2.1.

Equipment for UV Lithography

2.1.1.

Thorlabs UV curing LED system CS2010

“Thorlabs” UV curing LED system “CS2010” (Fig. 1) was used to investigate curing time of bio-based resins. The UV LED power was $P=270\mathrm{mW}$ (corresponding to data sheet), irradiation wavelength $\lambda =365\mathrm{nm}$, and divergence $\alpha$ from the optical axis 60 deg. The distance $D$ between a sample and LED chip was fixed at 4 cm. An exposed area was calculated by Eq. (1) and was equal to $150{\mathrm{cm}}^2$. Consequently, light intensity $I$ was evaluated by Eq. (2) and was $1.8\mathrm{mW}{\mathrm{cm}}^{-2}$

Eq. (1)

$$S=\pi {(D\tan \alpha )}^2,$$

Eq. (2)

$$I=\frac{P}{S}.$$

Fig. 1

Thorlabs UV curing LED system CS2010. LED is turned on and exposes $150\text{-}{\mathrm{cm}}^2$ area (in blue color). There is a sample at the center of the area.

2.1.2.

Autodesk 3-D printer Ember

“Autodesk’s” open-source 3-D optical printer Ember [Fig. 2(a)] was implemented for a dynamic projection lithography (DPL). The light source (projector “Wintech PRO4500”) uses 5-W power and 405-nm wavelength, an LED diode, and a digital micromirrors device, which consists of $1280\times 800$ (total 1.024 million) micromirrors spaced in the compact $9855\times 6161.4\mu {\mathrm{m}}^2$ footprint. One micromirror pitch size is $7.6\mu \mathrm{m}$, which creates a $50\times 50\mu {\mathrm{m}}^2$ projection of a single image pixel, which defines device resolution in $X$, $Y$ plane. Such an optical system ensures $I=18.61\mathrm{mW}{\mathrm{cm}}^{-2}$ light intensity projected through the UV light transparent polydimethylsiloxane (PDMS) window where the printing process occurs. The intensity is high enough to cure standard resins rapidly. Moreover, it can be modified by changing the electric current flowing through the UV LED. First of all, CAD model is sliced into cross-sectional layers that are transferred to the printer. Then, printing process starts and the images are projected on the PDMS window. The exposed area makes resin to cure one layer at a time and adhere to the build platform or previously formed layers. Single layer height can be from 10 to $100\mu \mathrm{m}$ and is controlled by $Z$-axis stepper and exposure time. The build platform dimensions are $64\mathrm{mm}\times 40\mathrm{mm}\times 134\mathrm{mm}$ in $X$-, $Y$-, and $Z$-axes. The print preparation software is called “Print Studio,” which allows to fix and prepare 3-D files and then delivers them wirelessly to the device. Device working principle is explained in Fig. 2(b).

Fig. 2

(a) An open-source 3-D optical printer Ember and (b) its principle working scheme. (I) Irradiation from the projector reflects to the mirror and exposes resin trough the PDMS window, (II) after exposure, the tray slides (1) and the build head rises up, and (III) the tray comes back in it primary position (1) and the build head lowers (2). (I) side view and (II and III) front views.

2.2.

Commercial Resin PR48 and Bio-Based Resins

Commercial resin Autodesk “PR48” was used as a standard printing material. The resin is acrylate based and mainly consists of methacrylate oligomers and monomers.²⁶ Exact components are EBECRYL 8210 (Allnex, 39.776% w/w) and SR 494 (Sartomer, 39.776% w/w). Being enriched with double bonds, they are suitable for polymerization reaction. Photoinitiator (PI) diphenyl(2,4,6- btrimethylbenzoyl)phosphine oxide (TPO, Esstech, 0.4% w/w) is added to increase photoreactivity. Usually PI used for tabletop SLA makes the resin sensitive to 420-nm wavelength light and lower, with a peak absorption around 365 nm. Also, reactive diluent (RD) plays important role in resin’s consist. For example, Genomer 1122 (Rahn, 19.888% w/w) is used to reduce viscosity of the base resin. What is more, RD can react with curing agents to become a part of the polymerization reaction and optimize cured resin’s properties, such as impact strength, adhesion, and flexibility.²⁷ UV blocker 2,2-(2,5-thiophenediyl)bis(5-tertbutylbenzoxazole) (OB, Mayzo, 0.16% w/w) is used in PR48 resin with the purpose to control the light penetration depth, which is needed to confine cured layer thickness. Additionally, various pigments can be mixed up to make resin colorful. However, dye concentration is important, because too much of the pigment makes it to settle out faster. This results in longer exposure time. PR48 initiating system is free radical polymerization proper for a UV lithography.

