Open Access is an initiative that aims to make scientific research freely available to all. To date our community has made over 100 million downloads. It’s based on principles of collaboration, unobstructed discovery, and, most importantly, scientific progression. As PhD students, we found it difficult to access the research we needed, so we decided to create a new Open Access publisher that levels the playing field for scientists across the world. How? By making research easy to access, and puts the academic needs of the researchers before the business interests of publishers.
We are a community of more than 103,000 authors and editors from 3,291 institutions spanning 160 countries, including Nobel Prize winners and some of the world’s most-cited researchers. Publishing on IntechOpen allows authors to earn citations and find new collaborators, meaning more people see your work not only from your own field of study, but from other related fields too.
To purchase hard copies of this book, please contact the representative in India:
CBS Publishers & Distributors Pvt. Ltd.
www.cbspd.com
|
customercare@cbspd.com
A highly reliable and reworkable underfill adhesive based on thermoset epoxy resin possessing thermally reversible dicyclopentadiene (DCPD) moiety is described. The adhesive can be cured rapidly at moderate temperatures resulting in high Tg cured network, which gives high reliability to the bonded semiconductor components. The inherent thermal reversibility of DCPD moiety causes network breakdown at high temperatures enabling easy removal of defective semiconductor chips. A discernible trend between loading level of the thermally reversible epoxy resin and high-temperature die shear strength was observed. Using this novel adhesive system, both high reliability and reworkability can be achieved concurrently, which is normally not possible with other thermoset adhesive systems. The epoxy resin used in the study was scaled up to multi-kg quantities demonstrating industrial applicability of the approach.
*Address all correspondence to: laxmishasridh@gmail.com and timothy.champagne@henkel.com
1. Introduction
The popularity of handheld display devices (HHDDs) has made their demand increase dramatically in recent years. Manufacturing throughput has consequently been challenged to meet the growing demand. One area that is particularly troublesome for manufacturers is the treatment and handling of defective semiconductor chips on a circuit board. For instance, during the manufacture of a circuit board subassembly, a multitude of semiconductor devices are electrically connected to the circuit board in chip scale packages (“CSPs”), ball grid arrays (“BGAs”), land grid arrays (“LGAs”), and the like [1]. The board may then be tested to evaluate function and sometimes the board fails. In such cases, it is desirable to identify the semiconductor device that caused the failure, remove it from the board, and reuse the board with the remaining functioning semiconductor devices. This would save cost for the manufacturer of HHDDs.
Ordinarily, semiconductor devices (also called as chips) are connected to electrical conductors such as Cu pads on printed circuit boards (PCBs) by solder connection as shown schematically in Figure 1A. The coefficient of thermal expansion (CTE) of the Si die and the solder balls is significantly different. When the resulting subassembly in a HHDD is exposed to mechanical shocks such as vibration, distortion, drop (Figure 1C), or rapid temperature change (thermal shock, for example, when a device is left in a car during winter or summer), the reliability of the solder connection between the circuit board and the chip often becomes suspect. There are other modes of failure that can occur in a subassembly such as electrical shorting or stress cracks in solder balls (B & D, respectively, Figure 1).
Figure 1.
A cross-sectional representation of Si die bonded to a circuit board through solder balls without an underfill adhesive.
Underfill adhesives are widely used to improve the overall thermal shock resistance, mechanical and electrical reliability of the assembly [2]. After a chip is mounted on a circuit board, the space between the chip and the PCB is filled with an underfill adhesive resin. Once the adhesive is cured, the stress is uniformly distributed throughout the bondline rather than just at the contact point between the solder ball and the chip. This enhances the overall thermal and mechanical shock properties and thus the reliability of the assembly [3, 4]. The adhesive formulation is typically a low-viscosity liquid (<1000cPs), which penetrates the gap between the chip and the PCB by capillary action when dispensed. The circuit board can also be slightly heated to about 50°C to lower the viscosity further and to accelerate the capillary fill process. In a high-throughput assembly, the dispensing and the fill process is completed in less than a minute, which requires low viscosity in the underfill resin. The adhesive resin is typically cured in the temperature range 120–130°C in less than 10 minutes. A picture of two CSPs with cured underfill adhesive is shown in Figure 2 (adhesive is shown in black color indicated with arrows).
Figure 2.
Two CSPs with cured underfill adhesive filling the gap between the chip and the PCB (indicated with arrows).
