1. Introduction

Multicore optical fibers have emerged from the need to satisfy the increasing internet bandwidth demand driven by applications such as high quality video streaming or Big Data analysis, among many others [1]. The research in the domain of Space Division Multiplexing (SDM) [2] has triggered the investigation and fabrication of optical fibers with multiple elements, or Multicore optical Fibers (MCF), namely optical fibers with a typical cladding diameter of ∼125µm, with more than one core, typically, 4, 7, or 19 cores [3]. Although the concept of the multicore optical fibers is not new [4], it was in the last two decades, that their unique advantages were exploited in various applications including bend measurements or shape sensing [5–7], distributed sensing [8], lasers [9], or even in optical tweezers [10] among many others [11]. However, a common issue in every multicore optical fiber application is the interfacing between each individual, closely packed core of the MCF to other optical fibers, external light sources or detectors.

The best performing device, in terms of insertion losses, for multicore optical fiber connection to individual optical fibers or to a single point is the optical fiber Fan-In / Fan-Out (FIFO) [12,13]. A FIFO is usually comprised of a bundle of chemically etched or tapered optical fibers, brought together in a lattice resembling the MCF core layout and glued or fused together [14]. The advantages of the FIFOs are numerous, including their high-power damage threshold and insertion losses as low as 0.3 dB [14]. However, manufacturing of such devices is a time-consuming process and requires precision single core placement and alignment which drive the total cost of the device up making it unattractive for some cost sensitive applications. Another method used for interfacing MCFs to single core fibers is by free space coupling. This approach can achieve overall insertion losses as low as 1 dB [15], however, it requires precise alignment and the overall size of the system is relatively large, even in more compact configurations [16]. A third approach towards MCF core interfacing is the use of ultrafast lasers for the formulation of waveguides that act as the fan-in and fan-out paths in bulk optics or other optical fibers [17]. The method shows practically unlimited scalability and insertion losses below 1.3 dB have been reported [18,19]. Apart from glass substrates, polymer based materials have also been exploited for the fabrication of FIFO waveguides where insertion losses as low as 2.36 dB can be achieved in the optimal configuration [20]. However, this method is limited by the substrate refractive index (RI) and the maximum RI changes achievable by the writing laser beam and requires expensive infrastructure and optics for the fabrication.

In several applications utilizing MCFs it is necessary to combine the optical signals of each individual core to a single point in space, or to a single optical fiber core. Such an application example is the fabrication of different wavelength Fiber Bragg Gratings (FBG) in the cores of an MCF and the combination of the signals to a single optical fiber that connects to the FBG readout unit. In this case, the information from each individual core is spectrally encoded and the signals can be combined in a single optical fiber negating the need for expensive fan-out devices. Currently, mainly FIFO devices are used for this application as in [6] and [7], but alternatively other fan-in techniques could be used such as fusion splicing. Optical fiber fusion splicing of MCF to a single core optical fiber has been used in [9], where an MCF was spliced to a standard SMF optical fiber with total insertion losses of 2 dB for a lasing application. Another application where optical fiber fusion has been utilized was the realization of optical tweezers by means of multiple optical fiber fusing and lensing [10]. Optical fiber fusion splicing is a cost-effective method for MCF interfacing; however, it lacks flexibility, cannot produce fanouts and requires specialized optical fiber fusion equipment.

In the present work, a novel approach on the fabrication of fan-in waveguides is undertaken by utilizing the two photon polymerization (TPP) technique [21] to manufacture free-form waveguides at the tip of a 4 core MCF. The TPP technique has proven to be an invaluable tool in waveguide fabrication due to its capability of easily producing complex structures in 3d space. Such waveguides have been used in numerous applications, ranging from hydrogen detectors [22] to optical interconnects for photonic neural networks [23]. A thorough review of such applications can be found in [24] and recently the capability to control waveguide refractive index by altering the laser exposure time has been demonstrated [25], which opens up new possibilities in precision waveguide parameter control. The TPP technique is used in this work to combine the individual signals of the 4 cores of the MCF into a single core, focusing mainly on sensing applications such as FBG interrogation in MCFs. The proposed solution is scalable to higher core count MCFs and can be potentially expanded to include not only fan-ins, but also fan-outs. Free-form waveguides are designed and manufactured using a TPP station and performance is evaluated by examining the insertion losses. The advantages of the proposed approach are the great design flexibility, the good scalability, low-cost and the relatively fast fabrication time. The free-form fabrication capabilities in 3d space combined with a large selection of commercially available resin materials for TPP, open up unique possibilities for refractive index and consequently waveguide engineering, which would be otherwise very difficult or impossible to achieve with other techniques.

