Over the past decade, the rapid evolution of electronics has led to a significant decrease in the dimensions of integrated circuits. As denoted by Moore's law, the size of the transistors has consistently decreased while the number of transistors on a chip continued to grow exponentially. On average, the number of transistors in microprocessor units has doubled in approximately two years. 1 This repeated miniaturization of computer components has sufficiently enhanced technological performance. Moreover, the downsizing of transistors has enabled computers to achieve greater storage capacity and speed because of the increased energy efficiency and faster operation of smaller transistors. 2 However, the strategy of semiconductor miniaturization to boost chip performance is now approaching its limits slowly, presenting substantial challenges in the precise control and manufacture of microelectronic devices with accurately regulated electrical properties. 3 One of the key barriers to further increases in the speed of microelectronic integrated circuits is the resistive-capacitive (RC) delay. As the feature sizes of the microchips decrease to increase CPU clock speeds, the gate delay is reduced but the distributed RC delay increases significantly. 4 The resistance grows more rapidly compared to the technology scaling factor and the measured capacitance of the device. This escalating RC delay increasingly overshadows the benefits of the new microelectronic device architectures, leading to a deterioration in circuit performance. 5 To minimize these delays, several strategies can be employed, such as replacing the aluminum conducting wires with copper to reduce resistance, altering the geometry of the wires, or using low dielectric constant (low-k) materials for the interlayer dielectric to decrease the capacitance of the chip. 6 These approaches help manage the growing impact of RC delays and sustain performance improvements in modern microelectronic circuits.
One promising method to reduce RC delay is the use of materials with low dielectric constant. 7 Recent research in low-k dielectrics has focused on polymeric materials compared to inorganic materials like glass and ceramics because of their inherently lower ARTICLE pubs.aip.org/avs/jvb electric permittivity, lower cost, lighter weight, and ease of processing. [8][9][10] There are two primary approaches to developing polymer materials with a low dielectric constant. The first method involves reducing the k value by decreasing the strength of dipoles or reducing the number of dipoles present in the material. [11][12][13] For example, the introduction of a nonpolar bulky adamantly group into the siloxane backbone has successfully produced a low-k dielectric polymer with constants ranging from 2.50 to 2.53. 14 Another example is a new structural polymer consisting of perfluorocyclobutane group and aryl ether group called fluoropolymer which has demonstrated a dielectric constant as low as 2.33, surpassing the performance of the adamant group. 15 While these materials exhibit low-k values, they also have relatively low stiffness and operating temperature.
Another approach is to introduce porosity into the material, where the presence of air reduces the effective permittivity of the material. 15 Given that the dielectric constant of air is 1, the inclusion of air-filled pores can significantly lower the overall effective dielectric constant of the material while simultaneously decreasing its density. 16 Recent work has demonstrated that nanoarchitected materials or nanolattices can have a low dielectric constant of 1.06 while maintaining high stiffness. 17 Furthermore, the refractive index of such nanolattice materials at optical wavelength can be designed to range between n = 1.025 and 1.360 by precisely controlling the lattice geometry. 18 Such material can also have improved mechanical stiffness at low density owing to the ordered architecture. [19][20][21] As a result, inorganic nanostructure materials with air pores offer a potential solution, as they can be engineered to have a low dielectric constant while maintaining mechanical robustness.
An area that can benefit from nanostructure materials is nanophotonics such as waveguides and other integrated devices with multilayer architectures. 17,[22][23][24][25][26][27][28] It has been demonstrated that a nanolattice structure can significantly enhance light trapping by utilizing a low-index nanolattice material. 22 Furthermore, the lowindex nanolattice can be integrated into a multilayer photonic crystal reflector to tune the reflectance band. This integration increases the index contrast, which in turn enhances the reflectivity of the structure. 23 These findings suggest that nanolattice structures can play a crucial role in advanced photonic applications such as signal processing, waveguides, and radiative cooling, [24][25][26][27] where precise control of optical behavior is essential. While porous nanolattice materials are attractive for various applications, their integration into devices still poses challenges due to their porosity. This work demonstrates the successful fabrication of microelectrodes on porous nanolattice structures, paving the way for advanced electronic and photonic applications.
