Research Update: Enabling ultra-thin lightweight structures: Microsandwich structures with microlattice cores

The development of small-scale insect-like robots, micro-air vehicles, and other small micro-scale machines requires ultra-lightweight, high stiffness materials to form structural frameworks, loading surfaces, and airfoils.^1,2 These small-scale engineered structures generally rely on solid composite materials, such as carbon fiber polymer matrix composites, because of their high specific stiffness.³ In contrast, analogous natural systems, such as insects and plants, have structures composed of cellular materials that provide multiple functions such as fluid or nutrient flow and optical properties, in addition to a primary structural function.^4,5 In many cases, these cellular materials are formed as sandwich structures, providing higher bending rigidity compared to architectures with a more uniform distribution of mass.⁴ Likewise, in large-scale engineering applications where high bending stiffness and low mass are of prime concern, sandwich structures composed of a lightweight cellular core bonded between two rigid facesheets are often utilized.⁶ Although the principles of sandwich theory with cellular cores apply at smaller scales, and the utility has been demonstrated in natural systems, there have been few attempts to create micro- or meso-scale sandwich structures due to fabrication complexity and cost.^7,8 Here, we introduce a new class of lightweight sandwich panels that are fabricated by forming microlattice cores in situ between two thin polymer facesheets. By simultaneously bonding all three layers together while forming the core, we can achieve sandwich structures with high specific flexural rigidity in a new areal density regime below 0.05 g cm⁻².

The bending stiffness, or flexural rigidity, of a beam constructed of a single monolithic material is the product of the Young’s modulus and the area moment of inertia (EI). Sandwich structures provide a significant increase in bending stiffness with a nominal increase in areal density by increasing the beam’s moment of inertia through the incorporation of lightweight cellular core materials, such as honeycomb, foam, or balsa wood, effectively redistributing the mass to the outer surfaces. Traditional, larger-scale sandwich structures are fabricated by bonding a cellular core between rigid, monolithic facesheets using an adhesive. This conventional construction technique is difficult to apply at smaller length scales because of fabrication and handling challenges associated with thin cellular materials (∼μm–mm thickness), as well as the complexity of securely bonding the core material to facesheets without adding excess mass. Techniques have been developed to create microsandwich structures (<2 mm thick) using discontinuous stainless steel fibers bonded or brazed between stainless steel facesheets; however, the minimum areal density achieved using this technique is still >0.15 g cm⁻².⁸ For applications demanding lower areal densities, few options exist aside from using very thin solid constituent materials—an approach typically accompanied by a sacrifice in performance or increased manufacturing cost.

To overcome these limitations inherent to conventional microsandwich structure production, we have developed a simpler new process for fabricating sandwich panels with ultra-low areal density. As shown in Figure 1, the process involves fabricating a microlattice cellular core between two polyethylene terephthalate (PET) facesheets, while simultaneously bonding both facesheets to the core. The microlattice core is formed from a three-dimensional interconnected pattern of self-propagating photopolymer waveguides.^9,10 Formation of the microlattice core involves exposing a volume of liquid photo-sensitive monomers to multiple beams of collimated ultraviolet (UV) light through a photomask patterned with an array of apertures. Ultraviolet radiation induces the formation of free radicals in the liquid precursor, initiating the polymerization reaction only in highly localized regions since its interaction with the monomer is confined to the small area directly below each aperture. Polymerization results in a change in refractive index relative to the uncured monomer, creating an optical interface between solid and liquid phases. This interface allows the incident light to couple into the polymer, resulting in the formation of self-propagating waveguides. By simultaneously exposing the photomask to multiple angled collimated light sources, an interconnected, ordered array of waveguides is formed. Here, we show that a microlattice structure can be formed in situ between two transparent facesheets, while simultaneously bonding the facesheets to the newly created lattice core. These microsandwich panels are formed in a single exposure step. One biaxially oriented PET facesheet is placed in contact with a patterned quartz mask (Photo Sciences, Inc.). The thickness of the facesheet can vary from 12.7 μm up to 300 μm. A second PET facesheet is temporarily adhered to a rigid substrate, e.g., steel plate via vacuum grease. Reconfigurable shims are coated with grease and placed on top of the second facesheet, arranged and stacked to produce a reservoir with depth equal to the desired core thickness. Liquid photomonomer is then poured until the cavity is completely filled, and the photomask with the attached first facesheet is laid overtop, taking care to avoid trapping or entraining air bubbles. The quartz mask is exposed to multiple beams of collimated UV light (∼9 mW cm⁻² intensity at mask surface) generated from a 2000 W mercury arc lamp (Bachur & Associates). The exposure times vary between 10 and 30 s, depending on the thickness, aperture size, and geometry of the microstructure formed. Adhesion of truss core to the upper and lower facesheets occurs inherently and simultaneously during the polymerization process. After exposure, uncured monomer is rinsed away in an agitated toluene bath at 60 °C. Finally, the microsandwich structure undergoes a thermal post-cure for 24 h at 85 °C, followed by a second, direct exposure under the UV lamp for 5 min to remove any remaining solvent and achieve maximum crosslinking in the final polymer truss. The result is a micro-scale sandwich panel, with a fully adhered, rigid, open-cell polymer microlattice core.

