From ancient times, glass has always been a uniquely useful material for an extremely wide variety of applications. In modern times, the material’s properties are desirable for advanced applications in such areas as microfluidics (biocompatibility), semiconductor packaging (thermal compatibility), optics (transmission spectrum), sensors (inertness), radio frequency transmission (low loss), and so on.
Until recently, however, it has remained a challenge to form glass into precise, complex shapes. The advent of femtosecond laser structuring of glass has enabled true three-dimensional (3D) designs to be realized. In a process called selective laser-induced etching (SLE), a series of femtosecond laser pulses is focused inside a piece of glass. With sufficiently high energy density, the transparent material becomes opaque and quickly absorbs the pulse energy. Because of this, the absorption volume can be made very small—a few microns in diameter. Within this volume, the glass rapidly melts and refreezes, undergoing a change in morphology. The new morphology, when exposed to potassium hydroxide (KOH)-based wet chemistry, etches much faster than the surrounding untreated glass. The differential etch rate can be as high as 1000:1. Using a series of pulses and 6-axis motion stages, material can be “written” pixel-by-pixel and then selectively removed with high precision. The process is unique in that, while it is a 3D printing technique, it is a subtractive method of creating a shape. Unlike additive manufacturing methods, this technique enjoys the advantage of built-in support material. Furthermore, the high modulus of glass allows for the fabrication of intricate, free-standing parts.
SLE is most cost-effective when removing small amounts of material. Large features and thick substrates can be expensive to produce. Thus, the ideal target for this technology is small, complex parts, such as microfluidics or advanced nozzles. Even within the microfluidics realm, there is a limit to which small channels may be etched, which is on the order of 10 to 15 linear millimeters for a single-sided etch (twice as much when etched from both ends of a channel). One recent advance in laser processing now allows LightFab (Aachen, Germany) to exceed this limit. In addition, this advance opens the door to low-cost, high-volume products.
Laser welding of glass
LightFab has developed a laser-welding capability that can be carried out on the same LightFab production system as standard SLE. In this process, two substrates can be welded together, provided that the interfaces are planar and within 2 µm of each other. Figure 1 gives cross-section and plan views of the welded interface. In this process, a high-power laser pulse is focused just above the gap between the substrates. The pulse melts the glass within a sphere approximately 50 µm in diameter and induces mass transport of the molten material across the interface, leaving behind a void that appears dark in the image. The glass refreezes, forming a joint with near-bulk properties.
A series of pulses can be used to form a stitch line with a continuous glass joint, as shown in the plan view of Figure 1. This view also illustrates how a long channel may be fabricated. In the first step, a serpentine trench is scanned and etched in the surface of one substrate. In the second step, a second substrate is used to cover the open channels and a weld line is formed on both sides of the channel, creating a hermetic seal. This is an example of 2.5D fabrication wherein a 3D channel is fabricated with a combination of two-dimensional (2D) processes. Since the first substrate can also contain 3D features, the combination of substrates is termed “3.5D.”
This approach has the advantages of unconstrained channel lengths and low cost. The reduced cost is a result of two factors: first, the laser has an increased scan field, which reduces the number of mechanical stage movements, thereby reducing the write time. Second, because the weld is typically composed of only one to three lines in a single plane, they can be scanned very quickly. While standard SLE write times are typically minutes to hours, weld times are usually only a few seconds.
There are several advantages to welding when compared to standard wafer bonding techniques, such as fusion or anodic bonding. The first is the relaxed interface requirements. Fusion and anodic bonding require ultra-clean surfaces with surface-roughness values below about 1 nm. While cleanliness is still important for welding, a good weld can be achieved across a gap of up to 2 µm, as illustrated in Figure 1. The second advantage is that, where anodic and fusion bonding can require temperatures from 400° to 1300°C, laser welding takes place at room temperature. This not only allows the presence of coefficient of thermal expansion (CTE)-mismatched materials, but the process can also accommodate temperature-sensitive devices. Finally, because it does not involve the required ramp-up and cool-down times of the bonding approach, welding allows for increased throughput and reduced cost relative to bonding.
