1. Introduction
In recent years, the integration of electronics in injection-molded parts to manufacture “smart” mechanical, durable and light products cost-effectively has gained traction [
1,
2]. One way is to adapt the in-mold decoration (IMD) process. Here, a sheet or film is inserted into an injection mold and overmolded by rapidly injecting a thermoplastic melt. As a result, a plastic part with printed graphics and/or high-gloss surfaces is manufactured. By additionally incorporating conductive inks and surface-mounted electronic components to electronic films, the injection molded structural electronics (IMSE) process is formed. The assembly is exposed to high temperatures as well as pressures and shear stresses due to the molten viscous melt during overmolding. For that reason, TactoTek has postulated a certification process for cthe omponents and surface mounting adhesives used in their IMSE designs [
3,
4,
5].
To reduce the prevailing forces on the components, Ott and Drummer proposed using thermoplastic foam injection molding. Through this technology, they could encapsulate epoxy-based printed circuit boards (PCBs) with significantly lower cavity pressures (about 1 MPa) compared to traditional injection molding [
6].
Structural electronics can also be manufactured by embedding flexible/stretchable PCBs and components via lamination processes between covering plastics sheets [
7]. The thus created flat stack might be thermoformed to a 3D-shaped structure before injection overmolding. The stack is then additionally exposed to complex elongations during thermoforming. This can cause the out-of-plane deformation of the laminated structures, such as the conductive tracks. Madadnia et al. [
8] were able to reduce this undesirable deflection by adding a fractal structure to the original meander-shaped conductors.
The application and merging of multiple materials to one (functional) object is certainly ambitious from a sustainability point of view. Välimäki et al. [
9] recently showed the potentials of incorporating bio-based and recycled-film materials and replacing metals and metal oxides with PEDOT:PSS, carbon and amino acid/heterocycles to save energy and reduce the depletion of limited material resources.
It is difficult to make predictions about the “moldability” of injection-molded parts as multiple physical variables have to be considered. For instance, the viscosity—the resistance towards flow—of most thermoplastics decreases with the increasing shear rate and temperature in a non-linear fashion. Hence, numerical injection molding simulation software has become a must-have tool to make assessments about ever increasingly complex molding tasks in advance.
Kololuoma et al. [
10] showed the feasibility of the hybrid integration of printed and flexible electronics in combination with conventional electronic components in a personal activity meter demonstrator. They overmolded their assembled hybrid activity badge foils with thermoplastic polyurethane (TPU) using two sets of molding parameters. The setting with a lower melt and mold temperature and slower injection speed resulted in broken electrochromic (EC) displays of their devices. Injection molding simulations using Moldex3D (CoreTech System Co. Ltd., Taiwan) revealed higher shear stresses and filling pressures for this set of parameters. The authors concluded that in this case the adhesion between the layers of the EC display was exceeded by the vertically acting shear stresses.
Numerous authors have looked at different aspects of IMD-produced parts to optimize the molding parameters to reduce high scrap rates inherent to this sensitive process [
11,
12,
13,
14,
15,
16,
17,
18]: The insertion of the film on one side of the mold can cause an asymmetric flow front advancement as heat transfer is retarded on that side [
12,
17]. The insulating effect of the film frequently causes warpage due to an unsymmetrical temperature profile across the part thickness during packing and cooling [
15,
16]. An asymmetrical cooling system with a lower mold temperature at the side of the film can significantly reduce the warpage [
12,
13]. Part geometry [
18] and film thickness [
13,
16] as well as processing settings, such as mold and melt temperature [
12,
16], will determine the final dimensions of the warped parts.
