Multi-layering of SU-8 exhibits distinct geometrical transitions from circular to planarized profiles

Abstract

The negative tone photoresist SU-8 permits the creation of micrometer-scale structures by optical lithography. It is also the most used photoresist in soft lithography for the fast-prototyping of microfluidic devices. Despite its importance, the effect of capillary forces on SU-8 multi-layering onto topographical features has not been thoroughly studied. In particular, the profile of the added layer has not been examined in detail. The overlaying process exhibits a set of distinct behaviors, or regimes, depending on the relative thickness of the overlay and the underlying rectangular pattern. We demonstrate how capillary effects control the profile of multi-layer microchannels in a predictable manner. We derive a simple static model to describe the evolution of the overlay as a function of dimensionless geometric parameters. Our study provides a critical understanding of the parameters that govern multi-layer spin coating.

I. INTRODUCTION

The negative tone photoresist SU-8 allows for a wide range of layer thicknesses with high aspect ratio and resolution.¹ It has been used in numerous micro-electromechanical systems (MEMS) applications.² It has also supported the advent of soft lithography, which has driven the development of microfluidics by providing a simple and accessible method to manufacture devices.^3,4 Microstructures made of SU-8 are easily patterned by optical lithography onto a wafer. The photoresist is first spin coated to create a layer of uniform thickness.^5–7 The microstructures can be either part of the final microdevice⁸ or serve as a permanent mold such as in soft lithography.³ For single-layer devices, the pattern obtained is planar, that is with rectangular cross sections. Patterns can be successively layered onto each other to create three-dimensional structures.^8,9 The vast majority of multi-layered structures have focused on planarized multi-layers, where each added structure possesses a rectangular profile. However, the behavior of multi-layered SU-8 is more extensive and can generate rounded shapes.¹⁰

Experimental evidences point toward two opposite behaviors where thicker overlays generate planarized structures and thinner overlays adopt a rounded cross section. Indeed, it can be challenging to fabricate a thin planarized overlay with the desired thickness.¹¹ In contrast, we demonstrated that rounded channels necessary for the manufacturing of efficient on-chip valves could be obtained from multi-layered SU-8 by harnessing capillary effects.¹⁰ Similar capillary-based approaches have also been adopted by other groups.^11,12 However, there exist no rules to predict the profile of the added layer, referred to hereafter as the overlay.

Multi-layer spin coating has been studied because of its importance in optical lithography for micro-electronic applications;^13–19 however, those studies have aimed at finding the conditions for creating planarized overlays. Rounded overlays can be clearly seen in some data such as in Fig. 1 of Stillwagon et al.,¹⁸ but it has attracted little attention. Most studies have also focused on planarization during the spin coating process itself where capillary forces are counterbalanced by centrifugal forces but disregarded the effect of capillary forces during the levelling period. The levelling time, required to reach planarization, can be as short as a few minutes in some circumstances,¹⁶ which indicates that the overlay does not necessarily reach a fixed state during spin coating but after a levelling period. A previous study reporting multi-layering of SU-8¹¹ explained the shape of the overlay by the force balance during spin coating. However, it did not consider the levelling period and did not document the precise profile of the overlay cross section. Overall, the rounding effect and the transitions between the rounding and planarized regimes of overlays have not been thoroughly studied.

To address this gap, we studied the overlaying of SU-8 onto rectangular patterns using bright-field microscopy. By exploring a wide range of overlay and rectangular pattern thicknesses, we uncovered a set of distinct behaviors displayed by the transition of the overlay profile from circular to planar. We could derive a model based on static arguments to describe this highly dynamic process. Our work contrasts with previous studies that solely focused on the overlay height and considered only dynamical arguments. Notably, we demonstrate how capillary effects control the profile of multi-layer microchannels made of the SU-8 negative photoresist in a predictable manner. Our study provides a critical understanding and rules to design unique three-dimensional structures with the SU-8 photoresist.

II. EXPERIMENTAL

A. Structure fabrication

We fabricated the SU-8 structures onto 3 inches single-side polish test grade silicon wafers of orientation 〈100〉 (University Wafer, MA, USA) by optical lithography.³ We designed masks with the DrafSight software (Dassault Systems, Paris, France) that were printed at 25 400 dpi resolution onto a Fuji transparency mask (CAD/Art Services Inc., Bandon, OR). We spin coated the SU-8 (Microchem, MA, USA) onto the silicon wafer. The silicon wafers were used without pre-treatment. Spin coating, baking, and illumination conditions were adjusted according to the desired thicknesses following the manufacturer's instructions. The amount of photoresist poured onto the wafer was not controlled but it was over the recommended amount of 1 mL by inch of wafer.⁵ No hard bake was performed. The SU-8 was patterned by UV light through the mask with an aligner (Newport 500 W UV-illumination system, CA, USA). This procedure creates planarized rectangular patterns after polymerization of the illuminated photoresist and development. The base layer was manufactured with the SU-8 2025 photoresist. The overlay was spin coated after development of the first layer. We created the overlays with either SU-8 2007 or SU-8 2025. To generate additional data points per experiment, the wafer was tilted by 0.6° during pre-exposure baking steps so that the photoresist accumulated slightly more on one side of the wafer, resulting in rectangular patterns within a range of thicknesses on a single wafer. The effect of the tilting was considered by directly measuring the rectangular pattern height.

The topographical structure consists of a series of rectangular patterns of varying widths attached to a large pad to minimize delamination (Fig. 1). The rectangular the patterns are 12.5 μm, 25 μm, 50 μm, 75 μm, 100 μm, 150 μm, 200 μm, 250 μm, 350 μm, 500 μm wide, and the distance between rectangular patterns was 500 μm. To image the overlay, we used a 400 μm wide rectangular mask that encompasses the series of rectangular patterns (Fig. 1). The rectangle was aligned at an optimal distance of 190 μm from the rectangular pattern tips: (1) it allows for the use of bright-field microscopy by being within the working distance of the objective, (2) it is beyond the curvature induced by the tip of the linear structures (as shown in Fig. S1 in the supplementary material), which would affect the thickness measurements. Each base structure includes two additional and independent pads to measure the thickness of both layers independently. The layout of the masks includes several radially-oriented base structures located at different distances from its center.

