1 Introduction

Plasticising polymers using ultrasound has been of interest for decades using established ultrasonic riveting and welding processes [1,2,3]. Recent advances in manufacturing processes for micro and nano-scale components are now realising the potential for ultrasonic melting and forming of polymer materials as an alternative to conventional moulding processes, leading to new micro-injection moulding machines designed primarily for high-value medical products [2, 4,5,6,7]. The first commercial ultrasonic micro-injection moulding (u-µIM) machine was launched in the late 2000s, and since then, there has been increasing interest due to its potential benefits in energy efficiency, material savings and superior feature replication capabilities [8].

The plasticisation of polymers using ultrasound differs from conventional processes and is dependent on a host of factors related to the material and process characteristics, which determines the energy absorption during sonication [9,10,11,12,13]. The melting in u-µIM relies chiefly on the interfacial friction (IF) and viscoelastic heating (VH) under rigorous sonication of the feedstock [14]. IF involves rapid heating of the polymer at surface asperities that are in contact with other granules, mould walls or the sonotrode itself. This results in rapid, localised heating at the contact points which initiates the plasticisation. VH, primarily caused by the absorption of ultrasonic vibrations within the polymer chains, is the second physical process for plasticisation. VH relates directly to the polymer’s mechanical properties under oscillatory induced strains and is reported as the main source of heating and energy transfer during the ultrasonic process [15, 16]. Both IF and VH are of great importance to facilitate the micromoulding of polymers under sonication, and their control becomes critical to avoid excessive heating or sonication frequencies which can hinder a successful u-µIM process. VH is usually a fixed property of the material; however, IF can be controlled via different feedstock shapes such as in rod, pellet or disc forms [17].

Work in the literature shows that IF can play a significant role in the initiation of the plasticisation using ultrasound. Zeng et al. reported that polymer particles that were too large resulted in insufficient heating to plasticise the whole feedstock in u-µIM [18]. According to Wu et al.’s research, IF within the feedstock can serve as a significant heating source for plasticisation in u-µIM [19]. The study showed a significant localised heating rate at the contact points of the particles reaching 160.8 °C in 0.078 s based on their simulations. Their results also show that sonication with higher amplitudes amplifies the contribution of IF to the heating. The work also focussed on the origins of IF finding that IF heating between typical polymer feedstock causes material interfaces to disappear after 0.5 s.

Controlling IF in u-µIM can be a promising feature to facilitate the efficient replication of functional micro-features [20, 21], nano-structured patterns [22] and high-aspect ratio microneedle cavities [23, 24] due to differences that can occur in feedstock shapes. Research within our group demonstrated that micro-feature replication can be achieved at lower pressures compared to conventional processes using disc-shaped feedstock [10]. It was also found that higher mould temperatures facilitated more consistent flow fronts, which can eliminate process variability at microscales [25]. A more recent study in the literature expanded our work and focussed on the functionality of microneedles of our design where microneedle patches made from polypropylene using the ultrasonic method showed better pig skin penetration behaviour, sufficient mechanical strength and chemical stability [26]. These reports show the capabilities and potential of u-µIM in the rapid development of microneedle arrays and process improvements, including the utilisation of IF, which can widen the process window of specific materials and specific microneedle designs. This is due to the wide range of feedstock shapes and polymer material formulations that are in the form of powders, discs, pellets or discs [13, 18, 25].

The importance of IF for ultrasonic plasticisation of polymers and the promising capabilities of u-µIM were studied and emphasised by various reports. However, work in the literature lacks the comparison of various feedstock shapes and their impact on the heating during u-µIM and on final product quality. We postulate that by changing feedstock shape systematically, viscosity reduction, improved micro-feature replication and product quality can be achieved and controlled in u-µIM. To address these challenges, the presented research uses disc-shaped preforms pre-moulded on a conventional micro-injection moulding machine with different thicknesses and compares their performance against industrial standard pellets. The discs gave control over the quantity of interfaces desired for a given shot size by changing the thickness of the discs with more accurate shot sizes. The disc experiments were compared with industry standard pellet shapes by varying the sonication time. The u-µIM events were captured using high-speed thermal imaging, and temperature data were analysed to quantify the effect of the feedstock shape on temperature evolution and product quality. Laser-scanning confocal microscopy analysis was also carried out on selected samples for quantifying the impact of feedstock shape on microreplication quality of microneedle cavities. The characterisation and comparison of disc-shaped feedstock against pellets provide essential information on the replication of microneedle mould cavities and the significance of interfacial friction in u-µIM process development.

