Thermal spraying gives a new look to substrate design

TWI Bulletin, July/August 1994

Andy Sturgeon joined TWI at the end of 1990 as a Principal Research Engineer in the Advanced Materials and Processes Department. After obtaining a degree in physics, he began his research career at the Centre for Advanced Materials Technology at Warwick University, where he gained a doctorate for his work on bonding mechanisms between glasses and ceramics to metals. After this he spent two years with the ceramics group at Rolls-Royce plc, where he was involved with projects aimed at introducing advanced ceramic components into aero-engines. This was followed by a further two years with Alcan International Ltd, where he worked in developing new and novel processing routes to ceramic and ceramic composite materials, in particular sol-gel and chemically bonded ceramics (for example high strength cements and reaction formed ceramics).

He has responsibility for a number of projects on ceramic coatings and surface engineering in general. Most recently he has been investigating HVOF thermal spraying as a process for putting down high quality coatings for wear, corrosion and electronic applications.

In recent years high velocity spraying has opened new doors in microelectronic substrate manufacture, as Andy Sturgeon explains.

A continual drive for miniaturisation and low cost volume production has led to use of hybrid technology for power electronic devices. The resulting increase in power density has placed demands for improvements to be made in the thermal dissipation characteristics of current hybrid power circuits. Thermal resistance from the power chip to the heat sink is now one of the main issues limiting further development of power electronic devices.

High velocity oxyfuel (HVOF) spraying is an emerging thermal spraying process which is becoming established as a technique for the deposition of high quality metallic and carbide coatings. In early 1991 TWI installed a TopGun HVOF system and has shown that the HVOF process is also capable of depositing high quality ceramic coatings of alumina. The process can spray ceramics directly on to metals and vice versa, suggesting that it is ideal for fabrication of new hybrid power substrate designs. These new designs may give considerable improvements in thermal dissipation characteristics.

HVOF Spraying

In HVOF spraying an internal combustion process rapidly heats and accelerates a powder consumable to high velocities. Suitable combustion fuel gases include propylene, acetylene, propane, hydrogen and MAP, giving combustion temperatures above 2800°C. The fuel gas and oxygen enter the gun (Fig.1) where they are premixed and introduced into the combustion chamber. Powder to be deposited is fed either directly into the chamber or down stream into the nozzle using an inert carrier gas such as argon. The powder is heated, then accelerated along the nozzle by the rapidly expanding gases to reach velocities of up to 600 m/sec for a typically sized powder of 10-40µm, and ejected from the gun as a stream of molten or softened particles. It is the high velocity of the process that results in deposition of high quality coatings. This is shown in Table 1 which compares the important characteristics of a number of thermal spraying processes. Coating thicknesses of over 1.5mm are possible with careful control of cooling to reduce residual stress. Typical spraying rates for HVOF are 2-4 kg/hr and deposition efficiency is well over 60% for many coating materials.

Table 1 Characteristics of thermal spray processes

Substrate technology

Alumina of 96% purity has become the most widely used substrate material and represents a low cost, mechanically stable refractory electrical insulator which is well suited to most hybrid applications. The 4% of additives present are selected to accomplish complete densification during fabrication by sintering without excessive grain growth and to maximise electrical properties.

The most common additives are magnesium oxide and silica. Inclusion of alkali oxides is avoided because of their detrimental effect on electrical performance. Alternative ceramic materials are considered for substrates where an improvement in thermal conductivity is needed to remove hot spots, or where a closer thermal expansion match to silicon is necessary.

In low frequency power hybrid applications, beryllium oxide and aluminium nitride are sometimes used instead of alumina as the substrate material. Beryllium oxide has a thermal conductivity which is substantially greater than alumina and about half that of copper. High strength combined with high thermal conductivity also gives beryllium oxide good resistance to thermal shock. This material is primarily used in substrate applications requiring rapid heat removal.

Of the newer materials with improved thermal conductivity, aluminium nitride offers potential as a non-toxic replacement for beryllium oxide in low frequency applications. Commercialisation of aluminium nitride substrates has, however, been impeded by the unavailability of reliable thick film conductor compositions. Present conductors often suffer from a lack of aged adhesion. Until suitable thick film conductor materials are developed, use of aluminium nitride will remain limited.

Metal core substrates

Porcelain steel substrates have found applications in the market place where the fragility of alumina is of concern, and where the penalties of extra weight and manufacturing complexity can be accommodated. Additionally, this strength allows them to be used as a structural member of an enclosure. Good examples are telephone handset bases and camera flash bars. The enamels have a softening point which is typically 600°C. This puts an upper limit on the temperature of any subsequent processing.

Conventional thick film technology for laying down conductor paths, resistors and capacitor components was developed for use with alumina substrates, and has a process temperature in the region of 850-1000°C. The upper temperature limit of 600°C demanded by the porcelain enamelled steel substrates precludes use of conventional thick film materials. In addition glass enamels have been found to have unacceptably large electrical losses at high frequencies.

There is a need to develop ceramic metal coated substrate technology which has a subsequent process temperature capability of 850°C or greater, thereby allowing use of conventional thick film technology. Development projects underway are looking at replacing porcelain enamels with a crystallisable glass known as a glass-ceramic. These materials are applied to the metal base as a glass using the enamelling processes, and then caused to undergo crystallisation to give a refractory crystalline enamel. Such glass-ceramic enamels have softening temperatures above 830°C and enable a subsequent thick film processing step to be undertaken. Glass-ceramic enamelled steel substrates are becoming available in the market place, and have found applications including strain gauges using over printed thick film resistors, to accelerometer circuits where high forces and mechanical shock preclude use of sintered alumina.

