Scalable and Facile Formation of Microlenses on Curved Surfaces Enabling a Highly Customized Sustainable Solar‐Water Nexus

Solar‐driven water treatment suffers from low efficiency due to the solar energy loss during the energy conversion, especially in the scale‐up operation. One promising solution is using microlenses (MLs) to enhance the photodegradation of organic contaminants in water. However, most MLs fabrications apply to 2D planar surface only, which restricts their potential applications. In this study, a flexible and scalable technology is presented to fabricate MLs on curved surfaces. Precursor microdroplets form in a dilution process and are converted to MLs by photopolymerization. Optical simulations and experiments are combined to establish the correlation between optical properties of MLs and the performance of ML‐functionalized reactors in photodegradation. It is demonstrated that surface MLs on all‐shaped reactors significantly enhance the photodegradation efficiency of organic contaminants under simulated solar light or natural indoor light, with a maximum improvement of 83 folds. The surface coverage and size distribution of MLs can be adjusted by varying the solution concentration and the dilution rate when generating microdroplets. In addition, fabrication of MLs on a larger scale is achieved over an area up to 250  cm 2 $\left(\text{cm}\right)^{2}$ . MLs on 3‐dimensional curved surfaces fabricated by the technique enable significantly enhanced, highly customized, and sustainable solar‐driven water treatment.


Introduction
10] Current limitations in the application of solar-driven technologies include limited types of solar-degradable organic contaminants, [11] small capacity, [12] energy loss in contaminated water, in particular with high turbidity in water sources, [13] and high dependence on sunlight intensity. [14]So far, to overcome the challenge of the high turbidity of contaminated water and the restricted sunlight availability, a possible solution is to redistribute and focus light with low intensity at local hot spots. [15,16]mong technologies to improve the efficiency of solar energy usage, microlenses (MLs) offer clear advantages in terms of flexibility, [17] adaptability, [15,18] and scalability. [19,20]MLs are able to redistribute and focus light and inhibit the loss of energy due to light reflection and scattering. [21,22][25] Dong et al. have successfully improved the desalination process of a solar-driven membrane device with an ML array (MLA). [15]In addition, surface MLs can enhance the photodegradation efficiency of organic pollutants in water. [26]29] MLs with strong focusing effects also have the potential to concentrate more energy from irradiation in solar-water reactors. [30]raditionally, collectors with different shapes have been applied to enhance solar-utilization efficiency in solar-water reactors.
The basic solar collectors, such as parabolic or flat plates made with absorber or reflector surfaces, are inserted into solar-water reactors by trapping more solar energy. [31]Inclined plate collectors are developed from the flat plates, receiving more irradiation due to the tilted angle for higher effectiveness. [32]With using tube receivers, parabolic trough collectors have higher effective areas and more concentrated solar-radiation inside than planar reactors. [6,33]Compound parabolic collectors combine the functions of multiple types of reactors and further improve the solar-harvesting efficiency. [34,35]Generally, reactors with curved surfaces are found to be more efficient for solar-driven water treatment due to their ability to concentrate solar.However, the implementation of MLs is still restricted to reactors with flat surfaces. [36,37]The combination of MLs and reactors with curved surfaces possibly can provide further enhancement in the efficiency of solar-water treatment, which deserves more exploration.
Currently, most of the fabrication technologies of MLs are largely limited to flat substrates, small surface areas, and a low total number of MLs, [38] including hot-embossing, [39,40] laser-writing, [41,42] soft-lithography, [43,44] drop-templating, [45,46] and high-precision 3D printing. [47,48]Multiple steps are required to fabricate MLs on curved surfaces.For instance, an MLA can be prepared on a flexible poly(methyl methacrylate) (PMMA) substrate through screen printing followed by UV curing. [20]By combining a flexible mask and the reactive-ion-etching process, MLs were fabricated on photo-curable resin with a curved surface. [49,50]In a multiple-templating method, MLAs on a planar substrate are first copied to a flat film of siloxane elastomer that was bent into a curved template for a second round of templating. [51,52]The technologies mentioned earlier are dependent on complicated devices [53,54] and precise control of the dosage of the materials of MLs [55] .In comparison, surface MLs can be fabricated without sophisticated equipment by a solvent-exchange process followed by local photopolymerization. [56,57]The advantages of solvent exchange include various available substrate materials, solution-based fabrication processes, and flexibility for adjusting the size and curvature of MLs. [58]However, surface MLs made by solvent-exchange process are still restricted on 1D fiber or 2D planar surfaces, which limits the development of surface ML-functionalized reactors for a broader range of applications in solar-water decontamination inside large reactors with curved surfaces.
In this work, we present a highly tunable and scalable approach for producing MLs on 3D topological surfaces using a solvent dilution process.We apply this method to design reactors with improved photodegradation efficiency of organic contaminants in water.The method proves to be feasible for the preparation of reactors having irregular shapes, and large surface areas, comparable to three times of a standard 4 inch wafer, or made with glass or plastics, allowing for the functionalization of a broad range of reactors with MLs.By adjusting the parameters in the solvent dilution process, we are able to alter the size distribution and surface coverage rate of MLs.With this technique, the strengths of solar reactors with curved surface and MLs with strong focusing effects are combined for optimized solar-photodegradation efficiency.The as-prepared MLs on curved surfaces may provide a simple and effective approach to enhance solar-driven photodegradation in water treatment.

