Implementation of On-Bonder-Curing to Maximize Array ... · 1992 has gained experience in Hong Kong as Operation Manager for China and the Philippines. At the end of 1992 he joined

Implementation of On-Bonder Curing to Maximize Array Package Manufacturing ProductivitySEMICON West 2000

SEMI 2000

A-1

Implementation of On-Bonder-Curing to Maximize Array PackageManufacturing Productivity

Stephen J. TaylorDexter Electronic Materials

15051 Don Julian RoadIndustry, CA 91746

René J. UlrichESEC

Hinterbergstrasse 32CH-6330 Cham, Switzerland

Stephen J. Taylor has held both technical andmanagement positions in companies supplyingmaterials to the semiconductor assembly industry.Beginning in 1979, he was with 3M Company’sceramic packaging group, subsequently GeneralCeramics, and Ablestik Laboratories. He joinedDexter Electronic Materials in 1996, and currentlyholds the position of Director, International BusinessDevelopment for the semiconductor die attachproducts group. He received his BBA in 1987 andwill graduate with an M.A. in InternationalManagement from the University of Texas at Dallas,School of Management in December 2000. He hasworked extensively throughout the world withmaterials and processes for the manufacture andassembly of semiconductors and hybrids.

René J. Ulrich was born in Switzerland in 1945. In1972 he graduated in M.Sc. Mechanical Engineeringin Zurich, Switzerland. 1973 he began to work in theSemiconductor Industry with Philips-Faselec. 1974he developed the world’s first chip recognition systemfor the ESEC Die-attach machines. Up to 1981 he wasemployed as manufacturing manager for SOT-23transistors and SO integrated circuits and 1981 to1982 manager photolithography for high integratedcircuits. In 1983 he joined ESEC holding positions asmanufacturing manager and process engineer. 1989 -1992 has gained experience in Hong Kong asOperation Manager for China and the Philippines. Atthe end of 1992 he joined again ESEC as ProcessSpecialist in R&D for new paste- and adhesive-dispensing technologies. Due to his wide rangeexperience, René J. Ulrich now is in charge of theInternational Customer Die Bonder Support for allDie attach applications.

Abstract

The rapid technical advances and increased usage ofhand-held devices throughout the world has

accelerated the acceptance and utilization of advancedarray packages, both Ball Grid Arrays (BGAs) andChip Scale Packages (CSPs). The reason for this isthe drive toward small geometry designs thatsubsequently reduce total system size and weightwhile incorporating increased functionality. Trendsand forecasts indicate the largest sectors of advancedarray packages over the next four years will bewirebonded Bismaleimide Triazine (BT) resinlaminate and polyimide (PI) Flex Tape arraypackages. The projected compound annual growth(CAG) of these packages is forecast to be in excess of30%.

With this increasing demand for advanced packages,as well as total IC market growth, manufacturingcapacity constraint is increasing at both originalequipment manufacturers (OEMs) and sub-contractpackage assemblers. Therefore, packaging engineersare being challenged to seek out, investigate, andincorporate “best practices” for advanced processes,both equipment and materials.

In analyzing wirebonded integrated circuit (IC)package assembly operations, die attach (DA) curing,either on a step-cure oven or in a box oven, is the leastefficient step in the total manufacturing process. Inaddition, delays between die placement and die attachcure can exacerbate bleed, as well as allow die shift,and subsequent die misalignment and wirebondfailures. It is the purpose of this paper to explore anddocument productivity enhancements, as well as costreduction, when an on-bonder cure process isimplemented.

In order to accomplish these advanced package-manufacturing results, both equipment (die bonder)and materials (die attach adhesive) must possessunique characteristics and capabilities. ESEC andDexter Electronic Materials have workedcooperatively to develop an on-bonder die attachcuring process for Chip Array Ball Grid Array


SEMI 2000

A-2

(CABGA) and Tape Ball Grid Array (TBGA)packages that offers improved device quality atreduced manufacturing cost.

Introduction

The topic of this paper focuses on two distinct, yetsynergistic technologies, processing equipment andmaterials. In most organizations, the individualsresponsible for each of these areas, (1) die attachmaterial, and (2) die bonding/curing equipment areindependent. Although the die attach adhesive isdispensed and cured with specialized equipment, eachhas the requirement for technological specialization.That is why two individuals have written this paper,one from the equipment side and one from thematerials side.

The paper has been segmented to allow the reader toreview only those areas of individual interest. Anequipment engineer may not want to review thebackground history of die attach materials, while theadhesive specialist may bypass details on machineoperations. However, regardless of an individualengineer's area of interest, some level of additionalinformation relevant to semiconductor packaging canbe obtained. The Sections include BusinessBackground, Die Attach Material, Equipment, andConclusion.

Section I – Business Background

Technology Drivers

As the semiconductor Industry moves into the nextmillennium, three key drivers are moving technologyforward. First, the drive for improved performance.This includes faster and more highly integrateddevices for increased performance capability, smallerpackaging size to fit into smaller form systems, morerobust packaging that eliminates baking prior toattachment, and reduction and eventual elimination ofhazardous constituents from the finished product.

