*2.1. Molding Tool Geometry*

The experimental setup was designed in a way that accommodates both research tracks related to the process and product fingerprints. To proceed with the approach of product fingerprint and in order to access the quality of on-part micro features in correlation with on-runner μ-pillar features, specifically developed tool inserts for the production of a biochip were manufactured. The mold used was a two-cavity mold as seen in Figure 1 and the manufactured geometry consisted of the two sides of a bio-fluidic microchip for drug testing. The biochip had the form of a 20 × 20 × 2 mm plate with on-part conical μ-pillar features with 600 μm nominal height, Ø250 μm base diameter, and Ø200 μm top diameter [6] as seen in Figure 2. The tool inserts were manufactured to accommodate pillar μ-features on the runner equal to those on the part, as it can be seen in Figure 3.

**Figure 1.** Half section view (**a**) and <sup>3</sup> <sup>4</sup> views of the movable (**b**) and stationary (**c**) sides of the mold used for the experiment.

**Figure 2.** The micro pillars' feature shape and the dimensions of the parts.

**Figure 3.** Molded geometry with fingerprint structures on the part and runners (**a**,**b**), and measurement positions on Cavity 1 (**c**) and Cavity 2 (**d**).

Figure 3 illustrates the geometry of the molded plastic parts and presents the positions of interest; PP1 close to the gate, PP2 in the middle of the parts, PP3 far from the gate and RP2 on the runner of the molding for both cavities. The pillars in the illustrated positions are used to assess the replication quality of the molded components for all treatment combinations in the experiments as presented in the following section of the paper. Figure 4 presents an example of the physical molded components.

**Figure 4.** (**a**) Molded component with fingerprint structures on the part and runners. The fingerprints at the front (top) and back (bottom) side of the components are visible. (**b**) Bottom (Cavity 1) and (**c**) top (Cavity 2) parts of the microfluidic system.

## *2.2. Injection Molding Process and Experimental Conditions*

The proposed product and process fingerprint concept is built on the hypothesis that the quality of the on part μ-features is correlated to the on runner μ-features and other quality indicators originated from process signals as is discussed in Sections 2.4 and 2.5. The concept requires an experimental validation to confirm the hypothesis of the micro features and extracted indices suitable to be used as quality indicators. The experiments were performed on an electric Arburg 370A injection-molding machine (Arburg GmbH + Co KG, Lossburg, Germany), with a hydraulically actuated clamping unit capable of a maximum clamping force of 600 kN and a screw whose diameter was Ø18 mm. A statistically designed 24 × 3 full factorial experiment was utilized in order to investigate the experimental process window. The parameters under consideration are: Tmelt (Tm) [◦C], Tmould (Tmld) [◦C], Injection Speed (InjSp) [mm/s] and Packing Pressure (PackPr) [bar] that, as from well-established research [18] and preliminary screening experiments are known to be the most significant parameters affecting the quality of injection molded components and surface replication. Table 1 presents the experimental treatments. The process parameter levels were selected by assessing the specification of the material (Figure 5), a commercial grade of acrylonitrile butadiene styrene (ABS, Styrolution Terluran GP-35, INEOS Styrolution GmbH, Frankfurt am Main, Germany), which is characterized by a relatively large processing window. Other parameters such as packing (tpack = 10 s) and cooling times (tcool = tpack + 10 s) were set on levels high enough to avoid their influence on the responses of the experiment.


**Figure 5.** (**a**) PvT and (**b**) viscosity plots of material Styrolution Terluran GP-35 (Acrylonitrile Butadiene Styrene—ABS) [19].

For every experimental treatment, the initial 20-molded parts from the start of the process were discarded, as the process was running to reach stability. Then the following 10 parts were collected for assessment and the three sample parts were measured (denoted as: part 1, part 5, part 10) for the assessment of the μ-pillars' replication quality and then placed both on the parts and on the runners. The sequence followed and the experiment is illustrated in Figure 6.

**Figure 6.** Flow diagram of the experimental sequence. The figure denotes the measurement areas on the part (i.e., PP1 = Part Position 1) and on the runner (i.e., RP2 = Runner Position 2) without the indication of cavity as seen in the text (i.e., Cavity 2 RP2 = C2RP2)

## *2.3. Pillar Dimensional Measurement and Uncertainty Evaluation Procedure*

The pillar height dimensional measurements were carried out by using a focus variation microscope (Alicona Infinite Focus from Alicona Imaging GmbH, Raaba, Austria). The focus variation method is suitable for the scanning of the 3D topologies as it can effectively acquire scans of features with high slopes. A full scan of the μ-pillars though, proved to be challenging due to the almost vertical slopes (88◦) of the μ-pillars. The settings used for the measurements are presented in Table 2.

**Table 2.** Alicona measurement settings for μ-pillars.


For the assessment of the process' stability, the effect of process parameter changes and the replication fidelity of the pillar μ-features for each experimental treatment, three pillars in each position were scanned to measure the μ-pillar height. The middle pillars in positions PP2 and RP2 of both cavities were measured five times in order to determine the repeatability of the measurements (standard deviation in the range of 0.1–0.2 μm was achieved) and provide sufficient data for measurement uncertainty calculations (see Section 3.1). The measurement data sets were consequently processed with the use of scanning probe image processing software (SPIP V6.4.1 by Image Metrology A/S, Hørsholm, Denmark) to extract the μ-pillar height from each scan. In SPIP, a procedure was developed to process the scans and prepare the files for pillar height calculations following the same steps for all four positions of interest by correcting the 1st order tilt in the scan as well as to set the zero background for all data-points as illustrated in Figure 7.

