‘Thick film’ (more correctly ‘printed-and-fired’) technology, uses conductive, resistive and insulating pastes containing glass frit, deposited in patterns defined by screen printing and fused at high temperature onto a ceramic substrate. The films are typically in the range 5–20µm thick, the range of resistivities is 10Ω/square to 10MΩ/square, there are considerable possibilities for building multi-layer structures. Figure 1 shows schematically some of the components of a thick film circuit.
In Figure 2, working from the left, we can see the progressive build-up of:
[ back to top ]
The substrate materials most commonly-used remain the ceramics, usually alumina, with particle sizes in the range 3–5µm, and 94–98 percent alumina content (the balance is of glassy binders known as ‘fluxes’).
‘As-fired’ ceramics are suitable for thick film processing (thin film technology required a much smoother surface finish. For this we may use polished ceramics or glass. The insulation resistance of the glass (including its behaviour at high temperatures) is important and glasses with low percentages of low free-alkali are required.
New substrate materials continue to appear, including ‘porcelainised steel’ (vitreous-enamelled steel), organic materials such as epoxies, flexible substrates, and even synthetic diamond.
Thick film technology is traditionally an additive process, that is the various components are produced on the substrate by applying ‘inks’ (or ‘pastes’) and are added sequentially to produce the required conductor patterns and resistor values.
Different formulations of paste are used to produce
and each ink contains
The resultant structure after firing is shown schematically in Figure 3: the metal particles are bound together and to the substrate by the glassy phase, and this is particularly important at the substrate-ink interface. Fired surfaces are usually not even or homogeneous on a micro scale, a fact which can lead to problems when wire bonding.
Compared with solder pastes, the particle sizes are much smaller, and the suspensions correspondingly more stable. Inks are designed to give an appropriate viscosity for the screen printing process, and range from being just solid to just liquid: there is a balance to be achieved between a low viscosity ink which will spread after printing, and a thick paste, which will show too many mesh marks, having failed to ‘level’.
In Figure 4, working from the left, we can see the progressive build-up of:
When using any printing method, there are obvious lower limits on feature sizes, which makes high density designs difficult. Also the edge definition of narrow tracks can be poor, which has an impact on high frequency performance. Some materials have therefore been developed for printing over the whole substrate area and subsequent etching; Figure 5 gives an example of what can be achieved.
The sheet resistivity is an important electrical parameter in specifying film materials. If the resistance of a square film resistor of unit length and width, measured between two opposite sides, is Rσ then the resistance of a square of, say two units length and width can be seen to be that of two resistors, each two squares in length, in parallel, that is, two resistors of 2Rσ in parallel, i.e. Rσ (Figure 6). All square resistors made from the same sheet of resistive material have the same resistance when measured between the two opposite sides and the resistivity of the sheet material may be specified in units of ohms per square.
A rectangular resistor (Figure 7), length l and width w, will have a resistance of Rσ (l/w), where l and w are measured in the same units of length and Rσ is the sheet resistivity. The resistivity of a ‘meandered’ resistor (Figure 8), having equal strip width and separation between strips can be shown to be Rσ (A.B/2w), where A, B are the dimensions of the bounding rectangle and w is the dimension of the strip and the spacing.
Gold is a good conductor material and allows thermo-compression gold wire bonding and eutectic die attachment. It is, of course, costly and has poor solderability.
Silver is lower in cost, and solderable, but is not leach-resistant with tin/lead solders. More seriously, silver atoms migrate under the influence of DC electric fields, both causing short-circuits and reacting with many of the resistor paste formulations.
Palladium and platinum alloyed to the gold and silver produce good conductor pastes, with good adhesion to the substrate, good solderability, and moderately good wire bonding characteristics. Silver-palladium conductor inks are the most commonly used materials, with both price and performance (primarily resistance to solder) increasing with palladium content.
Copper and nickel are examples of materials that have been proposed for paste systems as substitutes for noble metals. However, they pose special problems in processing, and require the use of totally different material systems, so real cost savings have been difficult to achieve.
The sheet resistivity of the fired paste structure depends on the metals used, and on the percentage of glass in the ink. Figure 9 shows how pure metals show massive changes in resistance with a relatively small change in metal content; the more gradual curves are for compounds that are better suited to providing controlled higher resistivity.
The following requirements apply:
For crossovers the lowest possible dielectric constant is to be preferred in order to minimise the capacitance associated with the crossover. For the fabrication of capacitors a high dielectric constant is clearly advantageous: for applications requiring any degree of stability, a dielectric constant of 10 is about the maximum which can be obtained; materials with higher K values have thermal coefficient problems. The electrical performance of thick film capacitors is generally not very good and the capacitance per unit area of substrate is low (typically 2,500pF/cm2).
The requirements are:
Modern resistor pastes are based on oxides of ruthenium, iridium and rhenium. These are less sensitive to variations in the firing profile than were the earlier pastes, and provide better TCR and stability performance. Generally TCR is a function of the sheet resistivity, with best performance from products in the 1kΩ/sq. to 10kΩ/sq. range (Figure 10).
