A 3D printed microfluidic device for centrifugal droplet generation

1.1 Droplet microfluidics

Droplet microfluidics represents a class of microfluidic systems that use various droplet dispensing and manipulation mechanisms to isolate and control liquid-phase droplets of samples and/or reagents surrounded by a gas or an immiscible liquid (Li, 2008). Over the past two decades, droplet-based platforms have attracted growing interest in microfluidics because they enable precise and repeatable control of discrete volumes in the nanolitre and picolitre range. Depending on the application, droplets can be used as individual biochemical microreactors, microcapsules, or drug delivery systems. Droplet microfluidic devices enable rapid and repeatable routine laboratory operations, such as mixing, sorting, separating or merging droplets on a mass scale (Chen et al., 2021a; Sun et al., 2023; Jiang et al., 2023; Teo et al., 2020; Teh et al., 2008). More interestingly, droplet-based systems are finding increasing application in life science, such as molecular DNA analysis, protein research and cell culture (Ding et al., 2020; Li et al., 2018; Chen et al., 2021b; Shao et al., 2022; Feng et al., 2018; Yu et al., 2021). Nevertheless, the most fundamental feature offered by these devices is the generation of homogeneous droplets of a dispersed phase in a continuous phase.

This mechanism is applied to obtain various colloids, which in most cases are two- or multiphase emulsions, such as oil-in-water (O/W) and water-in-oil (W/O) (Doufène et al., 2019; Chen et al., 2022; Kamnerdsook et al., 2023; Lashkaripour et al., 2019; Hettiarachchi et al., 2021; Bouzetos et al., 2022; Whitesides, 2006).

1.2 Droplet generation techniques

Droplet formation techniques can be divided into passive and active generation. The passive technique uses hydrodynamic flow pressure or capillary forces (Chang et al., 2011; Hitzbleck et al., 2011) along with three geometrical configurations of the microfluidic system. T-junction, flow-focusing junction and co-flowing device (Doufène et al., 2019; Chong et al., 2016). For the first arrangement, droplets of the dispersed phase liquid are formed at the T-junction by shearing forces of the continuous phase liquid perpendicularly flowing. The flow-focusing approach uses a similar shearing force mechanism in the cross-shaped junction, which provides more precise control over the droplet distribution. The co-flowing solution is based on interception of the dispersed phase liquid stream from the inner channel by the continuous phase liquid in the surrounding channel. Therefore, the size of the droplets can be controlled by adjusting the flow rate, redesigning the geometry or changing the physical properties of the liquids (Chong et al., 2016). The impact of contact angles has been thoroughly investigated (Dreyfus et al., 2003; Tran et al., 2013). Microdevices with hydrophilic inner surfaces are optimal for the generation of emulsions with an aqueous continuous phase (O/W), whereas hydrophobic structures are more convenient when the continuous phase is oil (W/O). Passive droplet generation requires standard and commonly available laboratory equipment, such as syringe pumps, pressure controllers, external valves, ferrules and fluidic connectors (Yao et al., 2019), but it should be underlined that they must allow extremely precise control over flow rate. Second, the relatively high fluidic resistance of the tubing and fluidic capacitance generated by the compressibility of the liquid result in a slow response of the microfluidic system. For this reason, the equilibration of the flow of dispersed and continuous phases in the channels requires an initial run-in time, which can be counted in seconds or minutes. Last but not least, significant dead volumes of the dispersed phase are hard to avoid, leading to the waste of the liquid (Schuler et al., 2015).

In comparison, the active droplet generation technique is based on the delivery of external energy to provoke instability at the phase-to-phase interface and form a droplet. Due to a more manageable process, various active control solutions have been developed in the past decade and can be classified according to the applied physical phenomenon on electrical, thermal, magnetic and mechanical, as well as their combinations (Chong et al., 2016). However, the active droplet formation technique usually requires complex instrumentation and more sophisticated chip technology, which reverses the concept of lab-on-a-chip into a “chip-in-a-lab” solution. This approach may be acceptable for research tasks but is in contradiction to the general concept of miniaturised and portable devices for point-of-care applications.