2. Waveguide design and fabrication

Initially, a waveguide structure that combines the optical signals from and towards the 4 cores of the optical fiber is designed and discussed. The starting point of the design is the measurement of the core separation distance and lattice shape of the 4-core optical fiber. The MCF used in this work is the SM-4C1500 (8.0/125) 4 core optical fiber from Fibercore with a core-to-core separation distance measured at 50 µm (see Fig.1a) and a core diameter of 8 µm. The cladding diameter is measured by the manufacturer at 124.8 µm and the scale in Fig. 1(a) is calibrated according to this value.

 

Fig. 1. a) Optical microscope image of the 4-core optical fiber with measurement representing a core-to-core distance of 50µm, b) During the optical manifold fabrication three main parameters were examined for their effect in performance. Optical waveguide diameter (3 different variations at 8, 10 and 12 µm diameter), optical waveguide termination (flat-end or spherical lens termination) and finally, total waveguide length.

Download Full Size | PDF

The waveguide design is comprised of 4 S-shaped elements of 8, 10 or 12 µm diameter each, that fuse together at the end to a single output port, as presented in Fig. 1(b), forming an optical manifold. The selection of the above diameters was chosen as a trade-off between mechanical stability and optical performance. The location of the output port is designed to reside in the center of the optical fiber end-face, where the core of a single-core, single-mode optical fiber is typically located. The diameter of each waveguide remains constant along its full length and the output port length is typically ranging from a few micrometers up to 50 µm, depending on the total structure length. The termination of the port output is initially designed to be flat. The total structure length ranged from 150 µm up to 400 µm. From a design point of view, longer lengths lead to larger bend angles of the S-shaped waveguide, which make it less prone to bend losses. However, at the same time, as the length increases, material absorption and scattering can have a negative impact on the waveguide performance. From a manufacturing point of view, the maximum structure length is defined by the z-axis piezoelectric stage travel range, which is in this case 450 µm and is the upper limit of fabrication for this work. Evidently, the design depicted in Fig. 1(b) is destined for use mainly in sensing applications where the information can be wavelength encoded and fusion of the signals is desired.

The fabrication of the waveguides is realized using a commercially available TPP station, namely microFAB-3d Advanced from Microlight3D (France). This TPP station utilizes a semiconductor nanosecond laser emitting at 532 nm to produce structures with a lateral and vertical writing resolution of up to 200 nm and 600 nm, respectively. The overall build volume available is 300 µm x 300 µm x 450 µm and the minimum surface roughness achievable is 20 nm Root Mean Square (RMS). For the fabrication on top of the optical fiber tip, a 20x microscope objective is used, which, according to the manufacturer specification sheet, can produce a maximum lateral resolution as low as 0.7 µm. The bare optical fiber is held in place for waveguide fabrication using an optical fiber chuck which is positioned in a custom housing provided by the TPP station manufacturer. Alignment of the optical manifold structure with respect to the optical fiber cores takes place using the supplied software of the TPP station and is realized with a 10 nm step resolution.

The material used for the fabrication of the waveguide is a commercially available, proprietary photoresist from Microlight3D named OrmoBio, which is a biocompatible organically modified ceramic (Ormocer), derived from the commercial product OrmoComp (Microresist technology GmbH). The refractive index of the OrmoBio at a wavelength of 587 nm is 1.52 and at a wavelength of 1.5 µm is 1.505. The OrmoBio resin is chosen as the fabrication material because of the very good surface roughness it exhibited in the fabricated waveguides. Following the completion of the TPP fabrication process the structure is developed inside OrmoDEV a commercially available developer, for typically 7-15 minutes. OrmoDEV is a solution consisted of 4-methylpentane-2-one, methylisobutylcetone and 2-propanol. The TPP fabrication process itself lasted from 7 minutes up to 18 minutes for the longer length manifolds. A summary of the fabricated structures is depicted in Figure S.1 of the Supplemental Document.