2 Materials and methods

2.1 Ultrasonic micro-injection moulding (u-µIM) machine

The experiments for studying interfacial friction effects were performed on a Sonorus 1G u-µIM machine (Ultrasion S.L., Spain) as shown in Fig. 1a. Figure 1b depicts the vertical injection axis where solid polymer feedstock is loaded between a sonotrode and an injection plunger with 8-mm diameter which accommodated the pellet and disc feedstocks in the experiment. During the moulding process, the piston moves upward while the sonotrode remains stationary (Fig. 1c and d), compressing the feedstock in the presence of sonication. It is important to note that the melting and injection of the polymer are simultaneous, and there is no shut off mechanism to carry out a two-stage melting and filling process. The polymer is plasticised using pre-defined process parameters and injected into the mould cavity in a molten state for replicating the micro-features (Fig. 1e).

Fig. 1
figure 1

Ultrasonic micro-injection moulding: a Ultrasion Sonorus 1G; b a rendered image of the machinery cross-section; c pellets placed in the sonication chamber; d sonication is active and plasticisation starts; e the mould cavity is filled. The cavity size shown is only for demonstration purposes and not for scale in this work

Full size image

The u-µIM machine plasticises the polymer by means of ultrasound energy which is generated by an ultrasonic power source (Branson 50DCX-S30VRT). The ultrasonic stack including the converter, booster and the sonotrode was tuned to a sonication frequency of 30 kHz. Ultrasonic amplitude of 100 µm was used for facilitating plasticisation in relatively short sonication times. Technical specifications of the Sonorus 1G U-µIM machine are provided in Table 1.

Table 1 Technical specifications of Sonorus 1G U-µM machine
Full size table

2.2 Mould design and materials

The study utilised a specifically crafted mould tool featuring a linear channel and a circular base moulding cavity to observe the flow of polymer for disc and pellet experiments. The mould was equipped with a sapphire window on the upper (mobile) half of the mould which was used in our previous work [10]. The recess or the base, measuring 17.5 mm in diameter and 0.5 mm in depth, is designed to house circular and interchangeable mould inserts. In this particular case, the mould insert comprised 25 conical microneedle cavities, each with a base diameter of 300 µm and a depth of 550 µm (Fig. 2a). Details of the mould tool design, including the channel and other dimensions, are seen in Fig. 2b and remain consistent with our prior research [25]. Figure 2c shows a rendered image of the microneedle product and a real photograph, detailing key features in the product design including the gate, runner and slug locations. The thin surrounding around the sprue of the moulded part is due to the mould parting design. The mould assembly did not incorporate a venting system since the microneedle and sapphire window provided sufficient venting with typical slide-fit tolerances.

Fig. 2
figure 2

Microneedle patch and mould insert design: a a rendered image of the microneedle insert and technical drawings of the insert cross section; b dimensions of the part design [25]; c a rendered image of the microneedles and an image of one of the moulded parts

Full size image

An injection moulding grade polypropylene resin (PP–GA12, INEOS Group Limited, UK) was used in the present study. Table 2 summarises the physical properties of this material.

Table 2 Physical properties of Ineos 100-GA12 PP resin
Full size table

PP feedstocks with a controlled geometry were prepared by moulding into disc-shaped preforms with 8-mm diameter and 0.5-, 1.0- and 1.5-mm thicknesses using conventional micro-injection moulding (µIM, Micropower 15, Wittmann Group, Austria) and a dedicated mould tool as shown in Fig. 3a and b [10]. Our previous work using differential scanning calorimetry (DSC) indicated that the difference in melting point between the discs and the virgin material was not significant [9, 10]. The shot size was set at 250 mg in weight corresponding to a stack of twelve 0.5-mm, six 1.0-mm and four 1.5-mm discs, respectively. Ten industry standard pellets were used to match the 250 ± 1 mg disc shot weight. This resulted in a 238 ± 5 mg shot size for pellet experiments, approximately 5% less than the disc experiments. The feedstocks were accordingly placed in the sonication chamber as shown in Fig. 3c. Table 3 summarises the shot sizes corresponding to the different feedstock shapes.

