BCB-Based Dry Film low k Permanent Polymer with sub 4-μm Vias for Advanced WLP and FO-WLP Applications

Drivers for 3D packaging solutions are manifold and each requirement calls for different answers and technologies. The main goal is miniaturization, but component density and performance, simplification of design and assembly, flexibility, functionality and finally, cost and time-to-market have been found to be the core drivers for going 3D as well. Besides die and package stacking and folded packages, FO-WLP is a key technology for heterogeneous system integration. There are two main approaches for embedded die technologies: Fan-Out Wafer level integration, where dies are embedded into polymer encapsulants and 3D vertical integration, where dies are embedded into the substrate. Fan-out Wafer Level Packaging (FO-WLP) is one of the latest packaging trends in microelectronics. Mold embedding for this technology is currently done on wafer level up to 12”/300 mm diameter. A very important step in the process is the RDL (redistribution layer) on top of the FO-WLP.

There is a strong demand to increase the routing density of substrates, interposers and on-chip routing to match the requirements for future microelectronic systems which are mainly driven by miniaturization and performance. Photo-resists for structuring the metallization or acting as a mold for electroplating are common for very fine lines and spaces supported by the developments in the front-end processing. For example chemical amplified photo-resists are now moving in the back-end and wafer level packaging process. The results are mainly governed by the performance of the equipment i.e. the photo-tool. This is different for the permanent dielectric polymer material. Starting 25 years ago there was a strong move to develop photo-sensitive polymers due to the cost and performance of existing technology like dry-etching (for example RIE, reactive ion etching) [1], [2], [3]. The major difference between photo-resists and dielectric photo-polymer are the unequal functions of the material systems. Photo-resists are only temporary masks for subsequent process steps like etching and plating. This is different for the photo-polymers which are a permanent part of the future systems. Therefore it has a strong influence on the reliability but also on the performance of the system [4]. Properties like high thermal stability, excellent mechanical properties, low water up-take, low dielectric constant and low loss have to be achieved together by the chemical synthesis of the polymer and can therefore limit the optical resolution [3].

Depending on the application the importance of the polymer properties may be different Figure 1:

Highly important are also the following properties:

• Cu-Compatibility

• Low temperature cure

• High Breakdown V

• Low cost by alternative processes

Mainly, photo-polymers are used for electronic packaging and MEMS applications today. Mask aligners and steppers with UV light are used to pattern the coating.

The lithography tools have a much higher resolution but the photo-sensitivity of the thinfilm polymers are limited which is not compatible with the roadmap of 4 μm lines and space of the metallization in the near future. Vias must have similar sizes as the metal lines to achieve a dense routing.

Dow Electronic Materials has developed 14-P005 Dry Film Dielectric, based on the well-known dielectric material Divinylsiloxane-bis-benzocyclobutene (DVS-bis-BCB or BCB) [6]. The construction of the film can be seen in Figure 2. Patterning may be achieved through photolithography or laser ablation, however improved aspect ratios are provided by the laser ablation process.

BCB, traditionally utilized in spin-on or spraycoated dielectric products, has been reformulated as a laminate to enable application to substrates such as large area organic or glass panels. Hot roll or vacuum lamination may be used to apply the film.

Toughening components have been added to the BCB-based film to provide necessary flexibility and tackiness while preserving the excellent dielectric properties of BCB. Low D_k, D_f, and moisture sensitivity have been maintained while breakdown voltage remains high. Elongation and fracture toughness have also been improved vs. traditional BCB (figure 3).

Films may be coated over a wide range of thicknesses (4μm–50μm) to achieve a specific target based on packaging design.

Pulsed-laser ablation is a process through which material is removed by a short, high-intensity laser pulse. Details of this technology have been published [5]. In general this direct ablation using an ultraviolet excimer laser is a dry, one step technique that was developed for structuring thin-film polymers over 20 years ago [7].

The process for structuring photo-sensitive polymers consists of multiple steps. First the polymer (in the pre-cursor state) has to be coated on the substrate, typically using spin-coating technique for WLP applications. Emphasis has to be taken into account that these polymers are temperature-sensitive. The shelf-life of some of the materials is limited to one week at room temperature. Therefore a sophisticated material support has to be established from the polymer supplier to the pump of the spin-coater module at the wafer production site. After the coating most of the solvent has to be removed from the coated polymer layer by heating the wafer on hot-plates. This process has a high influence on the whole process and has to be optimized for each polymer type. If the temperature (or time) of the process is too high (or too long) the photo-sensitive components will be damaged which will heavily limit the resolution of the via. If the temperature (or time) is too low (or too short) the vias in the polymer will not form well. It may even lead to a dissolving effect in the later developing step. After the baking process the photo-polymers are exposed to UV light using either mask aligner or steppers. Some of the photo-polymers then need a post-exposure bake to improve the UV-induced chemical reaction in the polymer. Then the polymers have to be developed i.e. the polymer has be washed out of the vias. There is a difference in the process for positive or negative-acting polymers which is shown in Figure 4:

The developing step is also very sensitive to the duration of the process. The developing solvent may swell the polymer and may lead to a strong film loss (i.e. reduction of film thickness) or even a complete delamination. The next step is then the polymerization process (i.e. called cure) which gives the mechanical properties for the later application. Some of the polymers will shrink to close to 50 % of the film thickness which also has an influence on the via shape. For some of the polymers a descum process is necessary to remove any residues in the via to ensure a low electrical contact resistivity for the next metallization layer. In summary, the photo-process is a quite complex process which has multiple time- and temperature-sensitive process steps.

This is totally different for the laser ablation process of non-photo-sensitive polymers. All of these sensitive process steps are not necessary. The thin film polymer is coated on the wafer similar to the photo-polymer step but the cure will be done as the next step. The final process is the laser ablation process described in detail in the next chapter. The process can also be used for photo-sensitive polymers if the cure is done directly after the coating.

The wavelength of the system used for this investigation is 248nm which is excellent to pattern a wide variety of dielectrics. The excimer is a powerful, pulsed, ultraviolet laser that is well-proven in microlithography,

The laser ablation process can be simply described as the removal of the polymer without damaging the surrounding areas. No cracks or any other heat affected zone should limit the high density applications.

When a high-energy UV-Laser pulse is focused onto a material so that the intensity (which is measured as the fluence) is above a material-dependent threshold value, then the high energy ultraviolet photons directly excite electrons and break interatomic bonds. This threshold is quite important because it can be used to structure polymers on top of inorganic materials without destroying the metal underneath, for example because the threshold of metals is mostly very different than the threshold of dielectric materials. Along with the subsequent shock wave, this causes material to be ejected at high velocity in the form of a fine powder. Each pulse lasts a few nanoseconds and removes a thin layer of the polymer. It is assumed that the process is relatively cold.

Unlike most other laser types, the excimer produces a large area beam that is usually rectangular in cross-section. This has the advantage of being highly compatible with the use of photomasks. The main requirement for the mask is the transmission of the glass material to the wavelength of the laser. Therefore quartz has to be used. For the light blocking metal Al is most suitable due to a very high threshold level which makes the mask stable for a long production cycle time.

The aluminum is patterned as the inverse of the pattern to be structured on the actual polymer layer. The openings in the metal on the mask define the pattern that will be laser ablated. The photomask output is then reduced through a reduction lens (2.5x) onto the target – the targeted area depends on the laser power and the ablation threshold of the target material. In this work a laser ablation system from SussMicroTec (ELP 300) has been used. The laser was a Coherent LXPpro 305 with a power of 40W. The laser characteristics are summarized below:

• wave length: 248 nm (KrF)

• shot repetition rate: 50 Hz

• pulse length: ~ 30 ns

• beam spot size: 6,5 × 6,5 mm²

• fluence range: 70–650 mJ/cm²

Therefore the throughput is very high (i.e., multiple vias per second) and at higher pitches, the number of vias per second actually increases. The reason is that the amount of material removed with each laser pulse, i.e., the depth of any hole or trench, is dependent only on the pulse intensity and the specific materials to be ablated.

Only the stepping of the mask is the limiting factor for the process besides the nature of the polymer and the thickness of the layer. The maximum scan speed is therefore 50/s × 6.5 mm = 325 mm/s.

Because ablation directly breaks the interatomic bonds with minimal thermal effects, it results in excellent surface quality, no microcracking and no recast (melted) debris. The only postlaser process is a cleaning step.

The smallest feature (x-y axes) that can be ablated with the laser depends on the laser wavelength, the optical resolution of the projection lens and the photomask.

The precision depth control in z-axis is controlled mainly by the number of pulses. The high pulse-to-pulse energy stability of the latest excimer lasers means that every pulse will remove an identical depth of material. Depth control is easily provided simply by programming a fixed number of pulses at each stepper site location. In addition, excimer laser ablation even allows control of sidewall angle, by adjusting the laser intensity (fluence) and other means. A higher fluence produces a via or trench with steeper sidewalls, whereas a lower fluence results in a shallower side wall which has been proven using BCB.

The main requirement for a successful ablation process is the absorption of the polymer to be ablated to the UV light of the laser system. This has been already tested for BCB by Tessier et al. [7]. The absorption spectrum of fully cured BCB is given in Figure 5:

Similar spectrum is available for dry-film BCB. For the evaluation of the laser ablation process dry-film BCB has been laminated on 200 mm oxidized Si-wafers.

One of the few disadvantages of laser ablation processes is the fact that some of the ablated material will condense as a kind of dust on the layer. Therefore a cleaning process is essential to avoid any interference with the subsequent process steps. There are no changes to this for the laser scanning process introduced here. Several cleaning methods have been evaluated by the authors.

