Vapor-induced phase-separation-enabled versatile direct ink writing

Versatile printing of polymers, metals, and composites always calls for simple, economic approaches. Here we present an approach to three-dimensional (3D) printing of polymeric, metallic, and composite materials at room conditions, based on the polymeric vapor-induced phase separation (VIPS) process. During VIPS 3D printing (VIPS-3DP), a dissolved polymer-based ink is deposited in an environment where nebulized non-solvent is present, inducing the low-volatility solvent to be extracted from the filament in a controllable manner due to its higher chemical affinity with the non-solvent used. The polymeric phase is hardened in situ as a result of the induced phase separation process. The low volatility of the solvent enables its reclamation after the printing process, significantly reducing its environmental footprint. We first demonstrate the use of VIPS-3DP for polymer printing, showcasing its potential in printing intricate structures. We further extend VIPS-3DP to the deposition of polymer-based metallic inks or composite powder-laden polymeric inks, which become metallic parts or composites after a thermal cycle is applied. Furthermore, spatially tunable porous structures and functionally graded parts are printed by using the printing path to set the inter-filament porosity as well as an inorganic space-holder as an intra-filament porogen.


Solubility analysis
The relative energy difference (RED) parameter 1 is used to evaluate the suitability of a given polymer candidate for VIPS (vapor-induced phase separation) printability from a dissolution perspective, which shall present a value lower than the unity (RED<1) if the polymer is to be fully dissolved in a given solvent.This RED value is determined based on Hansen's chemical affinity relationships 2 as defined by Eqs. ( 1) and (2).  2 = 4( 2 −  1 ) 2 + ( 2 −  1 ) 2 + ( ℎ2 −  ℎ1 ) 2 (1) where Ra is the distance in the Hansen space, δd is the energy from dispersion forces between molecules (in MPa 0.5 ), δp is the energy from dipolar intermolecular forces between molecules (in MPa 0.5 ), δh is the energy from hydrogen bonds between molecules (in MPa 0.5 ), and R0 is the interaction radius.Based on the published energy values, the calculated RED values for polyacrylonitrile (PAN) and acrylonitrile butadiene styrene (ABS) are listed in Supplementary Table 1, proving that these two polymers are dissolvable in dimethyl sulfoxide (DMSO).Supplementary Table 1

VIPS dynamics 1.2.1 Ternary phase diagrams
The ternary diagrams for the polymer, non-solvent, and solvent of interest herein are constructed to elucidate the phase-separation dynamics of vapor-induced phase separation 3D printing (VIPS-3DP).The Flory-Huggins (F-H) model 4,5 has been extensively used as a thermodynamic model to describe the behavior of and interactions from ternary systems formed by polymer, non-solvent, and solvent mixtures during the phase inversion process.Specifically, the Flory-Huggins equation for the Gibbs free energy of mixing of two components (i and j) 6 is shown in Eq. ( 3) in its adapted form 7 : ∆   =  1  1 +  2  2 +  3  3 +  12 ( 2 ) 1  2 +  13 ( 3 ) 1  3 +  23 ( 3 ) 2  3 (3) where R and T are the gas constant and temperature, ni is the relative molar volume,  2 =  2 /( 1 +  2 ), φi is the volume fraction, and gij is the inter-polymer interaction function of the compounds i and j.A ternary system is usually formed by a non-solvent (1), a solvent (2), and a polymer (3), which is plotted as a triangle.
It is important to determine the equilibrium boundaries that a ternary system presents as they define the demixing threshold conditions and demixing types.The equilibrium points are the binodal curve, which is usually determined mathematically by equalizing the potential energy as ∆ , = ∆ , , where A and B are the polymer-rich and polymer-lean phases, respectively.However, for successful modeling, it is required to accurately describe the inter-compound interaction parameter gij, which is material-dependent and concentration-dependent 8 .Thus, the determination of the equilibrium binodal curve has been empirically identified by the use of the titration method and expanded to all domains by the use of the linearized cloud point (LCP) curve 9 .The titration method aims to empirically determine the first point under non-equilibrium conditions, that is, via the identification of the component concentrations under which the first turbid product is obtained 10,11 .Specifically, different polymer-solvent solutions were prepared due to the unequal increase of viscosity for each polymer.ABS and PAN mixtures were 10, 20, 30, and 40% (w/v), and 5, 10, 15, and 20% (w/v) in DMSO, respectively.Controlled amounts of deionized water, acting as nonsolvent, were titrated against the polymer-solvent mixture at 25 °C until the first cloudy appearance showed up.Upon further stirring, additional non-solvent was titrated if the solution cleared out (metastable point); if the solution maintained its turbidity appearance, this was taken as a valid cloud point.The cloud points were later correlated using the LCP curve 9 using Eq. ( 4) that correlates the weight fraction ∅ of the non-solvent (1), solvent (2), and polymer (3), respectively: The obtained values of the LCP curves for both polymeric systems as well as the regression fitness value R 2 are summarized in Supplementary The relationships established by the LCP can be extrapolated beyond the measured points by iterating all possible weight fractions for all components, knowing that all weight fractions add to the unit (∅1 + ∅2 + ∅3 =1), so the binodal curves for both polymeric systems are plotted for all the feasible domains as shown in Supplementary Fig. 1.
Supplementary Figure 1.Ternary diagrams for ABS and PAN-based ternary systems and their stable and unstable regions.
As seen from Supplementary Fig. 1, the stability region of the PAN system is greater than that of the ABS system, leading to the deposited material being able to absorb more water (higher nonsolvent content) prior to demixing, that is, solidification.This fact explains the faster phaseseparation dynamics of the ABS filaments that yield a higher achievable deposition speed as well as the RED value of ABS closer to unity as shown in Supplementary Table 1.This supports the conclusion that ABS has a weaker affinity with the solvent (DMSO) than the PAN counterpart, and thus, phase separating and demixing faster.

