Simulation and Experimental study on IMS (Injection Molded Solder) bumping with expanded resist patterning for reinforcement of fine-pitch capability

The demand for high density integration and high-power efficiency of semiconductor chip has expanded with the reduction of the IC packaging size. To meet these requirements, technologies for fine pitch interconnect with micro-bumps have been improved significantly in recent years. It is widely used for variety of applications from high end computer to personal mobile edge devices. Micro bumps consisted of lead-free solder play a crucial role as they are responsible for the performance and the reliability of the electronic systems.

One of the most common method for solder bumping is the electroplating due to its capability of fine pitch solder deposition. However, electroplating suffers from some technical limitations in addition to its complex and costly manufacturing process. For example, there is a limitation of material selection of solder alloy such as Sn or binary solder Sn-Ag. The lack of flexibility in terms of material composition greatly limits the physical properties of the material and narrow down the potential applications as a consequence. In addition, the inhomogeneous undercut during the seed etching process and the variation of the thickness are an inherent disadvantage of the conventional electroplating.

An alternative solder bump forming technology to electroplating has been developed named IMS (Injection Molded Solder) technology [1]. It is a very simple process with molten solder directly injected into the resist opening to form solder bumps. It offers many benefits beyond its low-cost and clean process with no flux used. IMS utilizes the molten solder of any kind of metal compositions and allows great flexibility of alloy. There has been substantial works for the development of IMS technology [2–6].

Meanwhile, one of the critical issues in IMS technology is the limitation of the applicable injection pressure. Applying high injection pressure for the bump formation could cause resist deformation and harm the process robustness. Nevertheless, required pressure to fill the resist opening with solder increases significantly as the resist opening area decreases. This arises the need for reduction of the injection pressure to form micro-solder bumps. This study presents the details of novel resist structure for IMS process to reduce the injection pressure. It consists of upper layer with larger diameter than lower layer, so that the required pressure is adjusted to the size of upper resist opening while targeted bump size is controlled by the lower resist opening. The capability of this double-layered structure is evaluated both in numerically and experimentally in this paper.

IMS process is an advanced bumping technology to form fine pitch micro solder bumps while offering a great flexibility of the variation of solder alloy usage. The schematic image of IMS bumping process for Si wafer is shown in Fig. 1. IMS is applied after resist patterning (Step1.) and electro plating of Cu pillar (Step 2). This process utilizes a solder injection head that melts the desired bulk solder alloy composition. An appropriate pressure by nitrogen gas is applied to the molten solder to be dispensed into the locally vacuumed resist hole within N₂ environment. Subsequently, the injected molten solder contacts and wets the Cu pillar surface inside of the hole. As IMS tool head is scanned horizontally and the resist opening is exposed at the atmosphere, solder bump is formed on Cu surface consecutively by surface tension of liquid solder (Step 3).

Although IMS provides reliable performance, there is a limitation associated with the applicable injection pressure mentioned above. Fig. 2 shows the correlation between the bump diameter and the required injection pressure in case of the resist thickness with 10μm. It can be seen that the pressure is inversely proportional to the bump diameter. In order to fabricate bumps with such small size, an effective way to reduce the pressure is required.

In this section, two types of new resist patterning design Type-A and Type-B are presented as show in Fig. 3 respectively. Both of them are consisted of lower layer with smaller diameter opening and upper layer with expanded opening. The injection pressure is adjusted to the diameter of the upper layer resist opening since IMS tool head directly contacts to upper resist surface during the scan. This results in the reduction of pressure compare to the conventional structure with single layer of which the diameter is as small as the targeted bump size. The difference between Type-A and Type-B is that Type-A has an isolated opening whereas Type-B has openings that are connected adjacently and appear to be wavy shaped. As an extent, Type-B is expected to reduce more pressure than Type-A as it has larger injection area. However, it is associated with the risk of solder bridging across the adjacent bumps which is a serious defect to induce electrical failure. The pattern of lower layer determines the position and diameter of Cu pillar, while the thickness of upper layer determines the volume of the solder. Molten solder filled in the resist hole is expected to aggregate with size of lower resist opening to form bump shape due to the intrinsic surface tension effect (Fig. 4).

CFD simulation is conducted to simulate the behavior of molten solder bump formation and validate the feasibility of proposed resist design. A commercial CFD software ANSYS Fluent 19.0 was employed in this study. The molten solder and air are considered as working fluid and three-dimensional analysis of multi-phase model was performed.

In addition, the effects of wall adhesion are incorporated in the VOF model through the contact angle θ at the wall as following equation.

ň_w and ť_w are the unit vectors of normal and tangential to the wall. The contact angle that the fluid is assumed to form is used to adjust the surface normal in cells next to the wall.

We first evaluated the effect of double-layer resist structure of Type-A model described in the previous section. Fig. 5 displays the side view of its three-dimensional model. It is composed of lower layer with thickness t_l and opening diameter d_l, and as well as that of upper layer with t_u and d_u. Subscripts l and u denotes lower and upper respectively. The bottom face of the model is considered as the surface of Cu pillar, hence it is set to have wetting characteristic for solder to sufficiently wet and aggregates around. Other walls are considered as resist material surface and have non-wetting property. Those wetting and non-wetting properties of the surface condition are realized in the numerical model through the contact angle mentioned above (9). A no slip wall boundary condition was used in the calculation.