In this work, we present the fabrication of microscale structures onto porous nanolattice materials using shadow evaporation through a stencil mask. In this approach, the nanolattices are fabricated using 3D colloidal phase lithography and atomic layer deposition (ALD), which are then protected with a polymer planarization layer. The microstructures are then deposited through a stencil mask by physical vapor deposition and then released after removing the planarization layer using a thermal cycle. The key advantage of the stencil mask is that it is a noncontact method, making the process fast and versatile to deposit a wide range of materials such as metals onto the nanolattice structure. Experiments demonstrate that microstructures with 400 μm length and 100 μm width features can be successfully fabricated. The results indicate that residual stress in the film is an important parameter to consider, which leads to cracking of the microstructures for thicker films. This work facilitates the integration of microstructures on nanolattice materials without affecting the porosity of the underlying structures and can find applications in microstructured waveguides and electrodes with integrated nanostructures.
The goal of this work is to develop the fabrication technique to integrate a microstructure device on nanolattices, as illustrated in Fig. 1. The test microstructures have a "dog-bone" geometry that emulates contact electrode pads with dimensions in the 100 mm range. Initially, the silicon substrates are spin coated with a 100 nm thick antireflection coating (ARC, Brewer Science, ARC i-con-11) and 300 or 800 nm of positive photoresist (Sumitomo, PFI-88A2). The ARC is important because it reduces the back-reflections during the lithography and reduces line-edge roughness in the patterned photoresist. Next, polystyrene nanospheres with diameters of 390 or 500 nm are assembled on the photoresist to form a monolayer of hexagonally closed packed structures using Langmuir-Blodgett assembly. The monolayer acts as a near-field phase mask and is illuminated by a HeCd laser (Kimmon Koha, IK3501R-G) with 325 nm wavelength and a dose of 90 mJ/cm 2 to generate 3D intensity patterns in the photoresist through diffraction and interference effects as depicted in Fig. 1(a). 28,29 Next, the nanospheres are removed with ultrasonication, and the patterned resist is developed with CD-26 base solution to create a resist template as shown in Fig. 1(b). After development, thin Al 2 O 3 films are deposited onto the defined 3D resist patterns using ALD. This process utilizes trimethyl aluminum and de-ionized water as precursors, depositing approximately 1.1 Å per cycle. The film was deposited with 200 cycles to achieve a conformal coating with 20 nm thick shells, as shown in Fig. 1(c). This results in a porous, thin-shelled nanolattice with an ultralow refractive index.
To define the microstructure, a stainless-steel stencil mask with microscale features is fabricated using waterjet (Stencils Unlimited). Stencil lithography using shadow evaporation masks has been used as an attractive additive method to fabricate micro/ nanostructures 30,31 but has not been examined for patterning over porous films. To protect the underlying structures, a thick photoresist layer is spin coated over the nanolattice layer, as shown in Fig. 1(d). Then, the stencil mask was securely placed in direct contact with the underlying substrate using Kapton tape. Various thicknesses of TiO 2 from 30 to 560 nm are deposited through the stencil mask using electron beam evaporation, as shown in Fig. 1(e). The deposition conditions for e-beam evaporation include a system base pressure of 5:0 Â 10 À6 Torr with no substrate cooling and a deposition rate of 1.0 A/s. TiO 2 was chosen for this experiment due to its high refractive index, making it suitable for optical waveguides and future photonic structures. However, any material compatible with the e-beam process could be used.