This single step approach eliminates the need for an additional adhesive layer to bond the facesheets to the core, reducing the overall mass, and enabling the scalable manufacture of sandwich panels with unprecedented ultra-low areal density. For example, a sandwich structure with areal density down to 0.005 g cm⁻² has been successfully demonstrated (Figure 1(d)).

Figure 2 presents scanning electron micrographs (SEMs) of two representative microsandwich structure architectures fabricated for mechanical testing. Multiple samples of each architecture were fabricated for three-point bend tests to measure flexural rigidity. The first microsandwich architecture shown (Figure 2(a)) is composed of two 75 μm biaxially oriented PET facesheets, separated by a pyramidal lattice core. The microlattice for this structure was generated using a photomask with 225 μm diameter apertures spaced approximately 2 mm apart in a square pattern.¹⁰ The inclination angles of the lattice members in the configurations tested are approximately 58° and 58°, with an average core height of 1.6 mm. The areal density of these samples ranges from 0.042 g cm⁻² to 0.044 g cm⁻². The second representative microsandwich architecture (Figure 2(c)) utilizes 50 μm thick biaxially oriented PET facesheets with an octahedral-type unit cell. The octahedral-type unit cell is a series of layered pyramidal unit cells, stacked without basal plane members.¹⁰ For these samples, the microlattice core is formed using a mask with 75 μm diameter apertures in a square pattern with a 1 mm pitch. The inclination angle of these lattice members is approximately 50° and the average core thickness of 2.0 mm. This microsandwich architecture was designed to increase panel rigidity and simultaneously reduce areal density compared to previous pyramidal architectures, with parameters chosen based on preliminary optimization results (described below). The areal density of panels with this octahedral architecture ranged from 0.030 g cm⁻² to 0.037 g cm⁻². This variation between samples can be attributed mainly to excess polymer build-up around the nodes of the truss core and non-uniformities in the fine feature geometry of as-fabricated microlattice cores.

Three-point bend tests are commonly used to assess the flexural rigidity of materials and composites. For sandwich structures loaded in a three-point bending configuration, the relationship between the applied force (F) and resulting displacement (δ) is given by the following equation:^11,12

where (EI)_eq is the equivalent flexural rigidity of the sandwich structure and (AG)_eq is the equivalent shear rigidity. The flexural rigidity and shear rigidity of a sandwich structure can be estimated by the following equations:

where E_face and E_core are the compressive moduli of the facesheet and core, respectively, G_core is the shear modulus of the microlattice core, b is the sample width, c is the core thickness, t represents the facesheet thickness, and d is the separation distance between midplanes of the facesheets.

The microsandwich structure samples represented in Figure 2 were loaded in a three-point bend test fixture to quantify the flexural rigidity (Table I). Three-point bending experiments were conducted on a TA Instruments Q800 DMA. The ∼1 cm × 6 cm microsandwiches were placed on a low-friction three point clamp and loaded at constant displacement rate of 500 μm min⁻¹ until failure. Equation (1) was used to calculate (EI)_eq for each sample using the slope of the linear force-displacement data prior to the onset plastic deformation or failure. When determining (EI)_eq, from load-deflection measurements, the shear deflection term was dropped based on estimates from Eqs. (2) and (3) indicating a negligible contribution to overall deflection from the intrinsically rigid, truss-type architecture of the open cell core.^10,13

To make valid comparisons between different samples, materials, and geometries, EI_eq was normalized by sample width (b) and plotted as a function of areal density in Figure 3. Both the pyramidal and octahedral unit cell structures exhibited significantly higher flexural rigidity than the calculated rigidity of a unidirectional carbon fiber composite with the same areal density.¹⁴ 

