMBS. Boston, USA, 2011:2969–2972.
Guenther, T., Schuettler, M., Ordonez, J.S., Henle, C., Wilde, J., Stieglitz, T. A novel interconnection technology for laser-structured platinum-silicone electrode arrays. IFMBE Proc. 2008; 22:2392–2395.
Ha, D., Kim, B.G., Lin, T.-Y., Ouyang, Y., Irazoqui, P.P., Chappell, W.J. 3D packaging technique on liquid crystal polymer (LCP) for miniature wireless biomedical sensor. Proc. IEEE MTT-S. 612–615, 2010.
Harpster, T.J., Nikles, S.A., Dokemci, M.R., Najafi, K. Long-term hermeticity and biological performance of anodically bonded glass-silicon implantable packages. IEEE Trans Dev Mat Rel. 2005; 5(3):458–466.
Hassler, C., von Metzen, R.P., Ruther, P., Stieglitz, T. Characterization of parylene c as an encapsulation material for implanted neural prostheses. J Biomed Mater Res B Appl Biomater. 2010; 93B(1):266–274.
Heraeus, Bonding Wires for Semiconductor Technology, Brochure of W. C. Heraeus GmbH Available from. 2009. http://heraeus-contactmaterials.com/media/webmedia_local/media/downloads/documentsbw/brochure/HERAEUS_BondingWire_Brochure.pdf [[Accessed 12 December 2011]].
Hsu, J.-M., Tathireddy, P., Rieth, L., Normann, R.A., Solzbacher, F. Characterization of a-SiCx:H thin films as an encapsulation material for integrated silicon based neural interface devices. Thin Solid Films. 2007; 516(1):34–41.
Kenney, L.P.J., Hermens, H., Francis, D., Bultstra, G., Holsheimer, J., Verloop, A.J., Encapsulation materials for implantable FES systems – a case study June. Proceedings of the 5th Annual International Conference of the IFESS. Aalborg, Denmark, 2000:18–20.
Kim, S., Bhandari, R., Klein, M., Negi, S., Rieth, L., Tathireddy, P., Toepper, M., Oppermann, H., Solzbacher, F. Integrated wireless neural interface based on the utah electrode array. Biomed Microdevices. 2009; 11:453–466.
Kisban, S., Herwik, S., Seidl, K., Rubehn, B., Jezzini, A., Umilta, M.A., Fogassi, L., Stieglitz, T., Paul, O., Ruther, P., Microprobe array with low impedance electrodes and highly flexible polyimide cables for acute neural recording. Proceedings of the 29th Annual International Conference of the IEEE EMBS. 2007:175–178.
Lee, S.W., Seo, J.M., Ha, S., Kim, E.T., Chung, H., Kim, S.J. Development of microelectrode arrays for artificial retinal implants using liquid crystal polymers. Invest Ophthalmol Vis Sci. 2009; 50(12):5859–5866.
Ma, X., Jansen, K.M.B., Ernst, L.J., van Driel, W.D., vander Sluis, O., Zhang, G.Q., Regard, C., Gautier, C., Fremont, H., Moisture diffusion model verification of packaging materials. Proceedings of the International Conference on Electronic Packaging Technology & High Density Packaging. 2008:1–5. [ICEPT-HDP].
Ordonez, J., Keller, M., Schuettler, M., Wilde, J., Guenther, T., Suaning, G.J., Stieglitz, T. Hochintegrierte hermetische Durchfuehrungen fur implantierbare Neuroprothesen. Proceedings of the German MST-Congress. 626–629, 2011.
Otto Bock, ActiGait® Implantable D rop Foot S timulator, Surgeon M anual of Otto Bock HealthCare GmbH Available from. 2007. http://ottobock.nl/cps/rde/xbcr/ob_nl_nl/im_647g357_gb_actigait_surgeon_manual.pdf [[Accessed 12 December 2011].].
Rubehn, B., Wolff, S.B.E., Tovote, P., Schuettler, M., Luethi, A., Stieglitz, T., Polymer-based shaft microelectrodes with optical and fluidic capabilities as a tool for optogenetics. Proceedings to the Annual Conference of IEEE EMBS. 2011:2969–2972.
Saeidi, N., Flynn, M., Byun, K.Y., Yu, R., Ferain, I., Colinge, C., Demosthenous, A., Donaldson, N.D.N. Developing a wafer level gold-polysilicon eutectic bond process to protect sensitive electronic devices. ECS Transactions. 2010; 33(4):83–92.