The second technology driver is productivity. Theability to manufacture an increasing number ofdevices with diminishing resources, which areprimarily equipment, labor, and facilities. Theobjective is to optimize the facilities’ capacitycapabilities through design. This includes selection ofthe most efficient equipment as defined by the

throughput, uptime, machine cost, operator cost,maintenance cost, and power usage cost. Anadditional factor that must be included with the aboveequation is the floor space utilization of theequipment. Another important component of capacitymaximization is yield loss reduction in all areas ofproduction, which will rely on application of bestpractice models to equipment, materials, andoperational procedures.

The third and most significant technology driver forthe semiconductor industry is profitability. That is,the ability to balance sales and costs through effectivemanagement in order to meet the financial objectivesof the company. In its simplest form, reducing costscan always increase profitability. Therefore, improvedprofitability is the result of the effective application ofproductivity enhancing methods.

As technological innovations and advancements areultimately the result of market driven forces, it is ofvalue to understand those underlying motivations thatinstigate new technology. With the primary marketdrivers of computers, consumer electronics,communications, industrial electronics, andtransportation, the semiconductor industry isprojecting annual volume growth in excess of 10%through 2003. While a broad variety of plastic smalloutline (SO) and quad flat pack (QFP) IC packagedesigns will continue to grow during this period, theBGA and CSP package designs are expected toproliferate dramatically through the next five years.

Market Forecast

The use of BGA devices should grow at a CAGR ofover 35% while CSP units will exceed 71% volumegrowth, combining to represent in excess of 13% ofthe overall semiconductor package market by 2003.1The current challenge for semiconductor packagingfacilities is the proliferation of BGA and CSP designs,with no dominant design clearly leading the way. Thesubstrate types for these packages include FR4, BT,and flextape which are interconnected within thepackage by wire, flip chip, or a direct chip attach(DCA) scheme. However, projections of usage byvolume indicate rigid BT will account forapproximately 58% and flex tape 28% (total 86%) ofthe array substrates by 2003. More significant is theinterconnection type with wirebond dominating at78% followed by flip chip at 14% and deposition 8%.2


SEMI 2000

A-3

With the continued dominance of wirebonding for thenext generation of semiconductor packages commonlycalled arrays, the use of die attach adhesives willremain an important factor for both componentperformance and assembly productivity.

Section II – Die Attach Materials

History

A review of past die attach technology will establish afundamental understanding of the importance ofchemistry in the development of the newestgeneration die attach adhesives that will play animportant role in the majority of array type packages.

In the early 1970s the microelectronics industry waslooking for alternative materials that would allowattachment of electronic components at temperatureslower than traditional solder and eutectic compounds.The search led in the direction of organic polymersand focused on the 1930s invention, epoxy resin.Epoxy resin is converted from its monomeric stateinto a “cured” thermoset by the initiation of a cross-linking reaction. The reaction is either a catalytichomopolymerization or a heteropolymerizationinitiated by co-reaction of the functional epoxidegroups with a cure agent. Because the conversion ofan epoxy resin into a thermoset solid requires theaddition of a reaction initiator, epoxy compounds canbe categorized as having a step- addition cure.

The early epoxy component attach adhesives utilizedaliphatic amines, which although being highlyreactive did not completely react with the epoxy resin.The unreacted amine-curing agent then combinedwith moisture, which hydrolyzed the free Chlorideions (Cl-) and created corrosion on the metallicsurfaces within the device or package causingcatastrophic electrical failures. As early problemsslowed the potential implementation of epoxycomponent attach adhesives in the microelectronicsindustry, other polymers were evaluated for thispurpose.

Due to the low level of hydrolizable Cl-, pre-imidizedpolyimide had a distinct advantage over epoxyadhesives for microelectronic devices that weresusceptible to corrosion. This type polyimide adhesivewas used for semiconductor die attach beginning inthe late 1970s, and in some applications are still

currently being utilized. However, their requirementfor a long two-step cure profile (>1 hour) to evolvethe high level of solvent they contain reducesproductivity. Other typical properties of Polyimidedie attach adhesives are high modulus and hightemperature processing.

By mid 1980s the difficulties with epoxide systemswere being resolved by the use of “cleaner”constituents with the resulting improvement inChloride levels and subsequent reduction ofcomponent or package corrosion. This secondgeneration of microelectronic adhesives wasimplemented in the production of military hybridsconforming to MIL STD 883C Method 5011 and inplastic dual inline package (PDIP) semiconductors.At the end of the 1980s an ever increasing volume ofsemiconductors was being placed into plasticpackages and silver filled epoxy became the primarydie attachment adhesive for the industry. Theemphasis became high productivity rates whichchallenged the epoxy die attach manufacturers todevelop better rheology products that would notexhibit stringing or tailing of material during the veryhigh speed dispensing processes.

