Fabrication of a 3D-Printed Porous Junction for Ag|AgCl|gel-KCl Reference Electrode

Fused filament fabrication (FFF) is a 3D printing method that is attracting increased interest in the development of miniaturized electrochemical sensor systems due to its versatility, low cost, reproducibility, and capability for rapid prototyping. A key component of miniaturized electrochemical systems is the reference electrode (RE). However, reports of the fabrication of a true 3D-printed RE that exhibits stability to variations in the sample matrix remain limited. In this work, we report the development and characterization of a 3D-printed Ag|AgCl|gel-KCl reference electrode (3D-RE). The RE was constructed using a Ag|AgCl wire and agar-KCl layer housed in a watertight 3D-printed acrylonitrile butadiene styrene (ABS) casing. The novel feature of our electrode is a 3D-printed porous junction that protects the gel electrolyte layer from chloride ion leakage and test sample contamination while maintaining electrical contact with the sample solution. By tuning the 3D printing filament extrusion ratio (k), the porosity of the junction was adjusted to balance the reference electrode potential stability and impedance. The resulting 3D-RE demonstrated a stable potential, with a potential drift of 4.55 ± 0.46 mV over a 12-h period of continuous immersion in 0.1 M KCl, and a low impedance of 0.50 ± 0.11 kΩ. The 3D-RE was also insensitive to variations in the sample matrix and maintained a stable potential for at least 30 days under proper storage in 3 M KCl. We demonstrate the application of this 3D-RE in cyclic voltammetry and in pH sensing coupled with electrodeposited iridium oxide on a gold electrode. Our method offers a viable strategy for 3D printing a customizable true reference electrode that can be readily fabricated on demand and integrated into 3D-printed miniaturized electrochemical sensor systems.


Introduction
Due to the increased demand for decentralized health, food, and environmental monitoring, there is a growing interest in the development of low-cost miniaturized sensors [1]. In particular, miniaturized electrochemical sensors, such as potentiometric or voltammetric sensors, are attractive methods for decentralized sensing applications due to their sensitivity, selectivity, reproducibility, ease-of-use, and portability [2]. Advances in digital additive manufacturing technologies such as 3D printing have also led to improvements in the accessibility and cost-effectiveness of these miniaturized, integrated electrochemical sensors [3][4][5]. With 3D printing, custom miniaturized electrochemical sensors can be rapidly prototyped and fabricated reproducibly on demand with inexpensive equipment and materials [5].
A major challenge in the development of integrated electrochemical sensor systems is the miniaturization, manufacturability, and stability of the reference electrode. In two-and three-electrode electroanalytical setups, the reference electrode maintains a constant potential to control and/or was purchased from Ajax Finechem Pty. Ltd (Thermo Fisher Scientific, Inc., Waltham, MA, USA). Oxalic acid dihydrate was purchased from Himedia (Mumbai, India). Isopropyl alcohol and NaCl were purchased from RCI Labscan, Ltd (Bangkok, Thailand). Acetone and absolute ethanol were purchased from Pharmaco-Aaper (Greenfield Global USA Inc., Brookfield, CT, USA). Analytical grade glacial acetic acid, 70% HNO 3 , 36% HCl, and NaOH pellets were purchased from Macron Fine Chemicals (Avantor, Randor, PA, USA). Granulated agar was purchased from BBL Difco (Becton, Dickinson and Company, Franklin Lakes, NJ, USA). SU-8 ink (SU-8 2002) was purchased from MicroChem (Kayaku Advanced Materials, Inc., Westborough, MA, USA). Polyethylene naphthalate (PEN) film (Teonex Q65HA) was purchased from Dupont Teijin Films (DTF Singapore, DuPont Company Pte Ltd., Singapore). High-purity gold coin (99.9%) was purchased from a jeweler's shop and analyzed with an XRF (Horiba, Ltd., Kyoto, Japan) for purity.
Microwave hydrolysis and synthesis were done using a laboratory microwave synthesizer flexiWAVE (Milestone Srl, Sorisole (BG), Italy). Centrifugation was done using Sorvall Legend X1R (Thermo Fisher Scientific, Inc., Waltham, MA, USA). Ultrasonic dispersion of nanoparticles was conducted in a Q500 ultrasonic processor with a cup horn attachment (Qsonica, Newtown, CT, USA).
Potentiostatic electrochemical impedance spectroscopy (EIS) and galvanostatic electrodeposition were conducted using a Gamry Interface 5000E potentiostat/galvanostat (Warminster, PA, USA). Open-circuit potential measurements and voltammetric analysis were conducted using a Rodeostat Model 10 V Potentiostat with a seven-channel multiplexer installed (IORodeo, Pasadena, CA, USA). The potentiostat was controlled through a Python-based interface (potentiostat package version 0.0.4). Electrochemical data were filtered with the Savitzky-Golay filter for data smoothening [40]. Unless otherwise indicated, potentials were measured with respect to a saturated calomel electrode (SCE) from Shanghai INSEA Scientific Instrument Co., Ltd. (Shanghai, China). pH measurements and ORP (oxidation/reduction potential) readings were obtained using a Eutech pH 700 pH meter (Thermo Fisher Scientific, Inc., Waltham, MA, USA).

Fabrication of 3D-RE Components
The model of the 3D-printed reference electrode (3D-RE) casing was designed in AutoDesk 123D Design software. The 3D-RE casing incorporated a 7.0 mm × 7.5 mm × 3.5 mm cavity to hold the Ag|AgCl wire and the internal gel electrolyte. A 3.5 mm × 3.5 mm opening in the cavity facilitates the contact of the internal gel electrolyte and the sample. Surrounding the opening is also a 7.0 mm × 7.5 mm × 0.80 mm recess to fit a 3D-printed porous junction. The 3D-RE casing and the junction are 3D-printed separately using different 3D printing parameters and postprocessing.
The design files (STL) were exported to FlashPrint version 3.21.0 software to control the G-code slicing and 3D printing parameters of the 3D model. Unless otherwise indicated, 3D models were sliced with a perimeter shell count of 2 and 100% infill, which ensures that the filament is deposited throughout the entire internal volume of the model with minimal G-code generated gaps within the internal structure. The 3D-RE casings were sliced with a 0.18 mm layer height and an extrusion ratio (k) of 1.10. The 3D-printed junctions, on the other hand, were vertically oriented and sliced with variable extrusion ratios (k = 0.85, 0.95, and 1.05) and a fixed layer height of 0.40 mm. G-code files were then exported to the 3D printer. 3D printing was conducted with an ABS filament using a 0.4 mm nozzle, a nozzle temperature of 215 • C, platform temperature of 105 • C, base print speed of 30 mm/s, and retraction length of 1.3 mm.
After 3D printing, the RE casing was postprocessed using acetone vapor smoothening. Here, approximately 10 mL of acetone was heated to 80 • C in a loosely covered 100 mL beaker. The casings Chemosensors 2020, 8, 130 5 of 20 were suspended in the beaker for approximately 5 min and were then allowed to cool and solidify for 30 min prior to further handling.
The 3D-printed junctions, on the other hand, were water bath sonicated in 3 M KCl for 30 min to expedite the displacement of air bubbles in the junction pores with electrolyte solution. The junctions were then assembled and fused onto the RE compartment using an ABS slurry (1:1 weight ratio of ABS filament dissolved in acetone).