GDGE (technical grade, Sigma-Aldrich) and ELO (Chemical Point, Germany) were chosen as monomer sources for bio-based resins. Their absorption spectrum is shown in Fig. 3. Both GDGE and ELO can be obtained from some of the cheapest and most abundant nontoxic annually renewable natural resources available in large quantities (in accordance with statistical data, the total production of vegetable oils worldwide amounted to about 177.73 million tons in 2015/2016 year²⁸ and about 2 million tons of crude glycerol consistently reached the market every year²⁹). GDGE and ELO have a high content of reactive functional groups making them suitable components for the preparation of bioresins.^19,30 However, these monomers exhibit low photoreactiveness, and thus efficient PI is mandatory. To cure bio-based resins employing Thorlabs UV Curing LED System CS2010, the following composition was GDGE (or ELO), 3,4-epoxycyclohexylmethyl-3,4-epoxycyclohexane carboxylate (Sigma-Aldrich) as RD, and triarylsulfonium hexafluoroantimonate salts (mixed 50 wt. % in propylene carbonate, Sigma-Aldrich) as PI. Components were mixed accordingly to the weight content of 1 mol. %:30 mol. %:3 mol. %. Polymerization mechanism of such system was mainly cationic since exposure to the UV light of the systems containing no cationic PI did not lead to the appreciable hardening.²⁷ To cure the resins using Ember, the initiation mechanism was changed to free radical promoted cationic polymerization (FRPCP) following Lalevée protocol,³¹ due to the knowledge that the previous PI system was not sensitive to 405-nm wavelength. Chemical compounds were phenylbis(2,4,6-trimethylbenzoyl) phosphine oxide (BAPO, Sigma-Aldrich), diphenyl(2,4,6-trimethylbenzoyl)phosphine oxide (TPO, Rahn) as radical PIs, diphenyl iodonium hexafluorophosphate (${\mathrm{Ph}}_2{\mathrm{I}}^{+}$, Sigma-Aldrich) as a cationic PI, and N-vinylcarbazole (NVK, Sigma-Aldrich) as a promoter. The investigated photoinitiating systems were based on $\mathrm{PI}/\mathrm{NVK}/{\mathrm{Ph}}_2{\mathrm{I}}^{+}$ with weight contents of 3% and 2% for NVK and iodonium salt, respectively.

Fig. 3

GDGE and ELO absorption spectrum ($\lambda$ range from 250 to 1500 nm). ELO absorbs several times more irradiance than GDGE almost in the whole measured range.

All composites were mixed by these steps: weighted ingredients, poured into stirring bowl without addition of organic solvent, and stirred with magnetic mixer in the dark for at least 12 h (overnight) until the homogeneous solution was obtained. After it, the composites were ready to use.

2.3.

Other Equipment and Materials

Thermal power laser measurement sensor “OPHIR 3A-FS” was employed to measure irradiance intensity on the Ember printing area. To evaluate custom-made resins, absorption spectra spectrophotometer “SHIMADZU UVProbe” was implemented. “Q150R” rotary-pumped sputter coater was used to metallize photopolymerized samples as the additional metal layer provides higher conductivity and optical reflectiveness making optical profilometer (OP) and scanning electron microscope (SEM) measurements easier. For characterization, SEM “HITACHI TM-1000” was employed. An OP “SENSOFAR $\mathrm{PL}\mu$ 2300” was used to obtain formed features profiles and determine their height dependence on absorbed irradiance energy dose.