[ back to top ]
In classic thick film technology, the substrate is a flat piece of alumina normally between one inch square and six inches square, normally 0.025 in or 0.040 in thick. Substrate materials other than alumina are also used, but purely from a screen printing point of view they are not significantly different. All are abrasive, brittle and easy to mark unintentionally. These features have an impact on handling practices and the design of jigs: there are relatively few totally automated lines, especially in the high-reliability area.
Screens are made of highly tensioned stainless steel or polyester mesh, with a relatively open weave to allow the printing paste to pass through it, typically with 100 to 300 0.003 inch diameter wires per linear inch. Such screens have a ‘transparency’ (also known as ‘open area’) of about 40%.
Thick film paste manufacturers normally specify stainless steel mesh because it has the best dimensional stability and a greater percentage open area than polyester, allowing an easier passage of the paste through the screen. Polyester, however, is more resilient, less prone to damage and more easily deflected to conform to the surface onto which it is to be printed. High mesh counts enable finer detail to be resolved, but give thinner prints. Generally the paste manufacturer will suggest a mesh type to suit his paste, and this will always form a very good starting point: 200 mesh and 325 mesh stainless steel are probably the most commonly used.
Four useful ‘rules of thumb’ for screen printing are:
1 This explains why much coarser meshes have to be used with solder paste, as this has particles of 40–50µm diameter.
In contrast to stencil-making, screens for hybrids are often made in-house. The mesh spaces between the wires are initially filled with a photosensitive emulsion, which is patterned by exposure to ultraviolet light through a photographic mask. The areas exposed during this process become soluble in developer, which opens up apertures (Figure 11) through which the ink is squeezed during the printing process.
(stray reflections distort the fact that the bulk of the mesh fill is complete,
even though some holes at the edge of the pattern are only partially filled)
In section the screen will look like Figure 12: the emulsion layer is flush with the top of the mesh, but continues below the bottom of the mesh. This is to ensure that the stencil can seal onto the substrate whilst holding the mesh filaments clear of the substrate at the edge of the stencil apertures. This ensures that pastes can flow underneath the wires to the edge of the apertures thus producing clean print edges. This extra emulsion or stand-off is typically 10–30µm thick.
The general operation of a printer is much as for stencil printing (Figure 13): the screen is held above the substrate, paste is applied to the screen and the squeegee travels over the screen, pressing it down into contact with the substrate, pushing the paste through the screen, thus depositing paste onto the substrate surface. Tension is important, to ensure that correct ‘snap-off’ occurs.
To screen print successfully one needs to understand the fundamentals of the process, to work as near to conditions of good practice as possible and to be prepared to make sensible adjustments to obtain the best results, not simply to work to some pre-ordained settings and hope for the best.
Steven Webster, Napier University
The squeegee has four functions:
Note the first and last of these – screen and stencil printing are different!
Two squeegee designs are commonly used; the diamond squeegee and the trailing edge squeegee. Both present an edge at 45° to the screen, which has been found to be the optimum angle: too vertical a squeegee will fill the aperture very inefficiently, too shallow an angle will not perform the shearing action completely. A squeegee with a very worn edge will behave like one having too shallow an angle and will give erratic prints, as it fails to shear the top of the paste columns uniformly.
The squeegee pressure will probably be of the order of 0.2–0.4kg/cm of squeegee width. For repeatability, pressure must remain constant, once set, so its travel must be parallel to the substrate, and it must apply pressure uniformly onto the substrate across its width. This is achieved either by mechanical adjustment or by allowing the squeegee to pivot on its mounting (‘self-levelling’).
With the appropriate paste quantity in place on the screen should lift away from the substrate immediately behind the squeegee. Ideally, the screen-substrate gap should be adjusted to the minimum that gives this peeling action. Too small a gap will result in smudged prints caused by the screen suddenly snapping away instead of peeling away from the substrate. Too large a gap will cause distortion of the print and damage to the screen. A typical value is about 0.004in/in width of screen for stainless steel mesh and 0.006in/in for polyester.
The speed should be set in conjunction with the screen gap, as too high a speed will prevent the peeling action of the screen from taking place: a typical value is 5–20cm/s.
While screen mesh and emulsion thickness provide the main control of print thickness, some variation (perhaps up to ±20%) can be made by altering squeegee pressure and speed.
After printing the substrates must be fired, softening or melting the glassy frit to form a cohesive and adhesive film, carrying the conductor, resistor or dielectric materials. Firing profiles are specified by the paste manufacturer but are normally of the form of Figure 15, with temperatures ranging from 500°C to 1,000°C. Such profiles are achieved by passing the substrates on a continuous metal belt through a multi-zone furnace (Figure 16).
While it is not necessary to match the paste manufacturer’s profile exactly, it is advisable to follow it reasonably closely: as with solder paste, time and temperature are important More essential, however, is that the profile remains constant from day to day and week to week. This is especially necessary in order to achieve a good yield with resistors.
