[Skip to content]

TWI
Search our Site
.

An introduction to die attach

TWI Bulletin, August 1985

by G L White

Gillian White, B Eng, is a research engineer in the Microjoining Section of the Sheet and Precision Processes Department.

Some of the problems involved with die attach (the process of mounting a semiconductor die to a substrate) are discussed in this article, together with the methods and equipment currently used.




The process of mounting a semiconductor die to a substrate is known as die attach. It is the process which usually takes place immediately prior to wire bonding, during monolithic integrated circuit (IC) or hybrid package manufacture.

The Welding Institute has been involved, for several years, in the research of wire bonding techniques, including the development of Al ball bonding. [1] For such expertise in wire bonding to be exploited in automatic assembly, it is necessary for accurate positioning and attachment of the semiconductor dice and the substrate, between which wires form electronic connections. Current automatic bonding machine software may be very sophisticated but, since tighter tolerances are imposed on such things as the alignment of bonding targets on the die and the substrate (especially for military applications), an increase of tolerances on die attachment becomes necessary.

Manual wire bonding is less affected since an experienced machine operator can compensate for slight errors in the placement of dice on a substrate. However, an improvement in the consistency of both positioning and attachment of dice would serve to increase yields, regardless of whether automatic or manual wire bonding is being used. Another point to consider is that failure, initiated by poor die attach, may not cause production losses (e.g. failure during wire bonding), but may cause the device to fail in service. This in itself could have far more serious consequences.

It would be wrong to imagine that all die attach is inaccurate. It is just that, as wire bonding becomes increasingly automated, the thrust for automation of die attachment becomes stronger. Speed, accuracy and reduced dependence on the need for experienced wire bonding machine operators are even more favoured for high volume production of electronic circuits. Die attachments must keep pace with this.

Methods of die attach

Epoxy, polyimide and eutectic are the three most common adhesive media used in the attachment of dice. Adhesive media serve to conduct heat away from the die, i.e. they act as a heat sink to prevent degradation of the electronic functions of the circuit because of excessive temperature levels. Epoxy materials are typically loaded with Ag to improve overall electrical and thermal conductivity.

Ionic impurities such as Cl - and Na + bear relevance to epoxy die attach failure. NH 3 can also be given off after attachment and degrade dice. All epoxy-based adhesives contain solvent, driven off during curing. Recent developments in epoxy adhesives include the so-called 'low-chloride' epoxies. The presence of ionic impurities , although not entirely removed, is markedly reduced (from ~440ppm to ~200ppm when comparing an established Ag-loaded epoxy with a low chloride epoxy [2] ).

The solvent can constitute 5-10% of the total melange. Curing of epoxies is a fast, fairly low temperature (80-180°C) and, therefore, relatively simple process. Low chloride epoxies are fairly flexible, which can help to absorb, and thereby reduce, the effect of some of the stresses introduced during the operating life of the die and its package.

Conductive polyimides are, in comparison to low chloride epoxies, much less contaminated by ionic impurities. Levels of Cl - and Na + down to <10ppm have been recorded. [2] In common with the more established epoxies, conductive polyimides are Ag loaded, but possess an advantage in that the purity is usually much improved. This increased purity helps improve electrical conductivity. However, solvent content is higher than for low chloride epoxies, constituting between 18 and 30% of the total.

Curing of conductive polyimides is fairly complex, usually requiring a two stage process. Curing temperatures are generally higher than for epoxy type adhesives, reaching up to ~270°C during the second curing operation. However, this lengthy two-stage curing does ensure that most of the solvent is vapourised. The worry with solvent remaining in the adhesive following curing is that it could possibly be given off during package operation, causing subsequent corrosion. Fortunately no NH 3 is given off during the cure, so polyimide adhesives can reduce failure from chemical reactions. Polyimides are inherently stiffer than epoxies, however, and this can cause stresses and other mechanical problems during operation e.g. cracking and die deformation.

Eutectic die bonding is based on materials such as AuGe and AuSi. The temperatures used for preform placement are consequently high (up to ~350°C for AuGe) and the presence of Au in the eutectic preform means that semiconductor die often have to be Au backed. Curing is not required but the high temperatures used often prevent reworking or replacement of badly aligned dice. Electrical and thermal conductivity are good, as is the adhesion of the die to the substrate.