Solvent-polymer interaction
In this study, the kinetics of the polymeric VIPS are studied to enable the fabrication of polymerbased parts.Such kinetics, the resultant porosity, and the developed ternary system stability boundaries are shown to be heavily dependent on the material and printing conditions.It must be noted that such boundaries may not be exactly applicable to other evaporation or thermallyinitiated phase-separation processes (air-casting or TIPS) as their thermodynamics are based on notably different mechanisms, such as evaporation or heating.However, these stability boundaries may be applicable to other liquid-form non-solvent-induced phase separation (NIPS) mechanisms.Such stability boundaries are not affected by the significantly higher kinetics of the solvent-nonsolvent exchange dynamics found in liquid-form NIPS, given that the amount of non-solvent in a given volume is more if found in the liquid-form process (NIPS) than its vapor-form process (VIPS).Nevertheless, the distinct exchange dynamics between the solvent and non-solvent are exemplified by the different non-solvent diffusion rates 12,13 .In that regard, delayed demixing is expected to occur using VIPS due to the lower solvent extraction rates when compared to NIPS, being the final porosity morphology also significantly affected 14 .
Furthermore, the nature of the polymer chain and its composition affect the solubility in a desirable solvent, therefore affecting the printability of polymer-solvent-based inks using a VIPS-based printing process.This relationship is usually expressed using the interaction parameter between the polymer and solvent g23, which can be usually determined by scattering, osmosis, or vaporpressure methods 15 .
Herein, the parameter g23 can be calculated using Hansen solubility parameters 2 as noted in Eq. ( 5): where i=2, j=3, vi is the molar volume of component i, and R and T are the universal gas constant and absolute temperature.The energy values  are listed in Supplementary Table 1, and the molar volume of DMSO is assumed 71.2 cm 3 /g as noted by the manufacturer.By using Eq. ( 5), the interaction parameters g23 obtained for the ABS and PAN are approximately g23 (ABS) = 0.55 and g23 (PAN)= 0.35, respectively.Such differences in the polymer-solvent interaction parameters are the cause of the distinct behavior observed in the miscibility gap, as the g23 increase results in a narrower miscibility gap, and therefore, a reduced stability region [16][17][18] .

