Dual-wavelength volumetric stereolithography of multilevel microfluidic devices

Abstract

Microfluidic devices are typically fabricated in an expensive, multistep process (e.g., photolithography, etching, and bonding). Additive manufacturing (AM) has emerged as a revolutionary technology for simple and inexpensive fabrication of monolithic structures—enabling microfluidic designs that are challenging, if not impossible, to make with existing fabrication techniques. Here, we introduce volumetric stereolithography (vSLA), an AM method in which polymerization is constrained to specific heights within a resin vat, allowing layer-by-layer fabrication without a moving platform. vSLA uses an existing dual-wavelength chemistry that polymerizes under blue light (λ = 458 nm) and inhibits polymerization under UV light (λ = 365 nm). We apply vSLA to fabricate microfluidic channels with different spatial and vertical geometries in less than 10 min. Channel heights ranged from 400 μm to 1 mm and could be controlled with an optical dose, which is a function of blue and UV light intensities and exposure time. Oxygen in the resin was found to significantly increase the amount of dose required for curing (i.e., polymerization to a gelled state), and we recommend that an inert vSLA system is used for rapid and reproducible microfluidic fabrication. Furthermore, we recommend polymerizing far beyond the gel point to form more rigid structures that are less susceptible to damage during post-processing, which can be done by simultaneously increasing the blue and UV light absorbance of the resin with light intensities. We believe that vSLA can simplify the fabrication of complex multilevel microfluidic devices, extending microfluidic innovation and availability to a broader community.

I. INTRODUCTION

A microfluidic device is a system of microscale channels that allow for precise and predictable manipulation of minute amounts of fluids for a variety of chemical and biological applications.^1,2 Currently, microfluidic devices are produced from silicon, glass, and polymers using well-established microfabrication methods such as photolithography, etching, deposition, lamination, and soft lithography.^3,4 Conventional microfabrication enables precise control over the shape and size of micrometer-scale features, but it requires an expensive equipment.^5,6 Soft lithography improves the cost and ease of microfabrication, but the masters used for molding are often made by photolithography, e-beam lithography, or micromachining—again requiring an expensive equipment.⁷ Furthermore, microfluidic devices produced by soft lithography involve bonding the soft imprinted material to a flat substrate and are subject to leakage at the bonding interface.⁸

In response to the cost and complexity of microfabrication, additive manufacturing (AM) is becoming an increasingly popular choice for making microfluidic devices.⁹ AM has few manual steps and can produce robust, monolithic structures. Stereolithography (SLA) is a common type of AM, which has been used to fabricate microfluidic devices and microfluidic components, such as valves^10,11 and pumps.¹¹ In SLA, microfluidic devices are created from the successive addition of cured slices (layers) of a photopolymer resin formed using repeated patterned light exposures and a moving build platform. After SLA, two post-processing steps are typically performed before the microfluidic device is useable. First, the channels are flushed with solvent to remove the uncured resin, which ideally has a low viscosity. Then, the device is irradiated by high-intensity light to fully polymerize the resin.¹² While SLA improves on the cost and simplicity of microfabrication, it has its own limitations. The throughput of SLA is hindered by the slow repositioning of the build platform between exposures.¹³ Additionally, SLA is unable to form certain overhanging structures without supports, as they can deform during build platform respositioning.¹⁴

New volumetric, photopolymerization-based methods offer advantages over SLA, including faster volumetric print speed and the obsolescence of support structures. One such method, known as volumetric additive manufacturing (VAM), forms 3D objects by patterning the dose of light exposed to a resin vat, which selectively gels entire volumes of the resin in a matter of seconds.^15–17 Because objects remain submerged in the resin during fabrication, completely suspended objects can be formed, supported by the buoyant force alone.¹⁶ Despite these advantages, VAM is generally not suitable for microfluidic fabrication because of the requirement for high viscosity resins.¹⁷ One exception is a microscale VAM system that was used to make high-resolution, silica microfluidic devices. However, the fine resolution is tied to the overall small printable volume of these devices.¹⁸ Volumetric stereolithography (vSLA) is operationally similar to VAM except the object is formed layer-wise, making the process amenable to low viscosity resins. vSLA uses special resin chemistries and multiple wavelengths of light to control the location of polymerization within a resin vat.^19,20 Unlike conventional SLA, no build platform is used in vSLA, despite layer-by-layer fabrication. de Beer et al. demonstrated that vSLA could be done with a dual-wavelength activated chemistry in which polymerization is photoinhibited by UV light (365 nm) and photoinitiated by blue light (458 nm).²⁰ van der Laan et al.²¹ and Li et al.²² have also used this system to do vSLA of prismatic geometries on a mm to cm scale. In this work, we present the steps and considerations for using dual-wavelength vSLA to fabricate microfluidic devices.