Throughout the following decade, managers ofsemiconductor packaging facilities worldwide beganviewing productivity improvement as a means of costreduction and an increasingly important factor forimproving profitability and competitiveness. In theearly 1990s, as the quality and reliability of epoxy dieattach gained stability, interest was expressed aboutreducing the cure time for the epoxy from thetraditional one-hour minimum. The term for this newconcept was snap-cure, and is generally accepted tomean a cure time of less than one minute in duration.

Although epoxy die attach compounds have providedgood overall results for semiconductor die attachduring the last twenty years as modifications wereattempted to reduce moisture absorption, lowermodulus, and accelerate cure reaction time, the resultsare now falling short of advanced processingrequirements. Although epoxy polymerizationreactions can potentially be accelerated to a fewseconds, the dynamics of die attach rheology haveintervened. To assure optimum dispensability withsingle needle, multi needle, or pattern writetechnology, die attach adhesives are typicallyformulated to a viscosity between 6,000 to 10,000 cP


SEMI 2000

A-4

(Brookfield CP51 cone and plate @ 5 rpm) and athixotropic index (TI) of 3.5 – 6.0 (Brookfield CP51cone and plate .5 rpm ÷ 5rpm). To meet thesecriteria, epoxy die attach formulations typicallyincluded diluents to lower the viscosity to the target.Although current technology epoxy die attachadhesives utilize diluents that react into the crosslinkmatrix rather than evolve during cure, they must havesufficient time to react, otherwise voiding occurs.

The solution for overreacting an epoxy snap-cure dieattach adhesive, and generating voids, is to heat thematerial in a ramping, or stepped cure operation. Thistype processing requires the installation of specialdesigned “stepped cure” ovens into the factory asstand alone units. While this process slowly evolvedin the semiconductor packaging industry, moreadvanced solutions were being explored. Thesesolutions required a new perspective on the type ofmaterials, equipment, and processing that could bringrevolutionary change to the industry rather than theslow evolutionary paradigm from the past.

Die Attach Adhesive Curing on Die Bonder

The changes that are currently underway insemiconductor packaging directly address the threetechnology drivers discussed earlier in this paper,Performance, Productivity, and Profitability. Cureon bonder describes the process of dispensing dieattach adhesive onto the substrate, placing the die, andcuring the die attach adhesive in a continuous processself-contained within the die bonder unit; and withoutthe loss of die bonding speed. With this newprocessing technology, the cure of the die attachmaterial no longer requires ‘stand-alone’ curingequipment that takes clean-room floor space. Also,the process typically does not slow the die bonderunits-per-hour (UPH) rate. Therefore, with no floorspace consumed by curing equipment and noadditional time consumed for curing, the cure processbecomes transparent in the assembly operation. Onedie attach manufacture has identified this advancedprocessing as Skip-Cure®.

In order to cure on-bonder, without reducingthroughput, the die attach material must be capable ofcompleting a significant portion of its cross-linkingreaction in ≤10 seconds and develop a void freebondline. As previously described, epoxies with theirstep-addition reactions and diluent additives have

been unable to meet this objective. Therefore,alternate polymer chemistries have developed overrecent years, including systems containing acrylateand liquid bismaleimide (BMI). These new systemswhich employ monomers that will co-polymerize arebeginning to replace traditional epoxies due to severalunique characteristics they possess. Of greatestdifferentiation with epoxies is the curing mechanismthat is free radical initiated rather than the step-addition of epoxy. A free radical chain reaction, onceinitiated, proceeds at an extremely rapid rate. In Chart1, a series of time of reaction kinetics curves arepresented at temperatures from 110ºC to 170ºC for acommercial BMI based die attach adhesive.

Time of Reaction - Cure KineticsBMI based Die Attach Adhesive

0

20

40

60

80

100

120

0 2 4 6 8 10 12 14 16 18 20 22 24 26 28 30Reaction Time in Seocnds

% R

eact

ion

170°C 160°C 150°C 140°C 130°C 120°C

110°C

Chart 1

The free radical chain reaction of advanced polymersystems, in which all the chemical constituents co-polymerize, support extremely rapid reactions inwhich the majority of cured adhesive characteristicsare attained in a period of time as short as 5 seconds.These free radical curing systems typically contain“100% solids” which reduces the weight loss duringcure to <0.2% compared to a commercial epoxy snap-cure die attach adhesive that reports a weight loss of2.1% during a 60 second cure @ 200ºC. Weight lossat cure reflects the weight volume of cure reactionbyproducts that are released as a gaseous volume.This can be a factor of consideration when evaluatingthe percentage of cure following a cross-linkingreaction.

Historically, semiconductor packaging engineers have“required” die attach adhesive be 100% cured whenemerging from the cure oven. Their concern was to


SEMI 2000

A-5

reduce or eliminate the evolution of outgassedconstituents from the die attach adhesive duringsubsequent processing steps, including wire bonding,molding/encapsulation, and post mold bake. Inreality, polymer reactions never achieve 100%polymer chain cross-linking. With the very low levelof weight loss during cure from the new systems(<0.2%), any post-cure reaction that may take placeduring subsequent operations, has nominal impact inprocessing. Chart 2 is an isothermal DSC of acommercial production L-BMI die attach adhesive.The reaction on-set occurs at 73.4°C with a peak at84.8°C.