Evolution of the solidification front and effect of demixing rate on the filament diameter
Understanding of solidification dynamics is of great importance for a successful VIPS-3DP, which is investigated by studying the evolution of the solidification front after filament printing.For such a solidification dynamics analysis, the solidification timescale is estimated via optical measurements, and Supplementary Fig. 2 presents an example, where it is shown a typical 20% (w/v) ABS-based ink being printed on a glass substrate and observed (bottom view) over time to quantify the solidification rate during the VIPS printing process by capturing the opacification of the filament.For the 20% (w/v) ink under 40% RH, it takes approximately 90 seconds for the opacification to be fully completed.The solidification front travels inwards (white arrows, inset).
The printing speed needs to match the solidification rate for a good shape fidelity and proper interlayer fusion.For the 20% (w/v) ink under higher RH, the opacification is a much more rapid process (<10 seconds as this is the time it takes for the contrast to adjust with our imaging setup).
It should be pointed out that higher concentration ABS inks present significantly more opacity in the liquid phase, inhibiting any meaningful observation of the change in opacity over time.
During the VIPS process, the demixing rate is dependent on the polymer concentration and RH level, where a higher polymer concentration and a higher RH level lead to a higher demixing rate (Supplementary Fig. 3 ) increases from 0.02 µm/s to 0.26 µm/s for the 20% (w/v) ABS as the RH level increases from 40% to 99% (Supplementary Fig. 3(a-b)).A higher demixing rate allows the deposited ink to solidify faster and hold the shape promptly, while a lower demixing rate allows the ink to spread, leading to a flat shape with a larger diameter.As shown in Supplementary Fig. 3(b), the spreading of the t = 0s t = 90s t = 75s t = 60s t = 30s Nozzle initially deposited filament leads to a larger diameter of a solidified filament at low RH levels (such as 40% and 70%), which is due to the slower demixing rate.At 99% RH, the diameter variation between the nozzle diameter and solidified filament significantly decreases, indicating less ink spreading and a better printing resolution.Supplementary Figure 3. Effects of relative humidity on solidification front thickness and filament diameter.(a) Evaluation of the solidification front of 20% (w/v) ABS ink under 40%, 70%, and 99% RH over time using a gauge 22 nozzle (diameter of 0.41 mm).(b) Effect of RH (40%-99%) on the printed 20% (w/v) ABS filament diameter using a gauge 15 nozzle (diameter of 1.36 mm) and a printing speed of 1.5 mm/s and effect of RH on the solidification rate k.Relative humidity (%)

Printability and print shape accuracy
For a detailed printability study during VIPS-3DP, a printability factor ( =  2

16𝐴
) is used to quantitatively evaluate the filament fidelity using a lattice design 19 , shown in Supplementary Fig. 4(a) and (b), where L is the perimeter length of a single cell of the lattice and A is the measured blank area enclosed by the printed strands.Pr of 1 is the ideal value.Pr of less than 1 results in a smaller perimeter to area ratio than that is present in a square geometry.A value greater than 1 also indicates deformed filaments with an increased cell perimeter.Images of printed lattice structures of 20%-30% (w/v) ABS under 40% RH (0.10 mm nozzle at 5.0 mm/s) are shown in Supplementary Fig. 4(c).It shows that the 30% (w/v) ABS ink has the best shape fidelity, while 20% (w/v) ABS filaments spread and fuse together due to the low viscosity and slow solidification.
The measured Pr values are present in Supplementary Fig. 4(d), which are consistent with the print images.The Pr value is also affected by the RH conditions as seen from Supplementary Fig. 4(e) and (f).For the 25% (w/v) ABS ink being printed at 3.0 mm/s using a 0.10 mm nozzle, the 70% RH condition results in the best printability factor around 1. The ink cannot have a good printability factor under 40% RH even when the printing speed is reduced from 5.0 mm/s to 3.0 mm/s in order to let the ink have sufficient time to solidify and retain a good shape fidelity.Even though printing under RH values of 99% may theoretically yield Pr values close to 1, it must be noted that such conditions are not advisable as the ink may solidify too fast and clog and block the dispensing nozzle, resulting in deformed filaments.In terms of the print shape accuracy, resulting polymer parts may experience some post-printing shrinkage, given that solvent is extracted out of the deposited filament as a result of the phase separation process.The level of shrinkage depends on the RH value (Supplementary Fig. 5(a) and (b)), polymer concentration (Supplementary Fig. 5(a) and (c)), and polymer material (Supplementary Fig. 6).To study the effect of RH on the shape accuracy, a circular structure with an inner diameter (ID) of 10.0 mm, an out diameter (OD) of 10.9 mm, and a wall thickness of 0.45 mm (Supplementary Fig. 5(a)) is designed and printed to evaluate the print shape accuracy using a gauge 22 nozzle (diameter of 0.41 mm) at 4.0 mm/s.The 40% (w/v) ABS ink was printed under the 40%, 70%, and 99% RH levels, and the printing results are shown in Supplementary Fig. 5(b).The printed circles are smooth with good shape fidelity, and the measured ID, OD, and wall thickness are close to the design values (Supplementary Fig. 5(d)).Of them, the 40% RH level results in the best print shape accuracy with slight shrinkage.
To study the effect of polymer concentration on the shape accuracy, a circular structure was similarly printed while using the 20%, 30%, and 40% (w/v) ABS inks using a gauge 22 nozzle (diameter of 0.41 mm) at 4.0 mm/s under 40% RH (Supplementary Fig. 5(c)).The 20% ABS ink has severe spreading due to the low polymeric concentration, resulting in a higher OD and a lower ID than design values, while the 30% and 40% (w/v) ABS inks result in relatively accurate prints due to the combined effect of ink viscosity (for spreading) and solidification speed (Supplementary Fig. 5(e)).Of them, the 40% (w/v) ABS ink results in the best print shape accuracy with slight deviation.
As observed, the good combination of RH and polymer concentration is 70% RH and 30-40% (w/v) ABS during VIPS-3DP ABS printing.Supplementary Figure 5. Post-printing solidification-induced filament diameter variation as a function of RH and polymer concentration.(a) Shape design, (b) printing results of 40% (w/v) ABS under 40%, 70%, and 99% RH levels, (c) printing results of 20%, 30%, and 40% (w/v) ABS under 40% RH, and (d-e) measurements of the inner diameter (ID) and out diameter (OD) of (b-c), normalized to the design values (dashed lines).(Scale bars: 2 mm) To study the effect of polymer material on the shape accuracy, both ABS and polyacrylonitrile (PAN) were used as two example polymer materials with different chemical affinities with the solvent (DMSO).Because PAN has a better affinity due to a lower relative energy difference (RED) value than ABS (Supplementary Table 1), it leads to a slower phase-separation process.
The slow solidification rate of PAN allows it to shrink as it gets slowly solidified and thus, presenting a high post-printing shrinkage degree and smaller pore sizes.In contrast, ABS solidifies fast due to the rapid solvent extraction, having a low post-printing shrinkage degree.An example of the as-printed and as-solidified parts, 20% (w/v) PAN-based semicircles before and after solidification are shown in Supplementary Fig. 6(a) and (b), presenting a clear reduction in diameter as indicated by black arrows.For comparison in terms of the print shape accuracy, the shrinkage magnitude of the 20% (w/v) ABS and 20% (w/v) PAN inks is quantified to be approximately 3% and 30%, respectively, as seen in Supplementary Fig. 6(b), showing that polymers having a low RED value may shrink more significantly.