II. MATERIAL AND METHODS

A. Resin materials and preparation

The photoinitiator and co-initiator were DL-Camphorquinone (CQ, Fisher Scientific) and ethyl 4-(dimethylamino)benzoate (EDAB, Acros Organics), respectively. The photoinhibitor was 2,2′-bis(2-chlorophenyl)-4,4′,5,5′-tetraphenyl-1,2′-biimidazole (o-Cl-HABI, Hampford Research). The resin was a mixture of a proprietary aliphatic polyester-based urethane diacrylate (CN991, Sartomer) and 1,6-hexanediol diacrylate (HDDA, TCI America). 3-(trimethoxysilyl)propyl methacrylate (TMSPMA, Sigma-Aldrich) was added to improve glass adhesion. The concentrations and functions of the resin components are summarized in Table I. Resins were prepared by first dissolving the solid components (CQ/EDAB/o-Cl-HABI) in minimal tetrahydrofuran (THF, ACS Reagent grade, Sigma Aldrich). The blue light absorber (Epolight 5675, Epolin) was added from a stock solution of 0.1 wt. % Epolight:HDDA. The liquid components (TMSPMA/CN991/HDDA) were manually combined with dissolved solid components. The resin was sonicated for 5 min to mix all components and sparged with N₂ for 1 h to remove THF. After THF removal, the resin was kept under an N₂ atmosphere. Except for the case of oxygen-free vSLA (see Sec. II G), transfer of resin to the resin vat and vSLA was done in air. All resins had less than 0.5 wt. % residual THF and were prepared the day of use.

TABLE I.

Resin components and concentration used in this study.

Component	wt. %	Function
o-Cl-HABI	2.5	Photoinhibitor
CQ	2	Photoinitiator
EDAB	1	Co-initiator
Epolight 5675	0.005	Blue light absorber
CN991	40	Oligomer
HDDA	53.5	Reactive diluent
TMSPMA	1	Promote glass adhesion

B. UV-vis spectrophotometry

UV-vis spectrophotometry (UV-1800, Shimadzu) was used to determine the mass-based absorptivity of the resin components at 458 nm. Spectra were collected from 800 to 300 nm with 1 nm spacing using a 1 cm path length quartz cuvette. 0.1 wt. % o-Cl-HABI, 0.5 wt. % CQ, and 0.1 wt. % EDAB samples were prepared in THF. 0.001 wt. % Epolight 5675 was prepared in isopropranol (IPA, ACS Reagent grade, Thermo Scientific). 10 wt. % solutions of the liquid resin components (CN991/HDDA/TMSPMA) were prepared in IPA. Reported absorbance values were averaged from three independently made samples per component. $h_{blue}$ was calculated as the linear combination of each component's Napierian (base $e$) absorbance at 458 nm. UV-vis spectra of EDAB, CQ, and o-Cl-HABI are shown in Fig. S-1 in the supplementary material.

C. Dual-wavelength printer

vSLA was done using a previously described dual-wavelength 3D printer.^20,23 A DLP projector (ML750, Optoma) with red and green LEDs disconnected was used to pattern blue light (458 nm). The blue LED intensity was controlled by a constant current LED driver (HLG-120H-12B, MeanWell), with 1–10 V dimming. The blue light pattern was passed through a biconvex lens (LB1630, Thorlabs) to focus the pattern at the bottom of the resin vat. The pixel resolution of the DLP projection was approximately 130 μm. A UV LED (M365LP1, Thorlabs) was affixed at a 90° angle to the DLP projection path and controlled by an LED driver (BuckPuck 3023-D-E-1000, LEDdynamics Inc.), with 0–5 V dimming. UV light was focused with an aspheric condenser lens (ACL50832U, Thorlabs) and collimated (SM2F, Thorlabs). UV and blue lights were overlaid using a long-pass dichroic mirror (DMLP425L2, Thorlabs). An aluminum chassis was built around the light sources to support the resin vat at the focal plane. LED intensities were controlled by analog voltage signals to the LED drivers. The maximum wavelength-specific light intensities were 20.9 mW/cm² for the 458 nm LED and 22.7 mW/cm² for the 365 nm LED (measured with an ILT2400 light meter and an SED005 UV-visible GaAsP detector fitted with a QNDS2 100× attenuation quartz neutral density filter, International Light Technologies).