The underfill adhesives are typically thermosetting resin compositions that form cross-linked networks when cured. With conventional thermoset adhesives, it is difficult to remove the chip without damaging the subassembly in the event of a failure of a semiconductor chip on the circuit board. Several approaches have been published in the literature on reworkable, reversible, or degradable thermoset networks used in a number of adhesive applications [5, 6, 7, 8, 9]. Reworkability or removability is also a valuable attribute of adhesives used for electronic packaging applications including underfill adhesives [10]. It is highly desirable for an underfill adhesive to provide good electrical reliability, mechanical and thermal shock resistance, while allowing for the semiconductor chips to be easily separated in a defective assembly without causing damage to the circuit board. Several chemistry approaches have been explored to make the underfill adhesive reworkable at high temperature [11, 12]. For good reliability (mechanical and thermal shock resistance), high modulus and Tg are essential for underfill adhesives, which allow them to have lower CTE at the service temperature of the HHDDs or during T-cycle tests (hence reduced mismatch in CTEs between bonded substrates) since the CTE of thermosets increases rapidly above the Tg. However, with conventional thermosets, it is difficult to achieve both high reliability and reworkability without a built-in rework chemistry mechanism. The results shown below demonstrate that by using a carefully designed thermoset resin system, both high reliability and reworkability can be achieved concurrently.
The concept of highly reliable and reworkable thermoset underfill adhesive system is shown schematically in Figure 3. The adhesive needs to be a highly cross-linked system (high modulus and Tg) for high reliability at the service temperature of the HHDD or during thermal cycle (T-cycle) tests. The thermoset adhesive needs to undergo network break down at rework temperatures (typically around 220°C for this application) for easy removability of the faulty semiconductor chip. The design needs to be such that the rework temperature is sufficiently high so that there is no network breakdown occurring during adhesive curing or thermal cycling reliability tests. Typically for T-cycle tests, the bonded components are subjected to a temperature ramp from −40°C to 80°C (thermal shock) with a 15-minute hold time at each temperature. For good reliability performance, 1000 T-cycles without electrical failure are required, and hence, the need for a high-performance underfill adhesive system.
Figure 3.
Highly reliable and reworkable thermoset adhesive concept.
2.1 Resin system design
For the resin system design, the built-in reworkable or reversible chemistry was carefully chosen so that the cross-linked polymer network breakdown takes place at sufficiently higher temperature than that used for adhesive curing. The dicyclopentadiene backbone was chosen because of its high Tg and relatively high retro-Diels-Alder temperature. Based on previous literature report on activation energy required for retro reaction [13], we sought to incorporate electron withdrawing ester groups on dicyclopentadiene double bonds such that the network breakdown occurs at the correct temperature range for good reworkability. Thiele’s acid 1 (Figure 4) was chosen as starting material for the resin design. The synthesis of this acid was first reported more than a century ago [14], and design of thermally reversible polymeric system based on alkylation of potassium salt of this acid with dihalides has been reported [15]. A cyclic diester of this diacid, also known as “mendomer,” has been made and used in thermally remendable composite design [16]. The temperature used in this study for remendability was above 150°C, which is consistent with the expected retro reaction temperature of dicyclopentadiene-based systems. Prior to our discovery, the diglycidyl ester 2 and its use in reworkable underfill adhesive system have not been reported [11].
Figure 4.
Synthesis of diglycidyl ester 2 from diacid 1.
For the proof-of-concept study, the diacid 1 was made following a literature process by reaction of sodium cyclopentadienide (supplied as a THF solution, Boulder Scientific Company) with dry ice followed by acidification [17]. The diglycidyl diester 2 was synthesized by reacting the potassium salt of 1 with epibromohydrin in DMSO as solvent in 60–70% yield [11]. The literature process for making diacid 1 resulted in several isomers with the isomer represented by structure 1 being the major component. Since diacid 1 was a mixture of isomers, synthesis of 2 also resulted in a mixture of isomers. Both 1 and 2 were made in multi-kg scale in a production plant demonstrating industrial scalability. However, significant process change was necessary for the scale-up of 1 in the plant. The diacid 1 has been obtained as a single isomer before by a multistep process, but the synthesis used a different synthetic route [18]. During the process development for 1, a proprietary isomerization process was developed to convert mixture of isomers into mainly one isomer represented by the structure 1. This allowed for better characterization of the diacid by 1H NMR and other analytical techniques. The 1H NMR of 1 thus obtained (Figure 5) matches closely with that reported in the literature [18]. For better structural confirmation, the diglycidyl ester 2 was also made starting from single isomer of 1. While trace isomerization was observed during the synthesis, the 1H NMR of 2 matches well with the assigned structure and shows predominantly one isomer (Figure 5). During formulation and testing, no significant dependence of isomer ratio in 2 on reliability or reworkability performance was seen. This is likely because the reactive epoxy functionality is remote from the adduct forming double bond carbon centers (sterics) to be consequential for polymerization reactivity. Also, all of the reported main isomers of 1 have the two carboxylic acid groups on the norbornene and cyclopentene rings of the DCPD unit, and thus, they are not expected to significantly affect the network breakdown [19].
Figure 5.
1H NMR spectra of 1 (CDCl3 with two drops of DMSO d6) and 2 (CDCl3). 1H NMR spectra were run on Varian 300 MHz instrument.