Results and Discussion
Surface PMMA MLs are polymerized from the surface methyl methacrylate (MMA) microdroplets forming in a solvent dilution process.The dilution process is designed to generate surface microdroplets on the complicated surface, which is based on a three-component fluidic system.The ternary system is composed of water, ethanol, and MMA, and the phase diagram of this system is shown in Figure 1a,b.During the solvent dilution process, the concentration of MMA changed along the dilution curve in Figure 1a,b.The solvent dilution process can be conducted in arbitrary glass reactors with functionalized inner surfaces (Figure 1c,d).Surface microdroplets of MMA containing a photoinitiator were polymerized into PMMA under UV irradiation (Figure 1e).To evaluate the performance of ML-functionalized reactors, the photodegradation of methyl orange (MO) in the water matrix is conducted in the reactors (Figure 1f ).

Light Intensity of MLs Functionalized on the Curved Surface of Glass Reactors
Considering the 3D curved surface had complex geometric structures, MLs functionalized on the curved surface are classified into two situations, as illustrated in Figure 2a.The surface ML is immobilized on a convex surface or a concave surface in situation 1 or 2, respectively (Figure 2b,c).Through the optical simulation by Zemax OpticStudio, the representative cross-sectional light intensity profile of a single ML in situations 1 and 2 under light irradiation is shown in Figure 3a,b, respectively.The lateral radius (r) of the ML involved in the optical simulations is 30 μm, located on the inner surface of a glass reactor with the curvature (R 2 ) of 11 mm (Figure 3a) and À11 mm (Figure 3b).In situation 1, a focal point with the highest local intensity is found at the position of 164 μm below the top point of the inner surface of the reactor.In situation 2, no focal point exists and the light beams under the ML have higher intensity than other areas, contributed by the focus effect of the ML.In comparison, the maximum intensity of light under the ML with the same lateral radius (r) is around 10 5 higher in situation 1 than that in situation 2. The huge difference indicates that the ML-functionalized convex surface has a much stronger focus effect due to the shape of the reactor surface.
According to the optical simulations, the dependence of peak irradiance intensity (I Peak ) on the size of MLs and the curvature of the reactor surface varies in situations 1 and 2. As demonstrated in Figure 3c, I Peak continuously increases with the lateral radius of the MLs in situation 1 for the reactors with different curvature.When R 2 is 30 mm, the MLs with the same radius achieve the highest intensity at focal points, while the I Peak is the lowest when the R 2 is 15 mm.In situation 2, I Peak maintains a similar value when the size of MLs differs (Figure 3d).Slight enhancement I Peak is observed when the curvature value of the reactor surface increases, which is possibly attributed to the weaker diffusing effect of the concave surface with a larger curvature value.
The area occupied by a single ML determines the amount of energy received from light irradiation.To quantify the strength of the focus effect of MLs, i Peak is defined by normalizing I Peak by the area covered by the ML.The influence of MLs lateral size and reactor curvature on i Peak is revealed in Figure 3e,f.In situation 1 (Figure 3e), i Peak does not continuously increase but shows a maximal value when the lateral size of the MLs reaches a certain value.The size of MLs with the highest i Peak shifts with the curvature of the reactor surface.In situation 2 (Figure 3f ), i Peak decreases with the lateral radius of MLs due to the negligible change of I Peak .