Power hybrid substrates

The normal design of a hybrid power circuit consists of a stack of different materials, see Fig.2. The power devices are attached to an electrically conductive track, usually of copper, to allow for interconnection. The copper tracks are electrically isolated from a metal heat sink by a ceramic layer of alumina, or occasionally beryllium oxide or aluminium nitride. The ceramic is required to be as thin as possible, while still allowing for handling during fabrication, and to provide sufficient resistance to electrical breakdown during use.

More recent developments use braze alloys to bond aluminium nitride on to matched thermal expansion copper clad molybdenum laminates. The problem of metallising the aluminium nitride to give the interconnection tracks and the poor thermal conductivity of braze alloys, together with relatively high cost, has limited its exploitation. An alternative approach is to use glass-ceramic metal cored substrates as discussed earlier. Copper clad laminates rather than steel are used as the metal core to improve thermal dissipation. These materials are still at an early stage of development, and also suffer from metallisation and high cost difficulties.

New substrate designs

• Lower production costs through a simpler design and more rapid production;
• The substrate may be built directly on to parts of any material having a heat sink capability, such as a structural part of an electrical motor, generator or automotive component;
• The thermal sprayed hybrid substrate design removes the need for a layer of copper between the insulator and heat sink, and removes the need for an attachment medium between the substrate and heat sink. The result is a reduction in the number of interfaces between the power device and heat sink, and their associated thermal impedances. This may lead to improved thermal dissipation performance of the substrate.

These new designs offer potential for improvements in thermal dissipation capacity of the hybrid substrate and for improvements in the manufacturing process. Designs using thermal spray technology can reduce both the number of interfaces between power chip and heat sink, and also allow for a reduction in dielectric thickness. For traditional DCB hybrid power substrates, the dielectric has a thickness usually over 600µm. HVOF deposited coatings having thicknesses of only 200µm will still give a suitable DC dielectric strength for this application. There is therefore considerable potential for decreasing the thermal resistance between chip and heat sink, leading to a possible improvement in thermal dissipation characteristics.

Use of thermal spraying to develop new hybrid substrates for power device applications has to date achieved best results using a combination of both plasma and HVOF spraying. Firstly, plasma spraying is used to deposit an alumina layer on to a heat sink of aluminium. HVOF is then used to deposit a pattern of copper tracks on to the sprayed alumina. This design has been shown to have the capacity to carry a current of up to 1000A and have a dielectric strength over 5kV for a 200µm coating thickness. Its thermal resistance from chip to heat sink has been measured at 1K/W for a 4 x 8mm test chip. Such a value is sufficient to allow operation of circuits that generate 100 W/cm 2 . This result is also comparable with those values obtained for the same test chip on the most advanced competing substrates.

Recent work at TWI has demonstrated that the HVOF process can deposit coatings of alumina that have dielectric strengths of 50 to 70 kV/mm for a nominal 100µm thickness. Some results from this work are summarised in Table 2. These values match, if not exceed, the values obtained by air plasma spraying. The surface roughness of the HVOF alumina coatings is typically 1.6µm Ra. This value indicates a surface finish that is suitable for thick film metallisation, and is significantly better than plasma sprayed coatings where 4µm Ra is more typical.

Table 2 Properties of traditional substrate compared with those for plasma and HVOF deposited coatings

The HVOF process is known to deposit coatings of copper that are less oxidised and more electrically conductive than those deposited using plasma spraying. HVOF sprayed copper has an electrical resistivity measured at 4-6µ Ω/cm which is considerably lower than can be achieved by plasma spraying, and is about three times the value of bulk pure copper. This work has shown that the HVOF process can spray alumina coatings which match, or exceed, the electrical insulation performance achieved by plasma spraying. HVOF spraying therefore offers the ability to deposit both an electrically insulating ceramic followed by electrically conducting copper tracks using the same spraying equipment.

The HVOF process can be used to deposit first a high quality electrically insulating layer of alumina, typically 200µm thick, on to a suitable heat sink of aluminium or copper. This can be followed immediately by deposition of copper tracks using the same equipment. Track thicknesses of 200µm should be sufficient to carry currents up to 1000A.

HVOF spraying can therefore be seen to offer significant potential for design and manufacture of power hybrid circuits. Such new designs may allow for improvements in thermal dissipation from power chip to heat source which are above those achievable using present power hybrid designs. Applications for improved performance power hybrid substrates are numerous, and will include integrated power electronic products handling 20-1000A, and needing a breakdown voltage up to and above 5kV. The major application areas are likely to be within automated industrial electrical equipment and automotive electronics. Such areas will include motor control and power supply components. HVOF spraying may allow fabrication of cheap, robust hybrid power circuits for 'under hood' automotive use. The power circuit could be fabricated directly on, for example, engine casings which provide the heat sink source.

Conclusions

• The HVOF process can deposit alumina coatings which have electrical properties suitable for dielectric applications. Dielectric strengths of 70-50 kV/mm have been achieved.
• The dielectric strengths of alumina coatings deposited by HVOF will match, if not exceed, the values obtained for a plasma sprayed coating of similar thickness.
• HVOF offers potential for new power hybrid substrate and metal core substrate designs. These may offer performance and fabrication benefits over present designs.