Both 1 and 2 were analyzed for weight loss using thermogravimetric analysis (TGA). The onset of weight loss began around 160–170°C for both compounds consistent with the expected retro reaction temperature of the dicylopentadiene units. While the TGA weight loss progressed up to a temperature of about 250°C for both resins, it reached a plateau for 2 above this temperature (Figure 6A). It is likely that onset of homopolymerization of epoxy functionality in 2 at higher temperature stabilizes the weight loss while no such effect is possible for diacid 1.
Figure 6.
A) TGA thermograms for 1 & 2 (Tests performed using Discovery TGA 55 instrument at a ramp rate of 10°C per minute). B) DSC thermogram for neat resin 2 at a ramp rate of 10°C per minute (DSC was run using Q-100 DSC from TA Instruments).
To investigate if facile uncatalyzed homopolymerization of epoxy functionality in 2 takes place above 200°C, a differential scanning colorimetry (DSC) thermogram was run using neat resin without any added epoxy hardeners at a scan rate of 10°C/minute. The DSC thermogram shows an exotherm with a peak position at 272.86°C (Figure 6B). The magnitude of the exotherm indeed confirms facile homopolymerization of the epoxy group taking place above 200°C without the need for added hardeners. The onset of exotherm is around the same temperature as the weight loss was seen stabilizing for resin 2 in the TGA (Figure 6A).
2.2 Development of underfill formulations
Epoxy resin 2 was formulated in several epoxy-only and epoxy-acrylic underfill formulations (Table 1). Typical epoxy-only underfill formulations contain bisphenol-A epoxy resin and bisphenol-F epoxy resins, which contribute to high Tg and modulus of the cured adhesives. The use of reactive diluents such as 4-tert-butylphenyl glycidyl ether is essential to lower viscosity to below 1000cPs for better capillary flow. The epoxy formulations also contain a blend of epoxy-imidazole adduct hardeners to balance cure rate and work life. Several epoxy-only formulations were developed (F1, F2, F3, and F4) by using varying levels of the key epoxy resin 2. An epoxy-acrylic formulation F5 was also developed consisting of high Tg acrylic cross-linker such as tricyclodecane dimethanol diacrylate (saturated DCPD backbone), a high Tg acrylic diluent such as isobornyl methacrylate, and hybrid resins such as glycidyl methacrylate, which presumably links the epoxy and acrylic networks to form interpenetrating type networks (IPN). A radical initiator along with a combination of epoxy hardeners was also used for the curing of epoxy-acrylic hybrid formulation F5. Table 1 shows several epoxy-only and epoxy-acrylic hybrid formulations (F1–F5) where the amount of the key reworkable resin 2 was varied to study its impact on properties such as Tg, modulus, adhesion, and reworkability. In formula F5, 5% of 2 was used for direct comparison with the corresponding epoxy-only formula F3. Formula F5 also contained 5% of commercially available cyanate ester bisphenol E cyanate ester. Cyanate esters have been known to co-cure well with epoxy resins, and they help lower the viscosity and improve the Tg and modulus of cured networks [20]. The formulations also contain additives such as carbon black color, dispersants, inhibitors, and silane adhesion promoters.
Materials
Formula function
Epoxy-only formulas
Epoxy-acrylic formulas
F1
F2
F3
F4
F5
Bisphenol A epoxy
Resin
41.4
42.3
43.9
45.4
Bisphenol F epoxy
10.6
11.7
13.1
14.6
10.0
2
10.0
8.0
5.0
2.0
5.0
Naphthyl 1,6-diglycidyl ether
20.0
Trifunctional epoxy
5.0
Tricyclodecane dimethanol diacrylate
9.9
Isobornyl methacrylate
5.0
Bisphenol E cyanate ester
5.0
p-tert-Butylphenyl glycidyl ether
Diluent
15.3
15.3
15.3
15.3
6.0
Epoxy-imidazole adduct
Hardener
16.7
16.7
16.7
16.7
20
Cationic polymer
Dispersant
0.7
0.7
0.7
0.7
Ethyl/ethyl-hexyl acrylate copolymer
0.4
Epoxy silane
Adhesion Promoter
0.7
0.7
0.7
0.7
Carbon black in epoxy
Color
0.9
0.9
0.9
0.9
5
Dicyandiamide
Hardener
3.6
3.6
3.6
3.6
Barbituric acid
Inhibitor
0.1
0.1
0.1
0.1
Butylated hydroxytoluene (BHT)
0.1
Glycidyl methacrylate
Crosslinker
8
tert-Butyl peroxy-2-ethylhexanoate
Radical initiator
0.6
Total
100
100
100
100
100
Table 1.
Epoxy-only and epoxy-acrylic underfill formulations containing key reworkable resin 2.