D. Cure depth and dead zone measurements

Resin wells for cure depth measurements were made from 5 mm tall rectangular polylactic acid (PLA) 3D-printed borders glued to 50 × 75 mm glass slides using two-part epoxy. The resin well was filled to the brim with resin and covered with another glass slide. Six 8 × 8 mm squares of blue light (λ = 458 nm, 10.4 mW/cm²) were projected into the resin for times ranging from 250 to 1500 ms. The gelled polymer was rinsed using IPA and post-cured under a high-intensity white light for 5 min. The cured height of each square was measured using a micrometer (0.000 05 in. resolution, Tormach).

For dead zone measurements, resin was sandwiched between two 50 × 75 mm glass slides spaced 1 mm apart with glass shims, resulting in two open sides of the resin vat. Six 8 × 8 mm squares of blue light with intensities ranging from 2.1 to 10.4 mW/cm² were overlaid with a flood UV exposure at a constant intensity of 22.8 mW/cm² for 3 min. The squares were rinsed with IPA, post-cured, and measured with the micrometer. The dead zone height was calculated by subtraction of the cured material height from the shim height.

E. Microfluidic device fabrication

To fabricate microfluidic channels, resin was sandwiched between two 1 mm thick glass slides separated by shims, where the thickness of the shims depended on the geometry. Single-level and multilevel serpentine channels were made with 1 mm shims, and multilevel crossing channels were made with 1.75 mm shims. Exposure parameters for each geometry are described in the results. Uncured resin was flushed from the channels by gently pushing IPA through pre-drilled inlet and outlet holes in the top glass slide. The devices were briefly left to dry, filled with water to prevent the top channel wall from collapsing, and post-cured. Single-level devices were detached from the glass slides, while multilevel devices remained attached.

F. Optical microscopy of channel cross sections

Images of single-level straight channels were taken using optical microscopy (ECLIPSE Ti-S, Nikon) at 40× magnification. The channels were cut into shorter lengths using a razor blade and were stood upright using a 3D printed holder with a 1 mm gap. Microscope images were captured using Ocular Scientific Image Acquisition software. Channel dimensions were analyzed using ImageJ image analysis software and an open-source macro as discussed in Fig. S-2 in the supplementary material. Heights are reported as the average of the heights of nine different channels that were processed using the ImageJ macro.

G. Oxygen-free cure depth experiments

Oxygen-free cure depth experiments were done in an enclosed, N₂ environment. All materials and preparation, including transfer of resin to the resin vat, were contained within the inert enclosure. Resin was initially prepared as described in the resin preparation section. To conduct the comparison between oxygen-free curing and curing in air, the resin was divided into two equal parts. Half of the resin was placed within the inert printing environment and sparged for an additional 30 min with N₂, while the other half was left in air and sparged with air for 30 min. Before use, both parts of resin rested for at least 10 min until the bubbles generated from sparging subsided. Here, cure depth experiments were done using variable intensity and a constant exposure time to vary the dose, though different times were used for N₂ and air environments. 15 s was used for curing in N₂ and 60 s was used for curing in air.

III. THEORETICAL CONCEPTS

In vSLA, dose is used to predict the location of curing. A previously published model^23,24 describes the accumulation of volumetric dose from concurrent blue and UV light irradiation,

$$D(z,t)={\left({\displaystyle \frac{I_{blue}}{h_{blue}}{e}^{-\frac{z}{h_{blue}}}-\beta {\displaystyle \frac{I_{UV}}{h_{UV}}{e}^{-\frac{z}{h_{UV}}}}}\right)}^{m}t,$$ (1)

where $I_{blue}$ and $I_{UV}$ are the incident blue and UV intensities, $h_{blue}$ and $h_{UV}$ are the blue and UV Napierian absorbance heights (i.e., at a height of $h_{blue}$, the blue light intensity is $I_{blue}/e$), m is the termination constant, $\beta$ is the inhibition factor, z is the height into the resin, and t is the exposure time. Imposing the condition that curing (gelation) occurs if $D(z,t)$ is greater than a critical dose, $D_{crit}$, then Eq. (1) can be used to describe the transient position of the curing boundaries during vSLA.