A comparison between the three different terminations of the optical manifolds, namely, flat-top, ball-lens, and lens, is presented in Fig. 2. Optical manifolds designed with a flat end termination tend to exhibit defects at the terminal edge, whereas the manifolds designed with a lens termination exhibit a smoother termination surface, indicating the manufacturing limits of the technique for such small area structures when a flat geometry is designed (S250-10-1). Additionally, it is worth noting the difference in quality between the samples S250-10L-1 and S250-10L-2 which are both terminated with a plano-convex lens of the exact same geometrical design (lens and curvature radius = 4µm). However, the former sample (10L-1) is sliced prior to fabrication along the z-axis (longitudinal) as the rest of the manifold structure, while the later sample (10L-2) is individually sliced along axes x and y (transverse axes). The quality difference between the two is obvious, already from the microscope images (Fig. 2, top row) and highlights the necessity of proper slicing prior to manufacturing. The transversally sliced plano-convex lens (lens radius as well as radius of curvature equal to the waveguide radius) termination is smoother, without significant defects as clearly visible in the scanning electron microscopy (SEM) images (see Fig. 2 bottom row, right). On the contrary, the longitudinally sliced sample exhibited repeatedly a point defect of ∼ 4 µm at the edge, which severs optical performance, by acting as a scattering center. Finally, when the structure is designed to be terminated with a flat top surface at the edge, the manufactured structure exhibited a gap in the middle, which could be the result of structural compaction during the development process (shrinkage of the interior liquid structure creates this defect, most likely due to capillary forces). This defect is clearly evident in the SEM image in the bottom row of Fig. 2 and is always evident in samples that are designed with flat top termination. All slicing is performed in the software Luminis, supplied by the µFAB-3D TPP station manufacturer.

 

Fig. 2. Effect of the terminal end design geometry on the structure quality after fabrication. Top row: Microscope images. Bottom row: SEM images. Left: Flat-top termination design resulted systematically in defects at the output port of the optical manifold. SEM revealed a hole formation in the middle. Middle: Ball lens termination, when sliced along the longitudinal, z axis could not be properly formed during fabrication, resulting in a protrusion. Right: Lens termination, when sliced along the transverse axes resulted in a systematically smooth end-point termination (see SEM image at the bottom)

Download Full Size | PDF

Scanning electron microscopy imaging is performed in several of the fabricated samples to better understand the method limitations and possible improvements in the design. A SEM image of an optical manifold structure of 250 µm length and 10 µm waveguide diameter, with plano-convex lens termination (transversally sliced) on top of the MCF is presented in Fig. 3(a).

 

Fig. 3. a) Scanning Electron Microscopy (SEM) image of 250 µm long optical manifold with 10 µm waveguide diameter and plano-convex lens termination with R = 10 µm and independent transversal lens slicing., b) SEM image of a partially detached S-shaped waveguide at the contact point of the optical manifold with the 4-core optical fiber, c) SEM image of two partially detached S-shaped waveguides. The optical performance of this sample was found to be inferior to a normally attached one.

Download Full Size | PDF

The structure exhibits a smooth surface across its length and the minor defects observed are postproduction impurities, as the sample is extensively used prior to SEM imaging for performance characterization. From a manufacturing point of view, apart from the optical manifold termination design that was already discussed, another point should be noted during the fabrication process. In some samples, it was observed that some of the S-shaped waveguides of the manifold are not in good contact with the MCF, as shown in Fig. 3(b) and Fig. 3(c). This defect occurs when the optical fiber is not perfectly leveled during the beginning of the TPP process and can be remedied by applying a small offset at the focus of the laser for the first layer towards the bulk of the optical fiber. Another type of structural defect that is observed has to do with the presence of impurities in the OrmoBio solution during the fabrication process which resulted in some samples having irregularly shaped waveguides.

Since SEM imaging is performed after performance assessment of the fabricated samples, it is observed that the most taxing defect in terms of performance originates from the gap formed at the optical fiber tip when designed for flat top termination. A significant fabrication flaw is also observed for samples with detached feet, which -in addition- suffered from mechanical stability issues, contrary to samples with all 4 S-shaped waveguides firmly attached to the optical fiber surface. Such an example of performance degradation is presented and discussed in the next section.