1.3 Centrifugal droplet generators

One of the promising improvements toward complete miniaturisation of droplet-based devices and the necessary equipment is the use of the centrifugal method. The rotating disc technique, also known as laboratory-on-a-disc, has been applied in microfluidics for more than two decades as an accurate and relatively simple alternative to pressure-driven or electrokinetic fluid control methods (Tang et al., 2016). Centrifugal devices usually have the form of a flat disc, resembling a CD in shape and size, which contains a radial microfluidic system. When a disc is spun at a speed of up to thousands of rounds per minute (RPM), liquids are displaced toward the periphery of the device. This process depends on both the geometrical model of the microfluidic system and the combination of centrifugal, Euler, and Coriolis forces, which affect the liquid. The application of an adequate spin profile enables precise and repeatable flow control on the disc (Tang et al., 2016). Therefore, centrifugal devices have already found numerous applications in routine laboratory processes, as well as point-of-care applications, such as particle mass sorting (Morijiri et al., 2011), nucleic acid analysis (Schuler et al., 2016; Schaerli et al., 2009; Li et al., 2019), enzyme-linked immunosorbent assays (Thiha and Ibrahim, 2015) and the detection of biomarkers of tropical diseases (Hosseini et al., 2015).

The LOAD technique has also introduced significant improvements in droplet microfluidics, reducing some of the limitations of passive and active control methods. First, a complete fluidic system is usually integrated into a disc, including transport channels, valves and reservoirs for reagents and waste, so the adverse effects of resistance and fluidic capacitance known from the passive technique are minimised. Compared to the active droplet method, a centrifugal device requires relatively simple and easy-to-miniaturise instrumentation, which can be implemented in the form of a modular test station (Thiha and Ibrahim, 2015; Wang et al., 2020), standard laboratory equipment (Ding et al., 2022) or portable compact disc players (Doufène et al., 2019; Schuler et al., 2015). However, one of the challenges in the widespread adoption of LOADs today is their relatively complex technology. Until now, lab-discs have been mainly made of poly(methyl methacrylate) (PMMA) sheets using subtractive manufacturing methods (Doufène et al., 2019; Gorkin et al., 2010). PMMA layers are designed and fabricated individually and then laminated to obtain a sealed microfluidic structure. Such an approach is laborious, time-consuming and requires technological experience. Moreover, even small misalignments in the structurization or bonding of the layers may lead to disc imbalances that affect the repeatability of the centrifugal process.

Herein, a monolithic lab-disc for droplet generation, which was fabricated in a single additive manufacturing process, is presented. To the best of our knowledge, this is the first solution of a 3D printed device for centrifugal emulsification. The lab-disc model was prepared using CAD software and then the device was fabricated from a biocompatible polymer using the material jetting technique (also known as MultiJet or PolyJet) and post-processed according to the developed protocol. Finally, the disc was characterised in terms of mechanical and surface properties, and the parametric generation of oil-in-water emulsions was performed using a laboratory setup for centrifugal analysis.

2.1 Device design and operation protocol

The design of the device was inspired by the solution of PMMA disc for centrifugal step emulsification, developed by Schuler et al. (2015). The disc enabled fast and accurate generation of monodisperse W/O droplets, playing the role of miniature reactors for recombinase polymerase amplification (RPA) and absolute quantification of DNA. The solution of the centrifugal droplet generator can be implemented for other chemical, biochemical, or pharmaceutical applications, and this was the starting point for our design.

The device has a shape and dimensions similar to those of a standard audio compact disc (CD), and its 3D model was developed using CAD software. The disc was created as a cylinder with a height of 1.4 mm and a diameter of 12 cm. A central hole, which has 15 mm in diameter, is used to mount the disc in the spinning unit (as described below). The device contains four symmetrically arranged microfluidic units for droplet generation. Each unit consists of an input chamber to load the dispersed phase, a connecting channel ending with a nozzle for droplet production and a bottom output chamber to store the continuous phase and collect the generated droplets (Figure 1). The height of the input and output chambers (h₂) is equal to 600 µm, and their volumes are approximately 215 and 230 µL, respectively. The diameters of the connecting channel are 55 mm in length, 1,000 µm in width, and 200 µm in height (h₁). The end part of the channel narrows into a nozzle (1,000 µm in length and 200 µm in height) that opens to the terrace (200 µm in width t), where droplets are formed in the output chamber. The width of the nozzle (w) varies in every unit (200, 300, 400 and 500 µm, respectively), and this differentiation was implemented to verify the influence of the geometric parameters of the nozzle on the diameter of the generated droplets, as investigated also in other works (Schuler et al., 2016; Schuler et al., 2015). The top layer of the disc contains 1.0 mm diameter inlets for loading liquids and venting holes for equal pressure distribution. Alphanumeric descriptions of the units are extracted in the top layer to a depth of 0.2 mm.