The parameters $h_{UV}$, m, $D_{crit}$, and $\beta$ are resin-specific and obtained from fitting experimental data to two special cases of the dose model [Eq. (1)]. First, in the case of no inhibition, the cure depth, $z_{cd}$, is calculated by setting $I_{UV}$ equal to zero and solving for the height at which the critical dose is reached. Because there is no inhibition by UV light, $z_{cd}$ is the maximum height of curing,

$$z_{cd}={\displaystyle \frac{h_{blue}}{m}\ln \left({\displaystyle \frac{{\left({\displaystyle \frac{I_{blue}}{h_{blue}}}\right)}^{m}t}{D_{crit}}}\right).}$$ (2)

Note that ${(I_{blue}/h_{blue})}^{m}t$ is the blue light-only incident volumetric dose. Equation (2) is functionally similar to Jacobs' working curve in traditional stereolithography,²⁵ except for the addition of m and the volumetric instead of areal dose. To include contributions of secondary photochemical processes, the overall rate of polymerization is approximated as $1/m^{\mathrm{th}}$ order in the concentration of radicals, such that the intensity component of the dose scales to the $m^{\mathrm{th}}$ power.²⁶

In the presence of UV, Eq. (3) is used to describe the height to which polymerization is inhibited or the height of the dead zone. To obtain Eq. (3), the case of infinitely long exposure time is considered. Because the dose must remain finite in the dead zone at infinite exposure time, the term in the parentheses of Eq. (1) must go to zero. Then, the dead zone thickness, $z_{dz}$, approaches a steady-state value,

$$z_{dz}={\displaystyle \frac{ln\left(\beta {\displaystyle \frac{I_{UV}{h}_{blue}}{I_{blue}{h}_{UV}}}\right)}{{\displaystyle \frac{1}{h_{UV}}\text{–}{\displaystyle \frac{1}{h_{blue}}}}},}$$ (3)

where a larger value of $\beta$ indicates more efficient photoinhibition. Although the dose equation indicates that the dose and polymerization will increase indefinitely with time, vSLA is in fact a batch process with finite initiator and inhibitor concentrations that can be exhausted. Additional studies are needed to determine how long the dose equation accurately represents photopolymerization behavior during vSLA.

In dual-wavelength vSLA, layers are cured from the bottom to the top of the resin vat. This approach is used because the dose is cumulative; although there is no curing above the cure depth of a layer, there is still non-negligible dose that contributes to polymerization and reduces the remaining dose needed to gel the resin in successive layers. A layer is cured by constraining $D>D_{crit}$ to only the range of z that the layer occupies, and D is controlled using $I_{blue}$, $I_{UV}$, and t. Spatial curing within a layer is patterned by the blue light source, as shown in Fig. 1(a). Each layer has two boundaries that mark the transition from uncured to cured states. The height of the top curing boundary is governed by the blue light dose, ${(I_{blue}/h_{blue})}^{m}t$ [Fig. 1(b)], and the height of the bottom curing boundary is determined by the ratio of UV to blue light intensity, $I_{UV}/I_{blue}$ [Fig. 1(c)]. We will prove later that at sufficiently long exposure times, the top and bottom boundaries are accurately approximated by Eqs. (2) and (3), respectively. Using these equations, the appropriate values of $I_{blue}$, $I_{UV}$, and t per layer are calculated to create the exposure sequence for an object.

FIG. 1.

(a) Schematic of the resin vat during vSLA. The resin vat consists of resin sandwiched between glass slides and spaced apart by shims. Blue light is patterned to control spatial curing. The combination of UV and blue light intensities and exposure time determines the location of curing. (b) Dose controls the top curing boundary, where curing stops. (c) The ratio of UV to blue light intensity determines the bottom curing boundary, where curing starts.

IV. RESULTS AND DISCUSSION

The strengths of dual-wavelength vSLA are its speed and ease of use. Existing SLA and micro-VAM systems can fabricate high-resolution microfluidics.^12,18 However, they still take hours to produce useable devices. Our vSLA method can produce finished microfluidic devices in less than 10 min. Additionally, the dual-wavelength printer can be made from an ordinary projector and a UV LED, with no other translating or rotating parts. The layers of the device are formed as they appear in the slices, and all calculations are limited to scalar equations. vSLA and SLA operate identically in the horizontal projection plane, and the novelty of vSLA comes from the ability to control the location of curing in the vertical z direction. Hence, we primarily report results about the z heights of microfluidic features and not x or y dimensions.