Problems associated with die attach

The most common causes of die attach failure arising from the adhesive are either chemical, mechanical or electrical.

Chemical

Ionic impurities, present in epoxy adhesives, can cause the release of Cl -, Na + and possibly, NH 3 ions during device operation which can lead to corrosion. Vapours (including solvent vapours) which are emitted by an adhesive during curing do not cause problems, unless they continue to be emitted after curing or device packaging. If this occurs, it is quite likely to result in corrosion. This problem is less likely with conductive polyimide epoxies which since they contain aqueous solvents, give off water, rather than any other vapour during curing.

Epoxy resins possess excellent wetting properties on clean oxide and metal substrates. This can lead to the separation of the epoxy resin phase and the Ag particles, causing contamination of the substrate surface surrounding the die. This is less likely to occur with polyimide and higher solvent containing epoxies (i.e. > 5% solvent content) because evaporation of the solvent increases the viscosity of the epoxy resin phase, preventing its separation from the Ag particles. Another chemical/ metallurgical problem is that of Ag migration. In most cases where this causes problems, it can be overcome by addition of Pd to the Ag loading. 30% Pd mixtures can be used and, for US military purposes, will still conform to Mil Std 883B.

Mechanical

Mechanical problems arise, for the most part, from induced stresses in the die/substrate system. Wafer preparation can cause scratches and notches on the surface and sides of dice. These may provide a flaw, from which a crack can propagate when the dice are deformed ( Fig. 1).

Fig. 1. Propagation of cracks because of thermal stress in semiconductor die: a) initiating from surface crack; b) initiating from side notch

Fig. 1. Propagation of cracks because of thermal stress in semiconductor die:
a) initiating from surface crack;
b) initiating from side notch

The presence of thermal stresses may be related to several factors including die size, adhesive thickness and voids in the adhesive. Bolger and Mooney [3] describe these effects in some detail, and many researchers [4,5] have attempted to present mathematical models for the effect of thermal stresses. Figure 2 shows how die dimensions, etc, can be presented for such calculations and Fig. 3, which shows the effect of die length and adhesive thickness on the resulting thermal stresses, is based on results obtained by Manchester et al. [4] The approximate level of stress required to initiate fracture for given flaw sizes between 1-4µm is indicated. This highlights the need to improve wafer cutting and die preparation techniques. The smaller the flaw size, the less likely it is that the die will fail because of thermal stresses.

Fig. 2. Model of semiconductor die with cracks induced on cooling following attachment with epoxy or polyimide adhesive: x - thickness of adhesive layer; 2a - flaw size for crack initiation µm; L - length of semiconductor die, mm

Fig. 2. Model of semiconductor die with cracks induced on cooling following attachment with epoxy or polyimide adhesive: x - thickness of adhesive layer;
2a - flaw size for crack initiation µm; L - length of semiconductor die, mm

Fig. 3. Effect of die length and adhesive thickness on induced thermal stress for two semiconductor die lengths.

Fig. 3. Effect of die length and adhesive thickness on induced thermal stress for two semiconductor die lengths.

b2684e1.gif

Where m = constant for particular mode of cracking; a = half flaw size; K Ic = fracture toughness for Si, MN m -3/2

The effect of the occurrence and size of voids on die attachment is summarised in the Table and Fig. 4. Reduction of thermal conductivity is common when voids are present in epoxy type adhesives. This is not seen during eutectic die attach since the preform thickness is usually less than that of the layer of epoxy, hence the improvement in thermal conductivity.