3. Multi-core to single core optical fiber link performance testing

To assess the performance of the fabricated optical manifolds a free space link is realized between a single core optical fiber and a four-core optical fiber with an optical manifold structured on its tip. The single core optical fiber is attached to an adjustable current continuous wave laser source emitting at 1550 nm which is used within a power range from 0.58 mW up to 4.73 mW (-2.4 dBm up to 6.74 dBm). The output terminal of the single core optical fiber is placed at the stationary part of an x, y, z flexure stage capable of submicron positioning within a total travel range of ∼ 5 mm on each axis. On the moveable part of the flexure stage the 4-core optical fiber with the TPP manufactured optical manifold is placed. Positioning of the 4-core optical fiber with respect to the single core optical fiber is realized through the x, y, and z axis motion of the flexure stage, facilitated by using a camera attached to a single board computer (SBC). An achromat doublet lens is used (f = 25 mm) for the image magnification along with a 5 Megapixel camera (Raspberry Pi Camera v1.3, 5MP 2592 × 1944 pixels, Omnivision 5647 sensor in a fixed lens module). The link performance is monitored through an optical power meter where the output terminal of the 4-core optical fiber is placed using a bare fiber adaptor. The best performance is determined by several iterative adjustments of the flexure stage axes until the maximum optical power is achieved. Experiments are realized with different surrounding media for the optical manifold. First experiments are performed with air, then distilled water is used and finally a solution of acetone-PMMA is prepared as the optical manifold surrounding medium. A schematic representation of the setup is shown in Fig. 4.

 

Fig. 4. Performance testing of the fabricated TPP waveguides setup. A single core optical fiber is used for beam delivery towards a 4-core optical fiber with an optical manifold at its tip. SBC: Single Board Computer.

Download Full Size | PDF

The first optical manifold parameter that is examined for its effect on the overall optical link performance is the waveguide length. For these tests, a set of optical manifolds manufactured with a waveguide diameter of 8 µm and lengths ranging from 150 µm up to 400 µm are tested using the setup of Fig. 4. It is found that the best performing length is 250 µm, followed by the samples of 200 µm and 300 µm length (with ∼4 dB and ∼6 dB performance degradation, respectively), while the 150 µm and 400 µm long optical manifolds are the worst performers (with ∼ 9 dB and ∼ 8 dB performance degradation with respect to the 250 µm long sample). A performance degradation with increasing length could be the result of elevated material absorption. According to the manufacturer, the OrmoBio transmittance of a 20µm thick sample at 1550 nm should be more than 98%, indicating a possible material absorption impact as the length increases [26]. Finally, another source of losses as a function of length could be attributed to scattering losses, which add up the longer the waveguide is. Given the fact that the major loss mechanisms related to structure length are bend losses, absorption and scattering, with the bend losses inversely proportional to the other two, might explain the fact that neither the shortest, nor the longest structure exhibited the best results. In general, an optical manifold length of 250 µm exhibited the best results and is used as the reference length.

Taking into account that 250 µm long samples have exhibited the best performance, this structure length is used for testing the effect of waveguide diameter on the performance of the optical manifolds. Samples of 250 µm length are fabricated with waveguide diameters of 8, 10 and 12 µm. The setup of Fig. 4 is used to determine overall link performance and the output optical power results versus input optical power are plotted in Fig. 5. It is immediately evident, that the smaller diameter waveguide structure, i.e. the 8 µm one, performed better than the 10 µm and the 12 µm one. Within an optical power input range from 0.58 mW up to 4.73 mW the output is linear, showing lack of non-linear absorption or scattering effects in the developed OrmoBio material. Linear fitting of the data, with forced intersection at point 0,0 is performed with the slope α for each waveguide diameter presented in Fig. 5 (α₁₂, α₁₀ and α₈, for the 12, 10 and 8 µm waveguides, respectively). The slope is indicative of overall performance of the optical manifold and reveals the poor performance when the optical manifold surrounding material is air. Additionally, the performance of a sample with 10 µm waveguide diameter and two waveguides detached from the MCF (sample of Fig. 3(c)) was plotted to highlight the negative impact of manufacturing defects (sample designated in plot as ‘10 µm defect’). The 8 µm waveguide manifold managed an overall coupling efficiency above 1.5%, while the others are below 1%.

 

Fig. 5. Optical power output from the structured 4-core optical fiber versus input power from the single core optical fiber for three different structure waveguide diameters.