Isopropyl alcohol was used to leach out noncross-linked monomers from the objects printed out of standard resin PR48. Approximate rinsing time varied from 10 to 15 min. Acetone as a more chemically reactive solvent was implemented to dissolve uncured bio-based resins (rinsing duration 1 to 2 min).

Autodesk “AutoCAD 2017” student version software was employed for CAD models designing.

3.1.

Photostructuring Employing CS2010

Estimated exposure time for commercial resins Autodesk PR48 is a couple of seconds. Ember software Print Studio provides possibility to modify exposure time tuning it to the requirements dictated by any particular material. By default settings, it is set up 1.8 s. However, the first layer is very important for successful fabrication of the whole 3-D object. It must firmly adhere to the building plate, thus longer exposure usually is used (8 s). Naturally derived resins are less sensitive because of long alkyl chains of monomer molecules. For this reason, UV dosages required to polymerize commercial and bio-based resins were compared. Thus, a UV lithography experiment was employed [Fig. 4(a)]. $10\text{-}\mu \mathrm{l}$ droplets of the resins were cast on the $150\text{-}\mu \mathrm{m}$ glass. The hydrophilic interaction between the glass surface and droplets was strong enough to keep the samples upside-down. It ensured adhesion of the formed structure after exposition to the substrate. After exposing the material to UV radiation of 365 nm, the samples were immersed into the solvent (acetone) for 1 min. If the droplets were polymerized, it stayed on the substrate, and if not, it was dissolved. Also, selective UV lithography was performed [Fig. 4(b)]. For this, a micropatterned amplitude mask was used. The mask was the same $150\text{-}\mu \mathrm{m}$ glass slide, coated with 200-nm gold layer. The gold was selectively removed with a femtosecond pulsed laser to form a binary mask (method is described elsewhere).³² The width of abraded lines was 150 to $400\mu \mathrm{m}$, and the opaque slits between them were $150\mu \mathrm{m}$. After exposing the material to UV radiation through, the mask selective photopolymerization was observed. Acetone dissolved the UV unaffected area, leaving only the formed microstructures.

Fig. 4

Resin photocuring scheme: (a) UV lithography: (1) sample droplet exposed to UV ($\lambda =365\mathrm{nm}$), (2) polymerized droplet, and (3) adhered sample to the substrate after rinsing in acetone. (b) Selective UV lithography: (1) exposition through mask, (2) irradiation affected areas, and (3) dissolved sample, only formed microstructures.

First, commercial resin PR48 was tested. “Formlabs” “Clear” as additional reference was also investigated. The results showed that both resins are completely polymerized after 2 s of exposure, what corresponds to energy dose $E=3.6\mathrm{mJ}{\mathrm{cm}}^{-2}$, calculated by Eq. (2) times exposure time. The formed structures are demonstrated in Fig. 5. They consist of single lines, which are strict, defined, thin, and according to the shape of the mask. Keeping the same order, experiments with bio-based resins followed. The exact exposure parameters required to induce photopolymerization reaction in the custom-made resins were unknown. Consequently, the experiment was started with UV lithography to deduce the required dose of irradiation. It was determined that both compositions need to be exposed at least 150 s to UV light (which equals to $E=270\mathrm{mJ}{\mathrm{cm}}^{-2}$). Exposed droplets of the resins solidified, did not dissolve in acetone, and were easily removable from the substrate. Then, selective UV mask lithography was used. As the mask added additional glass layer, more light was absorbed before affecting the droplets resulting in even longer exposition. For composition with GDGE, time increased to 260 s ($468\mathrm{mJ}{\mathrm{cm}}^{-2}$), and for composition with ELO, time increased to 220 s ($396\mathrm{mJ}{\mathrm{cm}}^{-2}$). Despite this, it was possible so obtain selectively polymerized structures (Fig. 6).³³ Compared with objects made from commercial resins, these also had single lines yet with the tendency of merging into the one object. Also, the whole structure was noticeably thicker.

Fig. 5

SEM images: (a) micropatterned amplitude mask. Microstructures after 2-s exposure to UV: (b) Formlabs Clear and (c) Autodesk PR48.