Chart 2

To evaluate completion of reaction after various curemethods, 5mm and 7mm die were placed on BTsubstrates with a commercial Ag filled BMI material.Three cure schedules were utilized, 12 seconds @160°C and 12 seconds @ 175°C on the die bonder,and 15 minutes @ 150°C in a standard box oven.Following these cure profiles, the die were removedand the cured die attach adhesive collected. DSCkinetics were run on each residual sample todetermine if the cure kinetics had been completedduring the initial cure. Chart 3 (160C @ 12 seconds)is representative of the six scans taken. In review ofeach result, no additional reaction is present andindicates the reaction was completed during the initialcuring process.

Chart 3

These methods are typical for evaluation of the bulkmaterial properties. An additional area of importanceis the interfacial reaction boundary (IRB) and thephysical, chemical, and molecular interactions thatoccur in this critical area. It is typical of free radicalinitiated die attach materials to demonstrate increasedadhesion over time, following initial cure. It ishypothesized that these changes occur as a result ofinterfacial coupling within the IRB rather than as aresult of additional bulk reaction cure.

BGA Flex Application

The current growth in BGA flex tape packages can beattributed to both technical and market drivenadvantages.3 However, the packaging process of flextape devices can be challenging due to the thinflexible polyimide tape. This difficulty is exacerbatedas the tape thickness narrows to near 50 microns. Diebonding on flex tape is difficult due to the nature ofthe polyimide and its tendency not to remain flat. Toresolve this, Cu rails have been attached to somedesigns, while bare tape is used in others. However,the die bonding, handling, and transportation of flextape from the time the die is placed until it is curedcan cause bondline shift and die movement resultingin misalignment and die tilt. The delay betweenplacement and cure also allows time for resinousbleed to escape from the die attach materials andcontaminate wirebond pads on the substrate. Each ofthese phenomenon, bond line shift, die movement,and bleed can create the need for additionalprocessing steps such as cleaning, and may also createyield loss.


SEMI 2000

A-6

The cure on bonder is capable of eliminating each ofthese problems, as well as eliminating the floor spaceneed for cure equipment and cure time.

Experiment on Flex Polyimide Tape Array

Parts were assembled on an ESEC 2007 die bonderretrofitted with a cure on bonder heater blockassembly and cured at 160°C for 18 seconds. Inaddition to the heater block on the bonder is thevacuum hold down during die attach dispense, dieplacement, and die adhesive cure, maintaining theplanarity and stability of the polyimide flex tapeduring all critical die attach functions.

Photo 1 is the bondline of one of twelve devicesmounted with Ag filled BMI die attach adhesive on a79 micron thick polyimide 4 x 4 matrix flex BGAsubstrate. The die size is 5mm x 5mm x 355 micronsand the cross sectional photograph is taken at amagnification of 500X. The die was bisected equallyalong a centerline and the three sections shown are theleft end of the die, the center of the die, and the rightend of the die.

Photo 1

The bond line thickness is 21 microns in the center,and 21 at each end. Twelve devices were attachedwith the results in listed in Chart 4 below.

Chart 4

Photo 2 is the same substrate and die, with a “leading”high volume Ag filled oven cure epoxy die attachadhesive. The twelve devices were assembled andstaged for one hour, followed by an oven cure, with a30 minute ramp to 150°C and held at 150°C for onehour.

Photo 2

The results for these devices in Chart 5 indicate thedie attach adhesive has pulled back from the peripheryof the die creating an average bondline thickness atthe center of the die of 27.42 microns and combinedaverage of both ends of 14.48 microns. Alsoobservable in the photographed cross sections areresin rich areas (red arrows) in which the movementof the die, flex substrate, and die attach adhesiveduring handling and cure have displaced the Ag fillerin the adhesive and allowed the epoxy resin to pool inthe voided areas.

Chart 5

A third set of devices was assembled with the samesubstrate and die, using a quick cure Ag filled dieattach adhesive. The material was dispensed anddevices placed and staged for one hour. They werethen oven cured for 15 minutes at 150°C in a pre-heated oven without ramp. The results showed an

SKIP-CURE BDevice Left Center Right

1 23 23 222 21 22 223 19 21 224 21 21 215 20 21 206 22 23 237 22 23 248 22 22 219 23 26 2510 23 23 2411 20 22 2312 21 22 22

Average B/L 21.42 22.42 22.42STD DEV 1.31 1.38 1.44

EPOXY A - OVENDevice Left Center Right

1 14 22 152 19 24 123 16 23 144 20 20 205 24 32 246 19 31 217 19 29 198 18 27 219 24 29 2210 17 31 2211 17 31 2112 18 30 24

Average B/L 18.75 27.42 19.58STD DEV 2.93 4.12 3.90


SEMI 2000

A-7

increase in die attach adhesive shift during handlingand cure. See Photo 3.

