**1. Introduction**

Three-dimensional (3D) printing has gained great importance for rapid prototyping of microfluidic devices in the past decade, since it allows the fabrication of microfluidic chips on the laboratory scale, with the possibility to test a great variety of different chip designs early on in the development process [1–4]. Different printing technologies have been examined for 3D printing of microfluidic devices, such as stereolithography (SL), inkjet printing, multi-jet printing, two-photon polymerization, suspended liquid subtractive lithography and fused deposition modeling (FDM) [2,5–8]. On the laboratory scale these 3D printed microfluidic devices have been used in a great variety of applications: for mixers, droplet or gradient generators or active components like valves or pumps [9–12].

However, one major issue for most rapid prototyping methods in microfluidics, including 3D printing, is the conversion of a laboratory prototype to an industrial scale product [13]. 3D printing of microfluidic chips has so far been mainly realized using stereolithography printing, since it remains the method of choice for high-resolution 3D printing with affordable machinery. However, SL usually requires photocurable resins (mostly acrylic or epoxy based) which are strongly cross-linked thermosets, rendering them unsuitable for industrial replication processes like injection molding or hot embossing, which require thermoplastic materials. Due to the high surface-to-volume ratio in microfluidics, the chip material plays a major role. Therefore the transfer between lab-processes and industrial scale comes with a change in the final system behavior, which is a major issue [13]. One option to solve this problem is the use of 3D printing methods, which can process industrially relevant thermoplastic polymers already on the laboratory scale. FDM is such a 3D printing method in which a thermoplastic filament is melted, extruded through a nozzle and solidified by cooling. Furthermore, FDM is interesting since it is capable of multimaterial printing [14]. The complexity of FDM printed systems can be even extended by integration of components like membranes or electrodes by pausing the process, integrating the external components of choice before continuing the print (so called print-pause-print) [14]. Using FDM printing a variety of microfluidic concepts like mixers or chemical reactionware have already been realized [15,16]. Recently FDM printing has been used to fabricate simple channel geometries with sub-100 µm [17].

In theory, a wide range of thermoplastic materials can be processed using FDM, but only few materials have been studied. Most microfluidic chips are printed using poly (lactic acid) (PLA) or acrylonitrile-butadiene-styrene (ABS). FDM printing of cycloolefin copolymers (COC), thermoplastic urethane (PLU) or polypropylene being a notable exception [15,17]. Polymethylmethacrylate (PMMA) is one of the most important thermoplastic materials in microfluidics due to its high optical transparency and low autofluoresence combined with a high biocompatibility [18]. Furthermore, it is a rather hydrophilic material, making it interesting for capillary-driven microfluidic systems. However, PMMA is usually structured using industrial scale polymer structuring like injection molding or hot embossing [19,20]. Rapid prototyping of PMMA has been mainly conducted using subtractive processes like laser structuring or high-precision milling, as well as laboratory-scale imprinting technologies, such as solvent replication or room temperature imprinting [21–24]. PMMA prepolymers consisting of the monomer methylmethacrylate (MMA) and the polymer PMMA have been used for replication of microfluidic PMMA chips with high resolution from a variety of different master structures like polydimethlysiloxane (PDMS) or stainless steel [25,26]. Recently, we have introduced Liquid PMMA, a photocurable PMMA prepolymer which can be structured with tens of micron resolution using direct lithography [27]. However, all these rapid prototyping processes are only capable of fabricating open 2.5-dimensional microfluidic structures, which need to be closed during a subsequent bonding step. Furthermore, the integration of microstructures and biomolecular patterns is becoming increasingly important. Protein patterns with specific shapes and patterns are capable of inducing specific cellular responses and are important for the study of cellular behavior. Structured biochemical functionalization by means of photobleaching was established in the last decade and was already used for different substrate materials, like functionalized glass slides, polymer films and cellulose paper [28–30].

PMMA has so far not been studied as a material for FDM printing of microfluidic chips, which, however, could allow the direct printing of embedded microfluidic chip structures. In this work, we demonstrate that microfluidic PMMA chips can be directly printed using FDM with a minimum channel width of 300 µm. To ensure a high optical transparency in the region of interest, we evaluated two strategies: direct FDM printing on the print bed and printing on top of a commercial PMMA substrate. We further evaluate the influence of the nozzle distance on the transparency of the printed PMMA. Further, we show that high-resolution microstructures like microscale line-patterns can be integrated within the printed PMMA chip and that the microchannels can be easily functionalized with biomolecules. The FDM printing process allows the fabrication of a fully functional embedded microfluidic chip within 1 h, thus demonstrating that an industrially relevant thermoplastic polymer material can be structured using 3D printing. This effectively allows a direct transfer of results gained in the laboratory to an industrial scale.