FIG. 1.

Design of the structures for studying the evolution of overlays on rectangular patterns, both layers are made of the SU-8 negative photoresist. (a) The design consists of a series of rectangular patterns of increasing widths. A rectangular mask was used to pattern the overlay (in red). (b) The addition of a thin second layer creates a rounding effect on top of the underlying rectangular patterns. (c) The rounding effect of the second layer is visible on the 150 μm width rectangular pattern. The scale bars represent 100 μm.

B. Structure characterization by bright-field microscopy

To obtain the full profile of the rounded rectangular patterns, we developed an optical method. We designed a custom 3D printed holder to maintain wafers vertically above the 10× objective of a Nikon Diaphot inverted microscope. The configuration of the setup is appropriate only for structures close to the periphery of the wafers. To image structures closer to the center, wafers were cut into centimeter scale pieces following the cleavage planes of the silicon wafers (Fig. S2 in the supplementary material) using a glass cutter and glass cutting pliers (McMaster, IL, USA). The pieces were individually mounted onto the holder for imaging. Images of the structures were captured with a digital Sony XCV60 camera (Japan).

Using a macro developed within ImageJ,^20,21 we captured three points that defined the shape of the overlay. The two flanking points captured the corners of the rectangular pattern and the in-between point permitted the calculation of the inscribed circle and its radius. Using this information and the location of the top of the rectangular pattern given by the two flanking points, the added height of the overlay could be measured. In addition, we captured the bottom of the rectangular pattern in order to validate its height. We measured the height of each rectangular pattern of different widths and verify for their consistency to validate our measurements.

To evaluate the uncertainty of the microscopy method, we repeatedly measured the same set of 33 structures 6 times and reported the coefficient of variation of the measurement as a function of the averaged measured thickness for each structure. We selected structures with thicknesses that span the whole range (5–230 μm) observed in the experiments (Fig. S3 in the supplementary material). Data indicated that the coefficient of variation can reach 30% for the measurement of the thinnest structure (2 μm), but it is below 10% for the measurement of structures thicker than 10 μm. Increasing magnification would theoretically reduce the measurement variability; however, this would also greatly reduce the working distance thus making measurements impractical.

We also used a digital camera (Sony XCV60) mounted onto a horizontally oriented inspection zoom monocular microscope (Amscope, H800-CL, CA, USA) to image wafers placed onto a kinematic platform (KM200B, Thorlabs, NJ, USA) and illuminated with an optical fiber. This configuration allowed imaging rectangular patterns along their length to estimate the effect of boundary conditions on the overlay height (Fig. S1 in the supplementary material). All the conditions used for the micrographs were spatially calibrated with a stage micrometer (R1L3S2P, Thorlabs).

C. Imaging by scanning electron microscopy

We obtained images of the masters using a Scanning Electron Microscope (Hitachi S-4800 SEM, JEOL, USA, Inc., Peabody, MA, USA). Images were acquired using a 5-kV accelerating voltage, 10 mA beam current, 40–57 mm working distance, and a stage tilt angle of 45–57°. Because of the relatively large size of the structures observed, the low magnification setting was used (<350×).

D. Interfacial energy and contact angle measurement

As previously described,²² interfacial energies were estimated using the pendant drop method. Images of pendant drops of SU-8 2007 and SU-8 2025 were obtained using a digital camera (Sony XCV60) mounted onto a horizontally oriented inspection zoom monocular microscope (Amscope, H800-CL). The drops were illuminated with a backlight system (LED-SP, Amscope) and manually generated with a 1 mL disposable plastic syringe through a stainless-steel blunt needle (0.5 in. long, 23 gauge, McMaster-CARR). Images were saved using a program developed within Labview (National Instruments, TX, USA). The interfacial energy was obtained with the ImageJ plug-in “Pendent Drop”²³ after spatial calibration of the imaging system with a stage micrometers (R1L3S2P, Thorlabs).

We used the goniometer configuration with the digital camera (Sony XCV60) mounted onto a horizontally oriented inspection zoom monocular microscope (Amscope, H800-CL) to image wafers placed onto a kinematic platform (KM200B, Thorlabs) and illuminated with a backlight system to measure the contact angle of SU-8 2007 and SU-8 2025 on fully processed SU-8 2025. The angle was extracted using the ImageJ plug-in Big DropSnake.²⁴ Advancing and receding contact angles were measured using the cradle configuration with the goniometer station. We noticed that the value of the contact angles and interfacial energies varied with the amount of solvent in the photoresist. To fully account for all the steps in the layering process, we performed the interfacial energy and contact angle measurements before and after the pre-exposure baking step that consists in sequential incubations at 65 °C for 2 min and 95 °C for 6 min for both SU-8 2007 and SU-8 2025.

III. RESULTS

A. Characterization of the profile of SU-8 overlays

Figure 1(a) depicts the design we used to characterize the profile of SU-8 overlays. We spin coated SU-8 onto a series of rectangular patterns of varying widths. The rectangular patterns are linked to a large rectangular pad to minimize delamination that can occur with a thin and long pattern. The rectangular pad was oriented toward the center of the wafer. The 400 μm wide rectangular overlay is located 190 μm from the extremity of the rectangular patterns. Each layer included an independent 100 μm by 400 μm rectangular pad situated 900 μm away from the extremity of the rectangular overlay to allow measuring the thicknesses of both layers independently [not shown in Fig. 1(a)]. Figures 1(b) and 1(c) depict an overlay of 10 μm nominal thickness onto 25 μm thick rectangular patterns made of SU-8 2007 and SU-8 2025, respectively. Scanning electron microscope images show the overlay that encompasses different rectangular patterns.