Water Absorption Study
To estimate the open porosity of the FFF 3D-printed junctions and casing walls, water absorption studies were conducted using 7.0 mm × 7.5 mm × 0.80 mm 3D-printed samples [39]. The size of our test specimen corresponds to the dimensions of the printed junctions and the dimensions of the rear wall of the 3D-RE casing cavity. Five replicate print samples were evaluated per test case. After 3D printing, specimens were kept in a desiccating cabinet for at least 24 h and were then weighed to obtain their dry masses (m dry ). The 3D-printed filament density was estimated by dividing the dry mass by the expected model volume (42 mm 3 ). The samples were then immersed in DI water and water bath sonicated for 30 min to expedite the displacement of air bubbles in the specimen pores. The immediate surfaces of the samples were patted dry with lint-free tissue to remove excess water on the surface of the specimens. Samples were then weighed to obtain their wet mass (m wet ). The % water absorption was calculated using Equation (1). Water absorption corresponds to the ratio of the mass of water absorbed to the dry mass of the 3D-printed specimen.

3D-RE Assembly
Ag wire with 0.4 mm diameter was polished with 50 µm alumina powder and immersed in 0.1 M HNO 3 for 10 s to remove surface oxides. The wire was then sonicated in deionized water and cut to 15 mm pieces. A coating of AgCl was formed at the surface of the Ag wire by immersing the wire in commercial bleach [41]. An immersion time of 30 min was sufficient to form a stable, thick coating of AgCl. The Ag|AgCl wire was then rinsed three times with DI water, air dried, then inserted into the 3D-RE casing. The wire was sealed into the electrode compartment with ABS slurry.
To form the internal gel electrolyte layer, 2% wt. agar was prepared in 3 M KCl saturated with AgCl [13,23]. While at 80 • C, 180 µL of the agar solution was injected into the inner reference electrolyte compartment and allowed to cool and solidify. Care was taken to avoid the presence of air bubbles in the gel. The 3D-REs were then conditioned in 3 M KCl for at least 24 h before use.

Characterization of 3D-RE
The potential stabilities of the 3D-REs were analyzed by measuring the open circuit potential (OCP) of a two-electrode setup, in which the 3D-RE was connected to the WE terminal of the potentiostat and the master reference electrode, which was a commercial standard calomel electrode (SCE), was connected to the RE terminal. To evaluate the potential drift of the fabricated electrodes, the OCP between the 3D-RE and the master RE was measured in 0.1 M KCl aqueous solution over a period of 12 h, with OCP measurements taken every 10 min. The long-term potential stability of the RE was also tested by measuring the OCP of the 3D-printed electrodes for a period of 10 min in 0.1 M KCl daily for 30 days. Electrodes were stored in 3 M KCl when not in use.
Reference electrode impedance was measured using potentiostatic EIS in 0.1 M KCl electrolyte solution [8]. The evaluated RE was connected to the WE terminal of the potentiostat, a Pt wire connected to the CE terminal, and a SCE connected to the RE terminal. An EIS scan was ran from 100 kHz to 0.5 Hz, with an applied DC potential set as zero versus the OCP and an AC waveform of 5 mV amplitude. The potential stabilities of the 3D-RE in various electrolyte solutions were evaluated, in separate 10 −4 , 10 −3 , 10 −2 , 10 −1 , and 1 M solution of each of KCl, NaCl, KNO 3 , NaNO 3 , Na 2 SO 4 , and KBr. The effect of sample solution pH was also tested by measuring the potential stability of the reference electrode in 0.1 M standard buffer solutions at pH 3, 7, and 11 (phosphate buffers), pH 5 (acetate buffer), and pH 9 (carbonate buffer), whereas pH 2 and pH 12 solutions were also prepared using HCl and NaOH, respectively.

Application of 3D-RE
The 3D-RE was also tested as a RE in two different applications. First, it was tested as the reference electrode in the cyclic voltammetry (CV) of 4 mM ferricyanide in 0.1 M KCl solution using a carbon paste electrode as the WE and Pt wire as CE at a potential scan range of +0.80 V to −0.20 V and scan rate of 100 mV/s. Voltammograms were compared to those recorded using a Ag|AgCl pRE and the commercial SCE.
Second, the 3D-RE was tested as the reference electrode for potentiometric pH measurements using an inkjet-printed gold with electrodeposited iridium oxide (IrO x |Au) pH sensor ( Figure S1). The synthesis, formulation, and inkjet printing of the Au ink to fabricate the Au electrode were achieved following previous work [42]. Here, the Au electrode is made of a 1-mm diameter circle connected to a contact pad with 250-µm conducting line. The conducting line was then covered with an SU-8 dielectric that was then dried at 100 • C and cured with UV light for 1 min. Alternatively, covering the conducting line with a generic, clear nail polish also works as effectively.
The synthesis and electrodeposition of iridium oxide was adapted from the protocols of Zea [43] and Marzouk [44]. The electrodeposition solution was prepared by combining 75 mg iridium (III) chloride hydrate, 0.5 mL 30% H 2 O 2 , and 250 mg oxalic acid dihydrate to 50 mL deionized water with 30 min intervals between each addition of reagent. The solution is then titrated with dry potassium carbonate until the pH reached 10.5. This solution was then allowed to age for 4 days wherein the solution turned from a yellow-green hue to a deep blue color. Prior to electrodeposition of the iridium oxide solution, the inkjet-printed Au electrode was electrochemically cleaned following the study of Fischer [45]. This was done by performing a cyclic potential scan (−0.4 to 1.4 V, 100 mV/s scan rate) in a dilute sulfuric acid solution (0.050 M). The electrodeposition solution was then galvanostatically electrodeposited onto cleaned inkjet-printed Au electrodes with 1 mA/cm 2 of current for 10 min. The electrodeposited iridium oxide electrodes were then conditioned in a pH 7.00 buffer for 24 h.
To test the IrO x pH sensor using the 3D-RE, the electrodeposited iridium oxide electrode and reference electrode were immersed together in different pH buffers and the potential readings were acquired using a pH meter in the ORP mode (IrO x electrode connected to the signal terminal and Ag|AgCl electrode connected to the ground terminal of the BNC connector in the pH meter). The potential reading was allowed to stabilize, as indicated by a "Ready" indicator (~15 s of constant reading) in the pH meter before the value was recorded. Potential readings for the different pH buffers were compared to those recorded against a Ag|AgCl pRE and the commercial SCE.
Unless otherwise indicated, data are presented as the mean ± standard deviation.

