Quick release for flip-chips

17 Jun 2002

A laser-driven release device that was originally developed to assemble microelectromechanical system components has been adapted for transferring copper bond-pads onto a silicon flip-chip. Michael Hatcher discovers that it could soon be in use on the manufacturing floor.

From Opto & Laser Europe January/February 2001

The use of lead solder in electronic products is destined to become outlawed soon, forcing manufacturers to find an alternative.

Pioneering work by scientists at Imperial College, London, UK, has led to a collaboration with the electronics manufacturing giant Celestica of Canada that could solve the problems that are posed by the impending legislation.

Current European plans call for a lead ban in "most electronics" from 2004. However, Celestica's Phil Hamilton, who collaborated on the project with Andrew Holmes from Imperial College, says that with the legislation likely to go through in Japan this year, there is an overwhelming need for electronics manufacturers that have a global customer base to make the change sooner rather than later.

Hamilton is excited about the prospects for the new laser technique. He says that the company plans to develop it "all the way" to mass production. "We could see this laser process on the production line in 12 to 15 months' time," he said.

The new technique is an adaptation of earlier work on laser-assisted assembly of microelectromechanical systems (MEMS) by Holmes, Sabri Saidam and Changhai Wang at Imperial College.

The advent of MEMS in recent years appears to be just the beginning of a microengineering revolution. However, owing to the MEMS' small dimensions, conventional engineering and fabrication methods do not always scale down and manual handling can damage the devices. Consequently, an aim of current MEMS research is to develop a microtool set that can be used to produce a range of microdevices.

Thanks to the efforts of companies such as Exitech in the UK, some optical techniques have already reached a degree of maturity in the field. These include excimer laser micromachining and ultraviolet lithography and galvanoforming processing.

One of the remaining challenges is to find non-damaging assembly methods and to integrate these with fabrication schemes.

Holmes and his colleagues developed a laser-driven microassembly system that is based on laser ablation using a Lumonics KrF excimer source.

In this technique the various parts of a MEMS device are fabricated. Some of these are assembled on the final substrate and others are mounted on top of a sacrificial polyimide material, which, in turn, is mounted on an ultraviolet-transparent carrier. The components are aligned and an excimer laser pulse ablates the polyimide, which expands and cools. The pulse fires the component towards the final substrate and the device is assembled.

Holmes experimented with a number of simple arrays of 2 x 2 mm nickel pads, the height of which was between 50 and 200 µm. He found that increasing the pulse energy above the ablation threshold generated a similar increase in the initial velocity of the nickel pads.

It is crucial to minimize the kinetic energy of the released parts to reduce the chance of damaging the components, says Holmes. "The kinetic energy rises steeply with fluence above the ablation threshold. You have to work some way above this threshold to allow for pulse-to-pulse fluence variations. A longer-wavelength laser could help here, because it gives less momentum to the released parts."

Other approaches include using a thinner release layer to reduce the volume of ablated material and to structure the carrier/release system to relieve some of the ablation pressure that is generated.

The maximum size of the arrays that are transferred successfully is limited by the laser-pulse energy. Holmes calculates that a 1 J pulse could release structures over an area of 10 cm². This capability gives the technique an advantage over traditional assembly methods, which are time-dependent on the size of the released structure. Manufacturing speeds and yields should increase using this system.

Holmes extended the method to build microelectromechanical wobble motors. These devices consist of four layers of nickel components, including a circular array of eight electrodes, a spoked wheel, an actuator and a stator. When switched on, the wheel flips between the contacts.

Arrays of the component parts were placed on 3-inch silica wafers, each one containing 36 rotors on a square grid with a 6 mm pitch. Limited only by the travel of the stages, 4 x 4 component arrays were assembled and showed no sign of damage.

The method has three key advantages over conventional sacrificial-layer processing systems. First, it enables the step-by-step release of individual components from a single sacrificial layer - whereas wet-chemical processing requires a number of different sacrificial materials.

Second, the technique is faster than wet-chemical release, which is dependent on the size and shape of the transferred structure. Laser release takes a millisecond or less and the transfer time is independent of the geometry of the parts.

Third, there is no undesirable motion of the released components, which is often seen with wet-chemical treatment due to surface tension or fluid transport.Quick to see the potential of the technique in chip making, Celestica part-funded work to adapt the transfer technique for lead-free flip-chip fabrication. In this scheme, copper bumps, which provide the contact points between the chip and the circuit board, are built up on a quartz carrier layer. A sacrificial polyimide layer is sandwiched between the copper and quartz.

However, instead of being transferred through free space like the wobble motor, the copper bumps are bonded to a silicon flip-chip. The excimer laser ablates the polyimide layer, releasing the bumps from the quartz substrate and leaving them attached to the chip. This process is known as bond-and-release assembly.

In trials, sixteen dies arranged in 4 x 4 arrays, each containing 28 copper bumps were transferred simultaneously. Each die was approximately 1 cm² and the 90 µm-diameter bumps had a pitch of 127 µm.

A key advantage of the laser method for Celestica is that all of the bumps are transferred in parallel with a single laser burst, whereas current bumping methods are serial in nature - with the bumps produced by pulling away copper wires sequentially.

Rival systems to laser bumping are also in development, but they require more expensive bumps made of precious metals. A further advantage of the laser technique is that the bumps can be arranged with a finer pitch. Fabrication on low-cost carriers, such as quartz, also reduces the amount of chip processing that is required prior to bonding, offering a potential increase in yield.In trials at Celestica's Kidsgrove plant in the UK, the copper bumps were ripped from the silicon to assess the quality of the bond to the chip. "The average shear strength is good enough for commercial production, although it does vary across the chip. We think that this is due to non-parallelism between the chip and the carrier prior to bonding," said Holmes. His set-up now includes a parallelism monitor to ensure that all of the bumps bond with equal strength. He is about to start trials on bumps that have a pitch of 90 µm.

Holmes thinks that it should be straightforward to implement the laser-bumping process in a commercial environment: "Such a machine would bump individual chips in a single bond-release operation."

A key achievement will be to scale up the process for simultaneous bumping on arrays of dies and wafers. Holmes warns that this could be problematic. "For larger areas there will inevitably be difficulties in achieving uniform bonding." However, Hamilton says that such effects are usually confined to the corner studs of each die and that by using a larger-area wafer these edge effects can be minimized.

As Celestica brings an excimer laser on stream for in-house trials and development, Holmes - having demonstrated how pure research can create a real commercial application - is returning to the MEMS assembly work that first inspired the flip-chip application.