B. Bright-field microscopy provides the profile of SU-8 2007 overlays on SU-8 2025 base patterns

We further sought to observe and predict the shape of SU-8 overlays and characterize the evolution of their height as a function of the relative thicknesses of the overlay and rectangular pattern. In our experience, stylus profilometry is not adapted to derive the profile of the structures because the shape of the stylus tip affects the measured profile in a non-linear fashion. Deconvolving the exact profile is not straightforward. The profile is distorted if the scanning line is not perpendicular to the patterns. Separating the contribution of the overlay from the base pattern involves measurement at different locations and further assumptions. To alleviate those limitations, we opted for bright-field microscopy to directly observe and characterize the profile of the structures after dicing the master using cleavage planes of the silicon crystal (Fig. S2 in the supplementary material) and mounting the pieces on a custom-designed 3D printed vertical holder.

Figure 2(a) depicts an overlay with a nominal thickness of 21 μm (as measured on the side pad) over a 95 μm thick and 250 μm wide rectangular pattern. The micrograph reveals both the rectangular pattern and the overlay. The shape of the overlay is circular with the presence of a triple line at the top corners of the rectangular pattern. We further characterized the overlay profile by capturing the position of the wafer and three points of the overlay. The three points included the top corners of the rectangular pattern and a point of the overlay in-between. We used those points to calculate the inscribed circle [Fig. 2(b)]. By super-imposing the inscribed circle on the images, we validated the circular shape of the overlay. For each condition, we characterized the overlay by reporting its shape (rectangular or circular), measured its height defined as the maximum added height onto the rectangular pattern, its nominal thickness (measured away from the rectangular patterns) and the thickness of the rectangular pattern.

FIG. 2.

Optical characterization of overlays. (a) Micrographs were collected after cleavage of the silicon wafer to observe the cross section of the rectangular patterns. The cross section of the overlay adopts a circular shape, and a triple line is clearly visible at the top corners of the rectangular pattern (black arrows). Here, a 21 μm overlay was spin coated onto of a 95 μm thick and 250 μm wide rectangular pattern. (b) The profile is digitized using three points to fit the parameters of an inscribed circle (bottom left). The thickness of the rectangular pattern is measured by locating of the wafer (horizontal line). The overlay height is the maximum added height on top of the rectangular pattern. The scale bars represent 100 μm. (c) Design used to study multi-layer spin coating with red arrows denoting the orientation of the pads.

The design studied consisted of radially oriented structures with the pads facing inward and located at different distances from the center [Fig. 2(c)]. We established that the distance to the center of the wafer had a negligible incidence on the overlay height (data not shown) and subsequentially focused on the peripheral structures. To better scan the range of thicknesses, the hotplates were slightly tilted (0.6°) to obtain rectangular patterns of different thicknesses on the same wafer. Thus, each structure has a different pattern thickness, and we report results as individual data points and not averages of repeated experiments.

Figure 3(a) displays the combinations of overlay thicknesses of SU-8 2007 and heights of the rectangular patterns of SU-8 2025 we investigated. Each range of parameters, defined by a different color, comprises at least 12 different series collected over 3 different wafers. Those three ranges encompass different thickness ratios of the overlay and the rectangular patterns. The conditions range from low thickness ratios (red set) to high thickness ratios (green set). The use of a tilt enabled to explore a wider range of combinations while providing statistical repeatability. Overall, the tilting strategy is efficient at scanning the parameter space and identify transitions. All the explored conditions create overlays with circular cross sections.

FIG. 3.

Analysis of the evolution of the overlay height exhibits a stereotypical behavior and uncovers three regimes when varying the relative thicknesses of the overlay and the rectangular pattern. (a) The structures analyzed encompass different ranges of overlay thicknesses and heights of rectangular patterns. Each set, differentiated by color, is based on the measurement of 12 pads for the green and orange sets and of 14 pads for the red set. The measurements were collected over three different wafers. (b) The profiles include a regular increase of the height with increasing rectangular pattern width before reaching a plateau. In the first regime (red), the height increase is marked by a steep rise before an early plateau. In the second regime (orange), the height increases more moderately and reaches the plateau for a wider rectangular pattern width. In the third regime (green), the height increases more slowly and reaches the plateau for an even wider rectangular pattern width. For clarity purpose, some repeats of the first behavior (red) were omitted from the panel. The wetting-based model developed in Secs. III D and III E associates each regime with the advancing, resting, and receding contact angles. We superimposed the curves (black) derived from our contact-angle limited model. They are plotted for the advancing, resting, and receding contact angles of SU-8 2007 on SU-8 2025 after pre-exposure baking steps. (c) Data are separated by regime. A typical example of a 100 μm wide rectangular pattern is shown in the bottom panel. Values within parenthesis indicate the thicknesses of the overlay and the rectangular pattern. The scale bars represent 50 μm.

Figure 3(b) displays the measured overlay height as a function of the width of the rectangular patterns. We observe a general trend where the overlay height increases with the width of the rectangular patterns. The different ranges of thickness ratio exhibit several differences. First, the rate of increase of the overlay height with the width of the rectangular pattern is lower for higher thickness ratio. The set of green curves regularly increase at a smaller rate, while the orange and red curves exhibit a steep increase before plateauing. The red curves display markedly steeper slopes than the other sets. Second, the plateau values for the green set are within the range of the overlay thicknesses, while the plateau values are higher than the overlay thicknesses for the lower nominal aspect ratios (orange and red sets).

These data reveal a couple of unexpected features: the overlay height can be higher for thinner overlays for a given height of the rectangular pattern, and the overlay height can be much higher than the nominal thickness [red curves in Fig. 3(a)]. In summary, our data suggest that the smaller the thickness ratio, the higher the overlay height is.

We superimposed the curves corresponding to our wetting-based model developed in Sec. III E, which associates each regime with the advancing, resting, and receding contact angles. Figure 3(c) display a typical example of structures for each parameter range. The panel also highlights that microscope observation of the cross section enables direct validation of the results.

These data were obtained for overlay thicknesses smaller than the rectangular pattern thicknesses. This limit is due to the use of SU-8 2007, which cannot generate layers thicker than those made of SU-8 2025. In those conditions, planarized overlays could not be observed.