Fabrication of 3D-RE
The method for fabricating the 3D-printed reference electrode (3D-RE) is illustrated in Figure 1a. Two main components of the reference electrode are 3D printed: the casing and the junction, while the other components are readily fixed into the assembled 3D-RE body. Since the 3D-RE was designed for application in aqueous samples, we selected ABS filament as our printing material because of its resistance to degradation during prolonged water exposure [46]. This is an important consideration since the 3D-RE is meant to be stored for extended periods of time in aqueous electrolyte solution. Other water-resistant filaments, such as PLA, can also be applied, although PLA has been shown to be more prone to pore and crack formation upon extended exposure to water [47]. An additional advantage of ABS filament is that 3D-printed objects can be postprocessed using well-established solvent treatment methods to smoothen and seal the surface [38], granting an additional means of control over the properties of the 3D-printed samples.
solution. Since the potential of the Ag|AgCl|Cl − electrode is dependent on chloride ion activity (3), traditional macroscopic reference electrodes contain an internal electrolyte of known chloride ion activity, which is separated from the sample solution with a porous junction. Like this basic electrode structure, the 3D-RE also contains an internal electrolyte, but in the form of a gel to improve robustness of the electrolyte layer. We selected agar as the gelling agent since it is low-cost, widely accessible, and has proven use in Ag|AgCl|Cl − reference electrodes [13,21,23,48]. Contact between the gel-KCl layer and the sample solution is limited through the 3D-printed porous junction, so the 3D-RE casing must be watertight to ensure that electrical contact is facilitated only through the porous junction. Thus, to fabricate the casing and junction with their contrasting property requirements, these components were 3D printed separately using different printing parameters and postprinting treatments, and then assembled into a single 3D-RE body.

Fabrication of 3D-RE Casing
Since the 3D-printed reference electrode (3D-RE) casing will be stored and applied in aqueous solutions, it is imperative that the walls of the casing exhibit minimal water absorption. FFF 3Dprinted objects tend to be porous with gaps formed between print layers; nevertheless, printing parameters can be optimized to control porosity so that the 3D-RE casing is as watertight as possible [34,35,37]. In previous work by Gordeev et al., it was shown that the most significant parameter that affects the porosity of FFF 3D-printed objects is the extrusion ratio (k), or how much filament is fed through the nozzle [34]. A higher extrusion ratio results in an increased amount of filament exuded through the nozzle, which in turn leads to fewer air gaps between layers. Another parameter that affects porosity is layer height [35]. Previous work also showed that lower layer thicknesses resulted in fewer air gaps between print layers due to the tendency for the nozzle to compress the exuded filament to fill a smaller print volume, resulting in fewer voids [35,49]. With these factors in mind, we opted to print the 3D-RE casing with the lowest possible layer height on our 3D printer (0.18 mm) A cross-section of the 3D-RE showing the internal architecture of the electrode is shown in Figure 1b. A photograph of assembled 3D-REs is also shown in Figure 1c: a 3D-RE without a junction installed (Figure 1c(i)) reveals the internal structure of the electrode, which includes a Ag|AgCl wire and agar-KCl. This reference electrode structure maintains the basic operating principle of a macroscopic Ag|AgCl|Cl − reference electrode, which is based on the following half-cell reaction: The potential of the half-cell is determined by the Nernst equation, where E • is the standard reduction potential, R is the universal gas constant, T is the absolute temperature, F is Faraday's constant, and a Cl − is the activity of the chloride ion in the electrolyte solution. Since the potential of the Ag|AgCl|Cl − electrode is dependent on chloride ion activity (3), traditional macroscopic reference electrodes contain an internal electrolyte of known chloride ion activity, which is separated from the sample solution with a porous junction. Like this basic electrode structure, the 3D-RE also contains an internal electrolyte, but in the form of a gel to improve robustness of the electrolyte layer. We selected agar as the gelling agent since it is low-cost, widely accessible, and has proven use in Ag|AgCl|Cl − reference electrodes [13,21,23,48]. Contact between the gel-KCl layer and the sample solution is limited through the 3D-printed porous junction, so the 3D-RE casing must be watertight to ensure that electrical contact is facilitated only through the porous junction. Thus, to fabricate the casing and junction with their contrasting property requirements, these components were 3D printed separately using different printing parameters and postprinting treatments, and then assembled into a single 3D-RE body.

Fabrication of 3D-RE Casing
Since the 3D-printed reference electrode (3D-RE) casing will be stored and applied in aqueous solutions, it is imperative that the walls of the casing exhibit minimal water absorption. FFF 3D-printed objects tend to be porous with gaps formed between print layers; nevertheless, printing parameters can be optimized to control porosity so that the 3D-RE casing is as watertight as possible [34,35,37]. In previous work by Gordeev et al., it was shown that the most significant parameter that affects the porosity of FFF 3D-printed objects is the extrusion ratio (k), or how much filament is fed through the nozzle [34]. A higher extrusion ratio results in an increased amount of filament exuded through the nozzle, which in turn leads to fewer air gaps between layers. Another parameter that affects porosity is layer height [35]. Previous work also showed that lower layer thicknesses resulted in fewer air gaps between print layers due to the tendency for the nozzle to compress the exuded filament to fill a smaller print volume, resulting in fewer voids [35,49]. With these factors in mind, we opted to print the 3D-RE casing with the lowest possible layer height on our 3D printer (0.18 mm) and with the highest possible extrusion ratio (k = 1.10) that can be used without observing sample shape deformation from excess filament [34].
Even with careful printing parameter selection, however, solution can still seep between the sample's microcavities, potentially contaminating subsequent samples and degrading the lifetime of the object [36]. To further ensure that the 3D-RE casing is watertight, we applied acetone vapor smoothening as a surface treatment to seal the air gaps between print layers [38]. This method is a postprocessing technique used for FFF 3D-printed ABS objects in which acetone vapors are condensed on the surface of the 3D-printed object. The treatment results in the softening of the outer polymer layers, minor flow, and solidification of the polymer layers to form a surface with a smoother, less porous texture that also results in lower water absorption [38]. Figure 2 shows optical microscope images of 3D-printed samples before (a) and after (b) acetone vapor treatment. The layered texture that is characteristic of FFF 3D-printed objects ( Figure 1a) is no longer observable after surface treatment, as evident in Figure 2b. Water absorption tests also showed that the sample without surface treatment exhibited a water adsorption of 1.84% ± 0.58% (n = 5), while the surface-treated sample did not increase in mass after prolonged sonication and immersion in water, suggesting no observable water absorption. Even after 1 week of immersion in water, the surface-treated sample did not exhibit observable water absorption and was therefore effectively watertight. This result indicates that acetone vapor smoothening can serve as an effective treatment for fabricating watertight compartments and accessories. and with the highest possible extrusion ratio (k = 1.10) that can be used without observing sample shape deformation from excess filament [34]. Even with careful printing parameter selection, however, solution can still seep between the sample's microcavities, potentially contaminating subsequent samples and degrading the lifetime of the object [36]. To further ensure that the 3D-RE casing is watertight, we applied acetone vapor smoothening as a surface treatment to seal the air gaps between print layers [38]. This method is a postprocessing technique used for FFF 3D-printed ABS objects in which acetone vapors are condensed on the surface of the 3D-printed object. The treatment results in the softening of the outer polymer layers, minor flow, and solidification of the polymer layers to form a surface with a smoother, less porous texture that also results in lower water absorption [38]. Figure 2 shows optical microscope images of 3D-printed samples before (a) and after (b) acetone vapor treatment. The layered texture that is characteristic of FFF 3D-printed objects ( Figure 1a) is no longer observable after surface treatment, as evident in Figure 2b. Water absorption tests also showed that the sample without surface treatment exhibited a water adsorption of 1.84% ± 0.58% (n = 5), while the surface-treated sample did not increase in mass after prolonged sonication and immersion in water, suggesting no observable water absorption. Even after 1 week of immersion in water, the surface-treated sample did not exhibit observable water absorption and was therefore effectively watertight. This result indicates that acetone vapor smoothening can serve as an effective treatment for fabricating watertight compartments and accessories.