Initially, the system is filled with solder phase up to the height of upper resist thickness t_u and the remaining region is filled with air. In this study, t_u is used as the variable parameter to evaluate the relationship between the resist height and the formed bump height. All the used parameter values are listed in Table 1. In addition, material properties of solder and air are summarized in Table 2 that has been used in previous study [9].

Simulated results of equilibrium solder bump shape with different upper resist thickness t_u in the case of d_l = 10 μm are shown in Fig. 6. It was confirmed that solder bump having a hemispherical shape was successfully formed with base diameter of d_l. The geometry keeps stabilized when balance among intrinsic surface tension and gravity is achieved. The result also showed that as the resist thickness increase, the bump size including its height and width increase. The relationship between the upper resist height versus the fabricated bump height, as well as the bump height versus bump diameter are shown in Fig. 7 and Fig. 8 It is compared with the theoretical value using truncated sphere model (10), (11) [10], [11]. D and h represent the diameter or the largest width and the stand-off height of the bump respectively. In Fig. 7, simulated result appears to be slightly lower than the theoretical model. This may be due to the gravitational force which is considered only in the simulation. In the same way, Fig. 8 showed that bump diameter of simulated result turned out to be slightly larger than the theoretical model.

If the upper resist thickness is too large, in other word the initial solder volume filling the resist hole is too large, the final shape of solder does not stabilize and fail to form uniform sphere-like shape. Table 3 shows the result of maximum achievable resist thickness to form satisfactorily uniform and stable shape of solder bump. These results are used for determining the dimension of the Type-B model with connected resist openings in the following simulation.

Based on the result obtained from the previous simulation, Type-B structure with connected upper resist openings are evaluated subsequently. The case of d_l [μm] is reflected this time. When determining the pitch of two adjacent Cu pillar location, or where the bump is targeted to be formed, the risk of solder bridging occurrence must be taken into consideration. The distance between two bumps should be longer than the largest bump diameter with some margin. Therefore, the pitch is set as 20 μm so that neighboring bump does not interact each other. Configuration of the simulation model with four Cu pillar aligned next to each other as shown in Fig. 10. The calculation of molten solder dynamical behavior is analyzed in the similar manner as previous calculation.

The molten solder evolving behavior with developing time is shown in Fig. 10. It is indicated that solder starts to break apart due to the combination of surface tension force and non-wetting repelling force derived from the hydrophobic characteristic of resist material. Subsequently, solder forms sphere-like shape at isolated position and stay stabilized while bridging is not occurring in any location. From those results, the effectiveness of resist patterning design with type-B structure is verified.

However, if the repelling force of the resist material is insufficient, bridging phenomena was observed in the simulation. It is empirically known that surface condition of the solid including the roughness and chemical condition, largely affects the wettability tendency of the material [12]. Therefore, the contact angle between solder and resist material was varied from θ = 140° to 120° as shown in Fig. 11. In the case of θ = 140°, solder bumps are formed separately with uniform height and width while in the case of θ = 130° and 120° resulted in bridging between the center two and all the four Cu pillars locations respectively. Those simulation results demonstrated the viability of novel resist design as long as the resist quality of resist material is high enough to dewet the liquid solder.

Experimental demonstration was also conducted to examine the feasibility of proposed design of resist patterning within the actual IMS process. For the experiment, two types of resist patterning, Type-A and Type-B were prepared as in Fig. 12. The dimensions of those resist were designed to be corresponding to the simulation model that it has lower resist diameter of 10.5 μm, upper resist diameter of 15.9 μm and the pitch of 20 μm for both Type-A and B. IMS process was operated on those two types using Sn-3.0Ag-0.5Cu (SAC305) solder.

Fig. 13 shows SEM images of micro-bumps formed with two types of resist pattering respectively. It was observed that each bump was successfully fabricated without any failure in both resist patterns. Regarding the concerns of solder bridging between adjacent bumps and non-uniform solder distribution in the case of the connected resist opening, the optimized area of resist opening as well as the pitch suppressed those defects.

Moreover, filling rate of molten solder at relatively low injection pressure during IMS process was improved by utilizing new design of presented resist structure compare to the conventional structure. Even the Type-B structure showed better filling rate of solder than Type-A due to its wide-open surface area as expected. Type-B resulted in 100 % of filling rate without any failure whereas Type-A showed 95.2% of that, though the difference is almost negligible.

Those significant improvements proved the validity of the concept of the novel IMS process presented in this paper. In addition, the reduction of the solder injection pressure will lead to further scaling down into a few micrometer scale bumping.

Novel design of double-layer resist patterning with expanded opening to reduce the injection pressure in IMS process is proposed in this study. Its effectiveness is examined both in CFD simulation and experimental measurement perspectives. Two different types of resist structure are evaluated that one has isolated resist opening whereas the other has connected opening to reduce more injection pressure.

It is revealed that both of the structure enables to form micro-bumps with good uniformity with lower injection pressure than conventional IMS process. Even the connected model was thought to be associated with the risk of bridging, it is demonstrated that molten solder tears apart between the adjacent positions and form isolated bumps. However, high quality of resist material with sufficient hydrophobic property must be promised, otherwise the bridging phenomena may occur.