The shape of things to come
But still, the image of atoms, molecules or even macroscopic objects such as individual die spontaneously self-assembling into a useful structure is almost magical. Now some engineers are attempting to master such magic. They are developing techniques that require little more than a solder-bumped substrate, several thousand die immersed in a water-filled beaker, and a few minutes of agitation, to produce an assembly of die that have adhered, aligned themselves and formed electrical contact to the solder bumps. The successful implementation of such an intrinsically parallel (and therefore quite fast) process has the potential to eliminate the need for serially driven pick-and-place and wire bonding techniques.
Researchers at the University of Minnesota have recently detailed such a process in which 1600 Si blocks (300 x 300 x 400 µm), each representing a Si die, and 113 LED die (280 x 280 x 200 µm) were self-assembled onto cylindrical membranes (Jacobs et al.). The figure illustrates the detailed steps in the self-assembly process of LEDs onto a cylindrical membrane (suitable for cylindrical displays). Each individual LED die had two gold contacts, a small circular cathode contact on the front of the die and a large square anode contact covering the back. The target substrate consisted of an array of 280 µm copper squares that were coated with a low melting point (50 °C) solder. The self-assembly of the LEDs onto the solder-coated squares occurs by placing both the cylindrical substrate and the LED die into a container of water that is made slightly acidic (pH 3.0) in order to remove the oxide from the surface of the solder. This is heated to 90 °C to make the solder molten.
After one to two minutes of agitation, each solder point (referred to as a receptor) has a die attached to it, where self-alignment of the die to the center of the solder bump automatically takes place to minimize the overall interfacial energy of the junction. Furthermore, to ensure that only the larger anode contact mates with the receptor, the intensity of the agitation and the temperature of the water are optimized, so that if a temporary contact is made with the smaller cathode contact the resulting interface is unstable and breaks.
The second step of the self-assembly process was to form the top contacts to the LED die. This was performed by dip coating both the copper wires (supported in a transparent polyimide film) that were used as the cathode contacts and the cathode contacts themselves on the top surface of the LED die. In this prototype demonstration, the copper wire grid was placed by hand in roughly the right position on the LED array, and then the entire structure heated above the solder melting point. With the same process that centered the LED die on the solder bumps, the copper wire grid then realigned itself to self-locate to the center of each cathode. The total time required for all 113 die to self-assemble was less than four minutes.
In a more aggressive experiment, a 5 cm2 polyimide substrate containing a 40 x 40 array of solder bumps (for a total of 1600 receptors) was used along with 5000 Si blocks in the mixing container. After three minutes of agitation, self-assembly was achieved on 98% of the receptor sites. Analysis of the resulting 2% defects indicated that they were not caused by bonding errors between the Si blocks and the solder receptors, but by defects in the Si blocks and solder receptors themselves. These defects were typically due to the receptors only partially being covered by solder, or geometrical defects such as Si blocks that were rounded rather than square.
One of the most important aspects of these results is the ability of the receptor to only contact to the larger area anodes by utilizing its greater adhesive force, compared with the smaller binding energies possessed by the smaller cathode. This is in essence a size filtering process that chooses the desired die side for self-assembly. Further progress in this area will undoubtedly come about by not only tailoring the size of the contacts being used, and controlling the agitation and temperature, but also by the shape of the receptor and die contact.
While most of us consider only the inorganic aspects of chemistry (bond strengths and 2D crystal growth), the reactions that take place in organic chemistry rely on the ability of one molecule to geometrically lock in place to another, just like puzzle pieces coming together. When this type of behavior is exploited in the development of self-assembly, a wide range of different die shapes and sizes should be able to self-assemble over substrates with very complex topographies. This potentially opens up a whole new methodology to complement, or even replace, the serially driven pick-and-place approach that currently dominates packing. Because so much of the III-V community is now involved in the move toward multichip modules, such as handset PA modules and the integration of electronic and optoelectronic functions for optical communications, such self-assembly processes may some day become very important in reducing production times, improving yields and lowering costs.
Further reading
H Jacobs et al. 2002 Science 296 323.