Fabrication of 3D-Printed Junction
The RE junction has the important role of limiting the diffusion of the chloride ion from the internal electrolyte layer while maintaining sufficient ionic conduction between the internal electrolyte and the sample. For miniaturized reference electrodes, even small amounts of chloride ion leakage can result in significant changes in chloride ion activity due to the small volume of the internal electrolyte layer. This decrease in internal electrolyte chloride ion activity results in a shift in RE potential towards more a positive potential, as indicated by Equation (3). The level of porosity in the RE junction should therefore sufficiently limit the diffusion of the chloride ion. However, there is

Fabrication of 3D-Printed Junction
The RE junction has the important role of limiting the diffusion of the chloride ion from the internal electrolyte layer while maintaining sufficient ionic conduction between the internal electrolyte and the sample. For miniaturized reference electrodes, even small amounts of chloride ion leakage can Chemosensors 2020, 8,130 9 of 20 result in significant changes in chloride ion activity due to the small volume of the internal electrolyte layer. This decrease in internal electrolyte chloride ion activity results in a shift in RE potential towards more a positive potential, as indicated by Equation (3). The level of porosity in the RE junction should therefore sufficiently limit the diffusion of the chloride ion. However, there is a trade-off between minimizing chloride ion leakage and minimizing RE impedance. A junction of lower porosity can slow chloride ion diffusion, but this requires more restricted electrolyte flow, which in turn results in higher electrolyte resistance [23]. High RE impedance can result in potentiostat oscillation and/or overloads and noisy voltammograms [8]. It was therefore of interest in this work to control the porosity of the 3D-printed junction to sufficiently limit chloride ion leakage and to maintain an adequately low junction impedance for electroanalytical applications in aqueous solutions.
In contrast to the 3D-RE casing that was optimized for watertightness, the 3D-RE junction must be porous to a certain extent to permit electrical contact between electrolyte interfaces. Since we wanted to ensure that the junction was sufficiently porous, we explored a printing parameter set that would maximize porosity. Specifically, we limited the print layer height to 0.40 mm, which is the highest print layer height possible on our 3D printer, and started with a low extrusion ratio of k = 0.85. Extrusion ratios lower than 0.85 tended to be fragile to print and handle. In addition, since the porosity of FFF 3D-printed objects originates primarily from the gaps between the vertically stacked print layers, we printed the junction as a thin wall so that the vertically stacked layers served as the interface of the junction. The horizontal thickness of the junction wall was also set to two layers, which was the thinnest wall that we could print and handle robustly. We preferred a thin junction to maintain a compact form factor for the 3D-RE, and to ensure that there are fewer layers to transverse for ionic conduction.
To explore different levels of porosity, we varied the extrusion ratio to three different levels (k = 0.85, 0.95, and 1.05) while keeping the other parameters constant, as extrusion ratio has been reported to be the most significant factor affecting FFF 3D-printed specimen porosity given the same sample geometry [34]. As an initial study to characterize the 3D-printed junctions, we examined optical microscope images of the 3D-printed junctions fabricated with the same printing parameters but different extrusion ratios (k = 0.85, 0.95, and 1.05), as shown in Figure 3. While there appears to be no visible differences between the surfaces of the specimens printed with k = 0.95 and 1.05, samples printed with k = 0.85 exhibit observable gaps between print layers, which can be related to the decreased amount of filament exuded from the nozzle when printing each layer, resulting in some imperfections in the interlayer adhesion. We made a similar observation when we 3D printed the same specimens with different 3D printers and filaments ( Figure S2). For the samples printed with k = 0.95 and 1.05, it is likely that they also present similar interlayer gaps, but at a smaller scale that is not visible at the current magnification range. the porosity of FFF 3D-printed objects originates primarily from the gaps between the vertically stacked print layers, we printed the junction as a thin wall so that the vertically stacked layers served as the interface of the junction. The horizontal thickness of the junction wall was also set to two layers, which was the thinnest wall that we could print and handle robustly. We preferred a thin junction to maintain a compact form factor for the 3D-RE, and to ensure that there are fewer layers to transverse for ionic conduction.
To explore different levels of porosity, we varied the extrusion ratio to three different levels (k = 0.85, 0.95, and 1.05) while keeping the other parameters constant, as extrusion ratio has been reported to be the most significant factor affecting FFF 3D-printed specimen porosity given the same sample geometry [34]. As an initial study to characterize the 3D-printed junctions, we examined optical microscope images of the 3D-printed junctions fabricated with the same printing parameters but different extrusion ratios (k = 0.85, 0.95, and 1.05), as shown in Figure 3. While there appears to be no visible differences between the surfaces of the specimens printed with k = 0.95 and 1.05, samples printed with k = 0.85 exhibit observable gaps between print layers, which can be related to the decreased amount of filament exuded from the nozzle when printing each layer, resulting in some imperfections in the interlayer adhesion. We made a similar observation when we 3D printed the same specimens with different 3D printers and filaments ( Figure S2). For the samples printed with k = 0.95 and 1.05, it is likely that they also present similar interlayer gaps, but at a smaller scale that is not visible at the current magnification range. For further characterization, we estimated the junctions' open porosities using the water absorption method, in which water is used to displace any air in the open pores of the sample of interest. Five replicate 3D-printed samples were evaluated per extrusion ratio. As expected, we observed decreased water absorption in samples 3D printed with increasing extrusion ratio, as plotted in Figure 4. Estimation of the filament density, which is the mass of filament used to 3D print a sample per unit volume of the sample model (7.0 mm × 7.5 mm × 0.8 mm) also shows that with higher extrusion ratio, more filament is used to fill the sample volume. Notably, measurements of filament density yielded relatively low % CV (percent coefficient of variation) at 1.74% to 3.22%, whereas measurements of water absorption exhibited high % CV of around 20% per case. These For further characterization, we estimated the junctions' open porosities using the water absorption method, in which water is used to displace any air in the open pores of the sample of interest. Five replicate 3D-printed samples were evaluated per extrusion ratio. As expected, we observed decreased water absorption in samples 3D printed with increasing extrusion ratio, as plotted in Figure 4.
Estimation of the filament density, which is the mass of filament used to 3D print a sample per unit volume of the sample model (7.0 mm × 7.5 mm × 0.8 mm) also shows that with higher extrusion ratio, more filament is used to fill the sample volume. Notably, measurements of filament density yielded relatively low % CV (percent coefficient of variation) at 1.74% to 3.22%, whereas measurements of water absorption exhibited high % CV of around 20% per case. These results suggest that the amount of filament deposited during 3D printing is reproducible even at different extrusion ratios, but the interlayer structure that gives rise to different levels of open porosity can exhibit greater deviation. The high deviation in the water absorption measurements may also be due to the variations in the accessibility of the pores to water displacement.