For good mechanical and thermal shock resistance, the underfill adhesive needs to exhibit a stable storage modulus in the service temperature and T-cycle test temperature range (−40 to 85°C). Since underfill adhesives distribute stress along the bondline rather than at the point of contact between the die and solder ball, the mechanical reliability improvement by their use has been well established (1, 2). Higher Tg and modulus in the adhesive further improve the mechanical reliability due to lower CTE mismatch below the Tg of the cured adhesive. At rework temperatures (typically > solder melting temperature), a low modulus adhesive would facilitate easy removal of defective chips. The storage modulus vs temperature plots for cured samples of formulations F-1 to F-4 were run using dynamic mechanical analysis (DMA) and compared with controls (Figure 7). During a typical application process with low volume applied, the adhesives were cured at 130°C for 10 minutes. For material property testing, the same temperature conditions were used but with longer cure times to ensure full cure. Two control formulations, Loctite Eccobond UF 3800 (epoxy-acrylate hybrid underfill) and Loctite Eccobond E-1216M (epoxy-only underfill), were used for the initial screening work. UF 3800 has inferior reliability because of its lower Tg, and it shows rapid decrease in storage modulus beginning around 70°C (Figure 7), which is well within the temperature range used for T-cycle reliability tests (−40°C to 85°C). This formula however has very good reworkability as it displays low modulus at the rework temperature (220°C). In contrast, formula E1216M exhibits stable modulus in the temperature range −55 to 120°C (good reliability), but its high temperature modulus is higher than UF3800, which is consistent with its poorer reworkability. It is important to note that the good reworkability seen with UF3800 formula (discussed later in the chapter) is a manifestation of relatively low modulus at higher temperatures. This formula does not contain any resin with built-in chemistry that causes network breakdown at higher temperatures. All the new epoxy-only formulations (F1, F2, F3, and F4) show a storage modulus profile similar to the highly reliable E1216M in the temperature range −55°C to 85°C (Figure 7). As expected at high temperatures, these formulas show storage moduli much lower than the best reworkable control UF3800, which indicates superior reworkability. The drop in storage moduli at higher temperatures corresponds well with the relative level of epoxy resin 2 present in these formulas. F-1 with the highest level (10%) shows the lowest high temperature modulus while F-4 with lowest level (2%) shows the highest modulus. The magnitude of modulus decreases seen at higher temperature with increased loading of 2 is consistent with expected higher network breakdown caused by the retro Diels-Alder reaction.
Figure 7.
Storage modulus vs temperature plots for formulas F1, F2, F3, F4, and controls UF3800 and E1216M. ASTM D5023 method was used for DMA analysis.
The DMA storage modulus plot for formula F-4 with the lowest level of resin 2 (2%) was also compared with three benchmark underfill formulas, UF3810, UF3808, and UF3800. None of these benchmarks have a built-in chemistry feature, which causes network breakdown at higher temperatures. As discussed previously, UF 3800 shows good reworkability at high temperature and has inferior reliability among the controls. In contrast, UF3808 has highest reliability and lower reworkability. UF3810 has an intermediate balance of both reliability and reworkability. The reliability and reworkability performance of these benchmark formulas have been well established by their use in commercial products. The mechanical reliability (shock, vibration, and drop resistance) performance of UF3808 and UF3810 correlates well with their higher Tg and modulus profiles shown in Figure 8. The T-cycle reliability performance of these three formulas also correlates well with their storage modulus profiles and is discussed separately later in this chapter. As compared with the benchmark formulas, F4 shows stable storage modulus over a wider temperature range (Figure 8B), which suggests superior mechanical and T-cycle reliability. A closer examination revealed that in the temperature range 130–180°C, F4 exhibited sharper modulus decrease than benchmarks. This phenomenon cannot be explained entirely by expected modulus drop typically seen above the Tg of the cured networks and is strongly indicative of network breakdown occurring in this temperature range. A similar modulus trend was also seen with epoxy-acrylic hybrid formula F5 (Figure 8A). It appears that the impact of resin 2 loading in F5 (5%) on storage modulus at higher temperature is lower as compared with the corresponding epoxy-only formula F3, which contains the same amount of 2 (Figure 7). The likely reason for this is discussed separately in the following section.
Figure 8.
Comparison of storage modulus of formulas F4 and F5 with control formulas. ASTM D5023 method was used for the DMA analysis.
Table 2 shows comparison of Tg and storage modulus numbers (MPa) at different temperatures (−75°C, 25°C, 85°C, 125°C, and 220°C) for formulas F1–F5 along with those for controls. The relative level of epoxy resin 2 in formulas F1–F5 did not appear to affect the Tg of cured networks significantly as measured by DMA. Formulas F2, F3, F4, and F5 show similar Tg as the highly reliable UF3808 control. In addition, they also exhibit relatively stable modulus numbers in the temperature range −75°C to 125°C similar to or better than UF3808 indicating good reliability performance. The storage modulus numbers for F1–F5 formulas at 220°C correlate well with the relative loading of resin 2, as discussed previously. It is interesting to note that UF3800, which has proven reworkability in commercial HHDD products, has a higher modulus number at 220°C than formulas F1–F5. These results strongly correlate with superior reworkability for F1–F5 formulas arising from adhesive network breakdown caused by the retro Diels-Alder reaction. The reworkability tests performed on these formulas (discussed later) confirm this further.