A. Characterization of resin photoinitiation and photoinhibition behavior

The vSLA resin (Table I) was comprised of photoinitiator, photoinhibitor, co-initiator, oligomer, reactive diluent, and blue light absorber. The blue light-activated photoinitiator, CQ, and the UV light-activated photoinhibitor, o-Cl-HABI, are required for this dual-wavelength system. The amine co-initiator EDAB is also necessary to improve the initiation rate because CQ alone is an inefficient initiator of free radical polymerization.²⁷ The rest of the resin formulation is flexible, though several factors should be considered when formulating a new resin, including monomer reactivity, cured polymer properties, transparency in blue and UV wavelengths, and viscosity. Monomers must be susceptible to radical polymerization. Here, we used multifunctional acrylates for their high reactivity and mechanical properties. CN991, a high molecular weight oligomer, was used because of its optical transparency and hardness. To reduce the viscosity, HDDA was added as it was found to be more susceptible to photoinhibition by o-Cl-HABI compared to other reactive diluents.²³ Another reactive diluent, TMSPMA, was used as a silane coupling agent²⁸ to improve the adhesion of the polymer to the glass surfaces of the resin vat. Finally, a blue light absorber was added to reduce $h_{blue}$, which increases vertical resolution by decreasing light penetration depth.²⁹ The cured polymer was found to be hemocompatible, so microfluidic devices made with this resin could be used for the red blood cell analysis (see Fig. S-3 in the supplementary material for polymer hemocompatibility).

For this resin formulation, the resin-specific parameters of the dose model were found experimentally and listed in Table II. UV-vis spectrophotometry was used to obtain $h_{blue}$, which was calculated from the linear combination of individual resin component absorbances at 458 nm. Initially, $h_{UV}$ was also calculated from resin absorbance data, but a better representation of the experimental dead zone data was produced when $h_{UV}$ was fit with least squares regression to Eq. (3). Thus, the fitted value of $h_{UV}$ was used instead. The difference between the calculated and fitted values of $h_{UV}$ may be due to photochromism observed in o-Cl-HABI.³⁰ Photochromism is more significant during vSLA due to the high intensities of UV light used, skewing the value of $h_{UV}$ calculated from spectrophotometry data. The values of m, $D_{crit}$, and $\beta$ were also obtained by fitting Eqs. (2) and (3) to cure depth and dead zone data with least squares regression.

TABLE II.

Resin-specific dose model parameters from UV-vis spectroscopy, cure depth, and dead zone experiments.

Variable	Value	Units
h blue	450.5	μm
m	0.8910	…
D crit	37.75	(mW/cm3)ms
h UV	192.6	μm
β	0.2377	…

The cure depth and dead zone data that were used to generate the dose model parameters are plotted in Figs. 2(a) and 2(c). The standard deviation error bars associated with cure depth data are noticeably larger than those associated with the dead zone data, which we attribute to variable oxygen concentration in the resin. Dissolved O₂ is a powerful free radical scavenger that decreases the polymerization rate.³¹ Although resins were sparged with N₂, resin was transferred to the resin vat in air, and we suspect different concentrations of dissolved oxygen were introduced at this stage, leading to variability of cure depth data. Figure 2(b) shows the large difference in dose required to cure an equivalent height in air- and N₂-sparged resins. Ultimately, N₂-sparged resin was used despite the variability because more rapid polymerization allowed microfluidic devices to be fabricated in mere minutes.

FIG. 2.

(a) Cure depth as a function of dose (m = 0.8910). (b) Cure depth experiments done with air- and N₂-sparged resin and printing enclosures. The dose is the product of blue light intensity and exposure time (m is set to 1), owing to different termination constants that arise from fitting the two datasets to Eq. (2). Resin in the presence of O₂ from air requires significantly more dose than deoxygenated resin in an inert environment to cure an equivalent height. (c) Dead zone as a function of the intensity ratio, $I_{UV}/I_{blue}$. The dashed lines represent Eq. (2) in (a) and (b) and Eq. (3) in (c). All error bars show standard deviation per data point [n = 10 for (a) and (c), n = 4 for (b)].

The dose model parameters were used to predict the location of curing and the properties of the gelled polymer for any exposure setting. Figure 3(a) shows the dose as a function of the exposure time and height into the resin, which is plotted using Table II parameters and constants $I_{blue}$ and $I_{UV}$ of 3.21 and 8.65 mW/cm², respectively. When the dose is equal to or greater than $D_{crit}$, the resin becomes gels (orange curing region). For doses less than $D_{crit}$, the resin remains fluid enough to be washed away by post-process rinsing (beige region). If the surface plot of the dose is viewed from above [Fig. 3(b)], we see that as the exposure time increases, the top side of the curing region increases logarithmically in height, while the bottom side approaches a steady-state value. Because Eq. (1) cannot easily be solved for z, Eqs. (2) and (3) were used to approximate the top and bottom boundaries of the curing region for exposure times greater than 15 s.

FIG. 3.