Controlled Fabrication of MLs on the Inner Surface of Cylindrical Reactors
The representative photos of ML-functionalized cylindrical glass reactors prepared by solvent dilution and local photopolymerization are displayed in Figure 4a.As shown in the photo, the cylindrical reactor with surface MLs functionalized remains transparent.Surface MLs are immobilized all over the inner surface of the reactors.The difference in MLs density and size is observed in the vertical direction.Based on the difference, each reactor is divided into three regions, including the top, middle, and bottom regions, and the properties of MLs in each part are relatively identical.The difference in the fabrication conditions of samples 1-9 is quantified by the concentration change of MMA with time in the solvent dilution process, which is displayed in Figure 4b-d.According to the varied parameters, the concentration dilution curves of MMA are classified into three groups.For samples 1-3, only the initial concentration of MMA (C MMA,0 ) in solution A is altered, for samples 2, 4, and 5, only the initial volume of solution A (V SolA ) is changing, and for sample 3, 6, 7, 8, and 9, only the flow rate during solvent dilution varies.
The concentration of MMA (C MMA ) with dilution time (t) can be calculated by Equation (S1).Before the solvent dilution process, the initial concentration of MMA is C MMA;0 , and the initial volume of solution A is V SolA .The dilution process can be divided into stages based on the correlation between the total volume of liquid (V T ) and t.In the first stage, V T is smaller than the volume of the reactor (V R ), and t should be smaller than the critical time, t R .t R is the dilution time when V T is the same as V R .At this stage, V T is equal to the addition of the volume of V SolA and solution B (V SolB ) at the time of t.V SolB can be calculated by the flow rate (Q) and t, so V T can be calculated with Equation (S2).With Equation (S2), the value of t R can calculated according to Equation (S3).By combining Equation (S2) and (S3), C MMA in the first stage can be quantified by Equation (1).When t is larger than t R , which is the second stage of the dilution process, the addition of V SolA and V SolB is over V R , so V T becomes constant and equal to V R .In the second stage, the differential change of C MMA (dC MMA ) with differential time (dt) is demonstrated by Equation (S4), and by integrating Equation (S3) from time 0 to t, the C MMA function with t in the second stage is obtained by Equation (2).
Stage 1 (t ≤ t R ) The morphology and spatial distribution of surface MLs are demonstrated by the images obtained from the optical microscope in Figure 4e-g.In all three regions, surface MLs randomly distributed on the homogeneous hydrophobic surface.Size distribution variation among the top, middle, and bottom regions of each ML-functionalized cylindrical reactor is validated with size analysis in Supporting Information, Figure S5a-i.By comparing the size distribution curves in the bottom region and the top region, it is found that a single peak can be observed in the lateral radius range from 5 to 65 μm.However, the size distribution curve of MLs in the middle region does not present an obvious peak in the same range of size.The frequency of MLs with a lateral radius larger than 70 μm in the middle region is higher than that in the other two regions.Generally, the lateral diameters of MLs in the middle region are generally larger than the diameters of MLs in the top and bottom regions.
The spatial difference in the size distribution along the glass cylindrical glass reactor is attributed to the variation of MMA concentration gradient with the relative position in the reactor and time.The solvent dilution process in cylindrical reactors can be divided into three stages as illustrated by the sketch in Supporting Information (Figure S6).In the first stage, the ouzo effect can be observed after adding solution B to solution A. Small droplets (size: a few to tens micrometers) form and adsorb onto the hydrophobic surface.The microdroplets formed in this stage become the MLs in the bottom region after the UV curing step.In the second stage, as the ratio of water further increases, phase separation happened, resulting in the formation of larger droplets (hundreds of micrometers).In this stage, the droplets are transferred into the MLs in the middle region.In the third stage, the total amount of the liquid added exceeds the capacity of the reactor and starts to get out of the reactor.Due to the phase separation, the oil-rich phase sits on the upper part and is excluded from the reactor first.Therefore, the concentration of MMA drops significantly.The rapid decrease of oil concentration leads to the decrease of droplet size, so as the size of MLs size on the curved surface.
The adjustment in the parameters during the solvent dilution process brings about the variation of the overall size distribution of surface MLs on each glass cylindrical reactor as illustrated in Figure S7a-c (Supporting Information).The overall size distribution curves consider all the MLs on the inner surface of the reactor regardless of the variance among the three regions.By comparing samples 1-3 (Figure S7a, Supporting Information), the largest lateral radius among MLs in a reactor increases with C MMA;0 in solution A. Among sample 2, 4, and 5 (Figure S7b, Supporting Information), higher frequency of MLs with lateral radius over 80 μm of the sample is observed when V SolA is larger.By controlling the flow rate of filling solution B for samples 3, 6, 7, 8, and 9 (Figure S7c, Supporting Information), the width of the size distribution is enlarged at the higher flow rate.In addition, the portion of MLs with lateral size smaller than 50 um becomes higher, and the maximum size of MLs becomes larger as the flow rate increases, which is more obvious in the sample prepared with a flow rate higher than 8 mL min À1 .
The influence of each parameter in the solvent dilution process is analyzed based on the MMA dilution curves in Figure 4b-d and the corresponding size distribution in samples 1-9.Higher C MMA,0 in solution A contributes to the higher oversaturation level during the solvent dilution process, so larger surface MLs are generated in sample 3. Except C MMA;0 in solution A, the total amount of MMA determines the oversaturation level during the solvent dilution.For samples 2, 4, and 5, the concentration gradient of MMA is larger under the condition with a higher V SolA .Therefore, more MLs with a radius over 80 μm were observed in sample 2. By increasing the flow rate in the solvent dilution process, the oversaturation level decreases more rapidly.Simultaneously, the higher flow rate also results in a faster mass-transfer rate, contributing to the growth of droplets.Therefore, we could see a broader size distribution in the sample at a higher flow rate (Figure S7c, Supporting Information).
The surface coverage rate of all ML-functionalized cylindrical glass reactors is plotted in Figure 4h.The minimum surface coverage rate is 30.6% obtained from sample 1, and the maximum rate is 49.9% in sample 3.By comparing the samples fabricated with varied C MMA , it is found that the surface coverage rate increases with the volume ratio of MMA.Among the samples with the only difference in the V SolA , the surface coverage rate first decreases and then increases with the volume of solution A. By only adjusting the flow rates during the solvent-exchange process, the surface coverage rate first increases and then remains at a similar level when increasing the flow rate from 3 to 15 mL min À1 .The total intensity at the focal points of MLs in each sample (I Total ) can be calculated based on the optical simulations of MLs on the curved surface and the size The correlation of I Total and the parameters in the solvent dilution process is investigated for the optimization of the ML-functionalized reactors.The correlation between I Total and C MMA;0 in solution A is demonstrated in Figure 4i.As C MMA;0 increases from 4.7 vol% (sample 2) to 7.0 vol% (sample 3), I Total decreases first and then increases.As displayed in Figure 4i, I Total of the samples increases when the V SolA is improved from 5 to 12 mL.The improvement in I Total is the possible consequence of the higher surface coverage rate of MLs in the sample prepared with larger V SolA (Figure 4h) and more MLs with the lateral radius over 80 μm (Figure S7b, Supporting Information).The correlation between I Total and the flow rate in the solvent dilution process is illustrated in Figure 4j.When the dilution rate varies between 3 and 15 mL min À1 , I Total first increases and then decreases as the dilution rate becomes higher.Among the five samples fabricated with different flow rates in solvent dilution, sample 6 has the highest I Total .The possible reason for the high I Total in sample 6 is the relatively high surface coverage (Figure 4h) and more MLs with lateral size between 20 to 60 μm (Figure S7c, Supporting Information), which is the size range of MLs with relatively higher focus effect as shown in Figure 3e.
Among the three involved parameters in the solvent dilution process, the flow rate causes the most significant change of I Total .The effect of changing C MMA , 0, and V SolA is mainly revealed in MLs with larger radius (over 70 =mum), and those larger MLs play a less important role in the focus effect as indicated in Figure 3e.In comparison, the size distribution width can be adjusted by controlling the flow rate in the solvent dilution process.According to the experimental results, the I Total reaches the maximum value when the flow rate is 8 mL min À1 .However, if the flow rate further increases, C MMA changes too fast to ensure a stable concentration gradient.In this situation, some large MLs (r > 200 μm) form and escape from the surface due to the increasing buoyancy, and then the portion of small MLs (r < 20 μm) becomes higher, leading to a lower value I Total and the weakened focus effect.