Formula
E′(MPa) (−75°C)
E′(MPa) (25°C)
E′(MPa) (85°C)
E′(MPa) (125°C)
E′(MPa) (220°C)
Tg (tanδ)
F1
2757
2159
1871
662
11.6
140
F2
3069
2322
2024
994
16.5
143
F3
2713
2092
1869
1188
22.4
145
F4
2627
1986
1807
1326
33.7
148
F5
3241
2507
2005
809
40.6
147
UF3800
2459
1860
1110
185
57
124
UF3808
3320
2502
2059
945
106
145
UF3810
4160
3031
1369
451
72
145
E1216M
3750
2858
2653
1452
88
144
Table 2.
Storage modulus numbers at different temperatures and Tgs for formulas F1–F5 and those for controls. ASTM D5023 method was used for the DMA analysis.
2.3 Die shear adhesion properties
An important property of underfill adhesives is good die shear adhesion at room temperature (25°C) and the peak temperature experienced during an assembly reflow process (260°C). In contrast, reworkable underfill adhesives need to exhibit high die shear adhesion at room temperature for high reliability performance while that measured at higher temperature needs to be low enough for easy removal of faulty chip at the rework temperature (220°C), yet high enough to prevent any solder extrusion during a reflow step (< 1–2 min @ 260°C). The die shear tests were performed on Dage 4000 instrument from Dage Precision Industries following MIL-STD-883 2019.9 die shear method. The 25°C die shear tests were completed using 3 mm2 size SiN dies on bismaleimide-triazine (BT) substrate. Adhesion at 260°C using the same 3 mm2 size dies was too low among all underfills for direct comparison. Therefore, 260°C die shear tests were performed using 7.6 mm2 size dies for better response and comparison of underfill adhesion. The die shear adhesion of reworkable formulas F1–F5 was measured and compared with controls.
Figure 9 shows the comparison of die shear strengths at room temperature. Formulas F1–F5 show similar or superior die shear strength than the highly reliable underfill benchmark UF3808 while another control formula UF3800 shows the lowest strength among formulas tested. There appears to be no discernible trend between loading level of resin 2 and die shear adhesion at room temperature. However, results from the die shear tests performed at 260°C discussed next show a clear trend, which further confirms network breakdown happening at higher temperature in formulas F1–F5.
Figure 9.
Die shear strength comparison of F1–F5 with benchmark formulas (UF3800, UF3808, and UF3810) at 25°C. MIL-STD-883 2019.9 die shear standard was used for the die shear tests using 3 mm2 size dies on BT substrate. The tests were performed at a load of 0.5 kg and speed of 13.77 mil/sec at a height of 11 mil (settings on Dage 4000).
The die shear tests performed at 260°C further corroborate the storage modulus results shown in Figures 7 and 8. The 260°C die shear strengths for formulas F1–F5 and their comparison with benchmarks are shown in Figure 10. The shear strengths correlate well with the amount of reworkable resin 2 used in the experimental formulations. F1, which uses the highest (10%), shows the lowest while F4, which contains the lowest amount (2%), shows highest die shear values. These results further confirm that the network breakdown arising from the retro Diels-Alder reaction of the DCPD moiety takes place at high temperature resulting in lower bond strengths in the adhesive. The results are also consistent with the storage modulus results discussed previously for these formulas. The only slight deviation appears to be the hybrid epoxy-acrylic formula F5, which shows higher 260°C die shear adhesion as compared with the corresponding epoxy-only formula F3 both of which have the same amount (5%) of epoxy resin 2. While the formulation components are different, the higher 260°C die shear value seen with formula F5 is likely resulting from relatively lower network breakdown occurring at this temperature. Presumably, the norbornene double bond in 2 (more strained as compared with cyclopentene) undergoes some radical copolymerization with acrylic components in F5 effectively lowering the level of thermally reversible adducts in the cured network. To validate this, a DSC thermogram of neat resin 2 was run in the presence of 10% tert-butyl peroxy-2-ethylhexanoate radical initiator. The resulting thermogram (Figure 11) showed an exotherm with a peak at 132.96°C that can be ascribed to the radical homopolymerization of the norbornene double bond while no such peak was seen in Figure 6B discussed previously. While the magnitude of this exotherm is relatively small (116 J/g), it is likely that the norbornene double bond possessing an electron-withdrawing ester group (as present in structure 2) would exhibit significantly higher reactivity in radical copolymerization with other acrylic components. Radical copolymerization reactivity of norbornene double bond connected to electron withdrawing ester group has been reported before [21].
Figure 10.