The small dog-bone has dimensions of L = 400 μm and W = 100 μm and the big dog-bone has dimensions of L = 7 mm ARTICLE pubs.aip.org/avs/jvb and W = 2 mm. Then, the photoresist template is removed from the nanolattice through thermal desorption, which removes all remaining polymer on the sample. In this process, the furnace heats the samples up at a ramp rate of 1 °C per minute to 550 °C, which is maintained for 4 h before cooling back to room temperature at 5 °C per minute. This process results in the "dog-bone" microstructure on top of the nanolattices, as shown in Fig. 1(f ). A blank silicon wafer is placed next to the sample to measure the actual thickness of the TiO 2 layer using spectroscopic ellipsometry. The TiO 2 microstructure film is not expected to diffuse or migrate into the underlying Al 2 O 3 nanolattices since the latter was protected by a planarization layer during the PVD process. Similar processes have been demonstrated in the stacking of nanolattices, where EDS measurement indicates the chemical composition is clean at the interface. 32
Initial fabrication results are performed on silicon substrates without nanolattices to examine the structure patterned by the stencil mask. In this experiment, a thick layer of the 560 nm TiO 2 film is deposited through the mask on a blank silicon wafer. The TiO 2 dog-bone microstructure was successfully fabricated on the blank silicon substrate, as shown in the top-down scanning electron microscope (SEM) images shown in Fig. 2. Figures 2(a) and 2(b) depict the fabricated TiO 2 structure using a small dimension of stainless-steel mask with 400 μm length and 100 μm width. In addition, Figs. 2(c) and 2(d) illustrate the TiO 2 structure using a bigger dimension of stainless-steel mask with 7 mm length and 2 mm width. The geometry of the mask and the structure are visibly different. The patterned microstructure has less defined edges due to shadowing effects and nonuniform thickness from the e-beam evaporation process. Figure 2 shows the TiO 2 film deposited on a plain silicon wafer, without incorporating any nanolattice structure or undergoing a furnace process. However, both samples show visible crack formation in the films which strongly suggests that the e-beam evaporation process is the primary cause of the cracking.
The larger microstructure has film cracks that stretch farther than the smaller structures, as shown in Figs. 2(c) and 2(d). One of the primary factors contributing to the cracking of the film is the stress state of TiO 2 when it is deposited using the electron beam
pubs.aip.org/avs/jvb evaporation process. During this process, tensile stress typically develops due to the formation of voids within the grain structure that is characteristic of electron beam evaporation. This tensile stress progressively builds up as the TiO 2 layer increases in thickness, eventually reaching a point where it causes the film to delaminate from the underlying substrate. 33 This growing stress forces the TiO 2 dog-bone microstructure to contract which ultimately results in the formation of cracks. Therefore, the nanolattices used in this work are fabricated using 390 nm polystyrene nanospheres, as illustrated in Fig. 3. The exposure dose of 90 mJ/cm 2 is used to fabricate interconnected nanolattices for the 800 nm photoresist layer. A thinner resist layer of 300 nm is used to fabricate nanostructures that are not connected and resemble isolated nanopillars. The mechanical robustness of the nanostructure layer is important and the two different architectures can be used to examine their effects on the collapse of the microstructures. Figures 3(a) and 3(b) depict the structure with an 800 nm photoresist layer forming an ordered, interconnected hexagonal pattern with a consistent spacing between adjacent holes. These results further confirm the nanolattice heights are uniform, resulting in a stable and well-ordered assembly. Some of the samples with 300 nm photoresist thickness are overexposed slightly on purpose to create isolated features that are more isolated compared with the 800 nm resist. This overexpose leads to the absence of the hexagonal top layer and instead yields tilted "nanopillars" as shown in Fig. 3(c).
The structures then resemble isolated pillars that are not connected to one another. A small portion of the sample still retains a thin hexagonal top layer in the lower right of the figure, which remains periodic. However, most of the structures are isolated and tips over during the thermal desorption process. Microstructure fabricated on both interconnected nanolattice and isolated nanopillars, which will be discussed in Sec. III B.