Liu et al. [
14] systematically investigated the effect of injection molding processing parameters on the ink wash-off during the IMD process. To that end, they used 20 wt.% glass-fiber-filled polyethylene terephthalate (PET) to overmold cut PET films (250 µm in thickness) in a rectangular mold (80 × 20 mm
2). The film featured screen-printed grid patterns (2 × 2 mm
2) and the size of the washed area was evaluated. An experimental design using the Taguchi method was established containing the five parameters of melt temperature, mold temperature, melt injection speed, injection pressure and part thickness (0.5, 1 and 2 mm, respectively). They concluded that wash-off could be reduced by adopting a larger part thickness, a lower mold and melt temperature as well as a lower injection speed.The same tendencies were reported by Woyan et al. [
16] who evaluated the washed-out ink on polycarbonate (PC) films under the injection point when overmolded in a rectangular plate mold with PC. The analysis of variance (ANOVA) of their full factorial design however did not yield any of the factors as statistically significant. They further compared two PC grades, with the material with the better flowability demonstrating very low ink wash-off. They concluded that the wash-off mainly depends on the shear stress, and hence a thick part and/or low-viscosity melt would decrease the shear stresses on the ink. Furthermore, the melt solidifies on the film during filling, forming a frozen melt layer that connects with the ink. The layer that is stretched by the melt can cause ink delamination. The higher the melt temperature is set, the thinner the frozen layer becomes and the easier it stretches.
The harsh overmolding conditions can damage the structural electronics, lead to high scrap rates or even make the process unfeasible (cf.
Figure 1).
To the best of our knowledge, there are only limited studies in the literature related to the optimized injection molding process for flexible laminated electronics. This paper aims to gain a better understanding of how the molding settings impact the flexible laminated electronics, and hence to develop molding guidelines. This is achieved by combining the observations of performed experiments with numerical simulations.
Test film strips resembling a typical laminated flexible electronics stack were fabricated and overmolded with polycarbonate (PC) in a stepped plate mold. Two different TPU types were used as glue layers and the processing parameters, melt and mold temperature, as well as the injection speed, were investigated using a two-level full factorial design of experiments (DoEs). The produced parts were then visually examined for damage. Injection molding simulations using Autodesk Moldflow Insight 2021 (AMI, Autodesk Inc., USA) were performed to virtually inspect the films regarding the prevailing temperatures and shear stresses during overmolding. Finally, a shear distortion factor (fτ) was derived based on the simulation results.
2. Materials and Methods
2.1. Film Design
Laminated stacks, as shown in
Figure 2a, were prepared and cut into 115 × 30 mm
2 film strips (
Figure 2b) for overmolding. Two different layer structures were assembled and are shown in
Figure 2c,d, respectively. The outer layers consisted in both cases of 125 µm-thick PC sheets (Isosport Verbundbauteile GmbH, Eisenstadt Austria). Here, a 5 × 5 mm
2 grid was screen-printed on both sides of the melt facing sheet (bottom side) using the non-conductive ink NORIPHAN
® HTR N 990/010 NC—Tiefschwarz (Proell GmbH, Weissenburg, Germany). A Upisel SR1220 film (50 μm PI—18 μm RA Cu from UBE EXSYMO CO., LTD, Tokyo, Japan) was used as the middle layer that could serve as a flexible printed circuit. In the first stack, the 100 µm-thick TPU Bemis 3914 (Bemis Associates Inc, Shirley, MA, US) was utilized, and in the second stack, the 75 µm-thick Platilon U073 (Covestro AG, Leverkusen, Germany) was chosen as an adhesive between the layers. The Bemis 3914 and Platilon U073 TPUs in the following will be referred to as TPU B and TPU P, respectively. A Lauffer laminator press (Maschinenfabrik Lauffer GmbH & Co. KG, Horb am Neckar, Germany) was used to laminate the different layers of the stacks with lamination parameters as indicated in
Figure 2e.
Figure 3 shows the specific heat capacity curves (
cp) of the individual film layers measured using a differential scanning calorimeter DSC1 (Mettler-Toledo International Inc., Columbus, OH, US). The semicrystalline TPU P exhibits a more pronounced melting peak at higher temperatures compared to TPU B (enthalpy: 11.2 vs. 0.8 J/g and peak temperature: 163 vs. 123 °C). The melted fraction,
αm, of the melting peaks are also depicted. TPU B is fully melted at around 150 °C, which is ~40 K below the melting peak of TPU P. The amorphous PC layer yields a glass transition temperature T
g ≈ 150 °C.
