1. Introduction

In micromechanics, the assembly of minute components particularly with dimensions of - 100 μm still remains a challenge [1]. Search for new microassembly methods is therefore important as it may lead to the potential development of novel micro(opto)electromechanical-systems [1–3]. Fluidic microassembly, for instance, is increasingly pursued due to several reasons including the reduced undesirable effects of van der Waals forces and electrostatic interactions in liquid host environments. Most of the current methods of fluidic-based microintegration rely on the probabilistic nature of random assembly processes [2, 3]. At present, microscale random assembly in liquid is constrained by the trade-off between positioning accuracy on the bonding of microelements (normally to receptor sites on a template) and bonding yield. Possible yet cumbersome solutions for achieving reasonable yield are sample recirculation and template agitation. Bonding selectivity has nevertheless been enhanced by suitably matching the geometrical shapes of the tiny building blocks and the receptor sites [3].

Here we demonstrate an all-optical, directed microassembly scheme in liquid by tiling a plurality of microscopic structural elements on a planar substrate using real-time reconfigurable optical traps from a variant of the optical setup described in our previous work [4, 5]. The number of traps, their intensity profiles and spatial locations can be controlled either interactively or in an automated way by a computer. Under computer-automated control, the system demonstrates the capability for fully autonomous search-and-collect routines without need for user-intervention [6]. Results show that optical traps of few milliwatts are able to achieve good positional and rotational control of the microstructure. We also make use of shape complementarity among the micropuzzle pieces. The puzzle pieces have identical geometrical shapes and in-plane rotational symmetry. Furthermore, the puzzle pieces have elongated aspect ratio so that orientations are readily discerned by an image analysis subroutine and optically controlled by likewise elongated traps. The microfabrication of the puzzle pieces via femtosecond laser two-photon polymerization technique [7, 8] is also described.

2. Design and fabrication of micropuzzle pieces

2.1 Geometrical considerations for the micropuzzle pieces

Not unlike its macroscopic analogue, the tiling of micropuzzles we perform here known as tessellation follows a simple rule of bringing two-dimensional shapes together to cover a plane or substrate without overlapping and without leaving gaps. There exist a wide variety of tessellations to choose from but the particular one selected here is shown in Fig. 1(a). With this design preference, identically shaped puzzle pieces whose geometry is depicted in Fig. 1(b) can be fabricated, thereby simplifying both optical procedures of fabrication and manipulation. The target puzzle pattern exhibits a characteristic checkerboard arrangement of horizontally and vertically symmetric tiles. It falls under the “p4g“ category [9] of the seventeen possible plane symmetry groups (a mathematical classification of 2D repetitive patterns).

 

Fig. 1. Illustration of (a) the desired tessellation to be optically assembled and (b) a single micropuzzle element, which is a symmetric cutout of a circle of radius r.

Download Full Size | PDF

Other simpler tessellations are those that can be formed by tiling one type of congruent regular polygons such as equilateral triangles, squares or hexagons. For the optical manipulation part, the puzzle piece design in Fig. 1 seems a reasonable choice since laterally elongated microobjects have been oriented by similarly elongated optical traps with good rotational selectivity [10]. Here we straightforwardly realize the sufficient trap geometry for controlling both the position and the angular orientation of a puzzle piece by using adjacently paired optical traps with identical tophat intensity profiles. The size of the composite trap is adjusted to match that of the microobject. Our optical trapping graphical user interface (GUI) developed in LabVIEW readily offers these trap-grouping and trap-scaling features together with the necessary control over clusters of traps, i.e. xy-position and angular rotation about the centroid of each paired traps.

2.2 Microfabrication via 2PP technique

The micropuzzle structures are fabricated by the two-photon polymerization (2PP) method with the optical and mechanical setup described in a previous work [8]. We start out by spin-coating an approximately 15-μm thin layer of an epoxy-based SU8 negative photoresist resin (Michrochem, Newton, MA, USA) onto a microscope glass coverslip. The beam from a 100-fs pulse laser (790nm, 80MHz) is brought to a diffraction-limited focal spot by a 100× oil-immersion microscope objective in a resin layer about 5 μm from the glass substrate. The very precise localization of laser energy in both time and space allows a desired volume-confined polymerization to take place – making 2PP a powerful high-resolution 3D microfabrication technique. During the polymerization of the individual pieces, the photocurable resin is translated with nanometer precision relative to the focused laser beam by a piezo-driven sample stage and a microscope objective displacer. The path of the polymerizing laser beam (shown in Fig. 2) is chosen so the polymerized lines merge during the process and therefore the desired continuous structure is achieved. The laser power (measured before the objective) is set to ~3 mW with translation speed of 5 μm/sec. The pieces are polymerized in the bulk of the SU8 resin layer, with their plane parallel to the coverslip surface. The pieces are not in direct contact with the glass surface but are separated by a thin layer of unsolidified resin.