The initial demonstration of the dog-bone microstructure on the nanolattice structure is shown in Fig. 4. Here, the TiO 2 microstructure has a thickness of 560 nm and has been fabricated on the Al 2 O 3 nanolattices. The optical microscope view in Fig. 4(a) shows the microstructure on top of the film with a yellowish hue, which indicates the uniform height of the nanolattices. Looking closely, the dog-bone geometry is more rounded and not as distinct when compared to the stencil mask. This effect can be attributed to the shadow evaporation effects where the material deposits underneath the mask throughout the duration of electron beam evaporation. The larger feature sizes and gradually increasing thickness near the edges can be attributed to the halo effect, a phenomenon commonly associated with e-beam evaporation. This effect arises from the gap between the substrate and the stencil mask, stencil aperture size, thickness of the material, deposition rate, and substrate temperature leading to nonuniform deposition and edge thickening. 34,35 An important aspect to consider is that the gap between the stencil mask and the substrate was not precisely controlled during the experiment. While the gap is estimated to be approximately 4 μm with a standard deviation of 2 μm, variations likely occurred across different samples due to the inherent limitations of the experimental setup. These gap variations can influence the deposition characteristics and lead to the observed variations in film diffusion and uniformity. In future work, the gap distance will be better controlled using precision positioners and their effects on the uniformity of the deposited film will be examined.
As a result, the microstructure has a nonuniform thickness at the edges. Furthermore, the thicker the TiO 2 layer, the higher the chance of having a tapering thickness toward the edge of the microstructure.
The integration of the dog-bone microstructures on top of the nanolattices can be better observed in the top-view SEM images shown in Fig. 4(b). Figure 4(c) shows the nanolattice holes on the silicon substrate after the thermal cycle ran to remove the photoresist. Here, it can be observed that the nanolattice holes feature a uniform pattern throughout the surface. Then, the TiO 2 microstructure is placed on top of the nanolattice, as shown in Fig. 4(d), where the underlying nanostructures are visible under the cracked TiO 2 film. While the microstructure has been fabricated on the nanolattices, significant film cracks that appeared distinctively on the microstructure can be observed and need further investigation. TiO 2 dog-bone microstructures with different thicknesses were fabricated on the nanolattices to examine the size effects on crack formation, as shown in Fig. 5. Here, the selected TiO 2 thicknesses are 560, 190, and 146 nm. The results indicate that while the geometry of the microstructure is well defined, film cracks exist for all the samples. However, it can be observed that the film crack size is smaller for the 190 and 146 nm thickness compared to the 560 nm sample. This can be attributed to the internal stress of the TiO 2 film during e-beam evaporation, which leads to more crack formation for thicker films and smaller grain sizes. Some collapses of the nanolattices in regions without TiO 2 microstructure can also be observed, which highlights the instability of the underlying nanostructures for these samples.
The crack formation mechanism in the dog-bone microstructures can be examined further using higher-magnification SEM images, as depicted in Fig. 6. Figure 6(a) depicts the SEM image at a crack boundary of 146 nm thick TiO 2 , where several thin films can be seen to be suspended on top of the underlying nanostructures. Here, the nanopillars can be observed through the TiO 2 layer to be periodic and consist of isolated pillars. Note there is a TiO 2 flake on the upper left of the image that has been turned upside down, showing the underlying pillar structures. The SEM image in Fig. 6(b) depicts the region within an uncracked TiO 2 film, further highlighting the ordered arrangement of the nanopillars, which are distinctly visible through the thin TiO 2 film. Additionally, Fig. 6(c) provides a close-up view of the crack boundary, which shows nanopillars protruding away from the film. Films deposited by electron beam typically have tensile stress, 35 which results in film contraction during the thermal process to remove the polymer layers. Since the nanopillars are not interconnected, the in-plane strain of the TiO 2 film bends the nanopillars and leads to crack formation. This result indicates that while it is feasible to pattern a microstructure on a porous structure, the stability of the underlying nanostructure is critical.
The grain sizes of the TiO 2 film cracks and their uniformity are measured and plotted in Fig. 7. Here, the relative position is measured with respect to the center of the dog-bone microstructure, which is denoted as 0 μm. Next, the dog-bone picture was cropped every 30 μm in length, and then, all the grain sizes in the cropped picture were measured using IMAGEJ software. A total of ten measurements are taken per image and then averaged and plotted. The bigger grain size indicates less film cracks and lower grain size length indicates more film cracks. The error bar in the plot is defined by the standard deviation for each sample. For example, the biggest error is at position 60 mm for the 146 nm thick dog-bone structure, meaning that it consists of mostly big grains and few small grains causing the standard deviation of all the measurements to be greater.