Enhancement of Photodegradation of MO in ML-Functionalized Cylindrical Reactors under Simulated Solar Light
Based on the representative absorbance spectra in Supporting Information (Figure S8), the enhancement of photodegradation of MO by ML-functionalized cylindrical glass reactors is confirmed.Based on the UV-visible spectra, the photodegradation efficiency of MO under the simulation light is calculated and plotted with irradiation time as shown in Figure 4k.In the control experiment conducted in the bare cylindrical glass reactor, the degradation of MO during the irradiation of 3 h is negligible.For the photodegradation of MO in the ML-functionalized cylindrical glass reactors, the photodegradation efficiency obviously increases under the same irradiation condition.The enhancement factor calculated with Equation ( 6) helps to quantify the improvement in η of MO in ML-functionalized cylindrical glass reactors (Figure 4l).The enhancement factor of MO photodegradation in ML-functionalized cylindrical glass reactors decreases with the irradiation time.During the same time of light treatment, the difference in the enhancement of η is observed among samples 1-9.Sample 6 shows the highest enhancement factor of 73.5, while the least enhancement is found in sample 5.The temperature change of bare cylindrical reactors and ML-functionalized reactors is similar (Figure S9, Supporting Information), indicating that the photothermal effect from MLs is negligible, possibly due to the thermal equilibrium with the surrounding aqueous phase.The difference in the performance of ML-functionalized cylindrical glass reactors is attributed to the varied properties of surface MLs in each sample that are caused by altering fabrication conditions in the solvent dilution process.In brief conclusion, the size and number of MLs determine the strength of the focus effect, quantified by I Total , which is revealed in the variation of η of MO in the ML-functionalized reactors (Figure 4m).