2.2. Part Geometry and Molding Material
The film strips were overmolded within a stepped, rectangular, plate mold in which the wall thickness reduced from 3 to 1 mm, as shown in
Figure 4. The films were fixed within the mold at the fixed half by applying a temperature-resistant adhesive tape on the flow-front-facing film edge (this tape and the flexPCB are not shown in
Figure 4).
A polycarbonate PC Lexan OQ1028 (Sabic, Riyadh, Saudi Arabia) of high optical quality was used for molding. A total of 0.5 wt% of yellow CC10104356BG masterbatch (PolyOne Color & Additives Germany GmbH, Melle, Germany) was added to color the PC and thus make the interface to the overmolded film more clearly visible.
2.3. Injection Molding and Experimental Design
A fully electric Arburg Allrounder 470 A Alldrive (Arburg GmbH + Co KG, Loßburg, Germany) injection molding machine equipped with a 25 mm screw was used to perform the overmolding. A Wittmann Tempro plus D 160 (WITTMANN Technology GmbH, Wien, Austria) temperature-control unit was utilized to regulate the mold temperature. The dosing volume was set to 40 cm3 and the switchover point (velocity to pressure-controlled filling) was adapted for each setting. The packing pressure was set to about 80% of the filling pressure for 10 s and the residual cooling time was set to 50 s.
Preliminary tests were performed for both stacks with TPU B and TPU P using different injection speeds ranging from 10 to 70 cm3/s at a melt temperature of 340 °C and a mold temperature of 120 °C. Those tests already indicated an important role of the TPU-layer: Films using the TPU B showed damage below injection speeds of about 60 cm3/s, while good parts could already be produced at injection speeds of 30 cm3/s, if films were used with TPU P in their stack.
For further investigations, the 2-level full factorial designs of experiments (DoEs) were created. In such parameter studies, input variables (factors) are investigated at two set levels (low and high) and all possible factor (k) combinations are investigated in a total of 2k runs.
Individual DoEs for both TPU types with factors of melt temperature (A), mold temperature (B) and injection speed (C) were created and are shown in
Table 1. DoE I features films using TPU B in its stack (setting B02-B10) and DoE II films with TPU P (P02-P10). The low-level injection speed (C) of DoE I was chosen to correspond to the high-level injection speed (C) of DoE II. Thus, DoE III (B01-B05 and P01-P05) can be derived with factors of melt temperature (A), mold temperature (B) and TPU type (D).
Additional center-point (CtPt) runs were added in which all the factors of the corresponding DoEs ere set to the intermediate level (CtPt DoE I: B10, CtPt DoE II: P10, and CtPt DoE III: B01 and P01).
Three parts per setting were produced, which resulted in a total of 3 × 2 × (2
3 + 1 + 1) = 60 parts produced. They were evaluated according to the procedure explained in the following:
Section 2.4.
2.4. Molded Parts’ Analysis Procedure
It was not possible to “measure” the appearance of the films according to some measured quantity. In such cases, Kleppmann [
19] suggests to establish a graded evaluation scheme based on a subjective assessment of the parts. In this way, it should still be possible to perform statistically sound comparisons.
Hence, the overall appearances of each of the overmolded films were assessed by separating the produced parts into three groups: I—not or slightly damaged, II—clearly damaged and III—severely damaged. In addition, the— counted via visual inspection—fraction of distorted squares (fs) was determined. Finally, the damage of the parts was weighted by introducing a distortion factor fd = fs × c. Where c corresponds to the assigned damage category: cI = 1/3, cII = 2/3 and cIII = 1. This was conducted to capture the distortion of the produced parts in a more realistic manner.
Those (unitless) distortion factors (
fd) were then used as the output variables (responses) to evaluate the three DoEs presented in
Table 1. To that end, the statistic software Minitab (Minitab Inc., State College, PA, US) was used. It performs a linear regression, linking the investigated factors of the DoEs (input variables) with the distortion factor (
fd, output variable or response). The response of each factor was hereby modeled linearly. In order to detect possible nonlinearities, additional center-point settings were added as described above.