The grain sizes are plotted against the microstructure thickness at a relative position of 0 nm in Fig. 7(a), allowing for a comparative analysis of the grain size distribution across varying microstructure thicknesses on the nanopillars. The trend observed in the graph illustrates that there is an inverse relationship between grain size and the microstructure thickness. As the microstructure thickness decreases, the corresponding grain size also increases, indicating less crack formation. This trend is particularly evident at the thinnest microstructure of 146 nm examined, which exhibits the largest grain size among the different thicknesses studied. In contrast, the 560 nm thick microstructure shows the smallest grain size, indicating that the cracks are forming with higher density. This trend underscores the critical role of microstructure thickness in determining the grain size within nanopillars, with thinner structures leading to bigger grains meaning less cracks.
Furthermore, the overall trend observed in Fig. 7(b) indicates that the grain size gradually increases as the position moves from the middle narrow part of the dog-bone to the start of the larger square end. This suggests that fewer film cracks occur in this region. However, after reaching the peak at the start of the larger square end, the grain size immediately decreases as it approaches the end of the dog-bone. This trend can be interpreted to mean that the grain size distribution is influenced by the structural changes along the length of the dog-bone and the distribution of the TiO 2 coating through the stainless-steel mask. This effect is more visible especially along the edges of the dog-bone geometry due to the thickness variation from the shadow evaporation over time as denoted previously in Fig. 4(a). In addition, the decrease in the grain size could be attributed to the geometric constraints and increased internal stress at sharp corners. In conclusion, the trend depicted in Fig. 7 suggests that the overall grain length size decreases as the thickness of the film decreases. These findings are significant as they highlight the relationship between the film thickness and grain size distribution, which is crucial for understanding the mechanical properties of microstructure film cracks on nanopillars.
Two modifications are made to mitigate the crack formation. First, nanolattices with interconnected architecture are used to improve the mechanical robustness of the porous film. Furthermore, the thickness of the TiO 2 deposition has also been decreased to reduce the effect of the residual stress. The fabricated TiO 2 dog-bone microstructure with 65 nm thickness has minimal film cracking, as shown in Fig. Here, the top-view optical image is illustrated in Fig. 8(a), which shows a distinct color difference between the TiO 2 structure and the surrounding nanolattices. The iridescent effect due to thin-film interference and nonuniform thickness can be observed. The structure is also observed in the SEM image as shown in Fig. 8(b), where the microstructure appears darker with a lower secondary electron signal. Additionally, due to the relatively thin TiO 2 layer, the dog-bone The fabricated sample with a 32 nm thick TiO 2 microstructure also has less crack formation, as shown in Fig. 9. Here, it can be observed that the dog-bone has less uniform thickness at the edges compared to the 65 nm thick sample, as illustrated in Fig. 9(a). This discrepancy is primarily attributed to the uneven durations of electron beam evaporation throughout the process, which can lead to inconsistencies in the deposition rate over time. Consequently, the variations in the thickness become particularly apparent at the edges of the 32 nm microstructure, where the reduced deposition results in lower uniformity. This color difference is more distinct in the 65 nm microstructures where the TiO 2 layer is thicker, compared to the 32 nm microstructure where the thinner coating results in less contrast between the dog-bone and the outer nanolattices.
For better visualization, the cross section of a cracked nanolattice section is examined to identify the TiO 2 dog-bone microstructures, as shown in Fig. 10. Here, a crumpled TiO 2 film can be observed along the curved nanolattices, as indicated by the arrow in Fig. 10(a). Note the thin layer is darker and appears like a membrane over the holes of the nanolattices. A higher-magnification image shown in Fig. 10(b) provides further evidence that the 32 nm TiO 2 film has been deposited uniformly over the porous structures. Another SEM image of the cracked section shows the film from a different angle is shown in Fig. 10(c), where the TiO 2 membrane can be observed as well. Furthermore, Fig. 10 microstructure without any cracks, ensuring the integrity and uniformity of the overall structure.