MLs on the Surface of the Reactors in Complex Geometry: Properties and Performance in Photodegradation
Aside from cylindrical reactors, surface MLs can be immobilized on the inner wall of more complex reactors, such as snowmanshaped, petal-shaped, three-armed, and flower-shaped glassware.As shown in Figure 5a-d, the whole inner surface of glass reactors with irregular shapes are functionalized by surface MLs, and the ML-functionalized reactors remain transparent.By zooming on the smaller area, surface MLs are observed to be randomly distributed over the inner wall of the reactors.The details of MLs morphology and surface coverage rates are obtained from the representative microscope images of sample 11 (snowman shape) and sample 12 (petal shape) in Supporting Information (Figure S10a-d).The surface coverage rate of MLs in sample 12 is around 6% higher than that in sample 11.By plotting the size distribution curves of MLs on the inner surface of the two reactors in Figure S11c (Supporting Information), we can observe that the frequency of MLs decreases with the lateral radius of MLs in both reactors.The frequency of MLs with a radius smaller than 7 μm is higher in the petal-shaped reactor than that in the snowman-shaped reactor, while a larger portion of MLs with a radius larger than 50 μm are found in the snowman-shaped reactor.
The reactors from sample 11 to 14 are utilized for the photodegradation of organic pollutants under either natural light or simulated solar light.The experimental setup shown in Figure 5e enables the MO-solution-filled reactors to receive solar light through the glass window.As the irradiation time increases, an obvious decay of color is observed in the ML-functionalized reactors (Figure 5f ), but the color change of solution in the bare glass reactors is indistinguishable.The more rapid color change in ML-functionalized irregular reactors indicates the faster degradation rate of MO, which is confirmed by the photodegradation efficiency calculated by Equation (3).In Figure 5g,h, η of MO in reactors without surface MLs and with surface MLs is plotted with irradiation time, respectively.In comparison, η in the ML-functionalized petal-shaped reactor is the highest, reaching 71.0% after 11 d of irradiation, achieving around 2 times enhancement compared with that without surface MLs.
The improvement of photodegradation of MO by surface MLs functionalized on the wall of irregular reactors under simulated solar light is significant.With the experimental setup demonstrated in Figure 5i, the light source is set on top of the reactors.During the irradiation of 2 h, the color change of MO solution in the unmodified reactors is negligible (Supporting Information, Figure S11), while the decolorization of MO solution is more obvious in the ML-functionalized reactors after the irradiation, especially the ML-functionalized flower-shaped reactor (Figure 5j).The photodegradation efficiency of MO in the irregular glass reactors without and with surface MLs is plotted in Figure 5k,l.On one hand, η of MO is less than 7% in all bare irregular glass reactors after the irradiation under simulated solar of 2 h.On the other hand, η of MO in the ML-functionalized irregular reactors is substantially increased.Among all ML-functionalized irregular glass reactors, the flower-shaped reactor has the best performance, in which η of MO reaches 83.0% after the treatment of 2 h, which is more than 12 times higher than the η in the nonfunctionalized flower-shaped reactor.
As shown in photodegradation efficiency, we can conclude that surface MLs generally improve the photodegradation efficiency of MO under natural indoor light and simulated solar light.In addition, the effect of surface MLs varies correspondingly while the shape of the reactor changes.One of the possible reasons is the variance in the surface area to volume ratio of reactors with different shapes.In the experiments under the irradiation of office light, η of MO increases with the surface area to volume ratio of reactors.However, after changing the direction of irradiance, the actual area receiving the irradiation of light sources is also altered, further affecting the performance of ML-functionalized reactors (shown in Supporting Information, Figure S12).As shown in the photodegradation under the simulated solar light, the flower-shaped reactor presents much higher η than any other reactors due to the highest irradiated area with the light source set on top of the "multi-petals" side.Additionally, the properties of surface MLs, such as surface coverage rate and size distribution, vary with the reactor shape because different geometric structures of the reactor may disturb the formation and growth of microdroplets during the solvent dilution process.Last but not least, the curvature of the reactor surface also influences the strength of the focus effect, as shown in Figure 3.To optimize the design of ML-functionalized reactors with complicated shapes, further studies are required in the future.

Enhanced Photodegradation under Indoor Solar Light
The fabrication method of surface MLs with solvent dilution can be scaled up by 16.7 times larger in volume and 5.6 times larger in the surface area inside a cylindrical glass reactor.The scaledup ML-functionalized glass reactor (sample 10) and the zoomed-in image of MLs on the inside wall are displayed in Figure 5 m, showing the high transparency of the MLfunctionalized glass bottle and the morphology of surface MLs.With the experimental setup in Figure 5m,n, higher photodegradation efficiency of either MO or sulfamethoxazole (SMX) in the ML-functionalized bottle under natural indoor light is acquired than that in a bare bottle (Figure 5o,p).The curve of η with time for MO or SMX differs due to the variation of photodegradation mechanisms. [58]The η of MO under indoor solar light continuously increases because sufficient active species are generated under the visible light during the light treatment.Meanwhile, the photodegradation of SMX under irradiation in the visible range is restricted in the absence of photocatalysts, [59,60] so the η of SMX hardly increases after the irradiation of 50 days (d).
The improvement of photodegradation under natural indoor light first increases with the irradiation time and then decreases, which is observed in the degradation of both MO and SMX.The highest enhancement factor in η by the ML-functionalized bottle is 6.57 for the degradation of MO after 62 d of irradiation, while the highest enhancement is 1.22 for the degradation of SMX after 27 d of light treatment.In addition, compared with the enhancement of η achieved in the wide cylindrical reactors is much lower than that in the narrower cylindrical reactors (samples 1-9) and the reactors with irregular shapes (samples 11-14).One of the important reasons may be the much lower surface area to volume ratio of the wider cylindrical reactor (Table 1).Last but not least, the ML-functionalized glass bottle can be reused multiple times without observing any difference in the chemical properties of MLs (as shown in Supporting Information, Figure S13).

