3D Printing a Low‐Cost Miniature Accommodating Optical Microscope

This decade has witnessed the tremendous progress in miniaturizing optical imaging systems. Despite the advancements in 3D printing optical lenses at increasingly smaller dimensions, challenges remain in precisely manufacturing the dimensionally compatible optomechanical components and assembling them into a functional imaging system. To tackle this issue, the use of 3D printing to enable digitalized optomechanical component manufacturing, part‐count‐reduction design, and the inclusion of passive alignment features is reported here, all for the ease of system assembly. The key optomechanical components of a penny‐sized accommodating optical microscope are 3D printed in 50 min at a significantly reduced unit cost near $4. By actuating a built‐in voice‐coil motor, its accommodating capability is validated to focus on specimens located at different distances, and a focus‐stacking function is further utilized to greatly extend depth of field. The microscope can be readily customized and rapidly manufactured to respond to task‐specific needs in form factor and optical characteristics.


Introduction
There are growing interests in miniaturing imaging platforms for potential commercial, research, and educational applications. [1] However, dealing with optical and optomechanical components with increasingly smaller size requires high manufacturing precision and tight system assembly tolerances to achieve diffraction-limited performance. Traditional lens manufacturing relies on costly and time-consuming lapping, grinding, and polishing processes. Injection molding can produce low-cost polymer lenses, but still requires precision machined molds for the high-performance features needed for optical applications. [2] Such applications also necessitate equally strict tolerances for system assembly and the alignment for the OpenFlexure Microscope design which provides precise mechanical sample manipulation in a lightweight and compact device. [15] These advancements have made AM optical lenses more economically favorable, however, there are still unsolved challenges in precisely manufacturing and assembling the dimensionally compatible optomechanical components into a miniaturized optical imaging system.
Unifying the fabrication processes for optics and optomechanics using AM would retain the low-cost, small-size benefit of the previously mentioned methods while improving the scalability and design customizability through additive manufacturing. In this article, we demonstrate the capability to AM the essential optical and optomechanical components for a miniaturized imaging platform featuring an aspherical lens and voice-coil motor (VCM). System configuration can be tailored for a range of lens focal lengths or procurement sources. Furthermore, the VCM features a near-zero Poisson's ratio 3D-printed compliant foundation to minimize interference during focusing, preserving customizability in imaging magnification. All optomechanical components are 3D printed in a single batch within 50 min with minimal assembly requirements, at a system cost of less than $4. We have experimentally compared the imaging characteristics of 3D-printed lenses with commercial polymer lenses. Finally, we demonstrate the active focusing through actuating the built-in VCM, and further implement the focus.stacking function by combining multiple images taken while axially translating the imaging lens, to greatly extend the depth of field (DOF). Figure 1 illustrates the modular design of the miniaturized accommodating microscope. A focusing motor translates the imaging lens along the optical axis to form magnified images onto the complimentary metal-oxide-semiconductor (CMOS) sensor for objects A and B located at different positions ( Figure 1a). Optical magnification requires translating the imaging lens further away from the CMOS sensor, and thus, the planar configurations commonly found in the cellphone cameras are not applicable. While the static DOF is shallow, successively acquired frames along the optical axis can be digitally combined using a focus.stacking function, thereby extending the DOF. The system design ( Figure 1b) incorporates two primary strategies-PCR and passive alignment features-to simplify assembly (details in the Supporting Information). PCR strategy reduces system part count to only five 3D-printed components (lens, elastic lens mount, front cap, and clamshells) and four off-the-shelf components (IR-cutoff filter, ring magnet, coil, and CMOS sensor) (Figure 1b), with a full assembly chart of several customizable options shown in Figure 1c. The fabrication of all shown 3D-printed components is unified using the homemade µCLIP system [12a] (Figure S1, Supporting Information). AM offers customizable options in two progressive steps. First, imaging lens effective focal length (EFL) can be tailored for specific working distances and imaging magnifications.

System Design of the Miniature Optical Microscope
Two options (EFL = 4.3 mm and EFL = 3.0 mm) are shown in Figure 1c. The digital design model of the elastic lens mount can be readily modified, and 3D printed with a desirable offset between the lens and CMOS sensor. Second, taking the design for EFL = 3 mm as an example, 3D-printed lens (AM30) can be interchanged with off-the-shelf polymer lenses ("T30" and "E30" in Figure 1c). We intentionally constrained EFL and lens outer diameter to be compatible with commercial polymer lenses to improve design modularity. All the components are snapped into place using passive alignment features without the needs for a dedicate fixture and machines (Figure 1d, detailed assembly procedure in Movie S1 and the Supporting Information). The assembled miniature microscope shown in Figure 1e has the outer dimension of 8.10 × 8.10 × 29.85 mm 3 , comparable to a penny. Such methods also led to the dramatic cost reductions shown in Table 1 (detailed cost analysis can be found in the Supporting Information). The estimated unit cost of the whole system is $3.74, making it potentially disposable.