Morphology and porosity of polymeric filaments
The printed filaments present a dense skin in the outermost layer due to being the first surface to have water-induced phase separation.Such dense skin layer with low porosity has been reported in similar VIPS-based solidification cases 20,21 .Generally speaking, the complete formation of dense skin layers during VIPS 3DP is largely dependent on the phase-separation kinetics of a polymer solution.Faster phase-separation kinetics promote the formation of hardening skin layer that leads to consecutively deposited layers not being fused well due to the significantly different hardening degree of their contact surfaces.If a previously deposited layer has a notable solventextraction time in the outer surface layer under fast phase-separation kinetics, it may harden well and form a dense skin layer before a subsequent filament is deposited.The difference in hardening status and the presence of a thin layer of solvent and non-solvent may result in insufficient bonding between two consecutive layers and a visible interface (Supplementary Fig. 7(a)) using a scanning electron microscope (SEM, Hitachi S3000, Hitachi High-Technologies Co., Tokyo, Japan).This phenomenon is particularly pronounced when ABS is used.
In addition, faster phase-separation dynamics play an effect on the inner porosity morphology as the presence of macro-and micro-voids can be found in the innermost layers, which are the last ones to demix and solidify 21 (Supplementary Fig. 7(a), inset).On the other hand, if the phaseseparation kinetics are slow, the solvent/non-solvent exchange-related mass transfer may be reflected as a reduction of the filament cross section 22 given the lower hardening degree of the filament, leading to well-bonded filaments (Supplementary Fig. 7(b)) and unnoticeable pores at the micron level being present in the filament cross section (Supplementary Fig. 7(b), inset).This is the case when PAN is used.
From a morphological perspective, distinct porous morphologies can be observed during ABS and PAN printing.ABS does present a clear porous morphology (Supplementary Fig. 8(a)) where the pores are in the range of the hundreds of microns in both the intermediate and bottom regions (Supplementary Fig. 8(b)-(c)).The PAN counterparts do not show any obvious and observable pores in the micron range (Supplementary Fig. 8(d)).Further mercury intrusion porosimetry (Poremaster, Quantachrome, Anton Paar, Graz, Austria) reveals that the PAN-based structures are porous in nature whose pore size averages 140 nm in diameter (Supplementary Fig. 8(e)).Interestingly, the symmetry of the porosity observed for the fast-separating polymer in Supplementary Fig. 8(a) is an adaptation of the reported one-dimensional (1D) asymmetry of the thin membranes obtained via VIPS 23 .
If the ABS concentration increases to 40% (w/v), the asymmetric porosity morphology resembles the one reported elsewhere 24 since the phase-separation kinetics are higher due to the less amount of solvent to be extracted, and the results are shown in Supplementary Fig. 9.The printed filaments present a dense skin (Supplementary Fig. 9(a)) in the outermost layer due to being the first surface to be in contact with the water mist, which acted as a non-solvent.As such, this dense skin layer with low porosity is expected to be found in all cases 20,21 , and such a morphology is observed in 20, 40, and 60% (w/v) ABS.The intermediate region of the filament presents a cell-like pore (Supplementary Fig. 9(b)).
The presence of macro-and micro-voids can be found in the innermost layers, which are the last ones to demix 21 (Supplementary Fig. 9(c)) that are in the range of hundreds of microns.
As seen in Supplementary Fig. 10, for the selected ABS concentrations (40% and 60%, typical concentrations used for polymer printing in this research) and RH conditions (40%, 70%, and 99%) investigated, the overall porosity varies between 35% and 45%, and there is no significant difference observed among these samples (Supplementary Fig. 10 (a)).Most of voids found on the cross section of whole filaments present a pore size between 1 and 5 µm (Supplementary Fig. 10 (b-c)).Although no change in the overall porosity is observed as the RH level increases, the average pore size decreases as observed in the distribution plots (Supplementary Fig. 10 (b) (from 2.78 to 2.59 to 2.34 µm) and (c) (from 2.81 to 2.50 to 1.87 µm)) wherein the smaller voids tend to have a higher frequency.This phenomenon is observed for both 40% and 60% (w/v) ABS, and the decrease in the pore size is more pronounced with the 60% (w/v) ABS, meaning that higher concentration polymers such as ABS may be more prone to form smaller voids under higher RH levels.60% ABS, 99% RH