Conclusion
Complex reactors with curved surfaces are functionalized by surface MLs for higher photodegradation efficiency of organic contaminants in water.Surface MLs on the inner surface of reactors are fabricated through polymerization of surface microdroplets generated in a solvent dilution process.The solution components, solution amount, and flow conditions during the dilution process are flexibly controlled to obtain customized size distribution and surface coverage rate of MLs.Enhanced photodegradation efficiency attributed to focusing effects of MLs is demonstrated in ML-functionalized reactors with six types of shape.Both optical simulations and experimental results show that the focusing effect strength of MLs, quantified by the total intensity at focal points, is correlated with the dimensions and spatial arrangement of MLs and the curvature of the surface.Through strategic fabrication of ML-functionalized reactors, focusing effects of MLs are optimized, achieving over 80-fold improvement of photodegradation efficiency in maximum.Furthermore, the method is applied to fabricate MLs over the larger curved surface of a reactor with the capacity of up to 500 mL, demonstrating the scalability of the approach.The technology demonstrated in this work may be implemented to design compact and highly efficient reactors for sustainable solar-driven water treatment regardless of the complexity of topological shape of reactors.We envision that the fabrication technology of MLs on 3D curved surfaces developed in this work opens up a wide range of applications of MLs in novel optical devices and sensing, with potential impact far beyond photodegradation in sustainable solar-driven water treatment focused in the current study.