Design, Fabrication, and Characterization of Optical Imaging Lenses
We experimentally characterized each interchangeable lens option shown for Design 1 (FL = 3.0 mm): AM30 (Aspherical lens 3D printed in house), E30 Edmund Optics), and T30 (APL0303, Thorlabs Inc.) (Figure 2). Optical properties of photocurable resin (detailed recipe in the Experimental Section) were measured using a spectroscopic ellipsometer (J. A. Woollam M2000U). The wavelength-dependent refractive index of photocurable resin was extracted from ellipsometry measurements and was further used as inputs to optimize the aspherical profile of AM30 using Zemax Optic-Studio (optimized profile in the Experimental Section). The wavelength-dependent refractive index determines the aspherical coefficients of 3D-printed lens with a given aperture and focal distance. Figure 2a,b depicts the simulated ray diagram and point spread function (PSF), respectively. The optimized lens profile was loaded into µCLIP to fabricate AM30, whose scanning electron microscope (SEM, FEI Quanta 650 ESEM) image is shown in Figure 2c. For comparison, we show the SEM images of E30 and T30 in Figure 2d,e, respectively. All three lenses have smooth aspherical surfaces, but with additional surface asperities found on E30 and T30. The distinct structural defects along the circumferences of E30 and T30 are common to all the lenses (N > 10) we purchased from both vendors. The smooth surface profiles are further confirmed by optical profiler (Nexview, Zygo) measurement (Figure 2f-h). Images captured with a 240 µm diameter aperture for all three lenses illustrate smooth surface profiles. We extracted linear profiles (dashed lines I and II in Figure 2f-h) for quantitative comparison. AM30 has the best surface smoothness (Ra = 7.63 nm, Figure 2i), in comparison with E30 (Ra = 8.67 nm, Figure 2j) and T30 (Ra = 11.39 nm, Figure 2k). Notably, at a volumetric printing speed of 1.25 × 10 4 mm 3 h −1 , the 3 mm high AM30 was 3D printed in just 3 min.
We experimentally compared imaging resolution for all three lenses ( Figure S2, Supporting Information and description in the Supporting Information). Captured images of the USAF 1951 resolution target illuminated with green light (λ = 532 nm) for AM30, E30, and T30 are shown in Figure 2l-n, respectively. All three lenses exhibit a comparable imaging resolution capable resolving Element 3 of Group 7, corresponding to a 3.10 µm spatial resolution which optimizes the overall performance of the imaging system as the highest spatial frequency  c) Microscope components used in the assembly, including lenses with different focal lengths and the corresponding elastic lens mount, IR-cutoff filter, enclosure components, focusing motor components, and CMOS imaging sensor. d) Schematic illustration of microscope assembly process. e) Photo of packaged endoscope with 3D-printed components. Scale bar: 10 mm. of the lens shall not exceed the Nyquist frequency of the CMOS sensor (OV09734-H16A, Omnivision) to avoid aliasing. Images acquired using E30 and T30 show relatively poor contrast, which might be correlated to the aforementioned structural defects found in Figure 2d,e. Similar behavior is also observed under different illumination conditions ( Figure S3 and description in the Supporting Information). For the interest of customization, we also characterized the imaging resolution of AM43 with EFL of 4.3 mm ( Figure S4 and description in the Supporting Information).

Design, Fabrication, and Characterization of VCM Assembly
The VCM assembly (Figure 3a) was designed to focus the imaging lens through electromagnetic actuation. It contains an electromagnetic coil, a ring magnet, a compliant foundation, and the imaging lens mount. The components were carefully designed to ease the assembly process. To better illustrate the assembly process, relative positions of elastic lens mount, lens, ring magnet, and coil are illustrated in Figure 3a. The elastic lens mount consists of three components: lens mount, ring magnet mount, and compliant foundation from top to bottom, as shown in Figure 3b). Grooves on the top and at bottom of the mount were designed for mounting the lens and ring magnet, respectively. To allow for the use of tweezers during the assembly process, we designed the interspace on the top of the lens mount to provide clearance. Two large sliding blocks on the side of the mount correspond with the sliding rails in the upper shell, to calibrate the movement of the mount. Meanwhile, four half-cylinder lug bosses were designed to reduce friction during translating of the mount. We implement PCR design to integrate three distinct optomechanical components-lens mount, ring magnet mount, and compliant foundation-into a monolithic elastic lens mount. The axial translation of the imaging lens is determined by the balance between electromagnetic forces from the coil and magnet and resilient forces from the compliant foundation. Figure 3b shows the boundary conditions used in a simulated compression test in ANSYS (details in the Experimental Section). The primary design variable in the compliant foundation is the strut angle (θ), which was varied between −4.5° and +10° (inset, Figure 3b). We optimize θ to achieve sufficient axial deformation (>1.0 mm) while reducing radial deformation to minimize its interference during VCM actuation. An optimal 0° strut angle exhibits the lowest Poisson's ratio of 0.051, achieving a largest axial displacement of 1.17 mm (Figure 3c). The fabricated compliant foundation demonstrated a highly linear behavior during compression testing (Figure 3d), corroborating the numerical simulation. We further simulate the magnetic force (details in the Experimental Section) that an N50 neodymium ring magnet (R0545, Super-MagnetMan) bears as a function of axial distance and input voltage (Figure 3e) obtained by a previously reported algorithm. [16] ANSYS simulation suggests that the force of 52.96 mN-an amount readily available from the coil/magnet over the designed displacement range-is required to reach the largest axial deformation of 1.17 mm, validating the feasibility of our proposed axial translation mechanism. The simulated displacement fields in the lens mount during the cyclic contracting and extending motion are predominantly translational along the optical axis (Figure 3f).