3D printing of overhang filaments in air
Successful deposition of completely overhanging continuous polymer-based filaments in air is achieved by using 40% (w/v) ABS under the nebulized condition, with a nozzle speed of 1.0 mm/s.The printing conditions were fixed, so the resulting filament height was approximately 1 mm.The deposited filaments were approximately 1 mm in diameter and could withstand a maximum span of 35 mm without noticeable gravity-induced deflection as seen from Supplementary Movie 1. Solutions with lower polymeric concentrations and/or slower speeds resulted in discontinuities in the deposited line or significant filament deflection.

Thermogravimetric analysis
The dynamics of debinding process of green parts, resulting in the polymer binder being volatilized, are studied herein.The debinding performances of ABS and PAN are investigated using thermogravimetric analysis (TGA, Q5000, TA Instruments) to assess the mass loss during various thermal cycles using 5 mg samples.The TGA study was conducted with nitrogen flowing at 60 mL/min and heating rates of 5 and 10 °C/min.The TGA curves of ABS and PAN are shown in Supplementary Fig. 11(a) and (b).ABS rapidly decomposes in a single step from 345°C, and the polymeric mass decays to a remainder of 3% at 420°C and has no significant amount remaining (<1%) at 470°C.PAN decomposes on a significantly different behavior: no mass loss is observed until the point of 270°C.From that temperature point, three decomposition rates are observed: a rapid mass loss rate in the 270-300°C region, a moderate mass loss rate in the 300-470°C, and a low mass loss rate occurring from 470 to 1000°C.It is noted that PAN samples still present a remainder of 20% of mass at the highest operative temperature of 1000°C.The high withstanding temperature of PAN is the main reason for it being used as a precursor of carbon fibers given its high non-volatility at high temperatures, and this property is due to its cyclized structure and highcarbon content in 1000-1300°C 25,26 .Thus, the use of PAN as a binder implies the presence of high carbon contents that may impact its suitability to bind metals due to undesired carbide formation.
As such, ABS is considered a better candidate to act as a binder for VIPS-3DP purposes, in particular, for metal printing, given that its complete polymer burnout occurs in the 420-470°C range.

Sintering dynamics 1.8.1 Sintering cycle and conditions
The sintering cycle is detailed as follows: the heating and cooling rates were 5°C/min; a dwell time of 30 minutes at 300°C was used to avoid thermal cracking on the alumina tube; the debinding cycle was based on the TGA data and set at 500°C for 60 min; and the sintering temperature was set at 1300°C for 2 hours, where solid-state sintering of the 316L-based powders was induced under vacuum conditions.The temperature profile of the furnace is shown in Supplementary Fig. 12.
Supplementary Figure 12.Thermal profile for debinding and sintering of ABS-316L-based printed parts.
Supplementary Table 3