Experimental Section
Fabrication and Characterization of PMMA MLs on Curved Surface: The solvent dilution process was conducted in reactors with inner surface hydrophobized with octadecyltrichlorosilane (OTS, 98.9%, Acros Organics, Fisher Scientific).The coating procedure was the same as what was described in the previous literature. [56]During the dilution process, solution A with a volume of four-tenths of the reactor capacity was first added.Solution A contained 6.2 vol% MMA (≥ 98:5%, Alfa Aesar) and 0.62 vol% 2-hydroxy-2-methylpropiophenone (96%, Fisher) as a photoinitiator in 50 vol% ethanol aqueous solution.Afterward, solution B that contained 0.05 vol% photoinitiator in MMA-saturated Milli-Q water was pumped into the vertical-set reactor through two outlets placed at the opening of the reactor with a fixed flow rate of (shown in the sketch in Figure 1c).The total volume of solution B for the solvent dilution process was three times of the capacity of a reactor.The excess liquid in the solvent-exchange process was excluded from the top of the reactors.MMA microdroplets containing the photoinitiator formed on the inner surface of the reactors due to the oversaturation when solution B replaced solution A.
The reactor used was optional from the self-designed arbitrary glass reactors with different shapes shown in Figure 1d.Two of the reactors had cylindrical shapes but with different dimensions.The smaller cylindrical glass reactor had a volume of 30 mL (Class A clear glass vial, Fisherbrand).The smaller cylindrical reactor was utilized as a representative reactor for the adjustment of parameters in the solvent dilution process due to its stability in dimension and simplicity of structure.The adjusted parameters include the initial volume of solution A before the solvent dilution (V SolA ), initial concentrations of MMA in solution A (C MMA;0 ), and flow rates when adding solution B. A much larger glass cylindrical reactor was also used to verify the scalability of this method, which had an inner diameter of 8.0 cm, a height of 10.0 cm, and a volume of 503 mL.Considering the scale of the reactor, the OTS coating procedure of the inner surface was modified into a chemical vapor deposition method reported in the literature to save the use of chemicals. [61]uring the solvent dilution process in the larger cylindrical reactor, a solid cylindrical object with the same height as the reactor and a diameter of 5.0 cm was set vertically in the center of the reactor to decrease the required amount of solutions A and B. Three tubes instead of two were evenly set at the openings in the solvent dilution process to ensure the required flow rates.The other four irregular reactors included a snowman-shaped, petal-shaped, three-arms-shaped, and a flower-shaped reactor, and the volume inside was around 9, 11, 27, and 72 mL, respectively.The shape, materials, volume, surface area, and surface area to volume ratio (S/V) of all arbitrary reactors with surface MLs functionalized and corresponding conditions in the solvent dilution process are summarized in Table 1.
After the solvent dilution process, the glass reactors filled with the mixture of solutions were sealed and irradiated under a UV lamp (365 nm, Analytik Jena UV lamp).The reactors with microdroplets on the inner surface were irradiated for 20-40 min depending on the reactor size for photopolymerization.During the irradiation process, the glass reactors were rotated every 10 min to ensure that sufficient irradiation all around the reactors.After photopolymerization of the droplets, the reactors functionalized with MLs were rinsed with water and ethanol subsequently and dried by air for characterization.The glass reactors functionalized with surface MLs were observed with an optical microscope (Nikon H600l and Nikon DSFi3).Each of the MLs functionalized cylindrical reactor (samples 1-9) was divided into three parts, bottom (15 mm), middle (28 mm), and top part (15 mm) for microscope observation.To analyze the surface coverage rate and the size distribution of MLs in each sample, five photos were collected for each sample and were analyzed with Image J. Atom forces microscopy and scanning electron microscopy were applied to characterize the 3D morphology of MLs on glass vials (Figure S1a-c, Supporting Information).The inner surface of a plastic bottle without surface functionalization could be also utilized as the substrate of surface microdroplets and MLs in the presence of surfactant, which is demonstrated in Supporting Information (Figure S2a,b).
Photodegradation Inside ML-Functionalized Reactors: MO (Sigma-Aldrich, ACS reagent) was utilized as one of the organic pollutants to demonstrate the effectiveness of the ML-functionalized reactors.The 5 mg L À1 MO aqueous solution was prepared by using ultrapure water (Milli-Q Direct 16) as the solvent, and the pH value was controlled at 3.0 by adding sulfuric acid (98%).Similarly, the aqueous solution of (SMX (analytical standard, Sigma Aldrich), another typical organic contaminant in wastewater, with a concentration of 5 mg L À1 was prepared.The pH value of the SMX solution was not changed with acid or base and was measured at 7.0.
The ML-functionalized cylindrical glass reactors fabricated by altering the conditions in the solvent dilution process (samples 1-9) were first filled with MO aqueous solution and then horizontally set under the simulated solar light (1 sun, SS200AAA Solar Simulation Systems, Photo Emission Tech) for 1, 2, and 3 h.The ML-functionalized reactors with irregular shapes (samples 11-14) were used for the photodegradation of MO under both simulated solar light and natural indoor sunlight (Figure 1f ).The light spectrum of the light source is shown in Figure S3 (Supporting Information).The photodegradation efficiency, η, was calculated via Equation (3) to quantify the extent of photodegradation.According to Beer-Lambert law [62] (Equation 4), the absorbance value of MO at a certain wavelength was proportional to the concentration of MO in the solution.Therefore, Equation (3) could be transformed into Equation (5), where A i ni is the peak value of the absorbance curve at the wavelength of 504 nm and A a f t is the peak value at the same wavelength after the light treatment.The absorbance values of the MO solution were obtained from a UV-visible spectrometer (Thermo Scientific, Genesys 150).
A ¼ log 10 I 0 I ¼ εCL (4) A bare cylindrical reactor was used to conduct the same photodegradation process as the control group.The enhancement in η achieved by the surface MLs that were functionalized on the inner surface of cylindrical glass reactors was represented by an enhancement factor defined by Equation (6).
The larger ML-functionalized glass cylindrical reactor (sample 10) was also used for the photodegradation of MO and SMX in ultrapure water under indoor solar light.The 500 mL MO or SMX aqueous solution prepared with the same method as mentioned earlier was added to sample 10.The solution-filled bottle was set beside the window for indoor light irradiation in Room 12-380 of Donadeo Innovation Centre for Engineering, Edmonton, Canada.For each round of light treatment, a bare glass bottle filled with the same amount of aqueous solution was set under the same light source simultaneously for comparison.The photodegradation of MO in water was conducted from December 17, 2021 to March 16, 2022, while the photodegradation of SMX in water started on March 16, 2022 and ended on May 15, 2022.η of MO and SMX was calculated by Equation ( 5) by inserting the absorbance values obtained from the UV-vis spectrometer.
Optical Simulations of Surface MLs: The surface ML-functionalized reactors filled with MO solution were modeled in 3D space with Zemax OpticStudio.The whole system was illuminated by a plane wave source along the Z direction.The space within the reactor was filled with MO aqueous solution, and the yellow arrows indicated the direction of irradiation, as shown in Figure 2b,c.The source intensity was set to be the same as in the experiment (1 sun).To identify the focal point of the ML, an X-Z plane monitor was placed along the central axis of the ML to obtain a cross-sectional intensity profile.The point in the profile with the highest irradiance intensity was identified as the focal point.
The peak irradiation intensity (I Peak ) in the cross-sectional light intensity profile was defined as the intensity at focal points of MLs.The peak intensity normalized by the area covered by an ML (i Peak ) was calculated based on Equation (7), which eliminated the influence of the size of surface MLs and revealed the strength of focus effect of MLs.I Total is the total intensity at the focal points of MLs within an area of 1 cm 2 from a specific sample, which can be calculated by Equation (8).In Equation (8), N is the total number of MLs within an area of 1 cm 2 from a specific sample.
The thickness of a single ML, h, is the key parameter for the optical simulation, which is obtained by Equation (9) based on the geometric relationship shown in Figure 2b,c.In the equation, R 1 and R 2 are the curvatures of the spherical side of the ML and the surface of the 3D curved reactor surface, respectively.r is the lateral radius of the observed ML which can be extracted by the analysis with Image J. α is half of the central angle that is occupied by the single ML; θ is the contact angle of PMMA MLs on the OTS-coated substrate, which is 7.5 ðAE0.2Þ∘ . [63]To distinguish the difference between situations 1 and 2, R 2 had a positive value for situation 1 and a negative value for the other situation.