Performance Validation of Assembled Miniature Microscope
We experimentally validated the performance of the assembled miniaturized microscope using AM30 (Figure 1e). The test objects used in this study are a pair of biological specimens (NEW-BRE1003, NOCOEX Company) indicated by objects A and B in Figure 1a that are spatially offset by 1.0 mm axially and 0.8-1.2 mm laterally. As the axial offset is greater than system DOF, we can actuate the VCM to translate the imaging lens to bring the desirable object in focus (Figure 4a,b). The ability to axially scan the imaging lens further enables us to implement the focusing stack function to effectively extend the DOF (Figure 4c). We used the pair of housefly labellum (object A) and honeybee barsitarsus (object B) specimens as test Case I. Figure 4a illustrates the scenario where the VCM is used to bring object A in focus, while leaving object B out of focus. Only the network structure of the housefly labellum can be clearly resolved (Figure 4d). The VCM can then translate the imaging lens backward to acquire clear images of object B (Figure 4e) and the recorded image is shown in Figure 4d. A series of images can be acquired covering the axial positions of objects A and B and processed using the stack function in Adobe Photoshop with defaults settings (Movie S2, Supporting Information). After focus stacking, both objects are brought into focus simultaneously without any degradation compared with single frames, thereby extending system DOF (Figure 4f ).
Similar tests were performed on test Case II (labial palp/ honeybee barsitarsus) and test Case III (two pieces of onion epidermis) and resulting recorded videos are shown in Movies S3 and S4 (Supporting Information), respectively. The final processed images are, respectively, shown in Figure 4g-l.
To demonstrate system customizability, we repeated the same tests using the miniaturized microscope assembled with E30 ( Figure S6 and Movies S5-S7, Supporting Information) and T30 ( Figure S7 and Movies S8-S10, Supporting Information). Although low apparent image contrast led to reduced focusstacking reconstruction fidelity, we successfully integrated optical components from different fabrication methods, demonstrating the benefits of the unified AM fabrication and PCR strategies reported.

Conclusion
We have established a unified AM process implementing PCR strategy and passive alignment features, which led to improved ease of assembly and overall cost savings for miniaturized imaging systems. The PCR strategy we employed consolidates the lens mount, ring magnet mount, and compliant foundation into a monolithic part, greatly reducing component manufacturing and assembly costs. The 3D-printed passive alignment features further simplify system assembly, as all components snap-fit together without requiring any precise machinery for active alignment. Collectively, this leads to the demonstration of a miniature optical microscope at a unit cost less than $4 without compromising performance or customizability. The 3D-printed lens demonstrates a spatial resolution of 3.10 µm under 532 nm illumination, exhibiting a better contrast than its commercial counterparts. The assembled microscope employs an integrated VCM to focus the imaging lens on objects at different distances. Synergizing successive image acquisition and VCM actuation results in a greatly extended DOF through focus stacking. The microscope can be readily customized and rapidly manufactured to respond to task-specific needs in imaging magnification and working distance. The capability to manufacture the system from a single resin provides an opportunity for this system to serve as an exemplar device, readily available to transform the novel materials developed by the material research community into functional devices to demonstrate their advantages. This demonstrated flexibility associated with the digital manufacturing process enables this low-cost imaging platform to be readily customized and disseminated to broader user community as an open-source project and greatly expand the application scenario of AM.

