Magnetic Assembly: A New Option for Delicate 3D Integration?
Pick-and-place can damage sensitive parts and contactless methods are inefficient for high volumes. What about magnetic assembly?
Researchers are continually working around Moore's law—recently by stacking transistors vertically to increase transistor count while keeping the physical dimensions of the chip relatively the same.
But this method becomes difficult when designers are working with sensitive parts (for instance, miniaturized sensors) that can be easily damaged with pick-and-place methods. By contrast, current no-contact methods can be difficult to control for high-volume manufacturing.
A new research study shows designers how they might sidestep these pitfalls in vertical semiconductor construction—by using magnetic fields.
Researchers at the KTH Royal Institute of Technology in Sweden have discovered a potential solution for "going 3D." These engineers have successfully created 3D silicon devices without the need for growing epitaxial layers on top of silicon devices.
The solution is a multi-step design that uses magnetic fields to place chiplets and edge wire bonding to connect the chiplets to the main substrate.
A T-Like 3D Structure
The 3D structure uses small silicon chips that have a T-like design. The long section of the T has aluminum electrodes, which have trapezium cut-outs for each electrode. The bottom section of the T has the main silicon circuitry using standard 2D fabrication techniques. The backside of the chiplet is coated in strips of ferromagnetic material (nickel).
The chips have a T-shaped design, preventing upside-down entry into the holes. Image used courtesy of Federico Ribet, et al. and Nature [CC BY 4.0]
The main silicon substrate (connected to the chiplets) has cut-outs as wide as the narrow section of the chiplets and the bonding pads. The bonding pads are located on the substrate next to the cut-outs. This means that when the chiplet is inserted into the cut-out, it can fall to the top of the T section acting as a plug.
Step One: Use Magnetism to Orient Chiplets
Getting chiplets into the slots typically requires highly-intricate tweezers, which can be a costly and time-consuming process. This is why the researchers utilized the strips of ferromagnetic material on the back of the chiplets.
Diagram of magnetic assembly. Image used courtesy of Federico Ribet, et al. and Nature [CC BY 4.0]
They then used an external magnetic field to not only orient the chiplets but also to move them around the substrate and get the chiplets to sit in the cut-outs.
Step Two: Secure Chiplets Into Cutouts
The next stage of the construction is to secure the chiplets into the cutouts. The researchers did this by using standard off-the-shelf superglue. However, other techniques could be used to hold these chiplets. For instance, designers could rely on the bond wires between the chiplet and substrate (more on this later). If the superglue needs to be removed, acetone will do the job.
The top two images depict vertical lifting microchips in the presence of a magnetic field because of striped nickel coating on the backside. The bottom two images depict how the lifting direction of the microchips was controlled by striping of the nickel coating. Image used courtesy of Federico Ribet, et al. and Nature [CC BY 4.0]
The researchers also noted that in a mass production environment, designers can automate the process by using miniature pumps and stepper systems to apply superglue.
Step Three: Bond Chiplets to the Main Substrate
The last stage involves edge wire bonding the chiplets to the main substrate. The researchers made a bond using a standard wire bonder that is typically used for bonding dies to IC packages.
However, using a trapezium design entailed a few things: 1) the bond must be made at room temperature with low contact resistance (less than 0.2 ohms) and 2) forming the bond will not require ultrasonic vibration.
Edge wire bonding. Image used courtesy of Federico Ribet, et al. and Nature [CC BY 4.0]
Deforming the bond wire in the contact will keep it in place and form a larger contact area.
The construction method shown here demonstrates that 3D silicon structures built using traditional methods is possible. Using magnetism to move structures is (at least in some ways) a practical option in contrast to using micro tweezers.
The demonstrated parts constructed were thin-film resistors, which would typically be too large and costly to integrate into a standard die, such as a microprocessor. But this method could be applied for a number of components, including power transistors, capacitors, and inductors.
One problem with this method is that using ferromagnetic materials on silicon dies could potentially affect magnetically-sensitive circuitry. However, the presence of such material may be easily overcome by using offset values in magnetometers.