New developments in high performance solder products forpower die assemblies

M. Fenner, A. Mackie, G. Wilson

Indium Corporation

Phone: +441 908 580 400, Fax: +441 908 580 411, E-mail: Europe@indium.com

Abstract

The use of special purpose solder pastes in power die attach is well-established offering low voiding andreliable bonding in volume manufacturing. However these materials are designed around high lead alloys andapplied by dispensing. IGBT circuits are made by printing high tin alloys to multiple die sites, placing die andreflowing in a process more similar to conventional PCB or hybrid thick film assembly. This paper describeshow the opportunity was taken to make use of the latest developments in Pb-free SMT flux technology and re-optimize them to the different requirements of IGBT die attach.

We rehearse the attributes and requirements of IGBT circuitry and then go on to show how a highperformance Pb-Free solder paste has been developed to meet the requirements of large power die attachment(LDA) in IGBT module manufacturing processes. The paste has excellent print and handling characteristics androutinely returns less than 0.5% voiding under large die over a wide range of vacuum reflow conditions. Theflux vehicle chemistry offers ease of cleaning to be compatible with the next stage processes of wire bonding &circuit encapsulation.

Key words: solder paste, IGBT, die attach, void free, wire bonding, cleaning

Introduction

IGBT (Insulated gate bipolar transistor)modules are fast becoming one of the mostimportant electronic power switching tools in avariety of industries from trains to military vehiclesto trucks. They are being adopted because theyallow control of large currents from small actuatorcurrents, and also exhibit increased reliability overanalogue switching units such as relays and knife-switches.

In the ideal IGBT die, the high conductivityat high current density seen with bipolar gatejunction transistors (BJT’s) is combined with the high speed switching capability of metal oxidesemiconductor FET’s (MOSFET’s).

Figure 1 shows a typical cross sectionthrough the functional areas of an IGBT module,excluding the thermal interface material (TIM) andits associated heat spreader, and the dielectricpotting gel/compound.

Die attach requirements

Significant amounts of I2R heat aregenerated during operation of the module. Anymaterials used in the assembly of the module musttherefore have high bulk (material) thermalconductivity and high electrical conductivity.Processes used for assembly must also ensure thatthe measured thermal and electrical conductivitiesare maintained as close as possible to the bulkproperties. This means that voiding (bubbles in thefinished solder joint, usually found using x-rayinspection) must be reduced to as low a level aspossible. Standard solder-based power die-attachprocesses may generate up to 5% or even 10% totalvoiding, however IGBT die have a much highercurrent density (2 to 2.5 times that of standardpower MOSFETs [1]), so voiding reduction iscrucial. Typically, less than 1.0% or even 0.5%voiding may be desired.

The primary factors [2] for reducing voidingare given in figure 2. To ensure low voiding using asolder paste die attach process it is necessary tooptimise the paste chemistry, metal loading; reflowprofile and preferably vacuum processing.Bond line thickness has to be controlled to betweentypically 250–375 µm (0.010-0.015”). The need for a higher thermal conductance using a thinnersolder layer is balanced by the requirement for asufficiently high stand-off to reduce strain on thesolder joints induced by the CTE mismatch

Figure 1: IGBT cross section

between the substrate ceramic (typically alumina

Al2O3, or, more commonly, aluminium nitride,AlN) and its copper pad.

Wire bonding

Wire bond strength is also central to reliableperformance. Cracks in the bond, bond lift-offs,wire corrosion and coarsening of the aluminiummicrostructure have been identified as major causesof wire bond failure [3.]

Aluminium bonding wire is used instead ofthe higher-conductivity gold, to keep costs to aminimum. Typically there are three or four bondingwires diameter 500 µm (0.020”) or ribbon of equivalent area, for each die. As with allcomponents and interfaces in the entire module, theresistivity must be kept to a minimum, so in someassemblies each wire or ribbon is stitch bonded(multiple bond sights) to the die surface

It is well known that contaminants such asorganic materials or ionic compounds on thesurface of the die can cause a decrease in the pull-strength of the final wire bond [4]. A bond that maypass the pull-strength test, yet entrap a smallamount of oxide or similar material that acts as adielectric, creating a small capacitance that,together with the known inductance problems of thewire bonding wire itself, may increase the durationof a transient (LC) voltage spike.

In many assembly processes using fluxesand solder paste, the trend for the last 15 years hasbeen towards the use of no-clean materials.However, with the notable recent exception of clip-bonding assembly, most Power Semiconductorassembly processes, particularly those using wirebonding, still require cleaning. It is crucial that allcontaminants are removed not just to ensure goodwire bonding but also to ensure proper adhesionand curing of encapsulants.