Experimental Section
Magnetic Force Simulations: To calculate the magnetic force between the coil and ring magnet, previously reported algorithms and code were adopted. [16,17] The thick coil, which had multiple radial turns N r , was modeled as superposition of a group of coaxial thin coils which carried same current and had same number of axial turns N a . The ring magnet was modeled as a superposition of two cylindrical magnets carrying opposite magnetization, with their radii corresponding to the inner and outer radius of the ring magnet, respectively. The magnetic force exerted on a cylindrical magnet with radius r 2 by a thin coil with radius r 1 was given by [17]  (1) where the intermediate term was (2) where K, E, and Π are first, second, and third elliptic integrals, respectively. J 1 = μ 0 N a I/l c where I is the current and l c is the coil length; J 2 is the magnet remanence; a 1 is the distance between the centers of mass of magnet and coil and other parameters were Equation (1) was calculated for a cylindrical magnet of r o with +z magnetization and for a cylindrical magnet of r i with −z magnetization, where r o and r i are outer and inner diameters of the ring magnet, respectively. The total magnetic force was then obtained by adding those two results together. The ring magnet used in the simulations and experiments was a N50 neodymium ring magnet (R0545, SuperMagnetMan) with a 3 mm inner diameter (ID) and a 4 mm outer diameter (OD). The coil was manufactured upon customization from WireWinders and had an ID of 4 mm, an OD of 6 mm, and a wire diameter of 0.0799 mm (40 American wire gauge (AWG)). The heights of the ring magnet and coil were 1 and 3 mm, respectively. Lens Mount ANSYS Simulations: Lens mount designs were inputted into ANSYS and given a mesh of 64 100 nodes and 35 132 linear elements, typically. The as-printed 1,6-hexanediol diacrylate (HDDA) (no-UV flood exposure) was given a Young's modulus of 171.85 MPa, yield strength of 6.7 MPa, and ultimate tensile strength (UTS) of 18.04 MPa, with Poisson's ratio of 0.3. Loadings and constraints are shown in Figure 3a, which were applied to all models. Frictionless supports were applied to the base of the design and the largest diameter faces of the top portion of the lens mount. An applied loading meant to simulate magnetic actuation was applied on the top surface (Figure 3a). The force loading was ramped up from 0 to 60 mN, which produced the desired axial displacements before the struts contacted each other. Noted Figure 4. Demonstration of active focusing and extended depth of field via focus stacking. a-c) Schematic illustrations of three distinct imaging modes: a) lens focusing on the distal object A; b) lens focusing of the closer object B; and c) focus-stacking mode by capturing a series of images while axially scanning the imaging lens. Those images can be digitally combined to render a clear image of both objects A and B with greatly extended imaging depth. Case I: schematic illustration of object A (housefly labellum) and object B (honeybee barsitarsus) and the recorded images: d) focusing on housefly labellum, e) focusing on honeybee barsitarsus, and f) focus-stacked image that brings objects A and B simultaneously in focus. Case II: schematic illustration of object A (housebee glossae) and object B (honeybee barsitarsus) and the recorded images: g) focusing on housebee glossae, h) focusing on honeybee barsitarsus, and i) focus-stacked image that brings objects A and B simultaneously in focus. Case III: schematic illustration of object A (onion epidermis I) and object B (onion epidermis II) and the recorded images: j) focusing on onion epidermis I, k) focusing on onion epidermis II, and l) focus-stacked image that brings objects A and B simultaneously in focus. Scale bars: 200 µm.
deformations were the radial deformation and the axial deformation and the ratio of radial and axial (R/A, design-relation Poisson's ratio). Shown in Figure 3b are the largest axial deformation and Poisson's ratio with respect to applied force (ANSYS) for designs with different strut angles. The maximum R/A (design-related Poisson's ratio) of the system was determined either at the maximum applied force (F = 60 mN) or when the struts came into contact with one another (F < 60 mN).
Ellipsometry Measurement: A thin film of photocurable resin was cured for use as an ellipsometry sample to obtain its wavelength-dependent refractive index. The thin film was made by casting the photocurable resin onto a piece of glass slide which was partly covered by a layer of palladium (100 nm). Then, another piece of glass slide was used to cover on the first one and peeled off after the photocurable resin was cured. The palladium layer here served as a spacer to offer creation of a thin, repeatable film of photocurable resin with a desired thickness. The wavelength-dependent refractive index was measured by using a spectroscopic ellipsometer (M2000U, J. A. Woollam Co.) using the reflection mode. The spectroscopic ellipsometry (SE) method measured the complex reflectivity comprising reflection ratio Ψ and phase difference Δ. The SE data analysis process began by constructing parametrized layered optical model which corresponded to the sample structure. Each layer of the sample was parametrized by its thickness and optical constants. The retrieval process simultaneously determined the complex refractive index and the thickness of each of the constituting layers. The original data were obtained at incident angles from 55° to 75° with an interval of 5° and were further fitted with Cauchy-film model.

Supporting Information
Supporting Information is available from the Wiley Online Library or from the author.