A promising approach has been also evaluated which is common in laser ablation technology: The usage of a protection layer (Figure 6):

In this case a very thin layer of an additional thin-film polymer or resist is deposited on the dry-film BCB after the cure. It acts as a sacrificial layer. Any residues which will fall as a debris on the polymer not being ablated will be removed after the process using a simple stripping process. The main advantage is the usage of very mild stripper which will not affect the surface chemistry of the polymer layer. Even water-soluble protection layers are under investigation by the authors. The disadvantage is the additional process step which implies a coating process of polymeric layer after the cure of the polymer being ablated. This might require an additional coater module in production. Therefore a trade-off has to be made between the process line and the impact on the surface chemistry by the stripping solvents which may vary for different polymer.

The prime property to judge the quality of a new via formation process is the electrical contact resistance between the metallic layers. Therefore a test mask has been designed to give feed-back form an electrical point of view. A classical two-layer metallization has been designed which give information about single via resistance using 4-Point Kelvin structures. The metal layers are structured using full-field mask aligner technology. 200 mm wafer size was chosen for this evaluation. The equipment is fully compatible to 300 mm wafer technology but 200 mm was chosen. An overview of the mask is given in Figure 7:

There was no issue to adjust the alignment marks generated by the mask aligner to the mask of the laser ablation tool. The optical resolution of the tools is correlated to a precise alignment system for an accurate layer-to-layer registration. This is done by an automated alignment system – global and site-on - site with autopattern recognition. Front side alignment precision is much below ±1μm.

Details of the test structure are shown in Figure 8 and Figure 9:

The via sizes have been varied on the laser mask between 2 μm and 30 μm to see the limits of the process. A via with a mask dimension of 10μm in 15μm thick dry-film BCB is shown Figure 10:

The opening is around 10.3 μm at the top and 7.5 μm on the bottom. The result for the 5 μm mask dimension is shown in Figure 11:

For this mask the via size is less than 5 μm on the top and less than 3 μm on the bottom.

Such small vias have never been published for dry-film permanent polymers using mask aligner or stepper technology

Tessier et al. has measured the threshold of different sputtered metals [7]. Values of 1000 mJ/cm² threshold for sputtered Cu and Al have been published. To verify this data for this new process and expand them to plated metal the following three metallization schemes for the first metal layer have been tested in this evaluation: 1 μm AlSi, 1μm sputtered Cu (200 nm TiW underneath) and 1 μm plated Cu on 300 μm Cu with a 200 nm TiW layer as an adhesion layer. All tests were done on 200 mm Si wafer with an 100 nm Oxide layer.

A comparison of these metals for metal 1 in the laser ablation process is shown in Figure 12:

The FIB (focus ion beam) cuts show the clean interface between the metal 1 and metal 2. No metal was damaged by the laser process. This indicates the large difference in the threshold between polymers like BCB and metal like Al or Cu. No differences have been found between sputtered Cu and plated Cu. All samples had a protection layer for the removal of the debris. No descum-like process was used after the removal of the protection layer. Standard back-sputtering of Al was used before the sputtering of metal 2 (AlSi).

To further judge the laser ablation process electrical tests have been made with these test wafers. The electrical resistivity of the vias were measured to be between 1 mOhm for vias larger than 7 μm (mask) and around 1 Ohm for the vias of 4 μm diameter (mask). Values were similar for 4 μm and 8 μm thick BCB layer. The electrical tests were done using multiple test devices spread over the 200 mm wafer. In Figure 13 the successful opening of the 3 μm vias is shown. The metallization of the next layer seems to be limiting the process if purely sputtering is used. The AlSi metallization is very thin at the bottom of the via.

Further tests will be made by changing the metallization scheme for Metal 2 to plated Cu to ensure a better coverage of the vias down to the bottom. In addition via chains will be designed for the laser ablation process to get also a broader via yield result.

Lithography in general is a multi-step process involving developers and other wet chemicals making it increasingly unattractive. All the risks and cost associated with these chemicals, and their safe handling and disposal can be eliminated using the laser ablation process. Like every other industry, advanced packaging and interposer manufacturing is under pressure to use greener manufacturing that is less polluting and more energy efficient.

It is important to distinguish excimer ablation from laser direct imaging called LDI. Excimer laser ablation is a direct one-step subtractive process which does not need any developing agents for structuring. Only a cleaning step is required after the structuring process. LDI needs photo-sensitive material and is more linked to classical photo-lithography if the process flows are compared, except that a focused laser beam directly writes a pattern on a resist instead of using a mask. With this new process sub-4μm vias could be structured in 15 μm thick dry-film BCB.

Many thanks to all persons from Fraunhofer IZM and TU-Berlin involved in this work not being mentioned as co-authors.