The device uses a centrifugal step emulsification mechanism, which is based on the formation of droplets at the outlet of the nozzle as a result of the abrupt change in capillary pressure. In comparison to non-centrifugal droplet generators, this solution is less susceptible to variations of pressure or flow rate and requires only one microfluidic channel. More details on the benefits and application of this technique can be found in the literature (Schuler et al., 2016; Schuler et al., 2015). In our device, the centrifugal emulsification workflow can be described as follows: the output chamber is filled with a continuous phase using a standard laboratory pipette; the dispersed phase is pipetted into the input chamber; the centrifugal process is started; the dispersed phase is transported along the channel and droplets are generated in the output chamber; and the emulsification may be stopped when either the volume of the dispersed phase is low or the output chamber is filled with the emulsion. After the process, the droplets are monitored under a microscope, and the emulsion is collected.

2.2 Materials and reagents

A material jetting printer (ProJet 3510 SD, 3D Systems) was used to fabricate the monolithic device. Post-processing was carried out using a laboratory drying oven (CLN 15, Poleko) and a laboratory hot plate (SD160, Stuart).

The disc was printed with transparent and biocompatible photopolymer construction material (VisiJet M3 Crystal, 3D Systems) and wax-based support material (VisiJet S300, 3D Systems), which were purchased from a local 3D Systems distributor (3D Lab, Poland). The chemicals used for post-processing included canola oil (Olej Kujawski, Bunge Polska Sp. z o.o.) and a standard detergent solution (Ludwik, INCO S.A.) purchased from a local market (Carrefour). Deionised water (DI) with a resistivity of 15 MΩ·cm was produced in a laboratory water purification system (Kuna System Sp. z o.o.). ACS-grade reagents used for disc treatment, namely, methanol (≥99.8%, CAS: 67–56 - 1), acetone (≥99.5%, CAS: 67–64 - 1) and isopropanol (≥99.5%, CAS: 67-63-0), were purchased from Sigma-Aldrich Poland.

Droplet emulsions were prepared using fractionated palm oil (Palm Olein Cooking Oil, Avena) purchased from a local grocery store and freshly produced DI water. No emulsifiers or other additives were used in the emulsification process to make it simpler and more repeatable.

2.3 Device fabrication

The 3D model of the device was converted to a stereolithography format (STL), which was imported directly into the printer firmware. The model was flat-orientated in the centre of the build platform and then sliced. The printing resolution was set to XHD (750 × 750 × 1600 DPI for the XYZ axes) and the rest of the parameters were set to default. The printing process took approximately 2 h, excluding the warming and homing steps.

The post-processing protocol scheme is illustrated in Figure 2. First, the structure was detached from the build platform using a spatula and placed in a drying oven at 65°C for 1 h to melt the support material (wax). The structure was then immersed in a glass beaker filled with hot oil for 15 min to dissolve the remaining support material. The interior of the microfluidic system was flushed with hot oil using a pipette. The device was then submerged in a detergent solution, and the microfluidic units were flushed using the pressure of the pipette. Finally, the device was thoroughly rinsed in DI water and dried with compressed air.

2.4 Experimental setup

The spinning experiments were carried out using a laboratory setup for centrifugal analysis, comprising a brushless direct current (BLDC) motor (HF-KP43, Mitsubishi) with a screw rotor adapter, a BLDC motor controller (Melservo, Mitsubishi), a non-contact digital tachometer (RPM 82, Multimetrix) located above the periphery of the device, an illuminator (a standard compact fluorescence tube), a high-speed digital camera (piA640-210gc, Basler) placed on movable arm post, a data acquisition card (NI DAQ, National Instruments) receiving signals from the tachometer, and a standard personal computer with display (Figure 3). A dedicated control software (LabVIEW, National Instruments) enabled setting the desired spin profile of the BLDC motor, acquiring a video signal from the camera, synchronising snapshots with a tachometer signal, and saving the video sequence. During the initial tests, a micrometre screw with a precision of ±10 µm (SPW690, Mitutoyo) was used to measure disc deflection.