### **2. Materials and Methods**

### *2.1. Printing Materials*

PMMA filament was purchased from materials4print, (Bad Oeynhausen, Germany). Methylmethacrylate (MMA) technical grade was purchased from VWR, (Darmstadt, Germany). Phenylbis(2,4,6-trimethylbenzoyl)phosphine oxide (BAPO), biotin (5-fluorescein) conjugate (F5B), streptavidin-Cy3 in buffered aqueous solution (STV-Cy3), phosphate buffered saline (PBS) and 2,20 -azobis(2-methylpropionitrile) (AIBN) were purchased from Sigma Aldrich, (Taufkirchen, Germany). PMMA substrates were purchased from Röhm GmbH (Darmstadt, Germany).

### *2.2. Fused Deposition Modeling*

All designs were created using Autodesk Inventor Professional 2019 and exported as STL files. The STL files were imported into PrusaSlicer-2.1.0-rc for the slicing process. The FDM printer *Prusa i3 MK3S* (Prusa Research, Prag, Czech Republic) was used for printing. The printing parameters have been optimized to 3D print microfluidic channels with a minimum channel width of 300 µm. The parameters of the printing process were as follows: layer height: 50 µm (first layer: 100 µm), infill: 100%, printing speed: 30 mm/s and nozzle temperature: 230 ◦C. The bed temperature was increased significantly compared to other commercial filaments like PLA to 115 ◦C for the first layer to reduce warping effect, i.e., delamination from the print bed. During the print, the bed temperature was lowered to 110 ◦C for all following layers. A 45◦ angle was set for the infill orientation. Infill/contour overlap was set to 25% to reduce air gaps between the printed features. All parts were printed with a 0.4 mm nozzle. For high transparency of the printed microfluidic chips, the *z*-axis distance to the printing bed was reduced by doing a first layer calibration. The nozzle was moved 100 µm closer to the print bed compared to the standard printing distance. The specified standard *z*-axis distance of the nozzle is usually adjusted until the polymer sticks nicely to the print bed and is only slightly squished. Reducing the *z*-axis nozzle distance even more than specified results in broader strands, which make the FDM printed PMMA appear more homogenous and transparent (see Section 3.1). To reduce sagging during the 3D printing of embedded channels, the material extrusion was reduced by 60% while bridging the channel structures (bridge flow ratio: 0.4). By keeping the printing speed at 30 mm/s the bridging PMMA strands are stretched upon deposition which prevents excessive sagging. To evaluate different channel geometries, a series of embedded channels with different cross-section geometries were printed and compared to their original CAD geometry. The accuracy of the printable channel widths and heights were evaluated by printing open and embedded channels with varying channel widths from 0.2–1 mm and compared to the original CAD dimensions. Evaluation of the channel sizes and geometries was executed using a light microscope of type VHX 6000 (Keyence, Osaka, Japan).

To obtain microchannels with an improved transparency in the region of interest, the channels were printed directly on commercial PMMA substrates with a thickness of 2 mm. A similar strategy has been previously described for the fabrication of channels in PLA [31]. The channels were designed to be open at the bottom side, which is printed directly on the substrate. The printed PMMA structure bonds to the PMMA substrate to form sealed microfluidic channels. The *z*-axis printing height of the chip was adjusted during slicing to allow printing on the 2 mm PMMA substrate (z-offset: 2 mm). After heating and calibration, the printer was shortly paused to align and stick the substrates to the print bed.

### *2.3. PMMA Precursor Preparation*

A thermally polymerized PMMA precursor was prepared by adapting a protocol described by Qu et al. [25]. The thermal initiator AIBN was dissolved in technical MMA (0.3 mg/mL) and the mixture was polymerized by heating the mixture to 93 ◦C within 20 min. After 15 min at this temperature

the polymerization was stopped by cooling to room temperature with an ice bath. The mixture was blended with 1 wt% BAPO.