The software also estimates which of the model factors are statistically significant based on the concept of the analysis of variance (ANOVA). One output of this analysis is the
p-value. It is calculated for each factor indicating the risk to reject the null hypothesis (no relationship between factor and response) when in fact the null hypothesis is true. Frequently, a factor is considered as significant (and not of random origin) for
p ≤
α = 5%, with
α being the significance level [
19]. A
p-value for the center point (CtPt) is also calculated, thereby indicating if at least one of the factors included in the model behaves in a nonlinear manner.
2.5. Simulation Model Preparation
The commercial injection molding simulation software Autodesk Moldflow Insight 2021 (AMI) was used to numerically study the molded DoE (cf.
Table 1). It provides an insight into the temperatures and shear loads faced by the film during filling. As a consequence, a simulation can help in developing a more profound understanding of the process.
To model the filling phase, AMI numerically solves the conservation equations of mass, momentum and energy using the finite element method (FEM) [
20].
Two 3D FEM models were created featuring either the laminated B-stack or the slightly thinner laminated P-stack film strips. The Upisel layer was excluded from the modeling resulting in a three-layer film of PC–TPU–PC and the injection-molded part, as shown in
Figure 4.
The global edge length was set to 1 mm with a minimum number of 12 elements through the thickness for the injection-molded part (AMI property part). The film gate and the region of the part in contact with the film was modeled with a mesh size of 0.5 mm. A minimum number of 6 elements through the thickness and a mesh size of 0.5 mm were selected for each of the three film layers (AMI property part insert). The auto-sizing scale factor was set to 0.9 and the machine die was modeled as a beam hot runner.
The linear tetrahedral element counts for the B- and P-stack models were 2,766,738/893,954/973,914/924,070 and 2,759,495/896,252/991,487/924,095 (part/PC layer/TPU layer/PC layer), respectively.
The simulation material data (Cross-WLF viscosity coefficients, Tait pvT coefficients) for the PC Lexan OQ1028 overmolding material were provided by Sabic (Autodesk udb-file). The possible influence of the added master batch (e.g., on the viscosity) was not considered.
While the specific heat capacity (
cp) of the individual stack layers was measured (cf.
Figure 3), the thermal conductivity (
λ) and density (
ρ) of those materials also needed for the thermal analyses were not. Those quantities were approximated by choosing the values given in the AMI material data base for PC Makrolon 2805 and TPU Desmopan 487 (both from Covestro) for the PC and TPU B or TPU P layers, respectively.
A uniform heat transfer coefficient (HTC) of 5000 W/(m2∙K) (AMI default for the filling phase) was used for all interfaces.
A starting temperature of 25 °C was assigned to the films and a 10 s contact time with the hotter mold prior to start of injection was specified. This should be about the time it took the operator to start a new injection molding cycle after inserting the film into the mold.
Fill was selected as the analysis sequence and 20 intermediate results were set to portray the process conditions in a higher time resolution. A constant mold wall temperature according to the experimental design was assumed as the surface boundary condition.
2.6. Simulation Analysis Procedure
The TPU layer has a significant influence on the overall intactness of the overmolded films (as revealed by the preliminary tests). Distortion will presumably only occur when the TPU becomes molten due to the prevailing temperatures (melted fractions: 0 ≤
αm ≤ 1). The damage on the film will increase the higher the shear stresses (τ) introduced by the viscous melt become and for the longer they act (
t). To capture this, a shear distortion factor
was derived. It is used as a magnitude to assess and compare the shear-introduced loads on the film between the different settings.
A Python script was developed that made use of the Synergy Application Programming Interface (API) [
21] to access the AMI results and information about the node location and element allocation: The closest shear stress at wall (τ) results of the part mesh were projected to the TPU-layer nodes. Similarly, the temperature results of the TPU-layer nodes were used to calculate the melted TPU fraction (
αm illustrated in
Figure 3) during filling (
tfill). Hence, using Equation (1) and the Python scipy.integrate.quad numerical integration routine [
22], an individual shear distortion factor (
) for each TPU-layer node could be obtained.
The procedure for obtaining
fτ is illustrated in
Figure 5. Finally, the local shear distortion factors (
fτ) were drawn as shaded contour plots using the API. An overall averaged shear distortion factor () for the TPU layer was calculated as well.