(a) Surface plot of dose as a function of time and height into the resin. The orange region shows the combination of times and heights for which the critical dose is exceeded and gels the resin, while the beige region denotes a dose level below the critical dose and the resin remains fluid. (b) Top-down view of (a) demonstrates that the curing boundaries approach the cure depth [Eq. (2)] and dead zone [Eq. (3)] above ∼15 s. (c) Right side view of (a) at 50 s of exposure shows that the dose profile is not uniform across the height into the resin vat and a non-negligible dose is generated above the cure depth. The following parameters were used to generate the three plots: I_blue = 3.21 mW/cm², I_UV = 8.65 mW/cm², h_blue = 450.5 μm, h_UV = 192.6 μm, m = 0.8910, D_crit = 37.75 mW/cm³)^ms, and β = 0.2377.

The dose calculated by Eq. (1) can instead be used to understand the degree of polymerization of the cured polymer. Higher dose corresponds to a higher conversion of polymerizable groups, which increases the crosslinks between monomers. At a sufficient crosslink density, the resin gels, and as more crosslinks form, the storage modulus increases.³² While any dose above $D_{crit}$ will induce gelation, the degree of polymerization and the green strength of the gelled polymer vary with the height into the resin. For example, in Fig. 3(c), the magnitude of the dose plot suggests that the polymer gelled at a height of 500 μm would be more polymerized than at 150 and 2000 μm. Figure 3(c) also shows that there is no dose almost immediately below the dead zone, while there is non-negligible dose far above the cure depth. This is an important feature of dual-wavelength vSLA because the dose accumulates. Even if two exposures individually do not produce $D\ge D_{crit}$, if they are applied consecutively and the cumulative dose is larger than $D_{crit}$, the resin will cure. For vSLA, where multiple exposures are used, we can assume that no dose is added in the dead zone, while the dose accumulated above the cure depth is equal to the blue light-only dose ${\left(\frac{I_{blue}}{h_{blue}}{e}^{-\frac{z}{h_{blue}}}\right)}^{m}t$.

B. vSLA of microfluidic channels

To design the exposure sequence for microfluidic fabrication, the microfluidic device was divided into one or more layers, corresponding to the locations that new features appear in the z direction. For example, a microfluidic device with a single horizontal channel could be divided into three layers. The first layer spans from the bottom of the resin vat to the bottom of the channel, the second layer is spatially patterned to form a void that becomes the channel, and the third layer spans from the top of the channel to the top of the resin vat. Using Eq. (2), any combination of $I_{blue}$ and t that yield the dose needed to cure to the ending height of each layer can be used, although high intensity blue light and a short exposure time are preferred. Finally, the corresponding UV intensity that produces the required dead zone for the starting height of each layer is calculated from Eq. (3). For simplicity, we limited fabrication to only three exposures, with constant light intensity. However, more layers and exposures can be used to make smooth continuous features in the z direction by increasing the number of layers or by using grayscale patterning with blue light slices. Table III shows exposures used to make various microfluidic devices.

TABLE III.

Geometry print parameters per layer and predicted layer starting (z_dz) and ending heights (z_cd), calculated using Eqs. (2) and (3). Geometries are categorized by the location of the channels, which can exist on a single level or multiple levels. z_cd is based on the total blue light dose that includes the accumulation of dose contributions from exposures before or after the current layer. m = 0.8910.

	Geometry	Exposure	Time (s)	Iblue (mW/cm2)	IUV (mW/cm2)	zdz (μm)	zcd (μm)
Single level	Single exposure	1	30	0.72	0	0	>1000
	Straight or obstructed	1	3.5	0.72	0	0	297
		2	15	3.21	0	0	>1000
		3	120	2.14	26.9	654	>1000
	Curved	1	3.5	0.72	0	0	465
		2	30	3.21	0	0	>1000
		3	120	2.14	26.9	654	>1000
Multilevel	Serpentine	1	15	0.72	0	0	659
		2	35	3.21	8.65	136	>1000
		3	240	2.14	26.9	654	>1000
	Crossing	1	15	0.72	0	0	782
		2	30	3.21	8.65	135	>1750
		3	240	2.14	26.9	654	>1750

1. Single blue light exposure fabrication

Microfluidic devices can be made using a single blue light exposure, although this is not true vSLA because there is no control over the bottom curing boundary. The blue exposure is patterned to have black pixels in the shape of the channel [Fig. 4(a)]. Using a single 30 s patterned exposure, we formed channels that adhered to the top of a 1 mm thick confined volume of resin [Fig. 4(b)]. Because the black pixels used for patterning emitted a non-negligible amount of blue light, a thin layer was cured at the bottom of the channels, evidenced by the green tint of the blue dye. However, the low dose led to a low degree of polymerization, and parts of this layer were eroded during resin removal, which is seen in blue regions around certain inlet ports.

FIG. 4.