Photo #3

As with the traditional oven cured Epoxy A, thequick-cure processing also allowed the polyimide flexto deflect upward around the die edges whilemaintaining a relatively constant adhesive bondthickness (29.00 microns) in the center of the die. Theedges of the flex were pulled upward creating asubstantial bow to the bottom of the package.Although wirebonding was not a component of theseexperiments, it could be anticipated that thedeformation in the backside of the flex polyimide tapedue to the variation in the bondline thickness couldinstigate die cracking. This may occur as downwardpressure on the perimeter structure of the tape createda levering action across the package due to thefulcrum point under the center of the die. Furtherexperimentation will be necessary to confirm thishypothesis.

A compilation of the results from the three curemodes described above are shown in Chart 6 andinclude: cure on die bonder with a commercial BMIsystem, quick 15 minute cure in box oven, and 1 hourcure in box oven with widely used standard epoxy.The die width (5mm) is represented by the x-axis, andthe y-axis depicts the maximum average difference(∆) in bondline thickness between the polyimide tapeand the die back from the center to the end of the die.

Chart 6

Experiments on Rigid BT Substrate Array

A series of die attach adhesion tests were performedon four different BT substrate designs (A-D) and 325mm x 5mm die per substrate type. Each substratedesign was cured at 150°C for 10 seconds on the diebonder. The results indicate a variation in substratesurfaces that translates into adhesion variation up to35% as shown in Chart 7.

Chart 7

Substrate B was selected for further adhesion testing.Four legs, with 16 die each, were run using a standardcommercial Ag filled BMI cured on the die bonder.Each leg was exposed to the heater block on thebonder for approximately 18 seconds. In production,time on heater is reliant on the number of sites in thematrix, in this case 16. The objective was to determineadhesion at two temperatures with two die sizes.

The results at 160°C show the average adhesionstrength on a 5mm x 5mm die at 14.97Kg with STDDEV of 1.27 and on a 7mm x 7mm die at 33.38Kgwith STD DEV of 4.92. When the cure temperaturewas increased to 175°C the adhesion on the 5mm x5mm was 12.73Kg STD DEV 1.77, and on 7mm x 7mmadhesion was 31.06Kg STD DEV 5.22. When theseadhesion results are compiled with the results onsubstrate B in Chart 7 (cure @ 150°C), it indicates theadhesion properties of the material are acquired veryquickly at a relatively low temperature.

10 Seconds @ 150C on Bonder

0

2

4

6

8

10

12

14

16

A B C DK

g sh

ear f

orce


SEMI 2000

A-8

Section III - Equipment

On-Bonder-Cure (OBC) Overview

To achieve the optimum quality level for the adhesivelayer, between the chip and substrate, the followingpoints must be considered:

• Stable bond line thickness• Minimum die tilt• Regulated fillet with 100 % coverage• Homogeneous adhesive layer (no voids)• Precise and regularly dispense volume

To accomplish this it is necessary to use a chipassembly machine that includes a precise adhesivedispense system. But, not only the dispense systemshould work precisely. There are many more points,to take care about it.

The system needs a special, integrated cure system,which is controlled by exact timing and temperaturelevels to guarantee a stable and reliable process.

Before explaining the details of the integrated curingsystem, it is helpful to know more about theadvantages. The following points can make thedecision to implement such a system easier.

• No bleeding• Not uncontrolled adhesive flow• No surface contamination on devices• No transportation damages• No chip shifting during waiting time• Perfect planarity of flexible material• Controlled curing time for all devices• Precise bondline thickness• Minimum of die tilt.• Minimal thermal stress for device

Because of these quality aspects, the integrated on-bonder-cure becomes very attractive. To run such aprocess, it is necessary to focus attention on thefollowing points.

The footprint of the die attach machine should notchange because of the integrated curing system (costsaving). Therefore, because of the necessity for a fasttemperature ramp up, only extremely fast curing

material can be used for this process. The adhesivesystem has to be designed for this new method.

Beside the previously listed benefits, the new processalso opens the door for completely new manufacturingmethods. Multi-pass applications are no longer asdifficult to process. Using single chip machinesinstead of the more complicated Multi-chip machinescan increase both productivity and throughput. Costsavings overall will be one of the results.

The next few pages will explain in detail what isnecessary to implement the process properly.Therefore it is necessary to understand more about thematerial which has to be processed for the “Skip-Cureprocessing”.

Thermal response of substrates

To know how a die attach adhesive material reacts in

Diagram 1

the curing position, it is necessary to know thethermal response of the substrate on which it will beplaced. This allows estimation of the percentage ofthe curing reaction that may take place during the dieattach process. Will it cure 100 % or maybe only40%.

The above Diagram 1 shows in detail the rate at whichdifferent substrate materials allow thermal transferthrough their mass to the die bond site while on aclose contact hot plate at 150°C. The cooling curve

Thermal response at 150 °C

0

50

100

150

200

0 10 20 30 40 50 60 70 80

Time [sec]

Tem

pera

ture

[°C

] BGACUFLEGLACER


SEMI 2000

A-9

shows the temperature decrease when the substrate isonly surrounded by air at room temperature (24.°C).