Download Full Size | PDF

As will be discussed in the next section, the major loss mechanism in this case is the numerical aperture (NA) mismatch between the TPP fabricated structure and the single mode optical fiber cores. As the material refractive index could not be changed due to the single material that was used for the waveguide fabrication (OrmoBio), the effect of the material surrounding the structure is investigated by applying a water cladding and a PMMA solution cladding and repeating the power coupling experiment of Fig. 5. The results for a plano-convex lens terminated, 8 µm waveguide diameter optical manifold immersed in air, water and PMMA solution are presented in Fig. 6.

 

Fig. 6. Optical power output from the structured 4-core optical fiber versus input power from the single core optical fiber for three different waveguide cladding materials: Air, distilled water and PMMA-acetone solution.

Download Full Size | PDF

Evidently, from Fig. 6, as the surrounding material refractive index approaches that of the optical manifold (n_manifold= 1.503 at 1550 nm) the NA mismatch loss is greatly reduced providing coupling efficiencies as high as ∼33% (n_air= 1.00, n_water= 1.32 [27], n_PMMA= 1.49 [28]). It is also worth noting that power output remains linear both for water and PMMA waveguide immersion at these optical power levels. This significant increase of coupling efficiency proves that TPP fabricated waveguide structures have the potential to become a practical, low-cost alternative to more expensive methods, especially in applications where some losses can be tolerated such as sensing applications, or even short-haul telecommunications. Optical manifolds with sufficient mechanical strength, such as those fabricated in this work, can be introduced inside cold splice connectors, to establish an easy link between a multi and single core optical fiber.

4. Losses

The fabricated structures, when operated using air as the surrounding medium, exhibit elevated coupling losses as high as 21.7 dB, which would render the optical manifolds unsuitable for many practical applications. There are several loss mechanisms that deteriorate the performance of the fabricated microstructures, however, the construction of a waveguide without a cladding, results in very high normalized frequency V numbers and the waveguide is highly multimodal. A precise calculation of the power loss between multimode optical waveguides is a rather complex task, because it depends on the power content of each mode and its excitation probability [29]. For the sake of simplicity, it will be considered that all modes are equally excited and carry similar amounts of power inside the multimode optical manifolds. In this case, the coupling efficiency η between the multimode waveguide and the single mode recipient core of the MCF is defined as [30]:

The number of modes in an optical fiber is approximated by calculating the normalized frequency of the waveguide V through [31]:

The calculation of the normalized frequencies V, the number of modes M and the corresponding coupling efficiency η between the waveguide and the single mode optical fiber core are presented in Table 1, using air as the waveguide surrounding medium.

Table 1. Optical manifold fabrication parameters, modal number, coupling efficiency and losses for air surrounding medium

View Table

According to the experimental setup of Fig. 4, the light is launched from the single core, single mode optical fiber (M_emmiter= 1) towards the fabricated waveguide which is highly multimodal. At the launch point (single mode, single core optical fiber to optical manifold) the coupling efficiency, without taking into account Fresnel losses, should be 1, as both waveguides share a single common mode. However, at the point where the optical manifold attaches to the cores of the MCF the high numerical aperture (and hence the V number) of the optical manifold surrounded by air, creates a very high loss which is calculated using [30]:

In every case, it should be noted that for these theoretical calculations, it is assumed that all modes are equally excited in the multimode waveguides of the optical manifold. The resulting losses in dB are presented in Table 1. It is immediately evident that the produced waveguides, when surrounded by air, are characterized by their highly multimodal nature resulting in coupling efficiencies less than 1% and correspondingly high losses. In practice, coupling efficiencies slightly more than 1% (1.5%) are observed. There are many mechanisms that might contribute to this increase, such as scattered light entrance into the cores, or lack of excitation of all supported modes. The evident mitigation measure for these losses is the immersion of the waveguide in a surrounding medium that could drastically decrease the waveguide V-number and hence the excited modes, reducing the numerical aperture mismatch losses. As already mentioned, for this reason, the optical manifold is immersed in either distilled water, or a PMMA solution (0.1 g of 99% purity PMMA diluted in 0.1 ml of 99.9% purity acetone) and the coupling efficiency increased to 4.8% for distilled water immersion and 32.8% for PMMA solution immersion. The theoretical NA mismatch losses of the different waveguides as a function of surrounding medium refractive index are presented in Fig. 7, along with the values obtained experimentally.