Fig. 6

(a) Polymerized linseed oil-based resin droplets (after 260-s exposure to UV radiation), (b) selectively polymerized droplets (after 220-s exposure to UV radiation), and (c) enlarged view of the droplet (image obtained with SEM).

3.2.

Determination of Ember’s Spatial Resolution

Resolution is a very important specification for determining printer’s quality. It is defined as a distance among features yet commonly used as a feature size as well. The goal of determining the spatial resolution was to deduce the thinnest separate lines that can be printed using Ember. Thus, a special CAD model was designed (Fig. 7). It consisted of thin vertical walls attached to the base. To determine if there is a resolution dependence on the feature aspect ratio, the walls were made of various altitude $a$, which starts from 0.3 mm (counting from the top of the base) and increases up to $1.5\mathrm{mm}\times 0.3\mathrm{mm}$ step. The highest one was 2 mm. The range of thickness $d$ of the walls was from 50 to $250\mu \mathrm{m}$ every $50\mu \mathrm{m}$ (total five different models). The slit $l$ between walls varied from 50 to $300\mu \mathrm{m}$ every $50\mu \mathrm{m}$. Fixed thick walls with fixed width slits were arranged in arrays. Such arrays were separated from each other with 1-mm gaps. The CAD models were printed with Ember out of standard resin PR48. After manufacturing, the objects were immersed into a solvent. The printed samples were characterized using SEM. Figure 8 shows that the thinnest printed walls were approximately $d=110\mu \mathrm{m}$ (image a). $l=200\text{-}\mu \mathrm{m}$ slit between them was enough to have fully separated walls (image c). Features with a narrower slit between them used to merge to the one object (image b). Compared (a) and (b), it is noticeable that lower features tended to separate from each other better than higher ones with the same slit. It is visible that walls of 0.6-mm height stood apart better than 2-mm ones. It can be explained by overexposure, which occurs during a new layer formation. Light transmits deeper into the previous layers and additionally exposes it. For this reason, the polymerization reaction lasts longer, and the feature spreads wider and merges with the other closest feature. Also, in a similar way, it was deduced that Ember enables printing of $200\text{-}\mu \mathrm{m}$-diameter circle holes and allows the creation of $250\text{-}\mu \mathrm{m}$-width rectangular holes.

Fig. 7

3-D model of resolution test in “AutoCAD” software.

Fig. 8

Ember printed feature size test sample SEM images. $d$ and $a$ represent wall thickness and height, respectively, and $l$ is the distance between walls. (a) $d=100\mu \mathrm{m}$, $a=0.6\mathrm{mm}$, $l=150\mu \mathrm{m}$; (b) $d=100\mu \mathrm{m}$, $a=2\mathrm{mm}$, $l=150\mu \mathrm{m}$; (c) $d=150\mu \mathrm{m}$, $a=2\mathrm{mm}$, $l=200\mu \mathrm{m}$. Compared (a) and (b), higher features tend to merge more than smaller ones. Image (c) shows that features are well separated using more than $150\text{-}\mu \mathrm{m}$ gaps.

3.3.

Formation of Woodpile Scaffold Structures

A lot of results were acquired in collaboration with colleagues from the Institute of Biochemistry, Department of Biological Models (Vilnius University). One of the most frequent tasks was to print 3-D microporous woodpile scaffolds for cell proliferation.³⁴ Accordingly, it was necessary to investigate if it is possible to manufacture required structures with an available DLP 3DP. The scaffolding microarchitecture models were designed (Fig. 9). Scaffolds’ features sizes were set up according to printers’ resolution capabilities: $d=100\mu \mathrm{m}$, $l=200\mu \mathrm{m}$; $d=200\mu \mathrm{m}$, $l=150$ and $200\mu \mathrm{m}$. After fabrication, the objects were postprocessed according the standard protocol. From their SEM images (Fig. 10), it was assessed that structures with $150\mu \mathrm{m}$ and narrower slits between features were fabricated as one uniform object [Fig. 10(a)]. When $l$ was increased to $200\mu \mathrm{m}$, it was possible to manufacture scaffolds with separated walls [Figs. 10(b) and 10(c)]. Fabricated microporous structures consisted of even $100\text{-}\mu \mathrm{m}$ size features, which is finer to the ones achieved by Jonušauskas et al.³⁵ The results of cell growing on the 3-D printed scaffolds will be published in a forthcoming publication.³⁶

Fig. 9

3-D microporous woodpile scaffold model in AutoCAD software. $d$ is the wall width, $l$ is the gap width, and $T$ is the period ($l+d$).