Fig. 3
figure 3

Visual details regarding feedstock: a an image of the mould tool for making disc preforms; b moulded disc-shaped PP feedstock; c images showing how the sonication chamber is filled with pellet and disc feedstock

Full size image
Table 3 Shot sizes weighed corresponding to different feedstock types. The values are given with standard deviation obtained from measurements taken from 15 different shots
Full size table

A total surface area calculation was carried out for pellet and disc-shaped feedstock stacks. This included the contact areas between individual discs and pellets, the side walls, sonotrode and piston contact points. The discs were assumed to have a perfect cylindrical shape with resulting in cylindrical stacks with a 6-mm height. These contact areas are presented in Table 4, where 0.5-mm-thick discs resulted in approximately 3 times the surface area of the 1.5-mm disc stacks.

Table 4 Total surface area approximations for pellet and disc feedstock shapes
Full size table

Due to the varied shape of industry standard pellets and their reduced total weight in the set shot size, an approximation had to be made for the surface area of compacted pellets assuming 100% packing density as opposed to the reality shown in Fig. 3c which is expected to be much lower. The following calculations and assumptions were made for approximating the total surface area of the pellets:

  1. 1.

    The total volume of the ten pellets was calculated based on the 237 mg mean weight and a 0.9 g/cm3 density resulting in a volume of 263 mm3. At this point, due to the compaction of pellets, it was assumed that the pellets formed a cylindrical shape of the same volume. The resulting height of the feedstock cylinder was found to be 5.25 mm with 8-mm base diameter (Fig. 4a).

  2. 2.

    CAD software (Solidworks Premium 2023, Dassault Systèmes, France) was used to create a 5.25-mm-tall cylinder, and this was sliced into ten shapes, representing compacted pellets (Fig. 4b). Then, the total surface area calculations were repeated three times with different pellet boundaries for quantifying the standard deviation, and the approximation resulted in 548 ± 20 mm2.

Fig. 4
figure 4

Depiction of the surface area calculation for pellets: a a CAD image and overall dimensions of compacted pellets; b a photograph representing compacted pellets

Full size image

2.3 Experimental details

u-µIM experiments with varying feedstock shapes were designed so that the effect of different feedstock interfaces on the consistency and viscosity of the polymer melt can be determined using thermal imaging and micro-feature replication measurements. The u-µIM process parameters used are given in Table 5. The settings were kept identical except for the sonication time (ts) for quantifying the impact of the feedstock shape in prolonged sonication.

Table 5 u-µIM process parameters used in the experimental runs
Full size table

The experimental campaign was based on a multilevel factorial design with 2 levels of ts (4 and 6 s) and 4 different feedstock shapes (pellets (P), 0.5-mm discs (D), 1.0-mm D and 1.5-mm D). This resulted in 8 runs, and 15 samples were moulded for each of the runs resulting in 120 mouldings providing statistical significance of the results (Table 6).

Table 6 Details of the experimental runs. Abbreviations: P, pellets; 0.5 D, 0.5-mm-thick discs; 1 D, 1-mm discs; 1.5 D, 1.5-mm discs
Full size table