An intrinsic advantage of microsandwich panels fabricated using the present technique is the ability to design the structure through an optimization method. In this study, multiobjective optimization using the compromise method was performed to determine the Pareto-optimal surface that defined the microsandwich structure parameters to achieve maximum flexural rigidity, maximum failure load, and minimum areal density.^15–17 The design parameters include lattice strut diameter, node-to-node spacing, lattice member inclination angle, facesheet thickness, and core thickness. For an application with a set minimum failure load requirement, with an added constraint on the maximum allowed areal density, this optimization routine uses Eq. (2) to determine the particular microsandwich architecture with the highest achievable flexural rigidity for a panel that still satisfies both of the specified constraints.¹¹ Designed for this study, the routine was simplified to consider only the most probably failure modes observed specifically: failure by facesheet yielding, facesheet indentation, facesheet wrinkling, or failure by lattice member buckling—the latter being the dominant mechanism in the structures presented here. The adhesion strength between the microlattice core and PET facesheets is difficult to quantify in the as-fabricated microsandwich structures, so this failure mode was not included in the optimization routine. SEM and optical microanalysis of all samples pre- and post-test suggest that core lattice member buckling leading to localized facesheet delamination is the dominant failure mechanism for the sandwich panel configurations presented here. Electron micrographs in Figures 2(b) and 2(d) show two samples that exhibit particularly complex damage patterns, showing signs of the two main mechanisms identified above, as well as facesheet indentation in Fig. 2(b), and facesheet wrinkling with some evidence of possible facesheet yielding in Fig. 2(d). All samples displayed significant core strut buckling, often accompanied by some degree of facesheet delamination, with coupons 7 and 8 presenting no obvious indication of any other significant failure mode. Because lattice member buckling was the leading cause of failure observed in the sandwich configurations reported here, this model is adequate for extracting useful trends.

The 3-D Pareto-optimal surface is depicted as a series of 2-D curves in Figure 3, one for each set failure load value. The calculated optimization curves show the predicted flexural rigidity for different sandwich structure configurations over a range of areal densities. Each point on the curve represents a polymer sandwich structure with a unique combination of geometric parameters. All curves show decay in flexural rigidity with decreasing areal density; however, curves with higher specified minimum strength have lower flexural rigidity at constant areal density. From examination of the geometric parameters that yield the Pareto-optimal surface, we observe that without constraint on failure strength, optimal designs maximize the ratio of lattice member length to lattice member diameter, i.e., maximize the slenderness of the lattice members. Furthermore, optimal designs include more than one lattice unit cell through the thickness of the core. Both features increase the core thickness, thus maximizing the flexural rigidity of the sandwich panel. As the constraint on minimum load before failure increases, the optimal geometry changes in multiple ways: the slenderness of the lattice members decreases, as would be expected to increase the load carry capacity of a structure with failure dominated by lattice member buckling. Additionally, other geometric features change to keep the flexural rigidity high with less slender lattice members. In particular, the number of lattice unit cells through the thickness increases, the diameter of the lattice members increases, and the facesheet thickness increases.

Based on the optimization results and the initial pyramidal lattice core data, we designed octahedral lattice core coupons to have higher slenderness and more unit cells through the core thickness. As can be seen in Figure 3, this redesign yielded microsandwich panels with performance closer and quite near to the Pareto-optimal surface, i.e., higher flexural rigidity per unit width at lower areal densities.

Variation between coupons is attributed to inconsistencies in the fabrication process resulting in variations in the lattice member diameter, taper, and angles. With improvements in processing, we expect to better and more uniformly replicate the optimal geometric features in as-fabricated micro-sandwich structure samples and thus more consistently approach the Pareto-optimal flexural rigidity for a given areal density and failure load.

The simple, rapid, and scalable fabrication method described here for creating micro-scale sandwich panels enables near-optimal microsandwich designs in an ultra-low areal density regime. By taking advantage of traditional sandwich construction at a small length scale, these materials exhibit a flexural rigidity far exceeding monolithic materials and conventional composite materials with an equivalent areal density. These microsandwich structures could replace conventional materials in thin-walled structures, thereby allowing increases in system efficiency, payload capacity, and speed or range without compromising performance, function, or appearance. In addition, this integrated fabrication process may open up new capabilities for design and optimization of small-scale structures where stiffness and low weight are the concerns. If combined with hollow lattice member cores,^18,19 even lighter-weight or multifunctional^20–22 microsandwich panels would be possible.