Fig. 10

SEM images of microporous woodpile scaffolds fabricated using Ember. (a) $d=200\mu \mathrm{m}$, $l=150\mu \mathrm{m}$; (b) $d=200\mu \mathrm{m}$, $l=200\mu \mathrm{m}$; (c) $d=100\mu \mathrm{m}$, $l=200\mu \mathrm{m}$.

3.4.

Photostructuring of Naturally Derived Resins Employing Ember

As mentioned before, 405-nm wavelength light source is used in Ember. Using spectrophotometer SHIMADZU UVProbe, it was measured that custom-made resins do not absorb above 375-nm wavelength. Thus, photoinitiating systems were changed from cationic to FRPCP. The system was modified using two different PIs (BAPO and TPO) and varying their weight content from 1% to 5%. The other part of the mixture was GDGE (or ELO) monomers plus 30 mol. % of RD. The concept of the experiment remained the same as photostructuring with CS2010. Glass substrates were used to cast $10\text{-}\mu \mathrm{l}$ droplets on them. The substrates with samples were placed on the PDMS window. Then, selective exposition was turned on. It was projections of special designed CAD model. The model was created corresponding to Ember resolution test results and consisted of 50- to $400\text{-}\mu \mathrm{m}$-width lines with 50- to $500\text{-}\mu \mathrm{m}$ (both increasing every $50\mu \mathrm{m}$)-width slits in between. The light intensity through the PDMS substrate was increased to the maximum and reached $26.54\mathrm{mW}{\mathrm{cm}}^{-2}$ value. Changing exposure dose selectively, photopolymerized structures were obtained. It was observed that to cure custom-made resins with Ember increased energy dose was required. The best results were obtained when monomers source was GDGE and the initiating system was $\mathrm{BAPO}/\mathrm{NVK}/{\mathrm{Ph}}_2{\mathrm{I}}^{+}$ ($3\%/3\%/2\%\mathrm{w}/\mathrm{w}$). In this case, samples were cured after $E\approx 16\mathrm{J}{\mathrm{cm}}^{-2}$. Using less PI (1% and 2%), resulted in dose exceeding $24\mathrm{J}{\mathrm{cm}}^{-2}$. Adding more PI (4% and 5%) did not make any significant changes in curing time, showing that inhibition of polymerization started using such high concentrations of PI.³⁷ When TPO was used instead of BAPO, no samples were photostructurized in 16- to $32\text{-}\mathrm{J}{\mathrm{cm}}^{-2}$ energy dose range. After exposure, affected areas looked more transparent than unaffected. However, the samples used to dissolve in acetone leave nothing on the substrate. The same phenomenon was observed with both PIs when ELO was used as a monomer source instead of GDGE.