(a) Schematic of microfluidic channel fabrication using a single exposure of blue light. Schematic is not drawn to scale. (b) Image of the Burns lab logo channel fabricated from a single exposure. Due to low intensity blue light emitted by the black pixels of the projector, a thin layer of resin cured on the bottom side of the channel, which resulted in a green tinge to the blue dye within the channels. Part of this layer eroded during post-processing resin removal revealing un-tinted blue dye. Scale bar is 5 mm.

2. Dual-wavelength vSLA of single level channels

While the single blue light exposure device remained attached to the resin vat, standalone devices were made by introducing UV light for true vSLA. These standalone microfluidic devices were fabricated using three exposures [Fig. 5(a)] to form geometries as shown in Figs. 5(b)–5(d). In the third exposure, UV light was applied to cure the top layer without curing through the channel underneath, so the device could ultimately be detached from the resin vat. The accuracy of Eqs. (2) and (3) was analyzed using images of channel cross sections [Fig. 5(e)]. The average heights (±standard deviation) of layers 1 and 3 were 242 ± 54 and 321 ± 32 μm, respectively, and the average channel height in layer 2 was 386 ± 55 μm. The layer 1 height of 242 μm corresponds to the layer 1 $z_{cd}$. Subtracting the layer 3 height from 1 mm gives the layer 3 $z_{dz}$ of 679 μm. These values agree relatively well with the predicted layer 1 $z_{cd}$ of 297 μm and layer 3 $z_{dz}$ of 654 μm from Table III. From the microscope image, we also see that the channel has rounded corners, which is likely the result of incomplete resin removal and over-cure in the channel corners. During flushing, the velocities of IPA at the edges and corners would be lower, allowing resin to remain in those regions. Furthermore, light diffraction is known to increase dose at the intersections of features and cause over-curing, resulting in the smoothing of sharp edges.^33,34

FIG. 5.

(a) Schematic of single-level, standalone microfluidic device fabrication. Schematic is not drawn to scale. Images of (b) straight channels, (c) the curved channel, and (d) the obstructed channel filled with dyed blue water. Scale bars in (b)–(d) are 5 mm. (e) Cross sectional area of the middle channel in the straight channel design. Layers are denoted by dashed white lines. Scale bar is 500 μm. Image was taken at 40× magnification.

3. Dual-wavelength vSLA of multilevel microfluidic channels

vSLA was also used to fabricate multilevel structures, such as serpentine and crossing channels [Fig. 6(a)]. Both geometries are commonly featured in microfluidics. Serpentine channels are used for efficient mixing,³⁵ and crossing channels are used in compact microfluidic devices and for functional structures such as membrane microvalves.³⁶ Unlike single-level devices, these multilevel devices were not standalone as they were fabricated with only three exposures. There was also significant deviation between the predicted channel heights [Table III] and the real channel heights. For exposures used in multilevel devices, complete curing was predicted above 136 μm, which would have cured through the channels at the top of the device. However, open channels are clearly seen in Figs. 6(b) and 6(c). The breakdown of Eq. (2) in the case of the multilevel devices may be due to the unknown concentration of dissolved oxygen in the resin, but further studies are needed to verify this claim.

FIG. 6.

(a) Fabrication process for multilevel, overlapping channels. Schematic is not drawn to scale. (b) Schematic and photograph of the multilevel serpentine channel. (c) Schematic and photograph of multilevel crossing channels. (d) Time-lapse images of crossing channels being filled with dyed water. The absence of mixing of two streams demonstrates the integrity of the middle layer, which separates the top and bottom channels. Scale bars are 5 mm.

The crossing channel device highlights one of the advantages of vSLA, namely, the ability to form suspended objects. The middle layer of the device or the separator [Fig. 6(c)] was initially a freely suspended layer. From Fig. 4(b), we know the curing region grows both vertically upward and downward over time. Thus, the separator formed suspended in the resin until it grew downward and fused with the layer below it, turning the separator into an overhanging structure. For the separator to form correctly, it was paramount that the resin was stationary. Flowing resin exerts stresses that can deform or displace delicate features. For this reason, overhanging structures are generally printed with supports, and suspended features are impossible to make in conventional SLA. Although vSLA happens layer-wise, the ability to form a suspended feature classifies this system as a volumetric technology.