Once set, the furnace profile should be checked routinely, to ensure that it has remained constant: peak temperature should repeat within a degree from week to week. Profiling is usually done by attaching a sheathed mineral-insulated thermocouple to the belt, with its tip shielded between a couple of substrates (Figure 17). The thermocouple is connected to a chart recorder to record thermocouple temperature against time.
Normally the substrate will be fired after each print in order to bind the print permanently to the substrate. The one usual exception to this is that resistors are normally printed, dried and the next paste printed and dried, and so on, until all resistor pastes have been printed. This is because resistor pastes are designed to be fired only for a given time and temperature, and will be altered by re-firing.
An alternative way of firing thick film is using an infra red furnace. Substrates are able to absorb energy very rapidly from such sources and so it is possible to fire three or four times faster in an IR furnace. However, it is more difficult to select furnace settings that will produce resistor characteristics similar to those readily obtainable in conventional furnaces.
The ambient atmosphere2 within the furnace is critical, and the user is responsible for providing this through a system of filters and dryers. One of the most common disasters on thick film processing is accidentally supplying the furnace with contaminated air. Typical contaminants are exhaust fumes from lorries parked outside the air compressor intake, flux fumes, halogenated solvents and oil from unsuitable compressors. If such contaminants are introduced, resistor values will become erratic, gold conductors will fall off the substrate and palladium silver conductors will be blackened and rendered unsolderable.
The furnace manufacturer and paste supplier will advise on how much to introduce and where to introduce it. For instance BTU suggest one change of muffle air per minute, distributed 2:1 in the burnout and firing zones. A furnace 10m long with a 30cm belt and an internal muffle height of 10cm would require 200 and 100 l/min in the burnout and firing zone air flows respectively.
Du Pont advise the use of their ‘ PLAWS ’ formula.
V = P x L x A x W x S
V = Volume of air flow required for adequate burn out (l/min)
P = Ratio of printed area to total substrate area (<1)
L = Belt loading factor, i.e. ratio of substrate area to belt area (<1)
A = Amount of air needed per unit area of printed paste to completely oxidise the polymer binders contained in thick film compositions (0.4 l.cm−2 )
W = Belt width (cm)
S = Belt speed (cm/min)
The airflow to the firing zone should be about 10 to 20% greater than ‘ PLAWS’.
As a practical test, print long palladium silver conductor tracks on to a few substrates. Fire two or three of these in an otherwise empty furnace and the remainder in the furnace when fully laden. If the second group has a significantly higher resistance than the first, greater airflow is probably required.
When firing remember to pre-load the furnace with a number of scrap or dummy substrates so that the furnace can stabilise to the new thermal load before the production substrates arrive in the hot zones.
In both thick and thin film technologies, the number of process variables is such that it is not possible to obtain resistor values consistently within better than 10–20% of the nominal value. Usually this tolerance is inadequate for the circuit requirement, so it is necessary to adjust the resistance values later in the process by trimming. Almost all trimming processes operate by removing some of the material or by increasing the sheet resistivity, so that the design value of the resistor must be over 20% below the circuit requirement before trimming is carried out.
Thick and thin films are trimmed by cutting away parts of the film to adjust the resistor value, using either a high-velocity jet of air carrying abrasive materials (Figure 18) or (more commonly) a high-power pulsed laser (Figure 19). Transverse, L-shaped or longitudinal cuts may be made to give coarse, intermediate or fine adjustment of the resistor value (see Figure 20). Resistors may be trimmed to a specific value, or adjusted while the circuit is at least partially operational to give the required circuit performance, a process known as ‘functional trimming’.
Adjust cut A first, then allow to stabilise; then adjust cut B
The hybrid circuit combines film circuit elements and added discrete components, and the manner in which these are interconnected has been a central problem in hybrid manufacturing processes. The situation is made more difficult by the fact that hybrids tend to be used in custom low-volume applications, specifically for products which require high reliability in demanding environments such as military, automotive and space applications.
The options (alone or in combination) are:
Note that the processes used in hybrid assembly often differ from those used in assembling discrete semiconductors, so we must be careful to establish device reliability for the format and methods being used, and to ignore any experience which may come from the same devices in the form of conventionally packaged, discrete components.
Note also that using some form of chip carrier or TAB allows the device to be tested after the hazardous operations of die attachment and lead attachment have been completed. This makes it possible to reject failed devices before assembly, to maintain a satisfactory yield of completed hybrids, and also allows more exact characterisation of the device function.
Two main options are available, encapsulation in epoxies or other plastic materials, or hermetic enclosure, in packages with ceramic or glass-to-metal seals. When unencapsulated silicon devices are used, the hermetic package is generally advised, and an hermetic enclosure is usually mandatory for high-reliability applications. The final sealing process is technically demanding and may require sizeable investment in capital equipment if adequate throughput and acceptable yield is to be realised. Some sample circuits are shown in Figures 21 and 22.
Note that the decapsulation process has damaged the bonds, especially those on the left-hand hybrid
Note how the leads have been formed into a staggered array
[ back to top ]