2.5 Data analysis

After each centrifugation process, the disc was gently removed from the spin rotor and placed under an optical microscope (SMZ-168, Motic) equipped with a digital camera (Moticam 5, Motic) and a microscope camera lens (0.45× C-mount, Motic). Microscopic images of the emulsions were saved as BMP files and analysed using software developed in LabVIEW, which was applied to detect circular objects and calculate the dimensions and volume of the droplets. The results of the analysis were saved in a text file and visualised in Origin (OriginLab).

3.1 Device characterisation

The real dimensions of the main functional elements of the developed device were compared with their nominal values. Three copies of the disc were prepared to obtain statistical data. Microscopic images of the elements were recorded with a digital camera and analysed using image processing software (DLTC CamViewer, Delta Optical). Each element was measured at three locations, and the mean value, standard deviation and percent error were calculated. The results of the investigation are listed in Table 1 and illustrated in Figures 4(a)–(d). The real values differed from the designed model mainly due to limitations of the printer resolution, as expected (Walczak and Adamski, 2015). The percent errors are repeatable from disc to disc, so calibration curves may be prepared to obtain more accurate dimensions of the elements in the future.

Another study was devoted to determining the deflections and mechanical robustness of the disc during the spinning process. A dummy disc was mounted in the laboratory setup for centrifugal analysis with a spindle of a micrometre screw gauge that slightly touched the periphery of the device, where the highest deflections were expected [Figure 4(e)]. The maximum amplitude of the vertical deflection at the edge of the disc did not exceed ±30 µm for 600 RPM. In another experiment, the disc was spun to the maximum settable rotation speed of 3,000 RPM for 5 min, and no damage was observed to the device.

The next experiment involved chemical treatment of the disc prior to the emulsification process. As mentioned above, the generation of O/W droplets is improved when the inner surface of the microfluidic system is hydrophilic because it promotes repeatable flow of the continuous (water) phase. However, the wettability of most polymeric surfaces without additional treatment is usually low. The residuals of the supporting wax material used in the inkjet process were also expected to degrade the hydrophilicity of the surface. Therefore, chemical pretreatment trials were performed along with contact angle measurements.

Several disc duplicates were fabricated and sacrificed for destructive tests. The microfluidic units were filled with different organic solvents commonly used in microengineering for surface treatment, including methanol, acetone and isopropyl alcohol (IPA). One device was left untreated for reference. After 10 min, the solvents were removed from the microfluidic units using compressed air [Figures 4(f)–(g)]. The upper layer of the input and output chambers was then removed using a scalpel. Finally, the static contact angle (CA) of the inner surface was measured using the sessile drop method. A drop of distilled water was distributed on the surface of the sample, and the image was captured and analysed using a dedicated software [Figure 5(a)]. The procedure was repeated 3–4 times in other areas of the device. The results of the statistical CA analysis [Figure 5(b)] confirmed that the surface of the untreated structure was slightly hydrophobic (mean CA = 96.3°), whereas treatment with methanol or acetone increased CA to 115.9° and 103.1°, respectively. It was concluded that only the application of isopropanol significantly improved the wettability inside the microfluidic units. Thus, in further tests regarding droplet generation, only two versions of the discs were investigated: untreated and IPA-treated.

3.2 Centrifugal droplet generation

To prepare the optimal spin profile for the generation of O/W droplets, the burst speed values for oil and water were determined at first. In summary, when the rotation speed is increased to the burst speed, the resultant force imposes movement of the liquid to another part of the microfluidic system. In this case, the burst speed was determined for both the oil and the water samples located in the input chamber.

The input chamber was filled with 100 µL of liquid and the output chamber was left empty. The disc was then spun with a profile from 50 to 1,800 RPM with an increment of +100 RPM every 10 s. The disc was continuously monitored, and the burst speed value was noted when the liquid was moving toward the output chamber. The results of the burst speed tests, including the chemical pretreatment of the disc and the width of the nozzle, are summarised in Table 2. It was concluded that for the oil sample, the influence of the nozzle size on the burst speed is practically negligible for the more hydrophobic device, whereas for the water sample, an analogous result was obtained for the more hydrophilic structure. On the basis of these results, three of the most promising spin profiles were applied in a further study.