Though SLE is a true 3D process, it can, of course, be used for producing 2D shapes in glass substrates (see Fig. 2). After irradiation, the substrate is etched using the standard SLE wet chemistry in order to release the shape. This process can be used to create high-aspect-ratio vertical vias or cutouts. It can also be used to singulate chips in rectangular or odd shapes.
Example: multiwafer stack
Various combinations of standard SLE (3D), 2D SLE, and welding (0.5D) can be used to fabricate designs with higher performance and lower cost. In the following example, these approaches are combined into a general platform for high-function, high-performance applications. Figure 3 shows an example of a multiwafer stack with various levels of functionality. In Level 1 (bottom substrate), both 2D and 3D microfluidic channels have been formed, with the 2D channels open to the surface. Blind or through vias may be formed in this layer. This layer may also contain active or passive valves, filters, reservoirs, manifolds, and so on. A second substrate, Level 2, is welded on top of the first. Level 2 may contain additional 2D or 3D channels, functional elements, and input/output vias. All channels requiring a hermetic seal are fitted with weld lines (shown in red). Typically, this type of microfluidic functionality is achieved by stacking three or four processed silicon wafers and bonding them anodically. The all-glass approach represents a potential cost savings due to manufacturing simplicity and a reduced number of substrates.
Level 3, which is implemented on the top surface of the second substrate, allows for metal traces, wire bonds, and chip bonding. Chips may include MEMS, sensors, optoelectronics, and application-specific ICs (ASICS). Standard lithography, metallization, pick-and-place, and flip-chip bonding techniques are used to fabricate this layer. The wafer-level nature of the assembly process allows for the lowest possible manufacturing cost.
Level 4 is used for encapsulation/packaging. Any combination of mechanical and hermetic sealing can be achieved with the LightFab welding process. It should be noted that the welding process requires a clear view to the interface of interest. However, if obscuring elements are present, they can usually be spatially separated.
A protective cover for packaging of microelectronics or sensors can easily be fabricated monolithically with the standard SLE process. However, laser structuring of large areas can lead to unacceptably high costs. Take, for example, a 1 cm2 cover with 0.5 mm sidewalls. In the monolithic process, the recess floor and sidewalls are irradiated and then etched to release. The write time is approximately 7 minutes. Alternatively, the LightFab 2D SLE process can be used in conjunction with a bonding or welding process to create an assembled cover much more cheaply. In this process, the cover recess is cut out of one substrate leaving behind an open “frame.” The frame substrate is then laser welded to another substrate, with the weld process taking about 6 seconds/chip. Altogether the machine time would be less than one minute, a factor of 10 less than the monolithic approach. As a result, the 2D SLE/welded part would cost 10X less and allow for a much higher machine throughput—one that is suitable for high-volume production.
In summary, the LightFab printer process suite is expanding and now includes some very useful tools: laser welding and 2D SLE. Laser welding enables the wafer-level assembly of 3D structures from 2D components and removes the length constraint on fluidic channels imposed by single-sided etching. 2D SLE allows rapid structuring of vertical planes and enables fabrication of low-cost, two-dimensional features. In particular, it enables cost-effective printing of through glass vias and cost-competitive wafer or panel singulation.
The power of the combined suite of tools, 3D SLE, 2D SLE, and welding, can perhaps be best realized in wafer-scale chip manufacturing. Here, an example has been given of an all-glass, multilevel, highly functional chip that supports two fluidic levels, including such features as 2D and 3D fluidics channels, active or passive valves, filters, reservoirs, manifolds, a device level supporting metallization, wirebonding and chips, such as sensors, MEMS, optoelectronics, and ASICS, and a packaging level that can provide hermetic or mechanical protection for sensitive devices in demanding applications. Perhaps most importantly, the new tools provide a path to realize such chips at low cost and for high-volume applications.