Metallographic studies and chemical composition of sintered parts
Further characterization of the printed 316L parts was carried out to determine their final composition.The printed parts were cut using a diamond saw, mounted in epoxy resin, and polished using silicon carbide diamond pads using water as a lubricant.The polished surfaces were analyzed using energy-dispersive spectroscopy (EDS, Hitachi S3000, Hitachi High-Technologies Co., Tokyo, Japan).As shown in Supplementary Fig. 13(a), the use of ABS results in an approximately 8% porosity present, and the pores are located along the grain boundaries.The pore size is on the order of 4 μm.Interestingly, PAN-based samples do not have such pores (Supplementary Fig. 13(b)), which may be due to the significant shrinkage of the polymeric phase during printing (Supplementary Fig. 6(b)).From a composition perspective, the EDS results confirm the presence of the same elements as the initial composition of the 316L powders if ABS is used as a binder due to its easiness of volatilization (Supplementary Fig. 13(c)), and there is no noticeable residue from the binder composition.Differently, the PAN-based sintered parts do show a significant presence of carbon (>30%) (Supplementary Fig. 13(d)), which is due to the highcarbon content of the PAN structure in the temperature range of 1000 -1300°C 25,26 .
It is noted that under the current vacuum-based sintering, no noticeable loss of chromium is observed as expected due to the high vapor pressure of chromium at elevated temperatures 27 .This is caused by the current vacuum conditions not meeting the extremely high vacuum level of industrial powder metallurgy routes.1.9 Densification results

Mechanical properties of dense parts
Flat stainless-steel dog-bone specimens were printed using different material conditions, and their dimensions were approximately 30-40 mm long and 8-12 mm wide after sintering.Various mechanical properties such as ultimate tensile strength (UTS), were measured (6800 Series, Instron, Norwood, MA, USA) using a crosshead speed of 6-7 mm/min (ASTM E8-03).Three independent trials were conducted under the same conditions.The minor differences in UTS values are attributed to the presence of porosity, microcracks at the surface of the powders, impurities, and segregation 28,29 when compared to those made using the other fabrication routes.

Post-processing feasibility
To demonstrate the ability of the printed parts to undergo regular post-processing machining operations, some stainless-steel parts were sandblasted and tap threaded.Supplementary Fig. 14(a) presents a different view of the sandblasted part in addition to those shown in Fig. 3.The sandblasting operation was conducted using a sand blaster (Blast It All, Model #3636-7DC, Salisbury, NC), which successfully removed the outer layer of the sintered part formed by the debinding and sintering debris, presenting a smooth and shining surface without compromising the integrity of the parts that include thin blade-like features.The near-theoretical density of the steel allows the thin features to withstand the sandblasting process as shown by the lack of deformation of the thin blades.
Moreover, a tap threading process was applied to a 3D printed and sintered zig-zag tube (Supplementary Fig. 14(b)), where a thread of 1/8 NPT was added to one end of the tube (Supplementary Fig. 14(c)).A push-to-connect fitting was assembled using the tap thread process, and no liquid leak was observed when blue-colored water flew through (Supplementary Movie 2).

Horn antenna printing
With the development of modern wireless communications, wider and higher frequency bands are employed to provide more bandwidths for high data-rate transfer, which results in the requirement of more complex and smaller antennas 30 .Generally speaking, the behavior of working antennas is based on the assembly of various sets of parts, which include the parabolic reflector, feeding guidelines, and supports for stability purposes.Therefore, the manufacturing route of such antennas requires specialized labor and may present high costs and time-consuming assembling steps.Thus, an emerging trend of designing antennas from traditional two-dimensional (2D) to 3D has gained momentum in recent years, as it provides more design freedoms, improves space utilization efficiency, simplifies manufacturing routes, and reduces manufacturing costs.As such, 3D printing has been pioneered as a powerful tool for antenna manufacturing that easily adapts to the variability of required designs in a cost-effective manner.
Herein the design of the fabricated horn antenna and its measured results are discussed.The horn antenna is known as a radiating element that has the shape of a horn.It can be considered as a radio frequency (RF) transformer or impedance match between the waveguide feeder and free space, which has an impedance of 377 Ω.The behavior and efficiency of the horn antennas are based on their geometry, specifically on the dimensions of the opening to be able to absorb the electric and magnetic field radiations, which are named the E-plane and H-plane, respectively 31 .For instance, the general parameters to consider in the design phase for a horn antenna are shown in Supplementary Fig. 15.For a desired operating frequency of a horn antenna, the standard waveguide with specific dimensions of a and b can be found elsewhere 32 .The designed horn antenna has a rectangle waveguide with the dimensions of a and b (Supplementary Fig. 15 The gain G of such a horn antenna is given in Eq. ( 6): Furthermore, a steel-based horn antenna with the optimized dimensions was fabricated by VIPS-3DP (Supplementary Fig. 17).Due to the limitations of the sintering chamber, its final geometry is reduced as a1 = 21.5 mm, b1 = 14.9 mm, Re = Rh = 9.5 mm, ρ1 = 13.6 mm, ρ2 = 12.6 mm, ρe =15.5 mm, and ρh = 16.6 mm.The change of the dimensions notably modifies the irradiation performance, as the printed antenna has a -10 dB bandwidth of 4.4 GHz, and the lowest reflection coefficient (S11) is measured as -36.58 dB at 24.5 GHz (Fig. 3(o) in the manuscript, solid line) using a vector network analyzer (ZVA67, Rohde & Schwarz, Germany).This is higher than the ideal working frequency of 22 GHz, which may be due to minor dimensional differences.Further simulations based on the printed dimensions yield the modeling results (Fig. 3(o) in the manuscript, dashed line) matching these obtained experimentally.While an acceptable behavioral agreement between the measured and simulated results is observed, the small noise-based difference in the magnitude of the reflection coefficient S11 may be caused by two factors: the inner surface of the horn antenna presenting a step-wise morphology due to the hardening of adjacent filaments and the measurement-induced noise.In conclusion, the fabricated antenna presents good performance from 22.3 GHz to 26.7 GHz, which demonstrates the successful application of the VIPS-3DP technology for RF uses.
Supplementary Figure 17.Photographs of the 3D printed horn antenna.