Table - Effect of adhesive void size and content on die attachment

Void description Effect
Excessive number and large ( Fig.4a) Shear strength reduced, thermal conductivity reduced, die could lift off.
Small ( Fig.4b) Shear strength unaffected, stresses increased in large dice, thermal conductivity unaffected, electrical conductivity unaffected.
None ( Fig.4c) Could cause dice to crack, from high induced stresses.
Fig. 4. Effect of voids in adhesive on die attach failure modes: a) Die lift-off caused by presence of excessive number of voids in adhesive (note presence of large void); b) small voids - deleterious effects more noticeable with large die where high stress could be induced; c) Adhesive without voids

Fig. 4. Effect of voids in adhesive on die attach failure modes:
a) Die lift-off caused by presence of excessive number of voids in adhesive (note presence of large void);
b) small voids - deleterious effects more noticeable with large die where high stress could be induced;
c) Adhesive without voids


Electrical

Electrical problems can arise from reduction in conductivity because of the presence of voids in the adhesive ( Fig. 4). There are ways to reduce, if not eliminate, the failure modes described. These include:

  1. Reduce the level of ionic impurities. To some degree this has been achieved by using low chloride epoxies, but there will always be scope for further improvement.
  2. Optimise the ratio of the die length to adhesive thickness. This will help to reduce the level of mechanical stress introduced in a system, and must certainly be considered when large dice are used.
  3. Improve wafer preparation. Better sawing would eliminate the early introduction of scratches and notches in the die. This will help increase the operating life of dice, especially by reduction of the effect of thermal stress, introduced during cooling after the die attach operation.
  4. Increase the flexibility of adhesives. Again this should alleviate mechanical stress problems since, to some extent, the adhesive will be able to absorb, and effectively damp, some of the induced stresses.

This list could be extended to include developments such as new adhesives with improved thermal and electrical properties. 'Au Sub', manufactured in the United Kingdom by Johnson Matthey Chemicals, is reported to be the first inorganic adhesive for attaching dice to circuit boards. [6] 'Au Sub' ('Au substitute') is intended to act as a heat sink. It is also claimed that Au Sub does not contain solvents or ionic impurities, and as such, problems like corrosion which are caused by the presence of vapour in an hermetically sealed package, would be removed.

Equipment for die attach

Depending on the volume of production, manual or automatic die attach equipments may be used. Where die size varies, or production runs are small, manual equipment will probably be better employed. Automatic equipment, which can be set up to attach large numbers of identical dice, is more suited to large scale volume production. A typical recent design in automatic epoxy and eutectic die attach equipment is shown in Fig. 5. It features a large number of user facilities aimed at improving production levels. These include die pick-up, presentation, alignment and missing die or preform detection. Closed circuit TV helps the operator to align the dice and most equipments offer full pattern recognition facilities, an important feature where large numbers of identical dice are to be attached.

Fig. 5. Automatic die attach equipment, Dynapert Precima EDB520 (Courtesy of Dynapret Precima)

Fig. 5. Automatic die attach equipment, Dynapert Precima EDB520

(Courtesy of Dynapret Precima)


Advances in both wire bonding and die attach equipment and techniques are being made and have been discussed by Hueners at some length. [7] A seemingly simple process, that of attaching electronic circuits in the form of semiconductor dice to substrates, relies heavily on good equipment design. It is only by achieving continuous improvement that the high quality ICs and hybrids, which are wire bonded on sophisticated fully automatic bonding equipments, can be provided.

References

1 Edson D A, Scott M H, Johnson K I and Carter A W UK Patent Application 31346/ 77. Return to text
2 Bolger J C 'Adhesive related failure mechanisms in military hybrid packages'. International 7ournal for Hybrid Microelectronics, 1984 7 (4), 23-28.
3 Bolger J C and Mooney C T 'Die attach in iti-rel P Dips: Polyimides or low chloride epoxies'. IEEE Transactions on Components, Hybrids and Manufacturing Technology, 1984 7 (4), 394-398. Return to text
4 Manchester K E, Bird D W and Chavest H 'A contribution of polymer die attach system to the reliability of plastic encapsulated integrated circuits'. Proc 20th Int reliability Physics Symposium, April 82, 272-275. Return to text
5 Kessel C G M, Gee S A and Murphy J J 'The quality of die attachment and its relationship to stresses and vertical die cracking'. IEEE Transactions on Components, Hybrid and Manufacturing Technology, 1983 6 (4), pp 414-420.  
6   New Technology, 29th October 1984. Return to text
7 Hueners B W 'Advances in wire bonding'. Solid State Technology, 1985 28 (2), pp 189-195. Return to text