Oil-in-water droplets were generated in IPA-treated and untreated LOAD devices. The emulsification protocol was carried out as follows:

• the disc was mounted on the rotor using a screw adapter;

• 100 µL of DI water was loaded into the output chamber;

• 100 µL of palm oil was loaded into the upper part of the input chamber, paying attention not to provoke sample flow into the channel; and

• the disc was spun according to the selected spin profile.

All spin profiles started with a relatively slow rotation speed of 300 RPM for 10 s, promoting the equal distribution of oil and water samples in the peripheral area of the chambers. Subsequently, the speed was elevated to 1,000 RPM for 3 s, to initiate the oil burst. In the main step of emulsification, the speed was changed repeatedly 20 times between 300 RPM and one of three values of the droplet generation speed v_g, which were 500, 700 or 1,000 RPM, for 5 and 1 s, respectively. Finally, the disc was spun at 1,000 RPM for 1 s, to suppress the backflow of the emulsion into the channel. After centrifugation, the device was detached and the images of the output chamber were recorded and analysed under an optical microscope equipped with a digital camera (Figure 6).

The results of the statistical analysis of the emulsions produced on the IPA-treated disc using 200, 400 and 500 µm wide nozzles show (Figure 7) the relatively high homogeneity of the droplets with a similar volume range of 7–10 µL from the first to the third quartile (boxes) and a similar scope of minimum and maximum values (whiskers). However, the result for the 300 µm wide nozzle shows that the formation of droplets in this configuration was more chaotic. The influence of the droplet generation speed v_g on the volume was concluded to be negligible below the value of the burst speed, providing a practically homogeneous emulsion. The mean droplet volume decreased significantly with increasing rotation speed.

Unlike the IPA-treated disc, the emulsification results for the untreated device were considerably less repeatable (Figure 8), especially for smaller nozzle widths and lower spin speeds. Monodisperse droplets were produced only for the 500-µm nozzle, and a rotation speed of 1,000 RPM. The results indicate that the interfacial stability of the untreated disc is significantly lower than that of the IPA-treated device.

4.1 3D Printed centrifugal devices

One of the main innovations in this work is the fully monolithic construction of the lab-disc, which was fabricated using the material jetting process. The literature review shows that there are only several published examples of 3D printed devices for centrifugal technique, and to the best of our knowledge, none of them have been used for droplet generation so far.

One of the most widespread and low-cost 3D printing techniques is fused deposition modelling (FDM). Wang et al. (2020) applied FDM to develop a centrifugal jet mixer made of biodegradable polylactic acid (PLA). Two arrays of micronozzles were used to mix DI water with dye in less than 5 s. Another PLA-based 3D printed device was reported by Cristaldi et al. (2021). A pump-free spiral reactor-in-a-centrifuge was fabricated on a cylindrical PLA scaffold also using FDM in less than 6 h. The device was applied to produce liposomes and silver nanospheres. Zhang et al. (2019) applied FDM to develop a centrifugal microfluidic platform for the automatic detection of trace oil pollution in test samples. The device used an optical detection method based on ultraviolet-range excitation of the oil sample and measurement of the emitted fluorescence light in the visual range. The centrifugal platform contained separate microfluidic chambers for buffer, eluent, sorbent, sample and detection. The platform was sealed with an elastic membrane and then assembled with a valving structure that contained a plunger and a radial spring. The valve structure moved along the platform according to the centrifugal force, allowing the plunger to press on the valving membrane. The microfluidic platform, the valving structure, and some other elements of the system were made of PLA.

It should be noted that the main advantages of the FDM technique come from the relatively low cost of the printer and filament, the short fabrication time and simple post-processing. However, the resolution of printed details is often too low for devices that contain microscale parts. Therefore, other rapid prototyping techniques have been investigated to develop centrifugal microfluidic devices.

Ding et al. (2022) presented a monolithic LOAD for rapid multiplexed molecular detection of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). An isothermal amplification assay was implemented for the simultaneous detection of virus nucleoprotein and envelope genes. The structure was made of methacrylate-based resin using the stereolithography (SLA) technique, which provides a detail resolution of 50 µm and a smooth surface. Compared to the FDM technique, SLA requires post-processing, which in this case involved immersion in IPA, an ultrasonic bath in acetone and washing with DI water. Kim et al. (2016) reported a hybrid lab disc with individually addressable 3D printed diaphragm valves for quantification of the level of prostate-specific antigen in whole blood. The bulk of the disc was made of polycarbonate using a CNC milling process, the valve was made of elastic membrane and a pushpin with lock key was fabricated using inkjet printing.

