Advances in Temporary Bonding and Debonding Technologies for use with Wafer-Level System-in-Package (WLSiP) and Fan-Out Wafer-Level Packaging (FOWLP) Processes

Drivers for 3D packaging solutions are many and diverse with each requirement calling for different answers using various technologies. The primary goals are miniaturization with increased component density, improved performance, simplification of design and assembly, flexibility, and functionality. Lower cost and faster time-to-market are also core drivers. For die and package stacking including folded packages, embedding dies is a key technology for heterogeneous system integration [3].

There are two primary approaches for embedded die technologies: Fan-out Wafer-level Packaging (FOWLP) and Printed Circuit Board (PCB) embedding. The FOWLP approach uses polymer encapsulants that are embedded with known good die (KGD). The PCB approach uses printed circuit boards and embeds the boards with the die.

A lot of activities are taking place worldwide to best facilitate and optimize Fan-out Wafer-level integration. The main approaches are Embedded Wafer-level Ball Grid Array (eWLB) by Infineon [4], the InFO package by TSMC [5], and the Redistributed Chip Package (RCP) by Freescale [6].

For Fan-out Wafer-level Packaging, two basic process flows are encountered; the “Mold-first” or the “RDL-first” approach. For the “Mold-first” process flow, two options exist: a face-down (e.g. eWLB) and a face-up (e.g. InFO). Both variants are already in mass production for different applications. “Mold-first”, face-down, starts with die assembly on an intermediate carrier followed by over-molding and debonding of the molded wafer/panel from the carrier. The redistribution layer is typically based on thin film technology and applied on the reconfigured molded wafer/panel. The face-up approach also starts with die assembly on a carrier with a temporary adhesive layer. However, in this process dies have a Cu bump and are placed face-up on the carrier. After over-molding, grinding of the EMC allows access to the now exposed Cu bumps on the dies. The redistribution layer is applied and finally the wafer is released from the carrier and diced for package singulation. For the “RDL-first” approach, the redistribution layer is applied first on an intermediate carrier and the bumped dies are assembled by chip to wafer bonding on the RDL. The assembly is underfilled and over-molded and the molded wafer/panel including the RDL, has to be released from the carrier.

This investigation will focus on the “Mold-first”, face-down process flow. Typically, a dedicated thermo-release tape is used to fix the dies onto a carrier during molding. Release of the molded, reconfigured wafer is done using a temperature step that causes one side of the release tape to lose adhesion. As the release temperature is typically around 20°C above mold and post-mold temperature, the release step significantly influences the overall stress, warpage and die shift behavior. Therefore, a low temperature or even room temperature debond process is of interest. In this study different release and temporary bonding materials suitable for room temperature mechanical and laser debonding processes have been investigated for their suitability for reconfigured wafer manufacturing [7].

The following sections describe in detail the experiments conducted for this work.

This work is focused on a “Mold-first” face-down process. Hence, reconfigured wafer preparation followed this manufacturing sequence. BrewerBOND^® bonding and release materials are spin coated onto 200-mm glass wafers. Chips are then placed on the bonding material. Each wafer is over-molded with a standard epoxy mold compound (EMC) and thermally cured. Molded wafers are then debonded either mechanically or by laser and cleaned to remove residual adhesive material. For evaluation, wafers are finally characterized for die shift, warpage and die stand-off measurement (Figure 1).

Material evaluations were made using 200-mm glass wafers. Three dies were placed per package: one large die 9×9×0.2 mm³ and two smaller die 3×2×0.2 mm³. Partial assembly was done on the wafers guaranteeing enough information for process and material evaluation. Package and wafer assembly are depicted in Figure 2. Mold thickness was set to 300 μm resulting in a 100-μm over-mold layer on the chip backside.

An initial screening design of experiment (DoE) was conducted to find the best die attach parameters using the lowest assembly head temperature and assembly force which gave the highest adhesion and lowest die shift (Table 1). Evaluations were conducted on the:

• Assembly on three wafers with different bonding material film thicknesses (2 μm, 5 μm, 10 μm)

• 5×5 matrix with assembly parameters (Temperature, Force) on each wafer

• Substrate was continuously heated at 75 °C

In Figure 3 the assembly force and temperature variation on wafer is shown and parameters summarized.

In a second step, optimized assembly parameters were used to evaluate different release and bonding parameters. Reconfigured wafers manufactured using the different materials were analyzed for die shift, warpage and die stand-off. In addition to the bond and release materials a standard thermal release tape was used with the same assembly as a reference. Material combinations used are summarized in Table 2.

In the following sections materials used for the investigations are described in detail.

Glass carrier wafers were obtained from Corning Glass, Eagle XG with a CTE of 3.17 ppm/K as 200-mm notched rounds with a substrate thickness of 0.7 mm. Carriers were used as received from Corning without cleaning or a dehydration bake.

Spin coat glass carrier wafers using a manual, static dispense with release material as follows:

1. BrewerBOND^® 510 release material was spin applied at 1250 rpm using a 3000 rpm/sec acceleration for a time of 35 sec. Hotplate baked using a 0.2 mm proximity bake at 220 °C for 1 min. Target film thickness after cure is 50 Å.

2. BrewerBOND^® 701 release material was spin applied at 2500 rpm using a 5000 rpm/s acceleration for a time of 60 sec. Using a bake plate at 300 °C, contact bake for 5 min. Target film thickness after cure is 150 nm.