C. SU-8 2025 overlays on SU-8 2025 exhibit similar behavior as SU-8 2007 overlays and reveal the planarized regime

We explored the behavior of overlays made of SU-8 2025 onto rectangular patterns made of SU-8 2025 to confirm the previous observations with a higher viscosity photoresist and test the possibility of producing planarized overlays. Using SU-8 2025, we could investigate a broader range of thickness ratio of the overlay and rectangular pattern. Figure 4(a) depicts the ranges of thickness ratios tested; data for each range were collected from two different wafers. In contrast to the previous set of experiments, we sought to obtain intermediate thickness ratios by increasing the overlay thickness instead of reducing the thickness of the rectangular patterns. Figure 4(b) depicts the evolution of the overlay height as a function of the width of the rectangular pattern. The 2025 overlaying experiments were more challenging due to de-wetting effects. As a result, some measurement series are incomplete. We also observe more variability in the SU-8 2025 experiments compared to the SU-8 2007 overlaying experiments. However, the data recapitulate the observations obtained with SU-8 2007 overlays. The overlay height regularly increases before plateauing, exhibiting the two phases displayed by SU-8 2007 overlays. We also identified the three different regimes characterized by the rate of increase of the overlay height for narrower rectangular patterns.

FIG. 4.

Characterization of SU-8 2025 overlays on SU-82025 rectangular patterns with varying widths and thickness ratios. (a) The structures analyzed encompass different ranges of overlay thicknesses and heights of rectangular patterns. Each set, differentiated by color, is collected over two different wafers. The blue range corresponds to cases where the thickness of the overlay is higher than the thickness of the rectangular pattern. (b) Overlays thinner than the rectangular pattern adopt a circular cross section. In this condition, data exhibit three different regimes similar to the behavior of SU-8 2007 overlays on SU-8 2025. We superimposed the curves (black) derived from our contact angle limited model. They are plotted for the advancing, resting, and receding contact angles of SU-8 2025 on SU-8 2025 after pre-exposure baking steps. The right panel depicts typical examples of steep, fast, and slow behaviors for a 100 μm wide rectangular pattern (from top to bottom). The numbers indicate the thicknesses of the overlay and the rectangular pattern. The scale bar represents 50 μm. (c) Overlays thicker than the rectangular pattern exhibit a planarized shape with a flat surface. The right panel depicts typical examples of planarized overlays. The numbers indicate the thicknesses of the overlay and the rectangular pattern. The scale bar represents 50 μm. (d) The overlay height is equal to the difference in thicknesses of the overlay and the rectangular pattern, except when the difference is small (lower left quadrant).

Figure 4(c) depicts the results for thickness ratios higher than unity. In these conditions, the overlay adopts a flat or planarized profile. We also reported the overlay height as a function of the difference between the thicknesses of the overlay and the rectangular pattern in Fig. 4(d). The data show that the overlay height is equal to the difference of the overlay and rectangular thicknesses, except in cases where the difference is small. For small differences in thicknesses between the overlay and the rectangular pattern, the deviation can be very important. This discrepancy may be due to the measurement imprecision of the microscope observation for small overlay height. Overall, 2025 overlays confirm the data obtained for SU-8 2007 overlays and expand the observed behaviors to the planarized regime.

By investigating the shape of SU-8 2007 and SU-8 2025 overlays on SU-8 2025 rectangular patterns, we uncovered several behavioral transitions that depend on geometrical parameters that are the rectangular pattern width, and the relative thicknesses of the overlay and the rectangular patterns. For a given thickness ratio, we observed two phases: (1) a rapid increase of the overlay height as a function of the rectangular pattern width, (2) followed by a levelling off. When the thickness ratio increases, we also observed three main regimes: (1) unexpected height of the overlay, higher than its nominal thickness, with a circular cross section, (2) an overlay height in the range of its nominal thickness with a circular cross section, and (3) planarized overlay with a flat surface.

D. The shape of the rounded rectangular pattern is dictated by capillary effects

Spin coating consists of two phases of constant rotational speeds interceded with a shorter acceleration phase.¹⁵ At constant rotational speed, centrifugal acceleration is restricted to radial acceleration. The rectangular patterns studied here are radially oriented [Fig. 2(c)], so there is no cross-pattern flow due to centrifugal forces. The rotation thins the photoresist layer by ejecting the excess material along the main direction of the rectangular patterns. We can thus consider that the rounding effect does not depend on centrifugal forces here. The forces at play are reduced to gravitational, capillary, and viscous forces.

Both gravitational and capillary forces create a flow of photoresist. Gravitational forces will be downward while capillary forces will round the photoresist into a cap on top of the rectangular pattern (Fig. 5). By contrast, viscous forces oppose any flow. The dimensionless Bond number B_o allows the comparison of the relative effect of the actuating forces, it can be written as $B_o={\left(\frac{L}{{\lambda }_c}\right)}^2$, where L is the typical dimension of the system and λ_c is the capillary length that describes the relative impact of gravity and capillary forces and takes the form ${\lambda }_c=\sqrt{\frac{\gamma }{\rho g}}$, where γ is the interfacial energy, ρ is the weight density of the liquid, and g is the gravitational acceleration. The capillary length is estimated here at 960 μm for SU-8 2025 and 1970 μm for SU-8 2007 (using values tabulated in Table I in the supplementary material). It results from these values that the effect of gravity is negligible, the shape of the cap is thus dictated solely by capillary effects and its cross section should adopt the form of the section of a disk.

FIG. 5.

Simple geometrical model of the rounding effect. (a) Geometrical description of the model. (b) Capillary forces tend to round the overlay. Differences in curvature can drive photoresist flow from the cap to the meniscus.

