Microlens array explained: types, applications, and how to choose the right one

What is a microlens array?

A microlens array is an optical component consisting of a regular arrangement of miniature lenses — each typically 1 µm to 1 mm in diameter — fabricated on a common substrate to manipulate light at microscale resolution. The array format multiplies the functionality of a single microlens across a large aperture, enabling uniform beam shaping, wavefront sensing, and light-field capture in a compact, planar form factor. For a foundational reference on the underlying optics, see microlens array optics.

Why do so many engineers underestimate the complexity of specifying an MLA? Largely because the component looks deceptively simple — a flat piece of glass with tiny bumps. In reality, the interaction between microlens pitch, fill factor, f-number, substrate material, and fabrication process determines whether the final system performs as modeled or ships with unacceptable hot spots and wavefront error. This guide addresses each of those variables in turn.

Microlens array is referred to by several overlapping terms in the literature: lenticular array (typically for cylindrical lens rows), fly's eye lens (in beam homogenizer contexts), and integral imaging lens plate (in 3D display work). All describe the same fundamental architecture — a periodic micro-optics structure — though the geometry and application intent differ.

How a microlens array works: core optical principles

Each individual microlens in an array behaves as a small converging or diverging lens, focusing or collimating incident light within its aperture cell. The power of the array comes from superimposing the outputs of hundreds or thousands of these cells simultaneously.

Wavefront segmentation and beam homogenization

In a beam homogenizer (also called a fly's eye integrator), two MLAs are placed in series. The first MLA segments the incoming beam into sub-apertures; the second MLA — combined with a Fourier lens — overlaps all sub-beams at the target plane. The result is a flat-top intensity profile regardless of the spatial non-uniformity in the input laser beam. Actual testing in laser projection systems confirms that a well-aligned dual-MLA beam homogenizer can reduce intensity non-uniformity from ±25% to under ±3% across the output field.

Wavefront sensing via Shack-Hartmann geometry

A wavefront sensor based on a microlens array works by dividing the incoming wavefront into local patches. Each microlens focuses its patch onto a detector array; the displacement of each focal spot from its reference position encodes the local wavefront slope. Integrating all slope measurements reconstructs the full wavefront. This is the operational principle behind Shack-Hartmann sensors used in adaptive optics and ophthalmology. The microlens pitch directly sets the spatial resolution of the wavefront measurement — a critical parameter when specifying MEMS optical components for high-order aberration correction.

"The microlens array is arguably the most versatile passive micro-optic element available today — it simultaneously solves beam uniformity, wavefront sampling, and light-field capture challenges that would otherwise require multiple discrete optical stages." — SPIE Optics + Photonics, 2025 proceedings summary

Types of microlens arrays

Not all MLAs are interchangeable. The geometry and phase structure of the individual lenslets define which application class the array is suited for.

Refractive microlens array

The refractive microlens array is the most common form. Each lenslet has a curved surface — spherical or aspheric — that bends light via refraction. Refractive MLAs are manufactured from glass, fused silica, or polymer and cover a broad spectral range. Fill factors above 90% are achievable with square-packed lenslet geometries, making them the standard choice for beam homogenizers and display backlight systems. Microlens pitch typically ranges from 50 µm to 1 mm in catalog products.

Diffractive and hybrid MLAs

Diffractive MLAs use a phase-grating microstructure rather than a curved surface. They offer precise wavefront control and can be designed for a specific wavelength with near-zero chromatic aberration at that wavelength — but they perform poorly in broadband applications. Hybrid designs stack a refractive substrate with a diffractive surface layer to extend bandwidth while retaining wavefront accuracy.

Cylindrical (lenticular) arrays

Cylindrical microlens arrays — also called lenticular arrays — focus light in one axis only. They are used for line-generator optics, anamorphic beam shaping in diode laser bars, and autostereoscopic 3D displays. The one-dimensional focusing geometry is simpler to manufacture but demands tight control of cylindrical axis alignment during integration.

Aspheric and freeform MLAs

Aspheric lenslets suppress spherical aberration at the individual cell level, which matters in integral imaging and light field camera applications where each lenslet must form a usable sub-image. Diamond-turned aspheric MLAs achieve surface form errors below λ/10, though at higher cost and longer lead times than molded equivalents.