 

Fig. 2. Selected trajectories for the focused femtosecond laser beam in the 2PP fabrication of the puzzle pieces with characteristic radius r = 2.5 μm. Inset shows the corresponding SEM micrographs of the 2PP-fabricated structures.

Download Full Size | PDF

After the whole 2PP-writing process, a 2-min post-exposure bake is performed at 95 °C. The sample is cooled down and the portion of substrate with polymerized puzzle pieces is placed into a reservoir with the SU8 layer facing up. Then 100 μL of a developer solution (Micro Resist Technology GmbH, Berlin, Germany) was gently added into the reservoir to remove the unsolidified SU8. In this process, puzzle pieces slowly drift from their original locations and eventually settle onto the glass substrate. The developer is replaced with another fresh 100-μL solution, which does not cause further positional drift. Then the developer is removed and the reservoir is rinsed gently with ethanol and dried. To obtain fluid-borne microstructures from the substrate, we first add few microliters of 5% surfactant solution (Tween 20, Sigma-Aldrich) to the reservoir. Then, while viewing the sample under a low-magnification microscope, we detach and extract the puzzle pieces using a customized mechanical micromanipulator made from glass capillary tube connected to a microsyringe by silicon rubber tubing. The free and sharper end of the capillary tube has approximately 30-μm diameter opening and is controllably positioned by a motorized 3D manipulator arm. An SEM image of the micropuzzle pieces is shown in the inset of Fig. 2. The estimated 400-500 nm lateral resolution of the 2PP method based on the implemented parameters enabled us to produce the pieces with acceptable morphology and uniformity. We estimate the thickness of the structures to be 1.0 ± 0.2 μm.

3. Optical micromanipulation: results and discussion

3.1 Optical trapping performance for position and orientation control of microstructures

In testing the viability of our multiple-beam trapping system as a microassembly tool, we first measure the degrees of spatial localization and angular orientation that can be achieved for a single puzzle piece. To do so, we initially position a puzzle element at the center position, (x, y) = (0, 0), of the imaging field of view using horizontally aligned pair of optical traps (Fig. 3). With our LabVIEW GUI, we capture a video (10 frames/sec), whose frames are subsequently sent to an image analysis subroutine that measures the angle θ of the longer symmetry-axis and centroid coordinates (x, y) of the microobject. Plots of θ(t) and the radial distance ρ(t), ρ ² = x ² + y ², in Fig. 3 show that, in the presence of a composite trap with a total power of ~6 mW (i.e. a pair of non-overlapping tophat-profiled beams imaged onto the sample plane), translational control of the microstructure can be achieved with submicron resolution. Angular control is precise to within a few degrees. While the puzzle element is trapped for ~95 sec, the measured standard deviations for ρ and θ are 39.6 nm and 3.41 deg, respectively. Immediately after the laser trap is switched off, the puzzle piece observably wanders away with unstable angular orientation due to Brownian motion.

 

Fig. 3. Plot of the angular orientation and radial position of a puzzle piece as a function of time in the presence and absence of paired trapping beams. Relative positions of the paired traps are indicated by the two circular markers overlaid with the image of the trapped and horizontally oriented microobject (inset).

Download Full Size | PDF

3.2 All-optical assembly of micropuzzles

Several modes of micropuzzle assembly are demonstrated experimentally. First, we show the interactive disposition by which the user can manipulate and assemble the puzzle pieces using real-time drag-and-drop optical manipulation system. In this mode, the user does the tiling in a serial manner coupled with the control of the motorized sample stage to search for additional unassembled pieces (Fig. 4). Although this manner of assembly is relatively slow (~ 4 min to form a 4×4 array), it allows the user to react instantly to unexpected setbacks that may arise during the procedure. Fully interactive assembly is perhaps most suitable for moderate particle densities where some stage scanning is necessary to accumulate a sufficient quantity of micropuzzle pieces.