2.4 High-speed thermal imaging

A flow visualisation mould tool previously designed and used was modified to be suitable for a vertical configuration as shown in Fig. 5a [3]. The upper half of the mould incorporates a sapphire window allowing infrared emissions to be collected using a 45° angled first-surface mirror. A high-speed infrared (IR) camera (FLIR X6540SC) was used to observe IR emissions from the molten polymer during filling. The frame size used for thermal imaging was 640 × 512 pixels, and the pixel size was calculated to be approximately 13.7 µm. An integration time of 50 µs was used for the acquisitions, and calibration procedures for the equipment are described elsewhere [4]. For each of the mouldings, mean and maximum temperature profiles were extracted. Differences occurring in the temperature profiles for different feedstock shapes and ts were analysed quantitatively by extracting 3 thermal imaging indicators. The first indicator is the maximum temperature (Tmax), measured as a single value from maximum temperature profiles (Fig. 5b). This parameter is significant, as it is a measure of the IF contribution to the temperature increase early in the process. Since the profile changes significantly from moulding to moulding, the curve was also integrated, and a single integral of maximum temperature profile (∫Tmax) was calculated. This parameter quantifies if the interfactial friction is sustained after a significant portion of the cavity is filled (e.g. t = 3.94 and t = 5.29 in Fig. 5c). Similarly, the integral of mean temperature (∫Tmean) was quantified from mean temperature (Tmean) profiles which provided a more generalised measurement of the melt temperature during filling. For both integrals, the lower integration limit was chosen where the highest increase in temperature was experienced (dT/dt). The curves were then integrated until the end of sonication, which corresponded to a characteristic decrease in temperature values as shown in Fig. 5b by the dashed blue line.

Fig. 5
figure 5

Details regarding thermal imaging: a image showing u-µIM machine and thermal camera system with flow visualisation mould opening; b maximum and mean temperature profiles recorded from a 0.5 D 4-s sonication moulding event, c thermal images showing the progression of the melt

Full size image

2.5 Filling and product assessment

The filling assessment of the microneedle patch products was carried out under a laser-scanning confocal microscope (LSCM) (LEXT–OLS4000, Olympus-EVIDENT Europe GmbH, Germany). The 3D data were acquired from a stitched 5479 × 5503 pixel area that covers the 5 × 5 microneedle arrays using a × 10 objective lens for capturing the conical shapes and surface features. Due to excessive scanning and data processing times, 3 mouldings per run were selected and characterised under LSCM. This resulted in 75 microneedle measurements contributing for each experimental run. The depth of the microneedle cavities was inspected under the LSCM system, and the average cavity depth was found to be 556 ± 30 µm [28].

3 Results and discussion

3.1 Analysis of maximum temperature data

Figure 6 shows the maximum temperature profiles overlaid and aligned for each process setting providing qualitative assessment of the heating effects. Nine profiles out of 15 were chosen randomly and used in the plots for showing differences clearly. The time information given on the x-axis were not synchronised with the thermal imaging acquisitions, and a delay of approximately 1 s was present. One of the striking findings from these graphs is that the pellet experiments showed significantly lower fall-off in maximum temperatures after the cavities are filled for longer sonication times (ts = 6 s), which typically occurs after t = 5 s. Another important observation is that disc experiments resulted in higher initial Tmax values well above 250 °C, whereas this value is between 200 and 250 °C for pellets. This is evidence that the discs’ orderly shape and stacking contribute to early plasticisation and reaching higher temperatures earlier compared to pellets. This can be attributed to the rapid disappearance of interfaces and their contribution to the plasticisation. The early rapid temperature increases also suggest VE’s dominance in the bulk of the feedstock, leading to higher temperatures earlier in the process and more effective conduction of heat and sonication for plasticisation.

Fig. 6
figure 6

Maximum temperature profiles for varying feedstock shape and sonication times

Full size image

Higher maximum temperatures after filling for the pellet experiments shown in Fig. 6 are contradictory to their total surface area provided in Table 4 where they have comparable values to 1.5 D: this feedstock shape having the lowest total surface area amongst the discs. However, less compaction density in the beginning of the process is evident for pellets (Fig. 3c), and they were gradually compacted and conveyed from the sonication chamber (sprue) to the runner, and the cavity. This was proven by a test where only 2-s sonication time was used for charactering the state of the pellets and surge of the melt. Figure 7 shows images of three identical moulding trials for pellets and 0.5-mm-thick discs. Figure 7a clearly shows that the pellets are compacted and fused together with the apparent surge of polymer melt towards the runner and into the gate. The cloudiness in the runner section and the sprue support the view that conveyance of a partially melted polymer towards the cavity is present. Figure 7b depicts the status of the 0.5-mm-thick discs where 1.0- and 1.5-mm discs also exhibited similar behaviour, with only a partial compaction or occasional fusing of the discs being observed. The higher temperatures seen after t = 5 s for the pellet experiments could suggest that conveyance of these partially melted interfaces could still act as heating sources later in the filling process.