It is also informative to consider the timescales of the gravity and capillary actuated flows to assess the duration of levelling period necessary to reach a fixed state. Considering the parameters described in Fig. 5(a) and assuming a lubrication approximation, the flow of photoresist is given by $Q=-\frac{1}{3}\cdot \frac{1}{\eta }\frac{\mathrm{\partial }P}{\mathrm{\partial }y}\cdot t^3$.¹⁶ For gravitational effects, $\mathrm{\partial }P$ can be approximated to the hydraulic pressure due to a column of liquid of height H + t that can be written as $\mathrm{\partial }P=\rho g(H+t)$. For capillary effects, $\mathrm{\partial }P$ can be approximated by the Laplace pressure where we consider the radius of the interface at the corner of the rectangular pattern by t, hence $\mathrm{\partial }P=\gamma /t$. It results that the timescales for an amount t² of liquid to flow under gravitational or capillary effects are ${\tau }_g=\frac{3\cdot \eta }{\rho gh}$ and ${\tau }_{cap}=\frac{3\cdot \eta \cdot h}{\gamma }$ respectively. Using the values in Table I in the supplementary material, we obtain 138 s and 14 ms for τ_g and τ_cap, respectively, for SU-8 2025, and 4.3 s and 0.1 ms for τ_g and τ_cap, respectively, for SU-8 2007. It results from these considerations that the rounding phenomenon is driven by capillary effects that act at a shorter timescale than gravitational forces and the spin coating procedure. It is important to note that these values are valid for the viscosity of the photoresist before spin coating. The photoresist becomes more viscous due to solvent evaporation during spin coating. It is nonetheless interesting to note that a significant amount of solvent remains in the spin coated photoresist,⁴ and that based on the observed effect of hotplate tilting we can conclude that significant flow of photoresist occurs during the levelling period.

By overlaying SU-8 2007 or SU-8 2025 photoresist on a rectangular pattern made of SU-8 2025, we could observe that: (1) the overlay adopts a circular cross section, which can be explained by the dominance of capillary forces, (2) the added height varies non-linearly with the width of the rectangular patterns in two phases, and (3) the evolution of the added height is non-monotonous and reveals three different regimes when the ratio of nominal thickness over the height of the rectangular pattern increases.

E. The transition between cases is dependent on the relative amount of the overlay compared to a cap limit

In this section, we seek to develop a model to explain the evolution of added height as a function of the width of the rectangular patterns for the rapid and intermediate regimes (color-coded red and orange in Figs. 3 and 4). This evolution comprises two observed phases: first a rapid increase, followed by a plateau levelling off. The cap formed by the photoresist on the rectangular pattern is defined by its angle at the top corners of the rectangular patterns; this angle is set by the wetting property of the overlay on the rectangular pattern (Fig. 6). These principles imply that there exists a cap limit set by the wetting properties of the system. Any excess over this cap limit would flow down the sidewalls [Fig. 6(b)]. This situation occurs mainly for the narrowest rectangular patterns. In contrast, when the amount of photoresist does not reach the cap limit, the cap formed is limited by the amount of overlay [Fig. 6(c)]. For a very low amount of photoresist that cannot comply with the wetting rule, i.e., if the angle is lower than the receding angle, we can anticipate a de-wetting region with a structure of unpredictable shape [Fig. 6(d)]. These simple considerations permit one to calculate the rate of increase of the overlay height and the transitions between the phases.

FIG. 6.

The overlay behavior changes as a function of the amount of the material deposited onto the rectangular pattern (model). (a) Geometrical parameters. (b) Contact angle-limited case. (c) Amount-limited case. (d) De-wetting case.

1. Contact angle-limited case

The geometry of the cap is limited by the contact angle only if the amount of photoresist exceeds the amount required to form the circular segment. Using simple geometrical arguments (Fig. S4 in the supplementary material), we can calculate the height of the circular segment formed by the overlay as a function of the rectangular pattern width when the contact angle αcap is fixed. The height h follows the expression

$$h={\displaystyle \frac{w}{2}\cdot \left({\displaystyle \frac{1}{\sin \alpha }-{\displaystyle \frac{1}{\tan \alpha }}}\right).}$$ (1)

Interestingly, the contact angle value can effectively vary within the range defined by the advancing and the resting contact angles. It is critical to note that the contact angles significantly change if they are measured before or after the pre-exposure baking steps (41° ± 1.4 vs 29° ± 1.5 for instance for the resting angle of SU-8 2007 on SU-8 2025). This is most probably due to the evaporation of the solvent. Plotting Eq. (1) for the different contact angles (values in Table II in the supplementary material) permits to separate out the three non-planarized regimes [Figs. 3(a) and 4(a)]. This result suggests that the transition between these regimes is due to a change in contact angle and that the overlay profile is controlled by the contact angle at the corner of the rectangular pattern.

Equation (1) also permits one to back-calculate the thickness necessary for the overlay to be in the contact angle-limited case, assuming a flat layer of the nominal thickness [see Fig. 6(a), Table III in the supplementary material]. For instance, a 12 μm thick overlay is necessary to reach the cap limit for 100 μm wide rectangular patterns. This simple calculation allows for the estimation of the transition between the cases and indicates that the transition moves toward wider rectangular patterns for thicker overlays.

2. Amount-limited case

Below the cap-limit, the cap is simply formed by the amount of photoresist spin coated on the rectangular pattern top. Using geometrical arguments and considering conservation of mass, i.e., the rectangle t.w assumes the shape of a circular segment (Fig. S5 in the supplementary material), we can derive the following relationships:

$$t\cdot w={\displaystyle \frac{R^2}{2}\cdot \left(2\cdot \mathrm{si}{\mathrm{n}}^{-1}\left({\displaystyle \frac{w}{2\cdot R}}\right)-{\displaystyle \frac{w}{R}\sqrt{1-{\left({\displaystyle \frac{w}{2\cdot R}}\right)}^2}}\right)}$$ (2)

and

$$h=R\left(1-\sqrt{1-{\left({\displaystyle \frac{w}{2R}}\right)}^2}\right),$$ (3)

where R represents the radius of the inscribed circle; and Eq. (2) is first solved for R before calculating h using Eq. (3). Critically, the overlay height does not depend on the contact angle but solely on the overlay thickness and rectangular pattern width. After tabulating the overlay height as a function of rectangular pattern width for overlay thicknesses of 7, 10, and 12 μm, we can deduce that it varies very little with increasing rectangular pattern width in the amount-limited case (Table IV in the supplementary material). This slow increase can be explained with a scaling argument. The height of the overlay is determined thanks to the conservation of matter, and the amount of photoresist deposited on top of the rectangular pattern increases proportionally with its width. As a result, the height of the overlay varies little with the width of the rectangular pattern because it scales as the amount of photoresist divided by the width of the rectangular pattern in first approximation. We also selected data series that exhibit a steep and intermediate behaviors with a similar overlay thickness of 9 μm and plotted these together with the curve predicting the overlay height (Fig. S6 in the supplementary material). We observed that the predicted curve gives a good estimate of the overlay height for the widest rectangular patterns.