In the solution described here, an inkjet or more specifically a multijet process, was applied to develop a monolithic centrifugal device. The principle of the inkjet technique is based on the ejection of droplets of photosensitive polymer from the printing head, followed by the UV photocuring process. This technique ensures high accuracy and enables the fabrication of multi-material structures, which are expected features in the development of microfluidic devices (He et al., 2016; Niculescu et al., 2021; Gale et al., 2018). Due to the relatively large area of a build platform, it also allows for the fabrication of devices with wide planar dimensions, like those described here lab-disc. Material jetting has already found various applications in microfluidics, such as microvalves (Keating et al., 2016; Walczak et al., 2017), DNA chips (Walczak et al., 2018) and microdevices for plant study (Walczak et al., 2019).

Unlike other 3D printing techniques, the support material used in the inkjet process prevents microfluidic structures from deflection, enabling the fabrication of high-resolution microfluidic devices. However, the main challenge is the efficient and thorough removal of the support. The post-processing procedure depends on the design and expected features of the structure and, therefore, must be individually tailored to the model of the device. In this case, the support wax material was simply melted in an oven and then in an oil bath. Additionally, the microfluidic system was washed with a detergent solution and DI water to remove residuals of the support material. The fabrication protocol was optimised after a series of experiments.

4.2 Droplet generator model and results of emulsification

As mentioned above, the droplet generator model applied here was inspired by the PMMA disc solution for digital droplet recombinase polymerase amplification (ddRPA) (Schuler et al., 2015); however, several modifications were implemented and had an impact on the properties of the 3D printed device, which are addressed and discussed below.

First, the properties of PMMA are commonly known and predictable, whereas here the monolithic lab-disc is made of a relatively new biocompatible polymer. The material is being investigated in various microfluidic applications (Walczak et al., 2017; Walczak et al., 2018; Walczak et al., 2019), so the impact of chemical pretreatment on surface properties and the optimal device preparation protocol were studied.

Second, the operation protocol of the mentioned ddRPA disc was quite complex and required sequential loading of the continuous phase into the input chamber, followed by translocation of the liquid to the output chamber during spinning, stopping the disc for loading a dispersed phase into the input chamber and finally running another spin process to generate droplets. In the presented solution, emphasis was placed on simplifying the operation protocol, so both phases were simultaneously loaded prior to a single spin process of droplet generation.

Third, the ddRPA technique requires the formation of highly monodisperse droplets for the proper quantification of nucleic acids (Schuler et al., 2015). In our case, the homogeneity of the mixture was not so crucial, as the target application of the monolithic centrifugal device was the repeatable production of precisely diluted essential oils for application in oral cavity microfluidic dispensers and miniaturised detection platforms, which have currently been under scrutiny in other works (Kojic et al., 2022).

It was observed that the stability of the oil-in-water emulsion produced was distorted if the disc had not been handled gently after centrifugation. Spontaneous merging of droplets is, however, normal, as the homogeneity of the emulsion decreases over time due to interface tensions generated between the phases, which originate from coalescence, flocculation, sedimentation or Ostwald ripening. Therefore, appropriate surface-active agents may be applied to prolong the emulsion monodispersity. As the consumption of surfactants in the form of food additives today has opened a discussion on their indirect toxicity and role in a growing number of allergic and autoimmune diseases (Venhuis and Mehrvar, 2004; Csáki, 2011), in this solution, emulsifiers were not used and only pure oil and water samples were applied. The addition of active agents, e.g. surfactants, may be considered if the homogeneity of the emulsion must be improved.

The final remark is related to the limitations of the applied 3D printer, which allowed us to fabricate nozzles with a minimum width of about 200 µm, which is almost twice the value used in the ddRPA PMMA device (Schuler et al., 2015). However, rapid progress in additive manufacturing techniques is expected to soon increase the available resolution of printed microfluidic details to or even above the capabilities of commonly applied subtractive technology (Mao et al., 2017).