Fig. 7
figure 7

Images showing the moulding tests carried out for 2-s sonication time. a Pellets; b 0.5-mm discs. 1.0- and 1.5-mm disc images are omitted for brevity as they showed similar behaviour as 0.5 D samples

Full size image

The statistical analyses for Tmax and ∫Tmax are depicted in Fig. 8. The data in Fig. 8a and b for the disc experiments show similar standard deviation (SD) and no impact of having a different number of defined interfaces on Tmax. The disc-shaped feedstock provides a very orderly stack with almost 100% packing density in the sonication chamber, and it can be concluded that the interfaces contribute to the heating and viscosity reduction early in the process, near the sprue and the sonotrode in the light of the high initial temperatures shown in Fig. 6. This results in the dominance of VH, and as a result, no pronounced effect of different interface levels is seen on Tmax. Although having random industry standard feedstock shapes, pellet experiments showed less SD and more repeatable Tmax values. However, this is a single pixel value coming from a single thermograph, and the ∫Tmax values provided in Fig. 8d can be a better indicator of heating and manifestation of random pellet shapes due to the significant spread in the data for the pellet experiments. Moreover, keeping the partially melted feedstock’s gradual movement towards the cavity, IF can be considered more dominant for the pellet experiments, and more desirable, especially considering the significantly higher integral values given in Fig. 8c and d. The conveyance of partially melted pellets towards the cavity and high ∫Tmax values is also evidence that some interfaces are still available as heating sources later in the process for the pellet experiments. Albeit the maximum temperature readings are very localised and should be analysed in conjunction with mean temperature statistics.

Fig. 8
figure 8

Interval plots depicting the impact of ts and feedstock shapes on Tmax and ∫Tmax. a Tmax data for ts = 4 s; b Tmax data for ts = 6 s; ∫Tmax data for ts = 4 s; ∫Tmax data for ts = 6 s

Full size image

3.2 Analysis of mean temperature data

Mean temperature profiles follow the trends seen in the maximum temperatures though with less visual significance (Fig. 9). This seems plausible since the mean temperature measurements were averaged and smoothed over individual thermal imaging frames, while the maximum temperature data is measured from a single pixel which makes it very localised. Generally speaking, the wider temperature profiles for the pellet experiments are apparent compared to the disc experiments. This is especially valid after t = 5 s and 6 s sonication times, and this follows the same trend in the Tmax data in Fig. 6 and the benefit in the conveyance of partially melted pellets towards the region of interest. The differences for the disc experiments are not directly detectable through these qualitative representations; however, the initial temperature peaks reached near t = 3 s early in the process for the disc experiments follow the trend in maximum temperature profiles.

Fig. 9
figure 9

Mean temperature profiles for varying feedstock shape and sonication times

Full size image

Quantitative ∫Tmean values are expected to provide the most representative insights regarding the overall heating occurred, and these are provided in Fig. 10. Clear evidence is present in the data for both 4- and 6-s sonication times that the pellet experiments resulted in increased heating by 17% and 9%, respectively compared to 0.5 D, this disc level exhibiting the greatest surface area available for interfacial friction (Fig. 10a and b). Hence, it is a noteworthy observation that although the pellets have ~ 3 times less surface area than for the 0.5 D experiments, they provide more heating in the region of interest which does not correlate to their random shapes or lower total surface area. This also strengthens the view that the significant surface area for discs (especially for 0.5 D) is consumed early in the process and might not have a significant impact on the microreplication behaviour. The pellets have 5% less shot size compared to the discs, and comparable total surface area to 1.5 D which supports the findings in gradual conveyance of partially melted interfaces which are available for the creation of more IF where the measurements were taken. Amongst the disc experiments, no clear trends were seen with varying disc thickness for ts = 4 s. On the other hand, a 9% reduction is apparent in the integral value between 1.5 D compared to 0.5 D. This supports the view that longer sonication times are favourable in the utilisation of interfacial friction which can only become dominant much later in the process. Figure 9b also provides evidence that the surface area differences amongst discs can only be detected when using longer sonication times.