3. De-wetting case

The third case arises when not enough photoresist is deposited on top of the rectangular pattern, which results in de-wetting. We can infer that the amount of added photoresist must be larger than the cap defined by the receding angle to avoid de-wetting. This allows to calculate the minimum overlay thickness to avoid de-wetting (Fig. S7 in the supplementary material). De-wetting is dominated by the higher receding angle, which is the post-baking condition for SU-8 2007 and pre-baking conditions for SU-8 2025 both on a fully developed layer of SU-8 2025. It results that the overlay thickness should be at least 24 μm and 66 μm for SU-8 2007 and SU-8 2025, respectively, to avoid de-wetting for 500 μm wide channels (Table V in the supplementary material).

Experimentally, it has been challenging to observe de-wetting because we only focus on a section of the structure. Indeed, de-wetting can occur at any place and the photoresist will transfer from that location and accumulate to a different area that will then comply with the wetting rule thanks to the extra material. We could still experimentally validate our model by observing thin overlays on a thicker rectangular pattern using the cross section (Figs. S8 and S9 in the supplementary material), or even better using a low magnification image of the whole structure (Fig. S10 in the supplementary material). These data show that de-wetting starts occurring for 200 μm rectangular patterns for an 8 μm SU-8 2007 overlay and 250 μm rectangular pattern for a 22 μm SU-8 2025 overlay and validate our prediction. Another region where de-wetting may be observed is at the transition from rounding to planarized overlaying. Indeed, there should exist conditions where the overlay is very thin. Experimentally, we observed some de-wetting in that region (data not shown).

F. The transition between regimes is set by the interaction of the cap and the meniscus

1. Model

In this section, we seek to develop a model to explain the transition between the different regimes observed experimentally. We hypothesize that the transition between regimes is due to the interaction of the cap and the meniscus. Let us consider the geometry of the system for a low thickness ratio [Fig. 7(a)]: the overlay forms a cap on top of the rectangular pattern and a meniscus on its side. The meniscus height depends also on the overlay thickness. In the first regime with a steep increase of the overlay height, the cap and the meniscus are independent. In this case, the overlay height is controlled by the contact angle at the corner of the rectangular pattern that can reach the advancing contact angle [Fig. 7(b)—left panel]. In the next regime, the meniscus and the cap interact such that some of the cap photoresist could flow into the meniscus. The contact angle at the rectangular pattern corners is thus influenced by the meniscus [Fig. 7(b)—middle panel]. Using the same reasoning, we can anticipate a planarized phase when the meniscus and cap are fully merged and create a flat surface over the rectangular pattern [Fig. 7(b)—right panel]. Simple geometrical considerations, based on static arguments, can explain the rich behavior of multi-layer spin coating of the SU-8 photoresist.

FIG. 7.

Geometrical model of the overlay. (a) Parameter description of the system. (b) We hypothesize that the different regimes depend on the interaction between the meniscus and the cap formed by the overlay with the rectangular pattern.

2. Meniscus calculation

The transition between the cap-controlled and the meniscus-controlled cases occurs when the cap and the meniscus interact. This happens when the height of the meniscus is equal to the height of the rectangular pattern. Thus, we calculated the height of the meniscus to predict when the transition happens.

The meniscus in the cap-controlled regime is not fully developed and is limited by the amount of photoresist present between two adjacent rectangular patterns. In addition, the rectangular pattern cross sections are much smaller than the capillary length of both photoresists, so we can assume that all the photoresist in-between the rectangular patterns form a cylindrical meniscus at the base of the rectangular patterns. The angles of the menisci at the triple lines in those conditions are equal to the receding angles. We considered the fact that the contact angle of the SU-8 photoresist is different with the naked wafer and the developed SU-8 2025 and named them β and α, respectively. Under these assumptions, the height of the meniscus is approximated by

$$m=\sqrt{{\displaystyle \frac{\delta }{A}\cdot t\cdot channeldistance}}\cdot (\cos \alpha -\sin \alpha ),$$ (4)

where $A=\frac{4-\pi }{4}+\alpha -1-\sin \alpha \cdot \cos \alpha /2-\sin \beta \cdot \cos \beta /2+\cos \alpha \cdot \cos \beta$ and δ is a geometrical factor; the meniscus height depends not only on the overlay thickness but also on the distance between rectangular patterns (Fig. S11 in the supplementary material for derivation).

The meniscus and the cap interact when its height is at least equal to the height of the rectangular pattern. Also, the meniscus profile becomes asymmetrical when its height is larger than the height of the rectangular pattern. Using this condition, we measured the meniscus of the 100 μm width rectangular patterns of different conditions and reported the meniscus asymmetry, defined as the ration of its height over width, to observe when the meniscus is in contact with the cap. A ratio close to unity means that the meniscus is symmetrical and is isolated from the cap. A ratio lower than unity indicates that the meniscus interacts with the cap.