For further investigations, all obtained samples were sputtered 20-nm gold layer. From SEM images (Fig. 11), it was evaluated that after $E\approx 16$- to $18\text{-}\mathrm{J}{\mathrm{cm}}^{-2}$ lines used to be less ordered, discontinued, prone to be affected by the solvent, shifted, or tilted. 20- to $24\text{-}\mathrm{J}{\mathrm{cm}}^{-2}$ exposure ensures well-defined, continuous, and straight lines. Results are summarized in Table 1. Also, wider ones ($>200\mu \mathrm{m}$ and more) were better adhered to the substrate because of a bigger contact surface area. In all cases, fully separated lines were obtained when a slit in between them was $>200\mu \mathrm{m}$. When slit was 100 to $200\mu \mathrm{m}$, it was filled with the leftover of the uncured resin still attached to the produced lines. This made it difficult to distinguish boundaries of the lines. Narrower than $100\text{-}\mu \mathrm{m}$ slits were totally filled with uncured material or clogged up. What is more, small circular “craters” can be noticeable on the formed features (see Figs. 11 and 12). They were caused by air microbubbles that appeared when sample droplets were casted on the glass substrate. However, this issue can be solved putting droplets in vacuum. After several iteration of vacuuming, no bubbles were seen in hand-spread droplets. Microbubbles should not be a problem for layer-by-layer SLA 3DP, due to translation of build head, which pushes out all the bubbles in the resin as it is in standard use of commercial materials. Also, small stripes were noticeable in the slits oriented perpendicular to the main lines similar to repolymerization at nanoscale as explained in other reports.³⁸ They most likely appeared due to diagonal pixel matrix orientation to the PDMS window. Figure 12 demonstrates how perpendicular stripes could be formed. Red-yellow squares net represents pixel matrix with yellow ones symbolizing ON state pixels (which expose the material) and red symbolizing OFF state. As it is shown, only half of upper yellow squares overlap with the line (marked with a green cursor). In such case, allegedly, the pixels turn to ON state and illuminate additional area, which normally should not be exposed. The same happens with pixels overlapping with other lines. Stripes were periodically arranged and their locations coincided with pixels positioning. Distance between those stripes varied from 65 to $75\mu \mathrm{m}$, which correlates with square hypotenuse ($\approx 70.7\mu \mathrm{m}$). In our predictions, such a device working principle explains perpendicular stripes origin and narrow slits clogging. Varying and achieving optimal printable objects orientation in respect to the pixels alignment, it might be possible to achieve higher resolution.

Fig. 11

Selectively polymerized bio-based resin samples and their SEM images: given line width (a) $d=50\mu \mathrm{m}$ and (b) $d=400\mu \mathrm{m}$. Both samples were cured after $E\approx 18\mathrm{J}{\mathrm{cm}}^{-2}$. Yellow dotted rectangles show not fully polymerized lines. (c) $d=100\mu \mathrm{m}$ and (d) $d=400\mu \mathrm{m}$. Cured after $E\approx 24\mathrm{J}{\mathrm{cm}}^{-2}$. Lines were formed more precisely than after shorter exposure. Green rectangles show features, formed perpendicularly to the main lines. In the images (c) and (d), red nets represent pixel arrangement on the PDMS window. Their orientation may cause additional feature formation in the gaps.

Table 1

Quality of photostructured various width lines after different energy dose exposure.

Fig. 12

Enlarged part of the fabricated lines. It shows pixel matrix and its positioning in respect to the lines. Yellow squares represent switched ON pixels and green dashed cursor marks their overlap with line. Blue arrows show scale of the perpendicular stripes.

Furthermore, the cross sections of formed lines were obtained with an OP (Fig. 13). It allowed evaluation of their height depending on absorbed energy dose. It was assessed that after lower doses lines had less height, compared to those, which absorbed more energy. Usually, 100- to $130\text{-}\mu \mathrm{m}$ height lines were photostructurized after $E\approx 16$ to $18\mathrm{J}{\mathrm{cm}}^{-2}$, 190 to $220\mu \mathrm{m}$ after $E\approx 20\mathrm{J}{\mathrm{cm}}^{-2}$ and 250 to $280\mu \mathrm{m}$ after $E\approx 24\mathrm{J}{\mathrm{cm}}^{-2}$. It shows that more irradiance is absorbed, and more material is polymerized. From sections picture, it is noticeable that thinner lines were formed less accurately than wider ones.

Fig. 13

Measured formed lines cross sections. Exposure dose, measured (and given) width, and height: (a) $E\approx 16\mathrm{J}{\mathrm{cm}}^{-2}$, $d=88(100)\mu \mathrm{m}$, $a=121\mu \mathrm{m}$; (b) $E\approx 20\mathrm{J}{\mathrm{cm}}^{-2}$, $d=365(400)\mu \mathrm{m}$, $a=201\mu \mathrm{m}$; (c) $E\approx 24\mathrm{J}{\mathrm{cm}}^{-2}$, $d=219(200)\mu \mathrm{m}$, $a=252\mu \mathrm{m}$. Red dashed cursors mark given width.