 

Fig. 4. (AVI, 2.4 MB, 3x speed) User-interactive pick-and-place optical assembly of sixteen micropuzzle pieces into a 4×4 tiling. The linear and angular speeds at which the pieces are translated and rotated are ~2 μm/sec and ~20 deg/sec, respectively.

Download Full Size | PDF

At higher particle densities, parallel trapping and assembly becomes more feasible as more micropuzzle pieces can be found within a single microscope scene. Using this approach, tessellations of the microscale pieces have been formed in a 4×4-tiling configuration (see Fig. 5) using a single GUI execution that moves all the trapping beams, and thereby the constituent pieces, simultaneously along pre-defined paths as specified by the user. This mode results in shorter assembly time of a few seconds.

It is also noted that the larger tiled structure consisting of several constituents can be translated and rotated by the array of traps (~6 mW trapping power for each puzzle piece) as one entity at linear and angular speeds of approximately 1.5 μm/sec and 10 deg/sec, respectively, without compromising structural integrity. This means that the sample stage may be moved at similar speeds instead of the grouped traps and can serve as a means of bringing the tessellated pieces to a farther transverse location in the sample where other free pieces could be gathered and thereby increasing the size of the tessellation. Such procedure has been applied in the computer-automated mode of microassembly.

 

Fig. 5. (AVI, 0.9 MB) Larger tessellation of sixteen micropuzzle pieces optically assembled in parallel into a 4×4 arrangement. Once assembled (~7 sec from 1^st to 3^rd frame), adjacent elements remain intact while the superstructure is displaced and rotated.

Download Full Size | PDF

Fully automated micropuzzle assembly was accomplished with the aid of basic image analysis packages available in LabVIEW. In Fig. 6, we describe this mode where the computer is left to autonomously assemble the puzzle without user intervention. An image analysis subroutine obtains the coordinates, (x, y), and orientation, θ, of individual micropuzzle pieces within a selected region of interest (i.e. rectangular area on the left side of the scene). Another subroutine is then triggered to launch properly positioned and oriented traps over the detected pieces. The latter subroutine also assigns correct paths that link the individual pieces to their target positions and orientations in the puzzle immediately after detection. In this mode, the sample stage is set to move at a constant speed such that new pieces appear on the left side of the screen. This mode of assembly is ideal when particle densities are very low. In fact, this demonstration exhibits the usefulness of the system for general search-and-collect tasks in sparse samples spread over a relatively large area. Fully autonomous operation eliminates tedious manual scanning of the sample and reduces the amount of time a user needs to spend on the system. One may proceed directly to any further experiments or processes after a sufficient quantity of particles has been collected through automated control.

 

Fig. 6. (AVI, 2.2 MB, 2x speed) Computer-automated “hunt-and-collect” procedure for tiling micropuzzle pieces. The dashed rectangle highlights the selected detection area where pieces coming from the left due to constant-speed sample stage movement are automatically detected. Once detected, trapping beams with appropriate target trajectories are immediately assigned.

Download Full Size | PDF

Several methods are also available if microassemblies of permanently bonded components are desired. One example is the use of polymerizable liquid medium that enable a focused laser to locally fuse microelements either by one-photon or two-photon absorption at elements’ adjacent edges while they are held in place by optical traps [11]. Microelements may also be attached together by surface functionalization or critical point drying methods.

4. Conclusion

To address the need for microassembly methods particularly for objects with dimensions much below 100 μm, we demonstrated the use of real-time adjustable multiple optical traps to assemble micropuzzles with user-interactive or computer-automated control. The demonstration of a fully autonomous search-and-collect routine highlighted the potential usefulness of real-time reconfigurable optical traps for working with low-density distributions of microparticles. We characterized the accuracy with which the optical traps are able to position and orient microscale components. A few miliwatts of trapping power can achieve precise placement of microstructures with only tens of nanometers and few degrees of variations in position and angle, respectively. This can be enhanced further by increase in trapping power. The reliance on complementarity of shapes of the puzzle components, accurately fabricated by the two-photon polymerization technique, may also be applied to future extension of optical microassembly using microscale building blocks with more intricate geometries.