Optimization of 3D-RE Junction Porosity
To determine the optimum 3D printing extrusion ratio of the 3D-printed porous junction, we fabricated reference electrodes with different junctions and evaluated their performance. For comparison, we also prepared a 3D-RE with no junction-that is, the gel electrolyte layer is in direct contact with the sample solution (Figure 1c(i)). We initially assessed the ability of each junction to slow chloride ion leakage by monitoring the open-circuit potential (OCP) of the 3D-RE versus a commercial SCE in 0.1 M KCl over a 12-h period. For short-term analyses, the potential drift of the RE should remain within ± 2 mV within 5 min, and for more long-term applications, a potential drift within ± 5 mV is acceptable [23]. Figure 5 shows the OCP vs. time of the 3D-REs. At the start, all of the 3D-REs exhibited OCPs of −32.6 ± 0.5 mV, which is consistent with the expected literature OCP of a Ag|AgCl|3 M KCl electrode vs. SCE [8]. Over time, however, the potentials of the 3D-REs increased. The 3D-RE without a junction exhibited the fastest increase in potential, which reflects the fast rate of chloride ion diffusion: potential drift exceeded 2 mV after 30 min and 5 mV after 40 min. Over a 12-h period, this RE exhibited a potential drift of 50.2 ± 1.2 mV. Addition of a junction led to improvements in potential stability. As expected, junctions 3D printed with higher extrusion ratios resulted in lower potential drifts due to decreased junction porosity. Over 12 h, the 3D-RE with a k = 0.85 junction exhibited a potential drift of 11.5 ± 2.3 mV, whereas the 3D-RE with a k = 0.95 junction exhibited an acceptable potential drift of 4.6 ± 0.5 mV. The electrode with a junction of k = 1.05 exhibited the highest potential stability with a potential drift of 1.4 ± 1.1 mV over a 12-h period of continuous measurement. The possibility that trapped air in the 3D-printed junction could block ionic conduction pathways emphasizes the need to sufficiently displace entrapped air with electrolyte. Thus, prior to installation, junctions were sonicated in 3 M KCl, and the fully assembled electrodes were also conditioned for at least 24 h in same solution. To determine whether the open pores form continuous open pathways throughout the junction for sufficient electrolyte flow between solution interfaces in the 3D-RE and to optimize that level of electrolyte flow, we evaluated the electrochemical properties of the 3D-REs with various junctions installed, as described in the next subsection.

Optimization of 3D-RE Junction Porosity
To determine the optimum 3D printing extrusion ratio of the 3D-printed porous junction, we fabricated reference electrodes with different junctions and evaluated their performance. For comparison, we also prepared a 3D-RE with no junction-that is, the gel electrolyte layer is in direct contact with the sample solution (Figure 1c(i)). We initially assessed the ability of each junction to slow chloride ion leakage by monitoring the open-circuit potential (OCP) of the 3D-RE versus a commercial SCE in 0.1 M KCl over a 12-h period. For short-term analyses, the potential drift of the RE should remain within ±2 mV within 5 min, and for more long-term applications, a potential drift within ±5 mV is acceptable [23]. Figure 5 shows the OCP vs. time of the 3D-REs. At the start, all of the 3D-REs exhibited OCPs of −32.6 ± 0.5 mV, which is consistent with the expected literature OCP of a Ag|AgCl|3 M KCl electrode vs. SCE [8]. Over time, however, the potentials of the 3D-REs increased. The 3D-RE without a junction exhibited the fastest increase in potential, which reflects the fast rate of chloride ion diffusion: potential drift exceeded 2 mV after 30 min and 5 mV after 40 min. Over a 12-h period, this RE exhibited a potential drift of 50.2 ± 1.2 mV. Addition of a junction led to improvements in potential stability. As expected, junctions 3D printed with higher extrusion ratios resulted in lower potential drifts due to decreased junction porosity. Over 12 h, the 3D-RE with a k = 0.85 junction exhibited a potential drift of 11.5 ± 2.3 mV, whereas the 3D-RE with a k = 0.95 junction exhibited an acceptable potential drift of 4.6 ± 0.5 mV. The electrode with a junction of k = 1.05 exhibited the highest potential stability with a potential drift of 1.4 ± 1.1 mV over a 12-h period of continuous measurement. In Figure 5, we also show the OCP of a bare Ag|AgCl wire as a pseudo-reference electrode (pRE) for comparison. The Ag|AgCl wire is in direct contact with the 0.1 M KCl solution, so its equilibrium potential is related to the activity of the chloride ion in the sample solution. We note that the OCP of the Ag|AgCl wire tends to decrease with time due to the increased chloride ion activity in solution from the Cl − leakage from the SCE used during the experiment.
The second factor to optimize for the 3D-RE junction is RE impedance. Many potentiostat manufacturers recommend that the impedance of a reference electrode should not exceed 1 kΩ in order to minimize potentiostat oscillations and/or overloads [8,50,51]. A high reference electrode impedance can disturb the potentiostat's negative feedback system that controls the potential between the reference and working electrodes by slowing its response rate and destabilizing the system towards oscillation [6,51]. Figure 6 shows the Bode plots of the different 3D-REs and a commercial SCE for comparison. The Bode plots of the complete set of electrode replicates are shown in Figure S3. For reference electrodes, the phase angle should be at or near 0° at high frequencies, and the impedance measured at this high frequency corresponds to the RE resistance [8]. As expected, the 3D-RE with no junction exhibited the lowest RE resistance at 114 ± 6 Ω since the gel electrolyte layer Figure 5. Potential stability of 3D-printed reference electrodes (3D-REs) with junctions printed using different extrusion ratios (k = 0.85, 0.95, 1.05). Open-circuit potential (OCP) measurements vs. saturated calomel electrode (SCE) were taken every 10 min over a 12-h period in 0.1 M KCl. 3D-REs from three separate fabrication batches were used for each type of electrode. The black solid line corresponds to a potential of −32 mV, while the shaded region corresponds to the acceptable OCP drift region of ± 5 mV. For comparison, potential stabilities of a 3D-RE without a junction and a Ag|AgCl wire pseudo-reference electrode are also shown.
In Figure 5, we also show the OCP of a bare Ag|AgCl wire as a pseudo-reference electrode (pRE) for comparison. The Ag|AgCl wire is in direct contact with the 0.1 M KCl solution, so its equilibrium potential is related to the activity of the chloride ion in the sample solution. We note that the OCP of the Ag|AgCl wire tends to decrease with time due to the increased chloride ion activity in solution from the Cl − leakage from the SCE used during the experiment.