The flux cleaning process is a complexfunction of a variety of factors to ensure saferemoval of all residue components without damageto materials of construction. It is usually assessedand controlled by monitoring ionic residues, as they

are the principal cause of flux related in-fieldfailures. They are also the easiest to detect andmonitor in (near) real time on a production lineusing ROSE (Resistance of Solvent Extract) testing.In power die assembly this is insufficient asabsence of ionics is not necessarily absence of allresidues. Non-ionic residues may not be harmful inthe service life of the product (the usual concern),but could interfere with subsequent process such aswire bonding, or damage integrity of encapsulantsor their adhesion. More comprehensive evaluationis required.

Solder Paste Selection and optimisation

Table 1 summarises the assembly functionsoutlined above and relates them to die attachprocess requirements.

We surveyed users throughout Europe; thisshowed that most users screen print solder paste, asmaller number stencil print. The most widely usedsolder alloy is 96.5Sn/3.5Ag and post soldercleaning is virtually entirely semi-aqueous. Thusour objectives were to develop a paste capable ofbeing screen or stencil printed through 200μmemulsion 80 mesh, and then batched for vacuumreflow and cleaning.

Project Criteria

Product in combination with process willproduce less than 1% voiding

A minimum 8 hours print/open life Good response to pause (RtP) Post solder cleanliness levels able to

routinely meet or exceed establishedindustry criteria for wire bonding

Capable of being used in existing standardproduction equipment.

The project emphasis was on reflow andcleaning as these areas have the greatest impact onyield and in-service performance.

Candidate formulations

The recent imposition of Pb-free solderingin electronics assemblies has not been without itsissues and these have been thoroughly aired atvirtually every industry conference for at least the

Table 1: Assembly function vs. solderingrequirements

Function Solderingrequirement

Electrical Conductivity Void FreeThermal conductivity Void FreeWire bonding CleanEncapsulation CleanReliable Void Free and Clean

Figure 2, contributors to voiding

last 5 years. On a positive note this lead to a verylarge amount of research into flux and pastetechnology using high tin content solders in SMT.The increased knowledge from this workilluminated the development of the differentlyspecialised LDA (Large Die Attach) pastes used inIGBT soldering.

After review two material types wereidentified and went forward for development

Paste A, New technology Rosin/resin.Paste B Print variant of high Pb die attach type

Paste A is formulated to be minimum processchange material. As an established die attachmaterial, Paste B acted as a control and as a lowresidue material also offered possibilities ofsimplifying the cleaning process.

Printing trials

Modifications made to the pastes were notforeseen to have impact on print characteristics,nevertheless, candidate types were tested at Dek’s UK facility using the range of metal percents to beinvestigated in the reflow, voiding and cleaningtrials. These initial print trials were done toestablish that materials would print and thatresponse to pause was unaffected. All performedwith no measurable variation from their existingproperties.

The print pattern used was designedspecially for this project. It is a generic pincushionrepresentative of that used by many companies.The screen material was SD 245/65 mesh at 22 deg,plus 200 um emulsion. Print method was offcontact with 2mm gap /snap off.

Reflow and voiding Study

Reflow and voiding studies were carried outin cooperation with Pink, using a Pink VADU100oven. Materials used were plain, un-patterned

copper DBC tiles. Die were either Infineon 12 x 12mm or ABB 14 x 14 mm.

Paste was printed using the screen and printparameters previously described. Candidatematerials were reflowed in air, nitrogen or vacuumwith metal loadings ranging from 84.5 to 90 % andevaluated visually and by X-ray.

The principal purpose of this study was todetermine effects of metal loading on voiding usinga variety of profiles. ‘Short and long’ profiles are illustrated below, Figures 4A and 4B. Thesecondary purpose was to determine the effect ofdifferent atmospheres on void levels for possiblelater work.

Figure 3, Screen aperture used in trials

Figure 4A: long profile

Figure 4B: Short Profile

Table 2: Summary of Voiding % all profilesin Vacuum; measured by Dage XIDAT

XD7500VR, grey scale calculation

Metal % Paste A Paste B84.5 0.725 -

85.74 - -

86 0.45 -

87 - -

88.5 0.25 0.675

89 0.6 0.575

89.5 0.5 0.225

90 0.55 0.45

Examples of voiding results are given inTables 2 and 3. Table 2 shows the aggregate ofvoiding performance for all atmospheres forvarious metals loadings. Table 3 shows the in-vacuum performance of Paste A across a range ofmetal loadings according to profile. Figures 5a, 5band 5c show these results pictorially. Figure 5a isPaste type A, and 5b is paste B. These show voids<0.25%, compared to the die in 5C at a little over3% voiding.

Visual Appearance

All Work samples were assessed visually forflux spatter and discolouration/staining around diearea, and ranked on a scale of 1-15. (15 beinglowest spatter)

15–11 Lowest10–6 Medium

5–1 Highest.Paste A Lowest Spatter score 12 at 89.5% M/L.Paste B Lowest Spatter score 14 at 88-90% M/L.