On the same graph, we plotted the meniscus asymmetry and the result of Eq. (4) as a function of the overlay and rectangular pattern thicknesses (Fig. S12 in the supplementary material). The data indicate that the calculated meniscus separates the experimental series in which the cap and the meniscus are isolated [Fig. S12(a) in the supplementary material]. Among the data, we can observe a deviation from the model with four points with an asymmetric meniscus located above the curve generated by Eq. (4) (underlined green points in the upper left corner); however, the images show that there is no interaction between the cap and the meniscus. The apparent discrepancy stems from the fact that the asymmetry assumption does not seem valid for these data series. Furthermore, we color-coded the series with their regime similar to Fig. 3 [Fig. S12(b) in the supplementary material]. Overall, there is a good agreement between the separation given by Eq. (4) and the regime classification of the experimental data for SU-8 2007.

We generated similar plots for SU-8 2025 on SU-8 2025 (Fig. S13 in the supplementary material). The value of the meniscus height depends on the receding contact angle of SU-8 2025 on SU-8 2025, which varies greatly during the pre-exposure bake time (Table III in the supplementary material). As a result, the receding contact angle may be sensitive to the experimental conditions and may have been over evaluated in our measurement that were carried out at centimeter scale. To counter this effect, we measured the receding angles at longer baking time (30 min instead of 3 min) (Tables III and VI in the supplementary material). In Fig. S12 in the supplementary material, we reported Eq. (4) using the average and the standard deviation of the contact angle measured in those conditions. Equation (4) separates the different regimes confirming that the ratio of the meniscus to the rectangular pattern heights controls the transition between the cap-controlled and the meniscus-controlled regimes. We can deduce that some photoresist flows from the cap into the meniscus when they interact because of the lower overlay height.

3. Confirmation of the model with data of overlay heights

The transition between regimes manifests in an overlay height higher or lower than the overlay thickness. In order to further test our model based on the interaction between the cap and the meniscus, we report the overlay heights normalized by the overlay thickness as a function of the meniscus normalized by the height of the rectangular pattern and the overlay thickness normalized by the width of the rectangular pattern. Indeed, the model would be supported by the data if there is a transition in the value of the normalized overlay height when the cap and the meniscus start interacting, that is to say when the normalized meniscus height reaches unity.

Figure 8 depicts the results for SU-8 2007 overlays over SU-8 2025 rectangular patterns. We observe a significant change in the normalized overlay height when the normalized meniscus height reaches unity. However, the normalized height itself does not reach unity at this transition. Figure 9 depicts the results for SU-8 2025 overlays over SU-8 2025 rectangular patterns. The data also show a transition of the normalized overlay height when the normalized meniscus height reaches unity. In this case, however, the normalized height itself reaches unity. These results confirm the data based on the meniscus asymmetry and confirm the difference between the behaviors of SU-8 2007 and SU-8 2025 overlays. The cap-controlled regime is controlled by a contact angle that ranges from the resting to the advancing contact angle for SU-8 2007 but by a contact angle that encompasses the receding and the advancing angle for SU-8 2025.

FIG. 8.

The cap-controlled to meniscus-controlled transition for SU-8 2007 on SU-8 2025 is controlled by the ratio of the meniscus to the rectangular pattern heights (m/H). In the left panel, the overlay heights are normalized and plotted as a function of the overlay thickness normalized with the rectangular pattern width (t/w) and the ratio of the meniscus over the rectangular pattern height (m/H). The normalized height is color-coded with the jet color map displayed in the color bar. The right panel depicts the surface view of the data. The vertical red dashed line and the red arrow indicate where the meniscus height is equal to the rectangular pattern thickness.

FIG. 9.

The cap-controlled to meniscus-controlled transition for SU-8 2025 on SU-8 2025 is controlled by the ratio of the meniscus to the rectangular pattern heights (m/H). In the left panel, the overlay heights are normalized and plotted as a function of the overlay thickness normalized with the rectangular pattern width (t/w) and the ratio of the meniscus over the rectangular pattern height (m/H). The normalized height is color-coded with the jet color map displayed in the color bar. The right panel depicts the surface view of the data. The vertical red dashed line and the red arrow indicate where the meniscus height is equal to the rectangular pattern thickness.

G. Transition to the planarized regime

The meniscus is less pronounced when the overlay thickness reaches the rectangular pattern thickness. It is expected that the transition toward planarized overlaying is achieved when the thickness ratio of the overlay to the rectangular pattern reaches unity. To test the idea, we plotted the SU-8 2025 data as a function of the thickness ratio of the overlay and rectangular pattern (Fig. 10). We observed that the normalized overlay height is close to zero for a thickness ration of unity and increases linearly to reach unity. These data confirm the simple rule that sets the transition toward the planarized regime.

FIG. 10.

Transition to the planarized regime for SU-8 2025 on SU-8 2025 is controlled by the ratio of the overlay thickness to the rectangular pattern height (t/H). In the left panel, the overlay heights are normalized and plotted as a function of the overlay thickness normalized with the rectangular pattern width (t/w) and the thickness ratio of the overlay and the rectangular pattern (t/H). The normalized height is color-coded with the jet color map displayed in the color bar. The right panel depicts the surface view of the data. The vertical red dashed line and the red arrow indicate where the overlay thickness is equal to the rectangular pattern thickness and indicates the transition to the planarized regime.

H. Practical considerations

Many applications can benefit from a model that predicts the shape of a SU-8 overlay. We have already shown how to harness capillary effects to fabricate efficient on-chip valves with SU-8,¹⁰ but channel rounding would be optimal within the cap-controlled regime. However, the constraints on channel dimensions are rarely practical because it requires rectangular patterns with thicknesses higher than the meniscus height, which already reaches 60 μm for a 7 μm thick overlay. The rounding can still be optimized by careful designing the mask of the overlay to encompass a section wider than the rectangular pattern to obtain an enhanced rounding effect [Fig. 11(a), Fig. S14 in the supplementary material]. Transitions between layers of markedly different thickness can also be smoothed out by spin coating an overlay and using a mask aligned with the channel [Fig. 11(b)]. We already showcased such fabrication to prevent the trapping of microfluidic droplets at the transition between a 240 μm deep incubation line and a 10 μm interrogation module.²⁵ The exact shape of the transition could also be derived by considering capillary effects. Finally, planarized overlaying can be easily obtained in most cases by using an overlay with a thickness equal to the desired thickness and the underlay thickness.