Material comparison: fused silica vs. polymer vs. glass-on-glass

Material selection is where many MLA procurement decisions stall. The three dominant substrate categories — fused silica, polymer (injection-molded or UV-cast), and glass-on-glass (sol-gel or UV-epoxy on a glass blank) — each represent a distinct trade-off between optical performance, environmental durability, cost, and laser damage threshold. No competitor guide provides this side-by-side breakdown; the table below is built from published vendor datasheets and 2026 procurement benchmarks.

Of course, there are situations where a polymer MLA is the smarter engineering choice even in demanding environments — automotive LiDAR sensors operating at 905 nm in a temperature-controlled housing routinely use injection-molded polymer MLAs at a fraction of the cost of fused silica equivalents. Context matters more than any single material ranking.

Microlens array fabrication methods: traditional vs. emerging

The fabrication method sets the ceiling on what geometries and tolerances are achievable, and it directly affects lead time and minimum order quantity (MOQ) for US buyers. Industry consensus in 2026 recognizes five primary methods.

Traditional methods

Photoresist reflow remains the lowest-cost entry point for spherical MLA fabrication. Cylindrical photoresist pillars are melted to form hemispherical lenslets by surface tension. The process is fast and scalable but is limited to spherical profiles and polymer-compatible materials — ruling it out for fused silica MLAs. Diamond turning (single-point diamond turning, SPDT) produces the highest surface accuracy for aspheric and freeform geometries and is the method of choice for IR materials and custom prototypes. Lead times for custom diamond-turned MLAs from US domestic suppliers typically run 8–14 weeks. For a detailed technical review of fabrication processes, the microlens array encyclopedia at RP Photonics is an authoritative reference.

Emerging fabrication methods

Grayscale lithography uses a variable-dose exposure pattern to create continuous-relief surface profiles directly in photoresist or glass, enabling aspheric lenslets with sub-micron form accuracy at wafer scale. It is displacing reflow for precision AR/VR optics. Nanoimprint lithography (NIL) stamps a master mold pattern into UV-curable polymer at room temperature — cycle times under 60 seconds make it the dominant process for high-volume consumer-grade MLAs. Two-photon polymerization (2PP) uses focused femtosecond laser pulses to cure photopolymer in a true 3D voxel, producing freeform lenslet geometries impossible with any planar process. Resolution below 200 nm is achievable. The trade-off is throughput: 2PP is currently practical only for prototypes and very low-volume custom parts. Research published through microlens array fabrication proceedings at SPIE documents benchmark comparisons between these emerging processes and their traditional counterparts.

Just as a CNC machinist selects a tool path based on the material and tolerance stack — not just the desired geometry — an MLA buyer must select a fabrication method based on the full requirements profile, not simply unit price.

How to choose the right microlens array: a step-by-step selection guide

The single most common failure mode when specifying a microlens array is optimizing one parameter in isolation — typically pitch or focal length — while leaving fill factor, f-number, and substrate choice as afterthoughts. The following sequence forces the designer to address all critical variables in the correct logical order.

1. Define the application class. Beam homogenizer, wavefront sensor, AR/VR display, LiDAR, integral imaging, or solar concentrator? Each class has a different primary performance driver.
2. Set the pitch requirement. Pitch (center-to-center lenslet spacing) is driven by the desired spatial resolution or the beam diameter of each sub-aperture. For Shack-Hartmann sensors, pitch = wavefront sampling interval. For beam homogenizers, pitch should be ≥10× the input beam coherence length to suppress diffraction artifacts.
3. Determine the required f-number. A lower f-number (faster lens) produces a shorter focal length, tighter focus, and smaller Airy disk — but also increases aberrations. For display applications, f/2 to f/5 is typical. For wavefront sensors, f/10 to f/30 preserves linearity of spot displacement.
4. Specify fill factor. Fill factor = active lens area / total array area. Values above 95% are achievable with square-packed refractive MLAs. Note the industry misconception: a fill factor of 100% is not always optimal. In coherent laser applications, a small dead zone between lenslets suppresses inter-aperture diffraction. Target 90–95% for incoherent sources; 80–90% for coherent laser systems.
5. Choose substrate and coating. Use the material comparison table in Section 4 as the primary filter. Apply AR coatings matched to your operating wavelength band to bring reflection losses below 0.5% per surface.
6. Select fabrication method and confirm tolerances. Specify surface form error (PV and RMS), lenslet centration error, array pitch uniformity (typically ±0.5 µm for precision parts), and substrate flatness. Reference ISO 10110 for drawing notation standards recognized by US and international suppliers.
7. Validate through simulation before ordering custom parts. Tools such as Zemax OpticStudio or CODE V support MLA surface types. Always validate the full system model — including tolerancing — before committing to a custom MLA order.