The next Diagram 2 shows now the stretched heat upprofile.

Diagram 2

Thermal response at 150 °C

0

50

100

150

200

0 1 2 3 4 5 6 7 8 9 10

Time [sec]

Tem

pera

ture

[°C

] BGACUFLEGLACER

Legend for Diagrams 1 and 2:-BGA = BGA strip, 0.3mm thickness-CU = copper 0.21mm-FLE = Flex or tape BGA 0.11mm-GLA = glass plate 1.05 mm,-CER = Ceramic substrate 0.5mm

After only two seconds, the temperature of a copperleadframe has reach 100 °C.

To obtain these temperature profiles it is important tohave perfect thermal contact with the oven (heater)surface. Therefore, the vacuum holding system is oneof the essential parts for the on-bonder oven design.Based on the bonding layout, the vacuum area has tobe designed for this size to guarantee the necessarydownhold force for the highest thermal transfer. Ifpossible, the dimension of the vacuum panel shouldbe slightly bigger than the bonding area or exactly thesame size of the bond pad to have the most reliableprocess. If the substrate does not fit exactly thevacuum area, leaking air can cool down the heaterplate and damage to the vacuum valves can bepossible.

To speed up throughput and have a reliable, stableprocess it is important to give attention to the abovepoints.

Thermal transfer Vac. on off at 170°C

0

50

100

150

-10 0 10 20 30 40 50 60 70 80 90 100

Time [sec]

Tem

pera

ture

[° C

]

Diagram 3

From the Diagram 3 it is possible to see the influenceof the downhold vacuum. The blue curve shows thetemperature profile of a substrate without any vacuumhold down. This profile can change from substrate tosubstrate. If the substrate is perfectly flat, it can showa curve nearly like the red profile but it can also beworse than the blue profile.

The above Diagram 3 is the profile of a 16 Chip (4x4)matrix application. But if the matrix is smaller, 2x2(4 chip), which is shown in Diagram 4, the availablecuring time will become much shorter.

The next Diagram 4 shows very clearly this newsituation.

Thermal transfer Vac. on off at 170°C

0

50

100

150

0 10 20 30 40 50 60 70 80 90 100

Time [sec]

Tem

pera

ture

[° C

]

Diagram 4The blue curve does guarantee a reliable process andthe percentage of completed cure reaction varies in awide range. The thermal contact time betweensubstrate and oven heater plate in Diagram 4 is 11


SEMI 2000

A-10

seconds. The previous Diagram 4 is based on contacttime off at 31 seconds. The curing time is primarilygoverned by the actual process time. The equipmentcycle time has some influence on this. But if theactual bonding time is higher then the cycle time, thenthis will take the first priority for the curing cycle timeof a single device or a complete matrix.

In addition to these influences, it is also possible to setthe actual curing time as a value for each processedchip.

Example:36 chip matrixBonding time: = 90 msCycle time = 700 msAux. Mod.t. = 0 ms

Throughput = 3600 / 0.7 = 5142 UPHCuring time = 36 x 2 x 0.7 = 50.4 sec.

The throughput varies dependent on what functionsare activated on the die bonder. If full qualitychecking, optical dispense alignment, etc areactivated, the throughput goes down to 3600 to 4000UPH. Therefore it is important to take the actualprofile of an existing application, and to know moreabout the curing conditions.

Measuring method to collect the thermal profile

The handling of the measuring equipment should beeasy as possible. The best method is to use a personalcomputer with a digital multi meter that has thecapability to communicate through the serial port.These are standard instruments, available for areasonable price.

The next Photo 4 shows the complete measuringequipment to take thermal profiles on the die bonderheater assembly.

Photo 4

Very simple but efficient software allows collectingall the necessary information. When the machinestarts, the leadframe with attached thermocoupleproceeds into the feed area to begin the measurementcycle and sends the data to the computer. It isimportant to watch closely that the coax thermocouplecable as it passes through the die bonder so it will notcatch on any objects or crash.

After starting, the computer program should createautomatically the filename, related to the date andtime. The values are stored in a text file in a way thatit can be easily loaded back into any spreadsheetcalculation program and create the graphic of theprofile.

The screen is automatically scaled to the length of thechosen measuring time. All the necessary informationis available from the screen (file name, actualtemperature etc.).

The thermocouple is attached to the substrate with theAg filled ultra fast cure adhesive QMI-506. In orderto replicate the actual curing process during thermalprofiling, a populated substrate is use. This can beobserved in Photo 5 below.


SEMI 2000

A-11

Photo 5

From the text file that is generated it is possible toretrieve more information for further calculations.This is especially the case if registration index timingcapability of the software program is utilized to settiming marks, by using the number keys from 0 to 9.