Getters, Saes. SAES Getters – Our products. Available from http://www.saes-getters.com/default.aspx?idPage=1625, 2011. [[Accessed 12 December 2011]].
Schanze, T., Hesse, L., Lau, C., Greve, N., Haberer, W., Kammer, S., Doerge, T., Rentros, A., Stieglitz, T. An optically powered single-channel stimulation implant as test system for chronic biocompatibility and biostability of miniaturized retinal vision prostheses. IEEE Trans Biomed Eng. 2007; 54(6):983–992.
Schuettler, M., Henle, C., Ordonez, J.S., Meier, W., Guenther, T., Stieglitz, T., Interconnection technologies for laser-patterned electrode arrays. Proceedings of the 30th Annual International IEEE EMBS Conference. 2008:3212–3215.
Schuettler, M., Koch, K.P., Stieglitz, T., Scholz, O., Haberer, W., Keller, R., Meyer, J.-U., Multichannel neural cuff electrodes with integrated multiplexer circuit. Proceedings of the 1st Annual International IEEE EMBS Special Topic Conference on Microtechnology in Medicine and Biology. 2000:624–629.
Schuettler, M., Ordonez, J.S., Silva Santisteban, T., Schatz, A., Wilde, J., Stieglitz, T., Fabrication and test of a hermetic miniature implant package with 360 feedthroughs. Proceedings of the Annual International Conference of the IEEE EMBS. 2010:1586–1588.
Schuettler, M., Schroeer, S., Ordonez, J.S., Stieglitz, T. Laser-fabrication of peripheral nerve cuff electrodes with integrated microfluidic channels. Proceedings of the IEEE Neural Engineering Conference. 245–248, 2011.
Schuettler, M., Stieglitz, T., Gross, M., Altpeter, D., Staiger, A., Doerge, T., Katzenberg, F., Reducing stiffness and electrical losses of high channel hybrid nerve cuff electrodes. Proceedings of the 23rd Annual International Conference of the IEEE EMBS. 2001:769–772.
Schuettler, M., Vanhoestenberghe, A., Saeidi, N., Liu, X., Evans, J., Colinge, C., Demosthenous, A., Donaldson, N. d N., Realization of an active book for multichannel intrathecal root stimulation in spinal cord injury – preliminary results. Proceedings to the Annual Conference of IEEE EMBS. 2011:2965–2968.
Schulman, J., Mobley, J.P., Wolfe, J., Davis, R., Acros, I., An implantable bionic network of injectable neural prosthetic devices: The future platform for functional electrical stimulation and sensing to restore movement and sensation. DiLorenzo, D.J. Bronzino, J.D. CRC Press, Boca Raton, 2008. [Neuroengineering].
Seidl, K., Herwik, S., Nurcahyo, Y., Torfs, T., Keller, M., Schuettler, M., Neves, H., Stieglitz, T., Paul, O., Ruther, P. CMOS-based high-density silicon microprobe array for electronic depth control in neural recording. Proc MEMS. 232–235, 2009.
Smith, B., Tang, Z., Johnson, W., Purmehdi, S., Gazdik, M.M., Bucket, J.R., Peckham, P.H. An externally powered, multichannel, implantable stimulator-telemeter for control of paralyzed muscle. IEEE Transactions on Biomedical Engineering. 1998; 45(4):463–475.
Sukumaran, V., Cato, M.E., Fuhan, L., Tummala, R., Nasiatka, P.J., Weiland, J.D., Tanguay, A.R. High density electrical interconnections in liquid crystal polymer (LCP) substrates for retinal and neural prosthesis applications. Proc IEEE ECTC. 2011; 1308–1313.
Wong, C.P., Clegg, D.B., Kumar, A.H., Otsuka, K., Ozmat, B., Package sealing and encapsulation. Tummala, R.R. Rymaszewski, E.J. Klopfenstein, A.G. Microelectronics packaging handbook part II: Semiconductor packaging. Chapman and Hall, New York, 1997.
Xiao, X., Wang, J., Liu, C., Carlisle, J.A., Mech, B., Greenberg, R., Guven, D., Freda, R., Humayun, M., Weiland, J., Auciello, O. In vitro and in vivo evaluation of ultrananocrystalline diamond for coating of implantable retinal microchips. J Biomed Mater Res B Appl Biomater. 2006; 77B(2):273–281.