It was found that different atmospheres didnot significantly affect flux spatter. There was somecorrelation of spatter with metal loading (higherloading marginally lower spatter) on flux types A,but this was not strong enough to be able to drawany conclusions or utilise. Flux B showed nospatter variation with metal loading but gaveunacceptable discolouration in surrounding area asshown in. Figure 6. This is attributed to the lowerprocess temperatures of Sn/Ag soldering comparedto high Pb alloys for which the material wasoriginally formulated.

Figure 6: Unacceptable discolouration

0

0.5

1

1.5

88.25 88.5 88.75 89 89.25 89.5 89.75 90 90.25

%w/w Metal

%V

oid

ing

Are

a(T

ota

l)

Profile 1 - Long

Profile 2 - Short

A: Typical Die Paste A Paste Ba: Paste A b: Paste B c: Typical Industry

Figure 5: X–rays of voiding under die

Table 3: Paste A, metals loading and voiding

Cleaning Study

Post solder flux removal or “cleaning” is a complex function of a number of inter-relatedfactors, Figure 7. As discussed in the Introductionpost cleaning of die circuitry has to be carried outto a higher standard than is normal in electronicsassembly and different criteria are needed tomonitor its success. Most cleaning controlprocedures are centred on measuring residualionics, or monitoring conductivity of final rinsewater. Non ionics are of equal concern in die attachassemblies; non ionic residues can impact onsubsequent processes such as wire bonding orencapsulation. A series of tests has been developedwhich monitor total resides [5] and these wereemployed in this project.

The cleaning study was carried out atZestron’s Ingolstadt facility.

Pastes were printed using the projects screendescribed above using further die and substrates asused in the voiding study. Test assemblies werethen reflowed under nitrogen with ~500ppmoxygen. Profile was a straight ramp to a peaktemperature of 255C, time above solidus 70seconds. Nitrogen reflow is “similar to worse” in terms of its effects on flux pryolosis/cleanabilitycompared to vacuum,

Two cleaning processes were used for thetrials to represent standard industrial processes incommon use. Cleaning cycles 1 and 2 were used

with pastes A and B. Cleaning cycle 1, withcleaning chemical at 0% was used for water solublepaste C

Cleaning cycle 1: VA201Miele industrial dishwasher type IR6002

1. 10 mins 50°C 20% concentration ofVA201

2. 3 mins 21°C (room °C) descaled water3. Repeat 3 mins 21°C (room °C) descaled

water4. 3 mins 50°C DI water5. Repeat 3 mins 50°C DI water6. 45 mins 85°C air dryCleaning cycle 2:FA Ultrasonic Process1. 10 mins 50°C ultrasonic 40kHz2. 5 mins 21°C (room °C) DI water3. repeat 5 mins 21°C (room °C) DI water4. 10 mins 85°C dry

Results of the study are listed in Table 3 andshow the pastes meet established criteria forsuccessful implementation industrially.

Conclusion

Paste type A meets the requirements of largepower die attachment (LDA) in IGBT modulemanufacturing processes. The paste has excellentprint and handling characteristics and returns lessthan 0.5% voiding under large die over a widerange of vacuum reflow conditions. The fluxvehicle chemistry offers ease of cleaning to becompatible with the next stage processes of wirebonding & circuit encapsulation.

Acknowledgements

The authors gratefully acknowledge the provisionof facilities, assistance and advice by Dage, Dek,Pink and Zestron, and the supply of workingmaterials by ABB and Semelab.

Cleaning

Flux TypeFlux Type

ProfileProfile

AtmosphereAtmosphereOxygen

level

Forming gas(H2/N2)

Water WashWater Wash ResinMildly

Activated

VacuumVacuum

NitrogenNitrogen

Bath age

Temperature

Time

Amount of post-reflow residue

Amount of post-reflow residue

Cleaning solution /solvent type

Cleaning solution /solvent type

Cleaning equipmentCleaning equipment

Process

ReflowSolder PasteSolder Paste

Figure 7: Major Control Parameters for Post-Reflow Flux-Cleaning Process

Table 3: results of cleaning tests

References

[1] Covi, “IGBTs Challenge MOSFET’s in Switching Power Supplies”, pp. 28-29, Switching Power Magazine,Winter 2002[2 ] Dr Ning-Cheng Lee, “Reflow soldering processing and troubleshooting SMT, BGA, CSP, and Flip ChipTechnologies”, Newnes, pp.288, 2001.[3] Schuetze et al., “The new 6.5kV IGBT module: a reliable device for medium voltage applications” PCIM, August 2001[4] http://ap.pennnet.com/display_article/293844/36/ARTCL/none/none/1/Clip-Bonding-on-High-power-Modules/[5] Strixner, “Optimization due to proper cleaning wire bonding processes”, EPP Europe May/June 2007