The average pillar height was calculated with the use of four profiles that intersected the center of the pillars with the procedure utilized to scan of both mold and molded parts in order to calculate the height and height deviation (mold-part) as a measure of the molded features replication fidelity.

To verify the quality of measurements and procedures an uncertainty evaluation was conducted. The evaluated expanded uncertainty U is a parameter associated with the measurement results and describes the data dispersion always in connection to the respective measurand. The estimation of the uncertainty and its inclusion in the replication fidelity assessment of the micro features is of great importance as the measurement repeatability and instrument accuracy can be of similar magnitude. The uncertainty budget of the measurements of the pillar heights on the parts and the respective cavity features on the mold insert were estimated based on the ISO 15530-3 (Equations (1)–(4)) [20]. The method was developed for measurements conducted with a tactile coordinate measuring machine (CMM); however, it can be adapted and applied for optical measurements [21] using Equation (4).

**Figure 7.** (**a**) SEM 3D image of the pillars and (**b**–**d**) pillar height measurement procedure, (**b**) step 1: extracting cros-section profiles, (**c**) step 2: assessing pillar height from the four extracted profiles as indicated by different color, and (**d**) 3D representation of the pillar [5].

The expanded uncertainty was calculated with a coverage factor k = 2 to achieve a confidence level of 95%, and four uncertainty contributors were considered (Table 3) (see Equations (1)–(3)). Such uncertainty contributors are ucal which is the standard uncertainty as evaluated from a calibrated step height artefact to have traceable measurements, ub which is the standard uncertainty associated with the systematic error (b) of the measurement process, which is the measuring instrument bias. Thirdly, the uth is the standard uncertainty associated with the systematic error of the measurement process based on the heat expansion coefficient deviations of the material, since the measurements were not conducted at the reference temperature, and lastly up is the uncertainty associated with the manufacturing variation from either mold or parts (upmould and uppart), which is calculated using a square distribution in the modified ISO 15530-3 (Equation (4)). The measurement on individual pillars, features, and different molded parts are all affected by instrument repeatability. Thus, for uppart the maximum value of uncertainty contributor related to instrument and process is considered in order to avoid underestimating the uncertainty. These contributors are part of uppart, where: uppillar is the standard deviation of five repeated measurements on the same pillar; upfeatures, the standard deviation of repeated measurements on four different pillar areas to estimate feature repeatability in terms of polymer replication and upsample the standard deviation of repeated measurements on 3 different samples on four different pillar area. The uncertainty contributors are used to calculate the uncertainty of the mold (Equation (1)) and part pillar (Equation (2)) measurements, as well as the deviation uncertainty (Equation (3)). The values of the specific uncertainties per position and experimental runs are provided in Tables 4 and 5, respectively. Table 5 provides information on the expanded uncertainty for pillar height and height deviation measurements per run.

$$\mathbf{U\_{part}} = \mathbf{k} \times \sqrt{\mathbf{u\_{cal}^2} + \mathbf{u\_b^2} + \mathbf{u\_{th}^2} + \mathbf{u\_{ppart}^2}} \tag{1}$$

$$\mathbf{U\_{modid}} = \mathbf{k} \times \sqrt{\mathbf{u\_{cal}^2} + \mathbf{u\_b^2} + \mathbf{u\_{th}^2} + \mathbf{u\_{pmoud}^2}} \tag{2}$$

$$\mathbf{U}\_{\rm dev} = \sqrt{\mathbf{U}\_{\rm nucul}^2 + \mathbf{U}\_{\rm part}^2} \tag{3}$$

$$\mathbf{u}\_{\text{P}\_{\text{i}}} = \frac{\mathbf{data}\_{\text{max}} - \mathbf{data}\_{\text{min}}}{2\sqrt{3}}, \text{ i } = \text{pillar}, \text{feature}, \text{sample for part or model} \tag{4}$$

$$|\mathbf{u}\_{\text{ppm}} = \max(\mathbf{u}\_{\text{ppillar}}, \mathbf{u}\_{\text{pfeature}}, \mathbf{u}\_{\text{pannple}}).\tag{5}$$

**Table 3.** Uncertainty contribution for pillar height measurements.


**Table 4.** Expanded uncertainty for single pillar height and height deviations measurements.




#### *2.4. Product Fingerprint as Quality Indicator*

The concept uses the microfluidic system described in Section 1 as a case study. It is of particular interest as it builds upon past studies that used nano features (fingerprint) on the part, where a close correlation of the fingerprint on the part to the overall quality of the component was revealed [8].

The current paper considers the use of dedicated μ-features positioned on the runner of the molding that are equal or similar in size and shape to the features on the part [5]. The μ-pillars on the runner can be used as a product fingerprint as they can be quickly measured with an in-line process set up, separated from the main component and directly correlated to the overall part quality.