Fig. 10
figure 10

Interval plots depicting the impact of ts and feedstock shapes on ∫Tmean. a ∫Tmean data for ts = 4 s; b ∫Tmean data for ts = 6 s

Full size image

3.3 Flow front velocity and microreplication efficiency

Flow front velocity (Vf) was calculated for each moulding event by employing a previously developed edge detection method along the vertical distance of the viewing window of thermal imaging [10]. The data given in Fig. 11 shows significantly slower flow fronts for pellet experiments compared to discs. This supports the result where higher Tmax was obtained from disc experiments early in the process (t = 3 s), which allows the flow front to move rapidly in the early stages of filling, and this behaviour is validated with higher Vf experienced for 0.5 D which had significantly more surface area compared to other shapes. Overall, the disc experiments provide a flat, well-defined interfacial heating for rapid heat generation resulting in an instantaneous increase in temperatures early in the process. Interfacial area is also not dependent on applied force or pressure for discs, whereas for pellets, the contact is gradual, and it is likely that the applied force will generate more interfaces and friction during compaction, similar to a cavitation effect [29].

Fig. 11
figure 11

Interval plot depicting the flow front velocities (Vf) for different feedstock shapes

Full size image

Considering 0.5 D, disc interfaces melt and disappear quickly leading to rapid temperature and viscosity reduction, resulting in significantly higher Vf (14.1 ± 3.6 mm/s). Further evidence of this rapid viscosity reduction in the disc experiments is seen in the thermographs taken from a cycle with 0.5-mm-thick discs given in Fig. 12. The thermograph recorded the highest temperature of 306.8 °C when the melt entered into the cavity (t = 3.70 s), which led to the “hot plumes” in the melt causing inconsistencies in the flow front progression. This is supported by the high SD in the Vf for 0.5 D experiments.

Fig. 12
figure 12

Thermographs depicting the rapid temperature increases when the melt enters the

Full size image

3.4 Microreplication efficiency

Figure 13 provides details regarding characterisation of microneedle height for different sonication times and feedstock shapes. The 3D LSCM images given in Fig. 13a and b demonstrate the best and worst samples amongst all 120 samples. Due to the differences in the flow front melt temperature, significant variations were characterised by partially filled needle geometries as shown in Fig. 13b and c. The microneedle profiles for the best sample shown in Fig. 13c show needles with 525–550 µm height, suggesting a replication efficiency over 90%. The statistics for average microneedle height provided in Fig. 13d were established from three different mouldings for each run or process condition totalling a sample size of 75 needles. One of the significant outcomes of this experiment can be seen as the advantage of using pellets which yielded the highest microneedle height for both 4 and 6 s ts. Specifically, the longer (6 s) sonication time resulted in a significant reduction in SD from 124 to 38 µm compared to other feedstock shapes. This is clear evidence that the conveyance of the partially melted pellet interfaces near the microreplication region of interest (ROI) under sustained sonication is beneficial for obtaining a desirable, zero-defect micro-featured product in u-µIM, through a combination of IF and VE heating sources. This result also underlines the importance of the wider maximum temperature profiles sustained throughout filling and packing, and higher values of associated integrals (∫Tmax) from the same data that could then be directly linked to microreplication. This is in agreement with Zou et al. where simultaneous plasticising and filling during ultrasonic micromoulding resulted in better moulding efficiency of miniature tensile bar components made from polypropylene [30].