FIG. 11.

Practical considerations. (a) Optimization of the rounding effect for the fabrication of effective on-chip valve with SU-8 photoresist. (b) Schematic representation of a transition between layers of different thicknesses. (c) Implementation of a smooth transition with SU-8 photoresist. The scale bar represents 100 μm.

IV. DISCUSSION

We used a microscope method to investigate the shape and dimensions of overlays of SU-8 onto a rectangular pattern. The method enabled direct measurement of both the overlay structure and the rectangular pattern. Its limitation resides in its low reliability for thin structures with dimensions in the few micrometer range. We used a tilted hotplate to allow exploring a wider range of thickness ratios of overlay and rectangular pattern without losing repeatability of the measurements. The tilting strategy is efficient at scanning the parameter space and identify transitions; however, it limits quantitative predictions by not providing exact repeats.

This strategy allowed uncovering a set of distinct regimes as a function of the thickness ratio. Our study contrasts with a previous report¹¹ that has provided only a very partial view of SU-8 overlaying and has described only the slow regime. In our study, the evolution of the overlay height as a function of the width of the rectangular pattern exhibits an accelerated phase followed by a plateau. The accelerated phase is dominated by the wetting properties of the overlay on the rectangular patterns. The plateau values seem to correlate weakly with the thickness of the rectangular patterns. However, the data in the accelerated phase are highly reproducible, while the data at the plateau are consistent but show higher variability. This variability is unlikely due to the variable amount of photoresist poured onto the master before spin coating,⁵ but rather to the tilting strategy that does not provide exact replicates.

Our study has focused on radially oriented rectangular patterns, which simplifies the analysis by considering the nominal overlay thickness for the conservation of mass and does not need to consider directly the effect of centrifugal forces. Importantly, our experiments point toward the existence of photoresist flow during the levelling period where capillary forces also dominate. Thus, our results could be expanded to cross-radial rectangular patterns by assuming the amount of material initially deposited using the balance of capillary and centrifugal forces,¹⁸ and estimate the relaxation of the overlay due to capillary forces during the levelling period. This contrasts with a previous report¹¹ that explains the overlay shape of SU-8 photoresist with the sole balance of capillary and centrifugal forces.

Interestingly, the wetting properties of the SU-8 photoresist evolve with the pre-exposure bake. This hints at a two-step process where the amount of material initially deposited depends on the initial contact angles. Overall, the model is in excellent qualitative agreement with the behaviors observed. We observed a difference between SU-8 2007 and SU-8 2025 regarding the transition between regimes. Assuming that the interaction of the cap and the meniscus drives the transition between regimes, the accelerated phase of SU-8 2025 is wider and the contact angle encompasses the receding and the advancing contact angle. This difference hints to dynamic effects when the cap and the meniscus interact as the data point toward a flow of photoresist from the cap to the meniscus as evidenced by a smaller cap despite the same initial amount of initial photoresist on the rectangular pattern when the rectangular pattern is thinner.

This work also demonstrates how using a rectangular pattern allows the reliable fabrication of rounded channel within a regular optical lithography setting with a negative photoresist. The process contrasts with other methods that rely either on an ink fountain or do not use an underlying rectangular pattern,¹² or rely on modified optical lithography setups.^26,27

V. CONCLUSIONS

We documented the shape of an overlay of SU-8 on rectangular patterns of varying widths and thicknesses and uncovered a complete set of distinct behaviors displayed by the transition of the overlay profile from circular to planarized. These regimes are characterized by the evolution of the overlay height as a function of the width of the rectangular pattern and their relative height to the overlay nominal thickness. The transition between the regimes is driven by the thickness ratio of the overlay and the rectangular pattern. For low thickness ratios, the evolution of the overlay height can be divided into a first phase with a steep increase followed by a plateau. The overlay height for the so-called steep regime is higher than the overlay nominal thickness. For higher thickness ratio, the regime exhibits instead a slow first phase and the overlay height is lower or equal to the overlay nominal height. The regime becomes planarized with flat overlays when the thickness ratio is above unity, with an intermediate regime of de-wetting when the thickness ratio is about unity.

We used bright-field microscopy that revealed the circular profile of the overlays. It also revealed the interaction of the photoresist-air interface with the corner of the rectangular pattern indicating that the overlay profile was shaped by capillary forces. Simple scaling arguments indicate that the capillary forces dominate the rounding effect observed.

We propose a simple static model to explain the distinct behaviors observed experimentally with both SU-8 2007 and SU-8 2025 overlays. Our model assumes that the transition between the rapid and the plateau phases depends on the amount of photoresist spin coated on the rectangular pattern compared to a cap-limit. The cap-limit is set by the wetting properties at the corners of the rectangular pattern. The cap is amount-limited if the amount of photoresist is less than the cap limit. Finally, our model assumes that the transitions between the different regimes are driven by the extent of the interaction between the cap and the meniscus that develops on the side of the rectangular pattern. Our model qualitatively explains the different behaviors observed experimentally. Altogether, this work uncovers the fundamental principles that drive SU-8 overlaying and sets the foundation for a quantitative and predictive model.

From a general point of view, our report presents a capillary-controlled system that exhibits a wide set of behaviors that can be explained with simple scaling and geometrical arguments. From a practical point of view, we have exposed the rules to infer the shape of overlays made of the widely used photoresist SU-8. These advances will benefit the development of advanced microfluidic microfabrication such as for the fabrication of on-chip valves or blood capillary-like networks for on-chip microcirculation studies where the channel cross section cannot be rectangular.

SUPPLEMENTARY MATERIAL

See the supplementary material for supporting data, including figures that describe some experimental procedures, the derivation of calculations used in the manuscript, and additional data analysis, and tables that list the properties of SU-8 2007 and SU-8 2025 and tabulations of the models.

AUTHORS’ CONTRIBUTIONS

M.S. and L.L. contributed equally to this work.