Respirator composite printing 1.11.1 Printing
The preparation of the printing material, mixing the antiviral particles with a dissolved polymer solution and obtaining a homogeneous ink suspension for 3D printing, involves three steps.The first step consists of mixing the polymeric inks separately: filaments of acrylonitrile butadiene styrene (ABS, Stratasys, MI, USA) and thermoplastic polyurethane (TPU, Ellastollan, BASF, Germany) are mixed with the dimethyl sulfoxide solvent (DMSO, Fisher, NJ, USA) with concentrations of 40% and 20% (w/v), respectively.From a polymeric view, a mixture of dissolved ABS and thermoplastic polyurethane (TPU) is chosen due to the desired balance between the stiffness and elasticity brought by the ABS and TPU phases, respectively.As such, the printed structure is stiff enough to enable self-supporting printing while being flexible to accommodate different face profiles to ensure a good fit and sealing.The polymeric mixtures are continuously stirred overnight using a roller mixer (DLAB Scientific, CA, USA).In the second step, both polymeric inks are mixed at a 2:1 ratio (ABS:TPU).The last step includes the addition of copper powders (10% (w/w)) to the dissolved polymer mixture.In order to guarantee the homogeneity of the suspension, the ink is mixed using a centrifugal mixer (AR-100, Thinky, CA, USA) for 2-3 minutes prior to loading the dispenser syringe.
Using a 3D extrusion printer (Hyrel Engine SR, Hyrel, USA), the ink is dispensed for a layer-bylayer deposition.This printing process is carried out while water is nebulized to the printing zone where the antiviral particle-dissolved polymer composite part is being printed, partially solidifying the polymer and entrapping the powder in the deposited filament due to the induced phase separation process.
The in-process solidification method is controlled to occur partially only for a balance of good fusion between two consecutively deposited layers and enough stiffness to sustain the following printed layers in air due to the existence of solidified polymer.For complete solidification throughout a printed part, it can be immersed, if needed, in a coagulation bath to fully remove the solvent.The collected solvent-relevant solution from the printing chamber and coagulation bath are post-processed to reclaim the solvent for its reuse, minimizing its environmental impact.After a drying cycle, an ABS-TPU-based 3D printed respirator is obtained.

Antibacterial testing
Staphylococcus aureus (ATCC 6358, ATCC, Manassas, Virginia, USA) cells were cultured in 1 mL of trypticase soy broth (TSB) (BD Difco, Becton, New Jersey, USA) in 24-well plates, along with the printed lattice-based ABS-TPU-copper disks of 8 mm in diameter.Additional positive controls (without disks) and negative controls (without S. aureus inoculum) were used.All groups were incubated at 37 °C, 150 rpm for 24 h.Each test run contained a single specimen, and it was run in triplicate.The antibacterial performance of the printed structures was determined using the Colony Forming Unit (CFU) counting method.For the testing of the reused disks, the same sample disks used previously were washed twice with phosphate-buffered saline (PBS, VWR, Radnor, Pennsylvania, USA), and once with 75% ethanol prior to the day of the test, then soaked in 75% ethanol for 30 min.Sterile disks were tested as previously described.

