Kulicke & Soffa has shipped its first Liteq systems for back-end lithography. One to an Asian customer, the other one was delivered to I3 Microsystems, a US-based high-performance microsystem supplier.
The first sales of Kulicke & Soffa’s Liteq 500A systems are a much-needed boost for the back-end equipment provider. After buying the Netherlands-based startup Liteq in the summer of 2017, K&S had to wait four years for the first deliveries. Quite long, given the initially high expectations.
Prior to the sale to Kulicke & Soffa, Liteq gained significantly in credibility when litho gorilla ASML invested and became a shareholder. A few months before the acquisition, former Liteq CEO Gerrit van de Beek set his goal at “twenty systems by 2020” (link in Dutch). In Bits&Chips, he called this “ambitious but realistic.” At 2-3 million euros each, this would have meant fifty million euros in sales by 2020.
Penetrating the market proved more difficult, but K&S CEO Fusen Chen said last year that his company remains committed to a position in back-end litho. After the back-end specialist had waved off the first system for an Asian customer last summer, he expressed expectations during the discussion of quarterly results last November that more systems were to follow, driven by the increasing complexity in chip assembly.
Better cost of ownership
K&S, which has its origins in the US and is now based in Singapore, is a market leader in wire bonders, used by semiconductor manufacturers to connect the contact pads on chips wire by wire to the outside world. The company has been competing for leadership in this market with ASM Pacific Technology for decades. But innovation in stitching chips is running out of gas. Complex ICs have so many connections to the outside world that stitching wires is no longer manageable. The tipping point is approximately the 28nm node. The simplest dies of this generation can still be wired; newer generations require techniques such as fan-out and interposer.
It’s quite astonishing that there’s still so much room in the market for wire bonders. But the investments are flowing to the more state-of-the-art segments, where market leaders like Intel, Samsung and TSMC are starting to embrace advanced back-end processes. After following the ‘More Moore’ path of shrinkage on chips, they’re building the infrastructure needed for their ‘More than Moore’ products where systems of chips from different nodes get assembled in advanced packages.
Intel already has decades of experience in advanced microprocessor packages; Samsung and TSMC now see systems of chips as a natural extension of their core business. All three have the deep pockets needed for the billions of investments in advanced back-end fabs – also called “mid-end” to distinguish them from the traditional packaging market. Just like in front-end manufacturing, the name of the game there is high-precision processes in increasingly clean production environments.
Not surprisingly, K&S saw in Liteq a great opportunity to enter the market for advanced back-end interconnect technologies. One of the techniques used there is fan-out wafer-level packaging, in which contacts fan out through a lithography-engineered network across a substrate laid out above or below the chip. Embedding the silicon in a mold and applying structured metal layers to it creates a larger surface area for the solder balls to connect between the silicon chip and the PCB.
Companies active in such processes all develop their own recipes and packages. Standardization is almost non-existent in back-end semicon markets. That’s also why the litho equipment arena is still open.
As a startup, Liteq used a smart strategy by focussing on the light path in developing its stepper, buying as many off-the-shelf components as possible. By choosing a 355-nm Nd:YAG laser source instead of the g, h and i-line from mercury lamps that are common in the back-end market, the Eindhoven-based company was able to design a much cheaper aerial illuminator and lens. This resulted in a significantly better cost of ownership for customers. K&S says the cost of ownership for its Liteq machines is still only half that of the i-line steppers from market leader Veeco (formerly Ultratech).
Apparently, it has been extremely difficult for K&S to compete with Veeco’s installed base at OSAT (outsourced semiconductor assembly and test) companies such as American Amkor and Taiwanese ASE. But there’s a new challenge. Increasing investments in the emerging market for modules that package multiple chips have attracted several new players, including the world’s biggest semiconductor manufacturers.
Notable is the success of Onto Innovation, the company formed over two years ago as a joint venture between Nanometrics and Rudolph Technologies. Rudolph had entered the back-end litho market in late 2012, and its Jetstep system is now achieving great success. Last January, Onto revealed that it had received just under 100 million dollars in orders for its Jetstep X500 litho systems to be delivered this and next year. Customers will deploy the steppers for back-end processes on large square sheets.
With this panel-level integration, tiles as large as 500 by 500 millimeters are exposed in a lithographic process. With the Jetstep’s 250-by-250-mm exposure fields, four exposures will do. After that, the panel is put through various chemical processes. The larger exposure field and form factor should bring cost benefits up to 30 percent.
Unlike wafer-level processes, panel-level integration isn’t yet mature. The final yield in panel-level factory lines will show the extent to which this strategy is successful. However, the orders for Onto are a sign that panel-level integration is gaining traction.
Canon also seems to have its eyes on this market. Last year, it launched the FPA-8000iW, a semiconductor lithography system designed to support manufacturing using large panels. This i-line stepper offers a wide exposure field, like Onto’s Jetstep, but with a pattern resolution of just 1 micron, much better than Onto’s 3 microns. The FPA-8000iW was designed to meet the needs of manufacturers aiming for high production efficiency through the use of 515-by-510-mm organic panel substrates.
K&S Liteq hasn’t yet announced any activities in panel-level integration. At the moment, it seems to be focusing on flexibility, accuracy and processing different wafers. For I3 Microsystems, it adapted its stepper to expose 100 and 150-mm wafers in addition to 200 and 300-mm wafers. The Liteq systems are also capable of flattening and exposing highly curved (warped) wafers. The machine delivered to I3’s R&D facilities can achieve a focus depth of 16 micrometers.
The big question will be: how much room is there in the back-end litho market? How many players will be able to make enough progress to keep investing? Will we see drop-outs and consolidation anytime soon? If the current laws in back-end equipment are any clue, we can look forward to a long and bloody battle.
Main picture credit: Syafiq Adnan/Shutterstock.com