C. Concentrating dose by tuning absorbance height and light intensity

The separator between the crossing channels was exceptionally fragile and easily damaged during post-processing; however, the green strength could be improved by exposing the separator layer to higher doses, increasing the degree of polymerization. Figure 6(d) shows a successfully post-cured crossing channel device with no mixing of two different dyes, indicating that the channel separator was intact. According to Eq. (1), the dose is increased by (i) increasing the light intensities, (ii) decreasing the absorbance heights of the resin, or (iii) increasing the exposure time. Separately, increasing the light intensities or the exposure time widens the span of the curing region, while decreasing the absorbance height shrinks the curing region. However, if all three are adjusted together, then the dose can be increased while maintaining the span of the curing region. For example, Fig. 7(a) shows the dose profile for original exposure settings used in the separator layer, and Figs. 7(b) and 7(c) show that the dose can be concentrated by scaling the resin absorbance and light intensities. The scaling factors for $I_{blue}$, $h_{blue}$, and $h_{UV}$ were arbitrarily chosen, and $I_{UV}$ was calculated from Eq. (1) to maintain the dead zone location $(z=z_{dz})$. To maintain the cure depth (not shown in the example), the required exposure time would be calculated from Eq. (1) for $z=z_{cd}$. Concentrating the dose like this requires high intensity light sources, so the operational limits of the dual-wavelength printer should be considered when designing microfluidic devices and formulating the resin.

FIG. 7.

The effect of varying resin absorbance heights ( $h_{blue}$ and $h_{UV}$) and light intensities ( $I_{blue}$ and $I_{UV}$) on the transient dose profile at 30 s of exposure. (a) The original dose profile plot was generated from the parameters used to form the separator layer in the crossing channel design (i.e., I_blue = 3.21 mW/cm², I_UV = 8.65 mW/cm², h_blue = 450.5 μm, h_UV = 192.6 μm, m = 0.8910, D_crit = 37.75 (mW/cm³)^ms, and β = 0.2377). (b) Dose profile with half the original absorbance heights ( $h_{blue}$ and $h_{UV}$), double $I_{blue}$, and 3.13 $I_{UV}$. (c) Dose profile with a third the original absorbance heights, triple $I_{blue}$, and 7.23 $I_{UV}$. Dose is concentrated by both lowering the absorbance height and increasing the light intensity in blue and UV.

Figure 7 illustrates an important limitation about the print depth of our vSLA method. Dual-wavelength vSLA is specifically suited to flat, slab-like designs commonly found in conventionally fabricated microfluidic devices. The light absorbing resin, used to prevent excessive light propagation and improve z resolution, prevents the timely accumulation of the critical dose at depths far into the resin. Thus, while our vSLA prints can be as broad as the projector allows, they are no more than a few mm tall.

V. CONCLUSION

Each step of conventional microfluidic fabrication poses an operational, software, or time barrier to even the most experienced scientists and engineers. Our dual-wavelength vSLA system has the potential to make microfluidics more accessible to users unfamiliar with microfabrication and to expedite the design process of microfluidic devices by combining the simplicity of SLA with the speed of VAM. Fabrication with vSLA only involves a blue and UV light exposure sequence and a short post-processing step. Multilevel microfluidic channels can be fabricated in a matter of minutes, and preparation for vSLA is also simple. The digital image slices for vSLA can be generated using basic graphics software (e.g., Microsoft Paint), and the exposure sequence for any microfluidic design can be calculated with relative accuracy using equations based on resin-specific curing parameters. In vSLA, channels with rounded cross sections are generated, which are difficult to replicate with some popular microfabrication methods, such as soft lithography on photoresist masters.^37,38 Tube-like microfluidic devices, such as vasculature models, could be fabricated with relative ease and are one of many possible applications for dual-wavelength vSLA.

SUPPLEMEMTARY MATERIAL

See the supplementary material for (S1) absorbance spectra of resin components CQ, o-Cl-HABI, and EDAB; (S2) measurement of the height of microscope images of channel cross sections using an ImageJ macro; and (S3) hemocompatibility test of the fully cured polymer.

AUTHOR DECLARATIONS

Conflict of Interest

M.A.B., M.P.d.B., and Z.D.P. are inventors on Patent No. 11174326 for the use of photoinhibitors, including hexaarylbiimidazoles, in polymerization.

Author Contributions

Kaylee A. Smith: Data curation (lead); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review and editing (equal). Sanaz Habibi: Data curation (equal); Investigation (equal); Methodology (equal); Writing – original draft (equal); Writing – review and editing (equal). Martin P. de Beer: Conceptualization (lead); Investigation (supporting); Methodology (equal); Software (lead); Writing – original draft (supporting); Writing – review and editing (equal). Zachary D. Pritchard: Investigation (supporting); Methodology (equal); Writing – original draft (supporting); Writing – review and editing (equal). Mark A. Burns: Funding acquisition (lead); Resources (lead); Writing – original draft (supporting); Writing – review and editing (equal).

DATA AVAILABILITY

The data that support the findings of this study are available from the corresponding author upon reasonable request.