The second factor to optimize for the 3D-RE junction is RE impedance. Many potentiostat manufacturers recommend that the impedance of a reference electrode should not exceed 1 kΩ in order to minimize potentiostat oscillations and/or overloads [8,50,51]. A high reference electrode impedance can disturb the potentiostat's negative feedback system that controls the potential between the reference and working electrodes by slowing its response rate and destabilizing the system towards oscillation [6,51]. Figure 6 shows the Bode plots of the different 3D-REs and a commercial SCE for comparison. The Bode plots of the complete set of electrode replicates are shown in Figure S3. For reference electrodes, the phase angle should be at or near 0 • at high frequencies, and the impedance measured at this high frequency corresponds to the RE resistance [8]. As expected, the 3D-RE with no junction exhibited the lowest RE resistance at 114 ± 6 Ω since the gel electrolyte layer is in direct contact with the sample solution. 3D-REs with printed junctions demonstrate higher resistances that are consistent with their junction porosity: the observed RE resistance increased with decreased porosity in the expected order: Z k=0.85 < Z k=0.95 < Z k=1.05 . The results are also consistent with the observed potential drifts: 3D-REs exhibiting lower potential drifts exhibit higher impedances. Thus, while the 3D-RE with a k = 1.05 junction exhibited the most stable potential, it also exhibited the highest RE resistance at 1.0 ± 0.2 kΩ, which exceeds the ideal limit for a reference electrode. The 3D-RE with k = 0.95, however, exhibited an acceptable potential drift of 4.6 ± 0.5 mV and an impedance of 0.50 ± 0.11 kΩ. This observed resistance is comparable to that of the commercial SCE, which has a measured resistance of 0.66 kΩ. We therefore selected k = 0.95 as the optimum 3D-printed junction printing parameter. Thus, while the 3D-RE with a k = 1.05 junction exhibited the most stable potential, it also exhibited the highest RE resistance at 1.0 ± 0.2 kΩ, which exceeds the ideal limit for a reference electrode. The 3D-RE with k = 0.95, however, exhibited an acceptable potential drift of 4.6 ± 0.5 mV and an impedance of 0.50 ± 0.11 kΩ. This observed resistance is comparable to that of the commercial SCE, which has a measured resistance of 0.66 kΩ. We therefore selected k = 0.95 as the optimum 3D-printed junction printing parameter. We also note that since the 3D-RE compartment and junction were fabricated separately and later joined together with ABS slurry, we tested a control to ensure that electrical contact between 3D-RE and sample solution was established primarily through the fabricated junction pores and not through defects between the compartment-junction joint. Our control consisted of a 3D-RE with an acetone vapor treated junction that is joined to the RE compartment in the same manner as the test 3D-REs. Since acetone vapor treatment seals the junction pores, any observed electrical contact between the RE and the sample solution would have to be due to any joint defects. However, upon application of the control 3D-RE as a reference electrode, we immediately observed potentiostat oscillations, suggesting excessively high RE resistance, indeed suggesting that electrical contact in the test 3D-RE was primarily due to the junction pores with negligible contribution from the joint between the compartment and junction.
While we were interested primarily in the high-frequency impedance of the reference electrodes to estimate their resistances, investigation of the low-frequency response can also yield interesting insights and notable limitations. A major difference between the 3D-RE and the commercial SCE is their respective phase shifts at lower frequencies. While the phase shift of the SCE is close to zero throughout the frequency range, the 3D-REs exhibit phase shifts of around −30° to −50° at low frequency. This difference in behavior can be attributed to the differences in internal electrolyte structure of the commercial SCE and the 3D-RE. While the SCE has an aqueous inner electrolyte, the 3D-RE's electrolyte is trapped in agar, so ions in the 3D-RE have lower mobilities [52]. Thus, the 3D-RE inner gel electrolyte structure exhibits more significant capacitive and diffusional control in its We also note that since the 3D-RE compartment and junction were fabricated separately and later joined together with ABS slurry, we tested a control to ensure that electrical contact between 3D-RE and sample solution was established primarily through the fabricated junction pores and not through defects between the compartment-junction joint. Our control consisted of a 3D-RE with an acetone vapor treated junction that is joined to the RE compartment in the same manner as the test 3D-REs. Since acetone vapor treatment seals the junction pores, any observed electrical contact between the RE and the sample solution would have to be due to any joint defects. However, upon application of the control 3D-RE as a reference electrode, we immediately observed potentiostat oscillations, suggesting excessively high RE resistance, indeed suggesting that electrical contact in the test 3D-RE was primarily due to the junction pores with negligible contribution from the joint between the compartment and junction.
While we were interested primarily in the high-frequency impedance of the reference electrodes to estimate their resistances, investigation of the low-frequency response can also yield interesting insights and notable limitations. A major difference between the 3D-RE and the commercial SCE is their respective phase shifts at lower frequencies. While the phase shift of the SCE is close to zero throughout the frequency range, the 3D-REs exhibit phase shifts of around −30 • to −50 • at low frequency. This difference in behavior can be attributed to the differences in internal electrolyte structure of the commercial SCE and the 3D-RE. While the SCE has an aqueous inner electrolyte, the 3D-RE's electrolyte is trapped in agar, so ions in the 3D-RE have lower mobilities [52]. Thus, the 3D-RE inner gel electrolyte structure exhibits more significant capacitive and diffusional control in its electrode impedance at low frequencies [23]. Our EIS results are similar to those reported by Hassel et al., who compared the spectra of a commercial Ag|AgCl|Cl − electrode to that of a capillary agar-based Ag|AgCl|Cl − electrode. While the impedance of the commercial RE was mostly frequency-independent, the agar-based RE exhibited a low-frequency phase shift of −25 • [48]. Since the 3D-RE exhibits phase shifts at lower frequencies, use of the 3D-RE as a RE for impedance measurements at these frequencies may result in artifacts in the spectra [48]. Nevertheless, since the 3D-RE exhibits near-ideal frequency-independent behavior at a localized frequency range (10 3 -10 5 Hz), the 3D-RE may be applied as an RE at this range [23,48,53].