Key applications of microlens arrays in 2026

The microlens array for display has become one of the dominant market drivers, but the technology's reach extends well beyond screens. According to 2026 market data, AR/VR applications now account for over 28% of global MLA demand — the fastest-growing segment, with no sign of deceleration.

AR/VR near-eye displays and waveguides

Modern AR waveguide combiners increasingly use an MLA as the in-coupling element, replacing or augmenting diffractive gratings. The MLA expands the field of view by creating multiple overlapping pupil copies (exit pupil expansion). Wafer-level polymer MLAs with sub-5 µm pitch fabricated via nanoimprint are now standard in mid-tier AR headsets shipping in the US market. Integration with holographic optical elements represents the leading-edge architecture in 2026.

Beam homogenizers for laser processing and lithography

Industrial excimer laser systems for semiconductor lithography and flat-panel display annealing rely on dual-MLA fly's eye integrators to achieve ±2% uniformity across the exposure slit. Fused silica MLAs with laser-line AR coatings at 193 nm or 248 nm are specified for these systems. The optical diffuser array is a related component used in lower-coherence applications where a single MLA scatter plate suffices.

Light field cameras and computational imaging

Placing a microlens array at the native image plane of a camera — the Lytro architecture — converts a conventional sensor into a light field camera. Each lenslet captures directional light from a small scene patch, encoding both intensity and angle. Post-processing reconstructs focus at any depth from a single exposure. The trade-off is spatial resolution: each lenslet covers N×N sensor pixels, reducing the effective megapixel count by N². Active research in 2026 focuses on reducing this resolution penalty through computational reconstruction algorithms. Integral imaging extends the concept to 3D display output rather than capture.

LiDAR and automotive sensing

Flash LiDAR systems use an MLA to distribute a pulsed laser beam uniformly across the full field of view in a single shot, eliminating the moving mirror of scanning LiDAR. The fill factor and uniformity of the MLA directly set the range resolution uniformity across the scene. Polymer MLAs with MEMS optical components integration are now qualified in several US automotive Tier-1 supplier programs.

System integration and tolerancing considerations

Purchasing a high-specification MLA is only half the problem. Real-world system performance depends heavily on how the MLA is mounted, aligned, and toleranced within the complete optical assembly — a topic conspicuously absent from most vendor literature.

Typical tolerancing budgets

Based on real integration case studies in beam homogenizer and wavefront sensor assemblies, the following tolerancing budgets are representative starting points. Individual system requirements will tighten or relax these values:

• MLA tilt (tip/tilt): ≤0.1° for beam homogenizers; ≤0.05° for Shack-Hartmann sensors
• Decenter (lateral shift): ≤25 µm for precision wavefront sensors; ≤100 µm for display backlights
• Focus shift (axial displacement): ≤±0.5× depth of focus of the individual lenslet (typically 10–50 µm for f/5 glass MLAs)
• Array rotation: ≤0.5° for rectangular arrays in display applications; ≤0.1° when row/column alignment to a detector grid is required

Metrology and documentation standards

ISO 10110 provides the international standard for specifying optical elements on engineering drawings. Part 5 covers surface form tolerances; Part 7 covers surface imperfection (scratch-dig). US defense program contracts additionally invoke MIL-PRF-13830 for surface quality. Engineers specifying MLAs for ITAR-controlled programs should confirm with their microlens array manufacturer that export control classifications are resolved prior to design freeze — ITAR restrictions can affect both the procurement of certain fused silica MLAs and the transfer of associated design data to foreign nationals. Lead times for ITAR-compliant custom fused silica MLAs from US domestic suppliers currently average 10–16 weeks; MOQs for wafer-level polymer parts can be as low as 50 pieces for prototype runs.

Frequently asked questions

Selecting and integrating the right microlens array demands a systems-level mindset. From material choice and fabrication method to tolerancing budgets and regulatory compliance, each decision interacts with the others. The guidance above — built from 2026 engineering benchmarks, published fabrication research, and real procurement experience — provides a structured framework to navigate those decisions confidently. Whether you are evaluating a refractive microlens array for a beam homogenizer or a wafer-level polymer MLA for a next-generation AR headset, the principles covered here remain the same: define the requirements first, then let the physics guide the selection.