Die shear strength comparison at 260°C for formulas F1–F5 and benchmark underfill formulas following MIL-STD-883 2019.9 die shear standard. The tests were performed using 7.6 mm2 size dies on BT substrate at a load level of 0.5 kg and a head speed of 13.77 mil/s at a height of 11 mil.
Figure 11.
DSC thermogram for neat resin 2 in the presence of 10 wt% of tert-butyl peroxy-2-ethylhexanoate. DSC thermogram was run using Q-100 DSC from TA Instruments at a scan rate of 10°C/minute.
2.4 Reworkability study
Select experimental underfill formulas F1 (contains 10% of 2), F4 (2% of 2), and F5 (5% of 2, epoxy-acrylic) were tested for reworkability and compared with controls. An underfill board array containing 0.4 mm CSP (chip scale package or simply called chip, die or semiconductor or component) test board (ID# ACEM 94V0 1612, nomenclature 1502009) was used for the study. Sample pictures of the (A) unpopulated test board, (B) populated and underfilled board, and (C) zoom-in of a single underfilled die assembly unit are shown in Figure 12. When a faulty chip is identified after assembly in such a board with a multitude of semiconductors, the chip needs to be removed, residues cleaned up, and the chip assembly process is repeated so that board can be reused again. Figure 13 illustrates the whole rework process. (A) A typical rework station used for this study, (B) chip removal tool with a focused heating element, (C) the scavenger nozzle for removing glue and solder residues after chip removal, (D) diagram of the scavenger nozzle illustrating the heating and vacuum suction to remove residues. A typical rework process consists of localized heating of the circuit board on a hot plate (heated to around 180–200°C), and the faulty die on top is heated using a hot air nozzle (typically around 220°C, picture B, Figure 13) for a few seconds. The die is then removed using suction tool optionally with additional force using a metal tool. The adhesive and solder residues are then cleaned using suction nozzle (picture C, Figure 13) to make the board reusable again. When a good reworkable adhesive is used, the whole rework process is typically completed in under 2 minutes.
Figure 12.
Photographs of (A) unpopulated 0.4 mm BGA test board used for reworkability and reliability study; (B) test board array populated with 6 mm2 WLCSP (wafer level chip scale package) and underfilled; (C) zoom-in of a single, underfilled WLCSP on test board.
Figure 13.
Photographs of (A) a typical rework station; (B) chip removal tool with focused heating element; (C) the scavenger nozzle for removing glue and residues after component removal; (D) diagram of the scavenger nozzle illustrating the heating and vacuum suction.
Table 3 shows qualitative evaluation guideline and score card for the board rework process. The rating of the rework process considers a multitude of factors such as ease of die removal, amount of underfill left on the board after cleaning, number of pads and traces damaged on the board, total cleaning time, and solder mask damage, each with its own weightage (total adds up to 1). The ease of rework is rated on a scale of 1–10 where a rating of 1 indicates poor reworkability and 10 best reworkability.
Weight
Score
1
3
5
7
10
0.3
Ease of removal of die
Suction+high force, cannot be removed
Suction+high force to turn & remove
Suction+medium force to turn & remove
Suction+low force to turn & remove
Removes on suction
0.2
Amount of UF left after cleaning
>75%
50–75%
25–50%
0–25%
0%
0.1
# of pads damaged
>9
6–9
3–6
1–2
0
0.1
# of traces damaged
>9
6–9
3–6
1–2
0
0.2
Time to clean
>150s
120–150s
90–120s
45–90s
<45s
0.1
Solder mask damage
Significant damage (>15% area)
Damage 10–15% area
Damage 5–10% area
Damage to <5% area
No damage
Table 3.
Qualitative evaluation guideline and score card for board rework. Higher score indicates easier rework process.
Benchmark underfill formulas UF3800, UF3808, and E1216M were compared with F1, F2, and F5 formulas for reworkability performance. A test board similar to that shown in Figure 12 was used to bond the chips using benchmarks and the experimental formulas. Table 4 shows total reworkability score for these formulas involving a multitude of factors discussed before. As expected, UF 3800 showed a high rating of 8 while the other two benchmark formulas UF3808 and E1216 formulas fared poorer for reworkability. The reworkability score for formulas F1, F4, and F5 was similar to or better than UF3800. The hybrid epoxy-acrylic formula F5 showed slightly inferior reworkability score than F1 and F2. The likely reason for relatively lower rework score for F5 as compared with F1 and F4 was discussed in the previous section and is suspected to be from the norbornene double bond copolymerization, which would result in partially (depending on extent of copolymerization) non-cleavable networks.
Rework test scores for control formulas and F1, F4, and F5 formulas (higher total score indicates superior reworkability).
Highest score is 2. Higher number indicates lower amount of underfill left after die removal.
Highest score is 3, higher number indicates easier removal of die.
Highest score is 1, higher number indicates less damage to the pads on the board.
Highest score is 1, higher number indicates less damage to the traces on the board.
Highest score is 2, higher number indicates less time taken for cleaning.