Characterization of Optimized 3D-RE
The optimized 3D-RE with a junction printed with k = 0.95 was further studied for its long-term potential stability. Figure 7 shows the daily measurement of OCP vs. SCE of three 3D-REs (represented by different symbols) over a period of 30 days in 0.1 M KCl. Results indicate that the electrodes can maintain a potential within ±5 mV for at least 30 days when the electrode is properly stored in a maintaining 3 M KCl solution. Regular calibration can account for the small potential shifts for more accurate analyses. Further studies can be done to assess longer shelf-life stability of the 3D-RE. these frequencies may result in artifacts in the spectra [48]. Nevertheless, since the 3D-RE exhibits near-ideal frequency-independent behavior at a localized frequency range (10 3 -10 5 Hz), the 3D-RE may be applied as an RE at this range [23,48,53].

Characterization of Optimized 3D-RE
The optimized 3D-RE with a junction printed with k = 0.95 was further studied for its long-term potential stability. Figure 7 shows the daily measurement of OCP vs. SCE of three 3D-REs (represented by different symbols) over a period of 30 days in 0.1 M KCl. Results indicate that the electrodes can maintain a potential within ±5 mV for at least 30 days when the electrode is properly stored in a maintaining 3 M KCl solution. Regular calibration can account for the small potential shifts for more accurate analyses. Further studies can be done to assess longer shelf-life stability of the 3D-RE. The ability of a reference electrode to maintain a constant potential regardless of the sample matrix is an important property to consider. To assess the potential stability of the 3D-RE in different matrices, we monitored the potential of the 3D-RE and Ag|AgCl pRE while immersed in a range of concentrations (10 −4 to 1 M) of various electrolytes. Figure 8a shows the potential of the 3D-RE and Ag|AgCl pRE over time with increases in KCl concentration after every 5 min. With changes in solution chloride ion activity, the potential of the pRE decreases while the potential of the 3D-RE remains constant and dependent only on the chloride ion activity of its gel electrolyte layer. Figure  8b converts the potential response of the 3D-RE and pRE as a function of the logarithm of KCl concentration. Here, the Ag|AgCl pRE demonstrates Nernstian behavior, as it now essentially behaves as an indicator electrode of the second kind due to its sensitivity to Clactivity. Similar experiments were conducted with different electrolyte solutions (NaCl, NaNO3, KNO3, Na2SO4, KBr) and solutions at different pH levels. Figure 8c shows the Nernst plots of the potential response of the 3D-RE in different concentrations of the respective electrolyte. For clarity, each plot was offset by multiples of 0.05 V to prevent plot overlaps. The Nernst plots demonstrate the stability of the 3D-RE potential to different electrolyte solutions. Complete Nernst plots and equations for both the 3D-RE and Ag|AgCl pRE in each electrolyte solution are shown in Figure S4.  The ability of a reference electrode to maintain a constant potential regardless of the sample matrix is an important property to consider. To assess the potential stability of the 3D-RE in different matrices, we monitored the potential of the 3D-RE and Ag|AgCl pRE while immersed in a range of concentrations (10 −4 to 1 M) of various electrolytes. Figure 8a shows the potential of the 3D-RE and Ag|AgCl pRE over time with increases in KCl concentration after every 5 min. With changes in solution chloride ion activity, the potential of the pRE decreases while the potential of the 3D-RE remains constant and dependent only on the chloride ion activity of its gel electrolyte layer. Figure 8b converts the potential response of the 3D-RE and pRE as a function of the logarithm of KCl concentration. Here, the Ag|AgCl pRE demonstrates Nernstian behavior, as it now essentially behaves as an indicator electrode of the second kind due to its sensitivity to Cl − activity.

Application of Optimized 3D-RE
To test the performance of the optimized 3D-RE for electroanalytical applications, the 3D-RE was applied as the RE in the cyclic voltammetry of a redox couple standard, as well as the RE in potentiometric pH measurements. Figure 9 shows the voltammograms of 4 mM ferricyanide in 0.1 M KCl using the 3D-RE as a reference electrode. Voltammograms recorded using the commercial SCE and Ag|AgCl wire pRE are also shown for comparison. The observed difference between the cathodic peak potentials (Epc) in the voltammograms recorded using the SCE and the 3D-RE is −32.7 ± 0.9 mV, which is consistent with the measured OCP vs. SCE of the 3D-RE in 0.1 M KCl. This suggests that the potential of the 3D-RE remains insensitive to the presence of a redox couple in solution such as ferro-/ferricyanide. The difference between peak potentials observed for the SCE and the Ag|AgCl pRE, on the other hand, is 92.5 ± 2.9 mV, which is much higher than the observed OCP vs. SCE of the pRE in 0.1 M KCl (see Figure 5). The difference between the expected potentials may be due to the sensitivity of the pRE to the presence of the redox couple in solution. Similar experiments were conducted with different electrolyte solutions (NaCl, NaNO 3 , KNO 3 , Na 2 SO 4 , KBr) and solutions at different pH levels. Figure 8c shows the Nernst plots of the potential response of the 3D-RE in different concentrations of the respective electrolyte. For clarity, each plot was offset by multiples of 0.05 V to prevent plot overlaps. The Nernst plots demonstrate the stability of the 3D-RE potential to different electrolyte solutions. Complete Nernst plots and equations for both the 3D-RE and Ag|AgCl pRE in each electrolyte solution are shown in Figure S4. Figure 8d shows the potential response of the 3D-RE and the Ag|AgCl pRE in solutions adjusted to a wide range of pH levels. While the pRE potential varies significantly, the 3D-RE demonstrates a stable potential regardless of solution pH. These results demonstrate the suitability of the 3D-RE over a pRE to serve as a reference electrode in applications that deal with samples of variable matrices.

Application of Optimized 3D-RE
To test the performance of the optimized 3D-RE for electroanalytical applications, the 3D-RE was applied as the RE in the cyclic voltammetry of a redox couple standard, as well as the RE in potentiometric pH measurements. Figure 9 shows the voltammograms of 4 mM ferricyanide in 0.1 M KCl using the 3D-RE as a reference electrode. Voltammograms recorded using the commercial SCE and Ag|AgCl wire pRE are also shown for comparison. The observed difference between the cathodic peak potentials (E pc ) in the voltammograms recorded using the SCE and the 3D-RE is −32.7 ± 0.9 mV, which is consistent with the measured OCP vs. SCE of the 3D-RE in 0.1 M KCl. This suggests that the potential of the 3D-RE remains insensitive to the presence of a redox couple in solution such as ferro-/ferricyanide. The difference between peak potentials observed for the SCE and the Ag|AgCl pRE, on the other hand, is 92.5 ± 2.9 mV, which is much higher than the observed OCP vs. SCE of the pRE in 0.1 M KCl (see Figure 5). The difference between the expected potentials may be due to the sensitivity of the pRE to the presence of the redox couple in solution.  Figure 10 shows the potential responses of the IrOx|Au electrode measured against the 3D-RE, Ag|AgCl pRE, or commercial SCE when immersed in various standard pH buffers. The IrOx|Au pH sensor demonstrated linear super-Nernstian response when measured against a commercial SCE (Figure 10a). The linear super-Nernstian response coincides well with most literature data, which report sensitivity levels of −80 to −70 mV/pH [43,[54][55][56] and consistently high R 2 (> 0.99) [54][55][56] for IrOx-based pH sensors coupled with true reference electrodes. In contrast, the IrOx|Au electrode response measured against a Ag|AgCl pRE demonstrated sub-Nernstian response with poorer linearity (Figure 10b), which can be attributed to the potential instability of the pRE when exposed to the different standard pH buffers. This result agrees with the data collected by Huang et al. (R 2 ~ 0.95 and sensitivity of ~51 mV/pH) [57], wherein they used an evaporated silver and electrochemically anodized silver chloride as an Ag|AgCl pseudo-reference electrode for their flexible IrOx-based pH sensor. This result highlights the importance of using a stable reference electrode: the performance of a good indicator electrode can be limited by the performance of the reference electrode.
Finally, when the response of the IrOx|Au electrode was measured against the 3D-RE, the system exhibited a linear super-Nernstian response (Figure 10c) that is highly comparable to the sensor response when measured against the commercial SCE. The responses using the different reference electrodes are only offset by a constant potential that corresponds to the difference in potentials between the SCE and 3D-RE. Unlike the Ag|AgCl pRE, the 3D-RE exhibits a stable potential that makes it promising for use as reference electrode for potentiometric sensor measurements.   Figure 10 shows the potential responses of the IrO x |Au electrode measured against the 3D-RE, Ag|AgCl pRE, or commercial SCE when immersed in various standard pH buffers. The IrO x |Au pH sensor demonstrated linear super-Nernstian response when measured against a commercial SCE (Figure 10a). The linear super-Nernstian response coincides well with most literature data, which report sensitivity levels of −80 to −70 mV/pH [43,[54][55][56] and consistently high R 2 (> 0.99) [54][55][56] for IrOx-based pH sensors coupled with true reference electrodes. In contrast, the IrOx|Au electrode response measured against a Ag|AgCl pRE demonstrated sub-Nernstian response with poorer linearity (Figure 10b), which can be attributed to the potential instability of the pRE when exposed to the different standard pH buffers. This result agrees with the data collected by Huang et al. (R 2~0 .95 and sensitivity of~51 mV/pH) [57], wherein they used an evaporated silver and electrochemically anodized silver chloride as an Ag|AgCl pseudo-reference electrode for their flexible IrOx-based pH sensor. This result highlights the importance of using a stable reference electrode: the performance of a good indicator electrode can be limited by the performance of the reference electrode.  Figure 10 shows the potential responses of the IrOx|Au electrode measured against the 3D-RE, Ag|AgCl pRE, or commercial SCE when immersed in various standard pH buffers. The IrOx|Au pH sensor demonstrated linear super-Nernstian response when measured against a commercial SCE (Figure 10a). The linear super-Nernstian response coincides well with most literature data, which report sensitivity levels of −80 to −70 mV/pH [43,[54][55][56] and consistently high R 2 (> 0.99) [54][55][56] for IrOx-based pH sensors coupled with true reference electrodes. In contrast, the IrOx|Au electrode response measured against a Ag|AgCl pRE demonstrated sub-Nernstian response with poorer linearity (Figure 10b), which can be attributed to the potential instability of the pRE when exposed to the different standard pH buffers. This result agrees with the data collected by Huang et al. (R 2 ~ 0.95 and sensitivity of ~51 mV/pH) [57], wherein they used an evaporated silver and electrochemically anodized silver chloride as an Ag|AgCl pseudo-reference electrode for their flexible IrOx-based pH sensor. This result highlights the importance of using a stable reference electrode: the performance of a good indicator electrode can be limited by the performance of the reference electrode.
Finally, when the response of the IrOx|Au electrode was measured against the 3D-RE, the system exhibited a linear super-Nernstian response (Figure 10c) that is highly comparable to the sensor response when measured against the commercial SCE. The responses using the different reference electrodes are only offset by a constant potential that corresponds to the difference in potentials between the SCE and 3D-RE. Unlike the Ag|AgCl pRE, the 3D-RE exhibits a stable potential that makes it promising for use as reference electrode for potentiometric sensor measurements.  Finally, when the response of the IrOx|Au electrode was measured against the 3D-RE, the system exhibited a linear super-Nernstian response (Figure 10c) that is highly comparable to the sensor response when measured against the commercial SCE. The responses using the different reference electrodes are only offset by a constant potential that corresponds to the difference in potentials between the SCE and 3D-RE. Unlike the Ag|AgCl pRE, the 3D-RE exhibits a stable potential that makes it promising for use as reference electrode for potentiometric sensor measurements.