Mechanical properties and fractometry of porous parts
Specimens of the identical geometry as specified in Supplementary Note 1.9.1 (mechanical properties of dense parts) were printed at different material porosity levels.
Dense parts (∼92-100%, close to the theoretical density) present an ultimate tensile strength (UTS) of approximately 480 MPa, but the presence of pores detriments the mechanical properties due to the stress concentration on the boundaries surrounding the pores.Specifically, the increasing presence of porosity yields a decrease of UTS values down to 190, 120, and 95 MPa for porosities of 62, 52, and 42%, respectively.
From a fractometry perspective, the dense specimens present ductile trans-granular fracture as shown by the presence of dimples (Supplementary Fig. 20(a), red arrows), while the presence of high porosity level results in an unclear, non-obvious fracture mechanism 35 (Supplementary Fig. in an incubator with 95% humidity and 5% CO2 at 37 o C. The complete medium was replaced every 2-3 days.When the cells were 80-90% confluent, the cells were harvested with TrypLE Express (Fisher Scientific, Pittsburgh, PA).Scaffolds with dense and porous filaments were sterilized with 70% ethanol for 30 minutes and washed with PBS (Fisher Scientific, Pittsburgh, PA) three times.
To study the cell proliferation and migration over the printed scaffolds, 150 µL MC3T3-E1 cells (4×10 5 cells/mL) were seeded on each disc scaffold in a 48-well plate and incubated for 30 min, and then 0.5 mL complete medium was gently added in each well.Then the samples were cultured in the incubator with 95% humidity and 5% CO2 at 37 °C.The medium was changed every other day.The samples of the dense, mediumly porous, and highly porous discs were stained with a final concentration of 10 µg/mL Hoechst 33342 (Sigma-Aldrich, St. Louis, MO) to stain nuclei in blue and a final concentration of 10 µg/mL fluorescein diacetate (FDA) (Sigma-Aldrich, St. Louis, MO) to stain live cells in green on Days 1, 3, and 7, respectively.The images were captured using an EVOS FL inverted fluorescence microscope (Fisher Scientific, Waltham, MA).
The cell-fixing protocol used herein was the one previously used elsewhere 36 .Briefly, the samples were removed from the wells after culturing and rinsed using PBS three times.The samples were then immersed in a solution of 4% formaldehyde (Sigma Aldrich, St. Louis, MO, USA) in DI water for 25 min at 4 °C and 0.2% Triton X-100 (Sigma Aldrich, St. Louis, MI, USA) for 10 min to fix the cellular matter.The samples were further rinsed with an ethanol-DI water mixture with an increasing ethanol concentration (15%, 30%, 45%, 60%, 75%, 90%, and 100%, 10 min each).The samples were then air-dried using hexamethyldisilazane (HMDS, Sigma Aldrich, St. Louis, MO, USA).Finally, the cell-laden metallic samples were coated with Au-Pd for imaging using scanning electron microscopy (Hitachi S3000, Hitachi High-Technologies Co., Tokyo, Japan and Phenom XL, Thermo Fisher Scientific, Waltham, MA, USA).

Scaffold biocompatibility evaluation
As seen from the fluorescent images of Supplementary Fig. 21, the cell-seeded dense filament scaffolds are biocompatible with MC3T3-E1.It is noted that the cell density is visibly lower when compared to that of their porous counterparts.

Figure 11 .
TGA curves of (a) ABS and (b) PAN.

Figure 13 .
SEM images and EDS readings of cross sections of ABS-and PANbased parts after polishing and etching.(a) ABS part and (b) PAN part SEM images.EDS readings of (c) ABS-based and (d) PAN-based parts.
(a)), whereas the a1 and b1 variables denote the side lengths of the horn end at the E-plane and H-plane, respectively.Supplementary Fig.15(b) represents the antenna geometry in the E-plane, where the Re variable represents the distance between the top and bottom horn apertures.The horn in the E plane is determined by the diameter of ρ1 and ρe.Similarly, the horn in the H-plane is defined by the parameter of Rh, ρ2, and ρh as illustrated in Supplementary Fig.15(c).A SubMiniature version A (SMA) connector is employed to transfer the signal from the coaxial cable to the rectangle waveguide.Supplementary Figure 15.Geometrical configuration of horn antenna.(a) 3D view.(b) E-plane.(c)H-plane.

Figure 21 .
Fluorescent images of cells on the surface of dense scaffolds after culturing for (a) 1, (b) 3, and (c) 7 days.Scale bars: 1 mm.a b c . RED calculations.