Figure 1 .
Figure 1.a) The phase diagram of a ternary system composed of water, ethanol, and methyl methacrylate (MMA) and b) the zoomed-in plot, and the dilution paths involved in the microlenses (MLs) fabrication are displayed.c) The sketch of the fabrication process of a surface MLs functionalized reactor based on a vertical-oriented solvent-exchange process, using a snowman-shaped reactor as an example, and the opening of the reactor is set on top.d) The reactors with different shapes in the solvent dilution process, the opening of each reactor was labeled with a red-dashed box.e) The sketch of the UV curing process to transform surface microdroplets to surface MLs.f ) The experimental setup of the indoor light treatment of methyl orange (MO) solution in an ML-functionalized petal-shaped reactor, from 0 to 3 days (d).

Figure 2 .
Figure 2. a) Sketch of a 3D curved surface functionalized with surface MLs under light irradiation.Surface MLs can be classified into two types.The ML on a convex surface is defined as situation 1, while the ML on a concave surface is defined as situation 2. b) Zoomed-in sketch of MLs in b) situation 1 and c) situation 2.Here, R 1 is the curvature of the lens and R 2 is the curvature of the surface at the position of the ML.r is the lateral radius of the ML and h is the height of the ML.θ is the contact angle of the ML and α is the half-central angle of the area occupied by the ML on the curved surface.

Figure 3 .
Figure 3. Cross-sectional view of the light intensity profile of a single ML (r = 30 μm) at the curved surface reactor in a) situation 1 (convex surface, R 2 = 11 mm) and b) situation 2 (concave surface, R 2 = À11 mm).Z = 0 represents the top point of the inner surface of the cylindrical reactor.Peak irradiance intensity (I Peak ) of a single ML with different r on the surface with varied R 2 in c) situation 1 and d) situation 2. Peak irradiance intensity normalized by the area covered by an ML (I Peak ) calculated with Equation (5) of a single ML with different r functionalized on the curved surface with varied R 2 in e) situation 1 and f ) situation 2. The dash and solid lines separately indicated the trend of I Peak and i Peak with the lateral radius of a single ML.

Figure 4 .
Figure 4. a) Photos of ML-functionalized cylindrical glass reactors from sample 1 to 9. Each reactor is divided into three regions, including the top (15 mm), middle (28 mm), and bottom (15 mm) regions.The concentration of MMA (C MMA ) with the time in the solvent dilution processes with b) different initial C MMA , c) varied initial volume of solution A, and d) varied dilution rates.Zoomed-in plots of (b-d) are displayed in Supporting Information, Figure S4.Optical images of MLs on the inner surface of samples 1-9 in the e) top region, f ) middle region, and g) bottom region, and photos are color-coded as the sketch in (a).h) Surface coverage rates of MLs on the inner surface in samples 1-9.Correlation between I Total in an area of 1 cm 2 and i) initial MMA concentrations in solution A (samples 1-3) and varied initial volume of solution A (sample 2, 4, and 5); j) varied dilution rates (samples 3, 6, 7, 8, and 9).k) The photodegradation efficiency (η) and l) enhancement factor of MO under the simulated solar irradiation of 1-3 h with bare cylindrical glass reactor (control) and ML-functionalized samples 1-9.m) η of MO after the light treatment of 1 and 2 h with the I Total over the area of 1 cm 2 in samples 1-9.

Figure 5 .
Figure 5.The photo and zoomed-in photo of surface MLs-decorated a) snowman-shaped (sample 11), b) petal-shaped (sample 12), c) three-arms-shaped (sample 13), and d) flower-shaped (sample 14) glass reactor.e) ML-functionalized reactors filled with MO solution (5 mg L À1 , pH = 3.0) irradiated by the natural-light indoor treatment for 6 d. f ) The color change of MO solution after the irradiation of 11 d in samples 11-14.η of MO in g) bare and h) ML-functionalized arbitrary glass reactors with irradiation time.i) Experimental setup of the simulated solar-light treatment of MO in samples 11 to 14. j) The color change of MO solution in sample 14. η of MO in the k) bare and l) ML-functionalized irregular glass reactors under simulated solar light.Experimental setup of the natural light indoor treatment of m) MO (5 mg L À1 , pH = 3.0) and n) sulfamethoxazole solution (5 mg L À1 , pH = 7.0), and a photo of an ML-functionalized large cylindrical reactor (sample 10) and the zoomed-in image of MLs on the inner surface are attached on (m).η of o) MO and p) SMX under the natural light indoor with a bare glass bottle (control) and sample 10.

Table 1 .
Conditions of solvent dilution process for the preparation of ML-functionalized reactors.