Highest score is 1, higher number indicates less solder mask damage on the board.
Figure 14 shows pictures of the test board after removal of the bonded die for the control formulas UF3800 and UF3808. The images on the left side show the substrate after removal of the die before rework and cleaning while those on the right show substrate board after cleaning. As expected, UF 3800 results in a clean board after die removal, rework, and cleaning. In contrast, UF 3808 causes damage to the board during die removal and leaves lots of adhesive residue even after cleaning. Figure 15 shows images of the substrate board after die removal (on the left side) and after the rework process and cleaning (on the right) for formulas F1, F4, and F5. Consistent with the storage modulus profile and high-temperature die shear results discussed previously, all of these formulas enable easy rework process that is similar to or slightly better than reworkable UF3800 benchmark as evidenced by the clean substrate board obtained after the rework process. This result further demonstrates that resin 2 with built-in thermally reversible DCPD moiety enables good reworkability of underfill formulations.
Figure 14.
Picture of substrates after removal of semiconductor before rework and cleaning for controls UF 3800 (top left) and UF3808 (bottom left). Picture of substrate board after underfill and solder residue removal for UF 3800 (top right) and UF3808 (bottom right). The marks indicated by the arrow are caused by removal of the adhesive that is not reworkable (UF3808).
Figure 15.
Picture of substrates after removal of die before rework and cleaning for F1, F4, and F5 (on the left side). Picture of substrate board after cleaning (underfill and solder residue removal) for the same formulas (on the right side).
2.5 T-cycle reliability tests
Since the HHDDs with bonded semiconductor components can be subjected to thermal shocks as described previously, the bonded components were tested for T-cycle reliability using an underfill board array containing 15 bonded semiconductor components similar to that shown in Figure 12. The reliability test was performed by using a 30-minute temperature ramp from −40°C to 85°C per cycle with a 15 min hold time at both temperatures (−40°C and 85°C). Air-to-air thermal cycling where the bonded parts cycle between an oven chamber and a freezer chamber in air was used for the study. The cycle is repeated several times, and the semiconductors on the board were tested for electrical failure. For good T-cycle reliability performance of HHDDs, no chip failure up to 1000 cycles is required. Table 5 shows comparison of reliability performance of three benchmark formulas and the experimental underfill formulas F1, F4, and F5. The highly reworkable UF3800 formula started showing chip failures after 600 T-cycles (three failed chips) and complete failure of all the bonded chips after 2000 T-cycles (Table 5). In contrast, the highly reliable UF3808 and E1216 controls showed no failures up to 2000 T-cycles. As expected, none of the experimental formulas F1, F4, and F5 showed failures up to 2000 cycles indicating that the T-cycle reliability is similar to the highly reliable UF3808 and E1216M. The loading level of thermally reversible resin 2 did not appear to impact the reliability performance (compare F1 vs F4, Table 5). This result is consistent with reliability model established based on the storage modulus results discussed previously (Figure 7). As noted previously, the cured underfill adhesives also improve the mechanical reliability performance (shock, vibration, and drop) by distributing stress uniformly through the bondline. Higher Tg and modulus adhesives improve the reliability further similar to the reliability model shown before for T-cycle. Thus, the results discussed in this chapter demonstrate that using resin 2 with built-in reworkability feature, both high reliability and reworkability can be achieved concurrently with underfill formulations.
Fifteen bonded semiconductors were tested together for each formulation and the number of failures is indicated on the left side out of the total 15 components tested.
Synthesis of a new diglycidyl ester epoxy resin possessing dicyclopentadiene backbone was described. The resin was formulated in underfill adhesive formulations to provide cured adhesives with high Tg and modulus. Clear dependence of loading level of thermally reversible resin 2 on high-temperature die shear strength was seen. The formulated adhesives provided high mechanical and T-cycle reliability performance to the bonded semiconductor components consistent with the reliability model established based on storage modulus results. At high temperatures, the retro Diels-Alder reaction of the DCPD unit caused network breakdown enabling easy removability/reworkability of the bonded components. The network breakdown occurred at sufficiently above the cure temperature of the adhesive to not interfere during the cure of the adhesive. Thus, using this novel adhesive system, both high reliability and reworkability can be achieved simultaneously, which is generally not possible with other thermoset adhesive system. The key epoxy resin was scaled up to several kg’s demonstrating industrial applicability of the high-performance reworkable adhesive technology.