 

Fig. 7. Theoretical Numerical Aperture mismatch loss (dashed line) between the optical manifolds of various waveguide diameters as a function of surrounding refractive index medium. Experimental measurements as points for 3 different surrounding media: Air, distilled water, PMMA solution.

Download Full Size | PDF

Figure 7 highlights the importance of the waveguide surrounding medium on the manifold performance, but also reveals that more loss mechanisms come into play in practice. When the waveguide surrounding medium is air or distilled water, the experimentally measured losses are less than those theoretically predicted. Taking into account that for the case of air there were also Fresnel losses, apart from scattering and material absorption, this result would seem paradox, however, it stresses that the assumption of all modes excitation in the optical manifold with each mode containing a similar amount of optical power is an oversimplification. When the PMMA solution is used and the waveguide becomes fewer-mode, the losses become much smaller compared to the other two cases and the experimentally measured losses are higher than those predicted by the NA mismatch. In this case, the remaining of the loss mechanisms such as scattering, or absorption begin to manifest themselves while the fewer-mode nature of the waveguide makes the assumption of all-mode excitation more probable.

In terms of possible performance increase of the produced waveguides there are two possible pathways that can be discussed. It can be assumed, that the use of a surrounding medium with a refractive index value closer to that of OrmoBio than PMMA would further increase the coupling efficiency, as the fabricated waveguide would become fewer mode, or even single mode. Moreover, in this work, only one material was used for the optical manifold fabrication (OrmoBio). Given the fact that there is a large resist variety on the market for TPP, it is likely that experiments with different resists, selected in such a way that exhibit improved optical properties compared to OrmoBio (i.e., less absorption and scattering) will result in better performance. Assuming that the intrinsic absorption of the OrmoBio material follows the Beer-Lambert law and using the absorption parameters communicated by the manufacturer an absorption coefficient α = 0.001 µm^-1 is calculated. Using this absorption coefficient value for a 250 µm long waveguide a minimum theoretical loss of ∼ 1.1 dB is calculated, if the optical manifold and the connecting waveguides are matched in terms of numerical aperture and no other losses are present. The use of a different resin with lower intrinsic losses can potentially further reduce this theoretical value.

Apart from the effect of the NA mismatch loss, it is worth mentioning the differences between scattering across different optical manifold samples. The most pronounced differences are observed between the flat-top terminated samples versus the plano-convex lens samples that are sliced along the x, y (transverse) axes. Light scattering is investigated by launching visual (λ = 632 nm) light into the MCF and observing the scattered light by means of an optical microscope. In every sample examined, the flat-top terminated optical manifolds exhibit significantly higher scattering losses as captured by the optical microscope, across all rotation angles of the structure. On the contrary, the plano-convex lens terminated structures are systematically showing much less scattering, highlighting the importance of proper termination design. In addition, significant differences were observed between structures with lenses manufactured with different slicing. Figure 8 shows the visible light scattering from two different optical manifold structures with lens termination, one with z axis slicing and another with x, y axis slicing. Image intensity profiling along axes AA′ and BB′ show a larger area for the z sliced sample, which corresponds to more intense light scattering. The profile plot is at the bottom part of Fig. 8. Similar information is also conveyed from the figure histogram which is presented as an inset at the bottom left corner of the optical microscope images. Both images were captured using identical exposure times and gain.

 

Fig. 8. Optical microscope images of visible light scattering (top) and SEM image insets of lens terminated optical manifolds manufactured using z axis slicing (left) and x, y axes slicing (right) for the lens.

Download Full Size | PDF

5. Conclusions

The two-photon polymerization technique has been used for the fabrication of optical waveguides connecting the four cores of a multi-core optical fiber to the core of a single core optical fiber. It has been demonstrated that optical coupling performance was dominated by the numerical aperture mismatch between the fabricated structures and the single-mode recipient optical fiber cores. The immersion of the produced waveguide in media with increasing refractive indices, exhibited corresponding increase to the coupling efficiency. The best result in terms of coupling efficiency (∼33%) was experimentally observed when a PMMA solution was used as the cladding material. The method has the capability to produce fan-ins at a fraction of the cost of other solutions, making it an attractive alternative for use in applications such as sensing, tweezers, or lasing. Due to the free-form design capability of the method, the fabrication of fan-outs will be tested in the future to explore the potential of the method in other applications.