For example, if the epoxy writing position has thenumber 1, then the bond position can be No. 4, withthe first curing position number 5 and the secondcuring position number 6. The software code linesbelow demonstrate how the information is stored inthe text file “05041508.TXT”

31,36,"5"32,58,""33,110,""34,140,""35,153,""36,157,""37,159,"6"38,160,""39,163,""40,164,""41,165,""42,165,"7"

The first value shows the time in seconds, the secondvalue shows the temperature and the third columncontains the remark sign. Therefore, it is easy to havethe real contact time for entry into the first curingposition (“5”) and exit from the second curingposition (“7”). In our sample these time are 42seconds – 31 seconds = 11 seconds curing time.

On-Bonder Heater System

Photo 6

Due do the very sensitive flexible BGA materials theconstruction has changed from the fix oven heaterblock (see above Photo 6) station to the movableshuttle oven heater block (see Photo7 below). Theprimary reason for this change was to minimize anymovement of the substrate during the die attachprocess. The substrate is held in position until the full

.

Photo 7

matrix is completed. During the transportation, thevacuum hold down is momentarily switched off.

Because of the very short transportation time anddistance to the curing position after adhesive dispenseand die placement, the adhesive on a BGA flex


SEMI 2000

A-12

polyimide substrate remains in place with minimaldisturbance.

With the on-bonder-cure process, when the substratewith cured die exits the die bonder, storage time in themagazine no longer has an influence on diemovement, die tilt, or adhesive resin bleed.

Photo 8

This is particularly of advantage with thin polyimideflex film that may be only 50 microns thick and sagsin the magazine, as shown in Photo 8.

For rigid BGA substrates additional “down-holders”may be recommended to minimize any movement inz-axis during the transport cycle. In the next Photo 9it is possible to see the shuttle plate with the bondposition, curing position one, and cure position two.

Photo 9

Between the bonding area and the curing position is agap that interrupts the thermal flow back into the bondposition. This is necessary due to the extreme fastchain reaction die attach materials used in thisprocess, which could initiate curing if the gap werenot present.

From the profile in Diagram 5, it is possible to see theinfluence of this gap between the bonding and thecuring zone. No. 1 shows the arrival at the dispensingposition. No.2=Bond position. No. 3=curing position1. No 4=arrival at curing position 2.

Curing profile 36-Chip Matrix, 150°C

0

50

100

150

0 60 120 180 240 300 360

Time [sec]

Tem

pera

ture

[° C

]

1 2

3

4 5

6

Diagram 5

Bondline thickness and Die tilt

Due to immediate curing after the die attach position,

Diagram 6

DIE tilt analyse Blank BGA strip Adhesive: QMI-536Data file: 05031425.TXT 1. Pad (16 units)

x y zCHIPSIZE 4.000 4.000 0.731 Measurings (BLT) No. E1 µm E2 µm E3 µm E4 µm TILT µm BLT µm1 28 28 26 25 3 272 32 29 27 29 5 293 29 26 25 27 4 274 31 28 27 30 4 295 32 29 30 31 3 316 31 29 29 31 2 307 32 29 29 31 3 308 29 27 28 29 2 289 28 29 29 29 1 29

10 32 30 30 31 2 3111 30 29 28 30 2 2912 31 29 30 30 2 3013 31 29 29 31 2 3014 29 27 28 30 3 2915 31 29 30 31 2 3016 27 28 30 29 3 29

AVG 3 29STDEV 1.014 1.220


SEMI 2000

A-13

the bondline thickness is more stable and therefore itis much easier to measure and collect the values forsetup. An example of this data collection can be seenin Diagram 6.

To take the bondline thickness (BLT) and die tiltmeasurements, a mechanical micrometer with a stablestand and computer interface was utilized. The valueswere directly transmitted to the personal computer.From the next Diagram 7 it is possible to see themeasuring points.

E1 Y2 E4

X1 Chip X2

E2 Y1 E3

Diagram 7

To prevent any measuring errors the point E4 is alsocalculated in addition to the collected value, to checkthat there is no reading mistake. The tilt value isdistance between the lowest and the highest chipcorner.DIE tilt analyse Cu-LF, PLCC68 LF3, pad 1 & 2Data file: 05051334.TXT 9 units/pad

x y z Adhesive: QMI-536CHIPSIZE 4.000 4.000 0.727 Measurings (BLT) No. E1 µm E2 µm E3 µm E4 µm TILT µm BLT µm1 29 26 26 27 3 272 30 27 23 26 7 273 29 26 24 27 5 274 28 26 25 26 3 265 28 25 25 27 3 266 28 26 25 27 3 277 27 27 27 28 1 278 28 26 26 28 2 279 28 26 27 28 2 2710 28 28 26 25 3 2711 30 28 23 26 7 2712 29 26 24 27 5 2713 27 27 26 26 1 2714 27 27 25 26 2 2615 28 25 25 27 3 2616 26 26 26 28 2 2717 27 24 25 27 3 2618 27 26 26 27 1 27

AVG 3 26.6STDEV 1.811 0.382

Diagram 8

To obtain excellent results it is necessary to adjust thebond head (leveling). The tilt adjusting is done by theprior calculated correction value for the bond headadjusting screw. To ensure that this result is not achance, the test was repeated with a metallicleadframe.