Each custom dilution of BrewerBOND^® 305 bonding material is spin cast onto glass carriers that have been previously processed with one of the above release materials. Various spin speeds and dilutions were used to achieve the desired film thickness. Each carrier was then baked using bake plates with proximity pins at 0.2 mm. Bake at a temperature of 160 °C for 3 min followed by a bake at 220 °C for 3 min. As a reference, a standard release tape was used with a release temperature of 170 °C.

For the first evaluations a granular epoxy molding compound (EMC) was used. Processing followed product datasheet recommendations. A summary of properties is shown in Table 3.

The following sections describe the various equipment used for this work.

Glass wafers were manually coated with both release and bonding materials using a Cee^® X-Pro 2 work station equipped with a coat bowl and bake plates with programmable lift pins.

Chips were assembled using a Datacon 2200evo on carriers prepared with the release and bonding materials and as reference with thermo-release tape. Compression molding was done with a 120 Ton press from TOWA followed by debonding after post-mold cure.

A mechanical or laser release debond process was used to remove the glass carrier from the EMC. An EVG^® mechanical debonder and an EVG^® Solid State Laser system were used to debond the carriers.

Die positions were measured with a Mahr OMS 600 optical measuring system after carrier release. Warpage measurements were done using a cyberSCAN VANTAGE, a laser based, non-contact inspection system and die stand-off was measured using a HOMMEL Tester T8000.

Using the process flow, test wafers are manufactured and analyzed for evaluation.

The goal of the assembly parameter evaluation was to find the optimum assembly parameters which gave the highest die adhesion on the bond layer. Test parameters were varied as described previousy in Table 1. Photos of wafers after mechanical debonding is shown in Figure 4.

Best die attach result could be achieved with 100 °C assembly head temperature and 3.5 kg assembly force for all bonding material film thicknesses. With these parameters, low die shift without flying dies are visible. Based upon this data, these parameters were used for the assemblies in the main DoE.

3D wafer warpage was measured at room temperature after the debonding process (Table 4). For all material combinations, a patelliform wafer shape was obtained with a maximum bow between 700 and 800 μm. Only the laser debonded wafer showed slightly lower warpage. For this work, the CTE mismatch between the glass carrier and the EMC is a significant factor and demonstrated the highest influence on warpage. Temporary bonding material film thickness and the debonding temperature were less significant. These observations will need further investigation using higher CTE glass carriers and different EMC.

Die shift is one of the major challenges in FOWLP and PLP and describes the effect of chip movement after placement during compression molding, debonding and cooling of the reconfigured wafer/panel. Dies are placed at a certain temperature onto the carrier. The carrier is then heated to mold temperature and the assembly expands with the CTE of the carrier material. During molding and curing of the epoxy molding compound (EMC) the dies are fixed in the EMC. During crosslinking, the EMC material shrinks, leading to a change in volume of the molded wafer.

After performing a post-mold cure or a high-temperature debonding process, the reconfigured wafer/panel is cooled and die shift is dominated by the CTE of the EMC. Due to the different CTEs of the materials involved and in combination with the temperature profile of the different process steps, dies will shift from their originally assembled position. In addition to the described effect, sliding of the dies during molding can occur depending upon the size and adhesion of the dies on the release tape and the flow behavior of the EMC and the related forces on the dies during compression molding.

Linear die shift can be compensated for with an assembly adaptation. Dies are assembled at the “wrong” position and shift to the “correct” position during molding. However, the general placement tolerance of the assembly equipment and a random shift during molding cannot be compensated for using this method. After molding, the final position of the die is now defined and impacts the RDL yield and how it will meet the die pads.

In Table 5, die shift behavior and factors of the material combinations evaluated are summarized. In general, the temporary bonding material demonstrates significantly lower die shift than the thermal release tape. In direct comparison, die shift for BrewerBOND^® 701 release material in combination with BrewerBOND^® 305 bonding material is not influenced by the bonding material film thickness. The combination of BrewerBOND^® 510 release material with the BrewerBOND^® 305 bonding material gave the lowest die shift when using a 10-μm bonding material film thickness. When using this combination of materials with a room temperature bonding process, the improved adhesion has a greater influence on die shift than the CTE mismatch between the glass carrier and the EMC.

It is not unusual for die to stick a couple of microns out of the mold compound after the debonding process. This effect, known as die stand-off, depends on the material combinations used and processing [8]. Die stand-off is a challenge for the RDL as the dielectric layer has to cover and level out the die edge. Therefore, a lower die stand-off is preferred. In Table 6 the resulted die stand-offs are summarized.

Die-stand-off using the BrewerBOND^® 510 release material and BrewerBOND® 305 bonding material combination results in a significantly lower stand-off than what was observed by the thermal release tape. A slight increase was observed with increasing bonding material film thickness. For carrier combinations using BrewerBOND^® 701 laser release material and BrewerBOND^® 305 bonding material, a smaller standoff resulted with only a slight bonding material film thickness dependency noted. Data for a 2-μm bonding material film thickness resulted in a negative standoff, where the die were slightly sunken into the EMC.

Within this study, two temporary bonding material combinations were evaluated as an alternative for the thermal release tape used in “Mold-first” face-down fan-out wafer- and panel-level packaging. Different material and process parameters were used to manufacture reconfigured wafers and analyze them in detail for warpage, die shift and die stand-off.

Promising results were achieved with low die shift and very low die stand-off with improved performance as compared to standard thermal release tape. Additionally, no influence from the bonding material or release material combination on wafer warpage was observed.

Further investigation will include additional bonding materials, high and low CTE glass carriers as well as liquid and granular epoxy molding compounds.

The authors would like to thank Damian Freimund for supporting this study in die assembly, Thanh Duy Nguyen for wafer molding and Marcus Voitel for data evaluation.