1.Chen A, Lo RHY. Semiconductor Packaging: Materials Interaction & Reliability. CRC Press; 2012. DOI: 10.1201/b11260
2.Tong QK. Underfill adhesive materials for flip chip applications. In: Advanced Adhesives in Electronics. 2011. pp. 137-154. DOI: 10.1533/9780857092892.1.137
3.Fan L, Moon KS, Wong CP. Adhesion of underfill and components in flip chip encapsulation. Journal of Adhesion Science and Technology. 2002;16:213-223. DOI: 10.1163/156856102317293713
4.Zhang Z, Wong CP. Recent advances in flip-chip underfill: Materials, process and reliability. IEEE Transactions on Advanced Packaging. 2004;27:515-524. DOI: 10.1109/TADVP.2004.831870
5.Yang S, Chen JS, Korner H, Breiner T, Ober CK, Poliks MD. Reworkable epoxies: Thermosets with thermally cleavable groups for controlled network breakdown. Chemistry of Materials. 1998;10:1475-1482. DOI: 10.1021/cm970667t
6.Sridhar LM, Oster MO, Herr DE, Gregg JBD, Wilson JA, Slark AT. Re-Usable thermally reversible crosslinked adhesives from robust polyester and poly(ester urethane) Diels-Alder networks. Green Chemistry. 2020;22:8669-8679. DOI: 10.1039/D0GC02938F
7.Kloxin CJ, Scott TF, Adzima BJ, Bowman CN. Covalent adaptable netwoks (CANs): A unique paradigm in crosslinked polymers. Macromolecules. 2010;43:2643-2653. DOI: 10.1021/ma902596s
8.Sridhar LM, Slark AT, Wilson JA. Furan functionalized polyesters and polyurethanes for thermally reversible reactive hotmelt adhesives. In: Khan A et al., editors. Furan Derivatives-recent Advances and Applications. London: IntechOpen; 2021
9.Das S, Samitsu S, Nakamura Y, Yamauchi Y, Payra D, Kato K, et al. Thermo-resettable cross-linked polymers for reusable/removable adhesives. Polymer Chemistry. 2018;9:5559-5565. DOI: 10.1039/C8PY01495G
10.Malik J, Clarson SJ. The chemistry and technology of reworkable polymeric materials for electronic applications. Surface Coatings International Part B: Coatings Transactions. 2003;86:9-20. DOI: 10.1007/BF02699588
11.Henkel Corporation. US Patent 9,938,437. 2018
12.Wang L, Wong CP. Novel thermally reworkable underfill encapsulants for flip-chip applications. IEEE Transactions on Advanced Packaging. 1999;22:46-53. DOI: 10.1109/6040.746542
13.Franklin WE. Kinetics of the reverse Diels-Alder dissociation of substituted dicyclopentadienes. The Journal of Organic Chemistry. 1970;35:1794-1796. DOI: 10.1021/jo00831a016
14.Thiele J. Ueber Abkommlinge des cyclopentadiens. Berichte der Deutschen Chemischen Gesellschaft. 1901;34:68-71. DOI: 10.1002/cber.19010340114
15.Chen X, Ruckenstein E. Thermally reversible linking of halide-containing polymers by potassium dicyclopentadienedicarboxylate. Journal of Polymer Science: Part A: Polymer Chemistry. 1999;37:4390-4401
16.Park JS, Takahashi K, Guo Z, Wang Y, Bolanos E, Hamann-Schaffner C, et al. Towards development of a self-healing composite using a mendable polymer and resistive heating. Journal of Composite Materials. 2008;42:2869-2881. DOI: 10.1177/0021998308097280
17.Marchand AP, Namboothiri NN, Lewis SB, Watson WH, Krawiec M. Thiele’s acid revisited: Isolation and characterization of two minor products formed by carbonation of cyclopentadienide anion. Tetrahedron. 1998;54:12691-12698. DOI: 10.1016/S0040-4020(98)00770-4
18.Chen J, Kilpatrick B, Oliver AG, Wulff JE. Expansion of Thiele’s acid chemistry in pursuit of a suite of conformationally constrained scaffolds. The Journal of Organic Chemistry. 2015;80:8979-8989. DOI: 10.1021/acs.joc.5b01332
19.Dunn GL, Donohue JK. The Structure of Thiele’s ester, a dimethyl dicyclopentadienedicarboxylate. Tetrahedron Letters. 1968;9:3485-3487. DOI: 10.1016/S0040-4039(01)99089-6
20.Ueyama J, Ogawa R, Ota K, Mori Y, Tsuge A, Endo T. Rapid curing system of a cyanate ester resin/epoxy resin with a thermal latent polymeric hardener based on a phenol-amine salt. ACS Applied Polymer Materials. 2022;4:84-90
21.Ihara E, Honjyo S, Kobayashi K, Ishii S, Itoh T, Inoue K, et al. Radical copolymerization of methyl 2-norbornene-2-carboxylate and 2-phenyl-2-norbornene with styrene, alkyl acrylate and methyl methacrylate: Facile incorporation of norbornane framework into polymer main chain and its effect on glass transition temperature. Polymer. 2010;51:397-402. DOI: 10.1016/j.polymer.2009.12.008
Written By
Laxmisha M. Sridhar and Timothy M. Champagne
Submitted: 09 August 2022Reviewed: 23 August 2022Published: 24 October 2022
Open access peer-reviewed chapter