DIE tilt analyse Cu-LF, PLCC68 LF3, pad 5 & 6Data file: 05051344.TXT 9 units/pad

x y z Adhesive: QMI-536CHIPSIZE 4.000 4.000 0.727 Measurings (BLT) No. E1 µm E2 µm E3 µm E4 µm TILT µm BLT µm1 27 25 25 28 3 262 29 27 22 24 7 263 28 26 23 26 5 264 26 26 25 27 2 265 27 26 25 26 2 266 28 25 24 29 5 277 26 25 25 27 2 268 25 25 26 27 2 269 26 24 26 26 2 26

10 27 27 25 25 2 2611 29 27 22 24 7 2612 28 24 22 25 6 2513 26 25 23 24 3 2514 26 25 24 25 2 2515 28 25 23 26 5 2616 26 25 25 27 2 2617 28 25 26 28 3 2718 27 25 25 27 2 26

AVG 3 25.7STDEV 1.854 0.564

DIE tilt analyse Cu-LF, PLCC68 LF3, pad 3 & 4Data file: 05051339.TXT 9 units/pad

x y z Adhesive: QMI-536CHIPSIZE 4.000 4.000 0.727 Measurings (BLT) No. E1 µm E2 µm E3 µm E4 µm TILT µm BLT µm1 28 25 25 27 3 262 30 26 22 25 8 263 27 25 23 26 4 254 26 26 24 26 2 265 27 24 23 27 4 256 25 26 26 27 2 267 26 24 25 27 3 268 26 24 26 27 3 269 26 25 25 27 2 26

10 25 24 27 29 5 2611 28 26 21 23 7 2512 27 23 22 27 5 2513 28 28 22 24 6 2614 26 24 24 26 2 2515 26 24 23 25 3 2516 25 24 25 26 2 2517 25 23 25 26 3 2518 25 23 25 27 4 25

AVG 4 25.3STDEV 1.801 0.550

Diagram 9


SEMI 2000

A-14

From the three data charts above (Diagrams 8 and 9),it is possible to see that achieving good die tilt resultsare realistic and reliable. Several tests show a similarresult about the values. An important influence is alsorelated to the pick up tools. Stable and precise toolsare also necessary. The traditional soft rubber pickuptool makes it very difficult to obtain stable, repeatableresults.

The following values were used for the tests above:

• Die size = 4 x 4 mm• Bond force = 80 gr.• Bond time = 90 ms• Adhesive volume = 0.086 mm3

Section IV - Conclusion

At present, the liquid BMI and acrylate based dieattach systems have gained worldwide acceptancethroughout the semiconductor packaging industrywith an estimated 45% of the rigid and flex array dieattach market in 1999. It is anticipated that within thenext year, all four major die attach suppliers will havea broad offering of new materials that are eitherentirely free radical initiated, or hybridizedformulations of epoxy, polyimide, BMI, acrylate, andother unique monomers. The proven JEDEC andEIAJ reliability of these materials is no longer aquestion or issue. The focus is now on manufacturingprocess optimization in the semiconductor packagefactory in order to mine the full capabilities of theseadvanced technology materials.

Based on the experiments described in this paper theresults demonstrate the ability to improve both thequality and manufacturing productivity of advancedsemiconductor packages, such as tape and rigidarrays, by the implementation of both advanced freeradical curing die attach adhesives and advanced diebonding equipment.

The combination of material and equipment isessential to derive the maximum capabilities andbenefits of this advanced packaging model. Theauthors believe that on-bonder-cure will become thearray package processing benchmark for thesemiconductor packaging industry based on thesignificant benefits it offers. As the industrycontinues to be driven by the desire to improveperformance, productivity, and profitability, newmanufacturing systems will be embraced. Since theon-bonder-cure process addresses all three areas, it isanticipated it will receive wide spread interest andimplementation.

References

1 Advanced IC Packaging Markets and Trends – ThirdEdition, Electronic Trend Publications, Inc., San Jose,1999, p.3-14 & 15

2 ibid. pgs.4-6 & 5-10

3 Flex tape for BGA applications, Diorio, Mark,Advanced Packaging, May 2000

Acknowledgments

Stephen Taylor would like to sincerely thank Dr.Steve Dershem, Dr. Carl Hoge, Dr. James Huneke,Dr. Richard Jaeger, Dr. Ben Santos, and JeffMcConnell for reviewing the content of this paper andoffering ideas and thoughts on the subject matter. Inaddition, Mr. Ulrich and Mr. Taylor want to thankShawn Gregg, Jose Newman, and Dan Raymor for thepreparation, analytical testing, and data collectionutilized in this paper.

Documents

Implementation of On-Bonder-Curing to Maximize Array ... · 1992 has gained experience in Hong Kong as Operation Manager for China and the Philippines. At the end of 1992 he joined