Fig. 13
figure 13

Details regarding microneedle measurements: a rendered 3D LSCM image showing the microneedles from the best sample (P, 6-s sonication), dashed lines indicate the cross-section; b image for the worst sample (1.5 D, 6-s sonication); c microneedle profiles showing details regarding the measurement; d microneedle height statistics for different process settings. The error bars represent SD

Full size image

For the disc experiments, the data in Fig. 13d does not show significant differences with disc thickness: the microneedle heights and SDs are very similar for all three disc thicknesses and two sonication times. This is in correspondence with Tmax and calculated integrals (Figs. 7 and 9), where only subtle differences were present for the disc experiments. The comparable microneedle height and similar SDs also support findings where discs contribute to rapid temperature increase only in the early stages of melting and filling processes. However, the relatively high SD corresponds to a coefficient of variation of 7–9% in the disc experiments for ts = 6 s. One of the significant outcomes of this microreplication analysis is that the random shape or shot size of the pellets does not affect the microreplication results. On the contrary, the pellets resulted in the best microreplication experiments for prolonged sonication times. This signifies that, provided overload errors on the sonotrode are avoided which often seen in pellet experiments, pellets can be beneficial for directing a good combination of IF and VE onto the region of interest with the correct process parameters and will provide the best microreplication for miniature components.

4 Conclusions

This study presents a comprehensive characterisation of varied feedstock geometries within u-µIM and quantifies the influences of interfacial friction and viscoelastic heating on microneedle replication. This investigation utilised high-speed thermal imaging and advanced three-dimensional microscopy techniques for a thorough examination of these heating phenomena. The experimental work characterised the effects of industrial standard pellet-shaped feedstock, but also disc-shaped preforms, designed specifically for this experiment. Amongst a variety of process parameter and feedstock shape combinations, it was found that directing the contribution of interfacial friction near the microneedle features, in combination with viscoelastic heating after the cavity is filled, is essential for achieving the best microreplication behaviour [10]. The usage of more orderly and well-defined discs resulted in an early reduction of viscosity and heat generation, well before the melt reached the microneedle features which was less desirable for interfacial friction utilisation. The work fills a significant gap in understanding the process characteristics of u-µIM, by proposing new routes for widening the process windows through manipulating feedstock shape and directing the IF as a heating source for the best micro-product quality. Key outcomes of this research can be summarised as follows:

  • The process measurements showed more heat generation for pellet experiments as demonstrated with higher integral values compared to discs, however at the expense of less process consistency due to high deviations in the values. However, this inconsistency did not manifest itself on microreplication; on the contrary, the pellet experiments resulted in the best microneedle quality. Hence, it can be stated that the pellet feedstock shapes provide a good combination of IF and VE near the cavity or the ROI for prolonged sonication which is desirable for a zero-defect micro-featured product.

  • Disc feedstock shape provided better repeatability in heat generation and a potential to control the amount of total heat generated during u-µIM. This could be of interest to keep the overall heating under control when using active ingredients for microneedles or medical products that are sensitive to high temperatures or heat [31]. The calculated surface areas within discs with different thicknesses manifested themselves as an increase in the heating for longer sonication times; however, they had no noticeable impact on micro-feature replication.

  • Industry standard pellets provide a gradual plasticisation in u-µIM which could be desirable to convey a proportion of partially melted interfaces to the region of interest. In combination with prolonged sonication, this achieves a superior micro-feature replication capability in u-µIM.

Future work can be considered and undertaken for contributing to the state-of-the-art of u-µIM in conjunction with the results of this research:

  • Pre-compaction: Pellets were advantageous for directing the interfacial friction onto the moulding cavity in combination with viscoelastic heating in the work; however, their initial packing density can be improved (Fig. 3c). With a fixed sonotrode position and blank piston strokes, pellets can be compacted, and this can have significant improvements on the filling behaviour of micro-features. The packing density could also be improved by a bi-modal distribution for the pellet size.

  • Preform interfaces: The disc-shaped preforms can have textured surfaces on the milli-, micro- or nano-scales, which will provide more surface area and increase the contribution of interfacial heating. This could give the advantage of generating more heat, but with slower injection rates, and the usage of orderly disc stacks for better replication and process repeatability at the same time [5, 20].

  • Surface area quantification: The total surface area of the feedstock (pellets) can be quantified more accurately by techniques such as X-ray computerised tomography or microscopy [32]. Hence, more accurate computational models and the impact of different feedstock shapes onto heating can be studied.