Outlook
The 3D-RE can be readily customized to suit the application of interest. The 3D-RE casing, for example, can be incorporated into a watertight 3D-printed electrochemical cell along with other 3D-printed electrodes for an all-in-one, fully integrated system. Other shapes and sizes can also be adapted. Most commercial 3D printers have sufficient resolution to fabricate 3D-RE casings as small as 3 mm × 3 mm × 3 mm, but the minimum size that the 3D-RE can adapt is limited by other considerations such as the volume of the gel electrolyte. The potential stability of 3D-REs with smaller form factors may be negatively affected by a smaller volume of the gel electrolyte, which becomes more sensitive to chloride ion leakage due to the limited supply of chloride ion [23]. Previous work by Huang et al. studied the effect of different volumes of gel-KCl on the potential stability of packaged Ag|AgCl|gel-KCl reference electrodes and found that a volume of at least 3.39 µL can yield a stable potential over 30,000 s [23]. On the other hand, if an even smaller form factor is desired, the trade-off to consider is whether the expected lower potential stability is tolerable for the target application.
The 3D-RE in this work was designed for use in aqueous solutions. For applications that involve organic solvents, however, the current 3D-RE setup may not be compatible because the 3D-printed ABS components can decompose or dissolve when exposed to organic solvents such as acetonitrile, methanol, and dimethyl sulfoxide [46,58]. The current gel electrolyte layer is also aqueous-based with KCl as the electrolyte, so factors such solvent cross-contamination and the differences in KCl solubility and mobility in the organic solvent should be considered [8,59]. The 3D-RE can be adapted for use in organic solvents by changing the 3D printing material and modifying the gel electrolyte layer. Nylon, for example, is a possible FFF 3D printing filament that has excellent stability in many organic solvents [46,60]. To further improve compatibility, the gel electrolyte layer can also be modified using an electrolyte that is also soluble in the organic solvent of interest [8,59]. An alternative strategy to adapt aqueous-based reference electrode for use in organic solvents is to use a double junction filled with electrolyte with near-equal mobilities in both the aqueous and nonaqueous solvents [8,59]. Using a solvent-compatible filament, a double junction can be readily 3D printed.
The current 3D-RE fabrication method involves multiple 3D-printed parts that need to be processed and assembled. To simplify the fabrication, an interesting avenue to explore for future studies is to 3D print a porous body that serves as both the casing and the junction. Here, the internal gel electrolyte is housed in a porous casing, and electrolyte flow between the internal electrolyte and sample solution can occur through all walls of the porous body. In this design, the porosity of the casing should be adjusted to ensure that excessive chloride ion leakage is avoided since electrolyte can now leak out from all walls of the 3D-RE body. This simplified fabrication strategy can also make it easier to design and fabricate more complex reference electrode structures, such as a double-junction electrode. The flexibility and capability of 3D printing for customization and rapid prototyping opens many possible design modifications to the basic architecture of the 3D-RE described in this work.

Conclusions
In this work, we have demonstrated one of the first successful fabrications of a true 3D-printed Ag|AgCl|Cl − reference electrode that exhibits low impedance and potential stability in various sample matrices. The novel feature of our reference electrode is the 3D-printed porous junction, which serves to bridge the internal gel electrolyte layer and the sample solution, and to protect the internal gel electrolyte layer from sample contamination and chloride ion leakage. Here, we utilized the porous property of FFF-printed objects that is typically viewed as a defect and tuned this property to serve a key function in our fabricated reference electrode. The assembly of the miniaturized reference electrode is simple and low-cost, and it requires commonly encountered laboratory materials. The reference electrode casing and junction can also be readily fabricated in less than 10 min using a commercial desktop 3D printer and commercially available ABS filament. The developed 3D-printed reference electrode has the potential to be readily fabricated on demand to serve as a reference for use in traditional laboratory-based electrochemical analyses or as an integrated reference in 3D-printed electrochemical sensor systems. It is our aim that our simple but elegant approach to fabricating a true reference electrode can broaden the application of 3D-printed electrochemical sensor systems to more complex samples that require a stable, matrix-independent reference electrode potential.
Supplementary Materials: The following are available online at http://www.mdpi.com/2227-9040/8/4/130/s1, Figure S1: Photographs of inkjet-printed gold electrodes (a) before and (b) after IrO x electrodeposition, Figure S2: Optical microscope images of 3D-printed junctions fabricated with 0.40 mm layer height and extrusion ratios k = 0.85, 0.95, or 1.05 using different 3D-printers and polymer filaments, Figure S3: Bode plots of 3D-REs fabricated in three different batches (each replicate shown in either red, green, or blue) with (a) no junction, (b) 3D-printed junction with k = 0.85, (c) 3D-printed junction with k = 0.95, and (d) 3D-printed junction with k = 1.05. For reference, the bode plot of a commercial SCE (in grey) is also included in each plot. Impedance spectra were recorded in 0.1 M KCl, Figure S4