This work has successfully demonstrated the integration of microstructures on porous nanolattices. The key factor contributing to the successful coating of 32 and 65 nm thick microstructures is the enhanced stability of the underlying nanolattices. This stability led to less film cracks in the dog-bone microstructure that were observed in the nanopillars. The improvement in stability is a result of the interconnection of the nanolattice, which ensures a more robust and well-aligned nanolattice framework. On the other hand, the nanostructures resembling isolated pillars have far less mechanical stability and result in crack formation in the patterned microstructure. The increased stability allows the thin TiO 2 layer, ranging from 32 to 65 nm, to be applied without cracking, as the more aligned and stable architecture of the nanolattice provides a solid foundation for the coating. However, the cracks that form in the microstructure for a higher thickness remain a challenge. The results indicated that reducing the thickness of the TiO 2 layer is effective in reducing film cracks, which reduces the effect of the residual film stress. However, further work is necessary to find a consistent process to completely remove these cracks on the top of the nanolattices for thick films. Film cracking can occur due to thermally induced mechanical stress during the furnace experimental process. To address this issue, alternative methods of removing the photoresist from the nanolattices using n-methyl-2-pyrrolidone (NMP) will be explored. NMP is a solvent that can dissolve the photoresist template and planarization layers without affecting the Al 2 O 3 structure deposited using ALD. This method is promising because it removes the photoresist as effectively as the furnace method but requires significantly less time. 36 The film cracking will be investigated further by looking at other patterning approaches such as contact lithography. In this approach, a lift-off process can be used to pattern the microstructure, which can lead to less thickness variation when compared with a stencil mask. Other microstructure geometries such as finger electrodes and slab waveguides will all be examined. The current dog-bone stainless-steel mask consists of sharp edges, and the geometric constraints have led to thickness nonuniformity and stress concentrations during TiO 2 deposition using e-beam evaporation.
In addition, the photoresist transitions from a solid to a liquid phase during the thermal desorption process, allowing the TiO 2 layer to sink into the nanolattice structure. In its liquid phase, the TiO 2 layer is held in place by surface tension and gradually settles as the planarization layer evaporates. This results in a gradual placement of the TiO 2 film on the nanolattice structure. However, a key challenge lies in the lack of a strong chemical bond between the TiO 2 film and the Al 2 O 3 nanolattice. In this experiment, the TiO 2 film is thin and has a limited volume, which ensures high adhesion to the nanolattice layer. However, this issue could become problematic for thicker films with greater volume. To mitigate this effect, the planarization layer can be etched back, allowing the TiO 2 film to directly contact the Al 2 O 3 layer and establish a stronger bond. By exploring different mask designs, photolithography techniques, and etching methods, the future experiment will aim to better understand and address the factors contributing to film cracking.
This work demonstrates that TiO 2 dog-bone microstructures can be successfully fabricated on porous nanolattices, making a step forward in addressing the integration challenges for nextgeneration photonic devices. One of the remaining challenges is cracks in the microstructure layer for thick films, which can be attributed to residual film stress during electron beam evaporation. The experiment indicates that the overall grain size increases as the thickness of the film decreases, meaning less crack formation for thinner TiO 2 layer and underscores the critical relationship between the film thickness and grain size distribution. The stability of the underlying porous structure also plays a critical role, where the number of cracks is significant for periodic nanopillar and can be significantly reduced for nanolattices. The crack formation can be further mitigated by exploring different types of masks, nanolattice geometry, along with using solvents to remove the photoresist at room temperature. Future research will focus on using this process to integrate patterned waveguides and microelectrodes on nanolattice structures for next-generation optoelectronic devices.
J. Vac. Sci. Technol. B 43(2) Mar/Apr 2025; doi: 10.1116/6.0004054