The Chemistry Of Semiconductors

At each new process node, the chemistry of chip manufacturing has become much more complex than at previous nodes. But at 5nm and below, it’s going to get orders of magnitude more complex.

For the first few decades, the chemistry of chips was largely shielded from view for most of the industry. Caustic gases were relatively well understood because they are a potential health hazard, but the slurry for polishing a chip was barely given a second thought by anyone outside of the CMP process. Now, as the critical dimensions on chips is reduced to the point where a single atom can make a difference for whether a signal is corrupted, or whether a logic gate will fail faster than expected, the whole industry is suddenly dependent on some very complex chemistries.

There are a number of areas where this is becoming more apparent and more difficult. One involves the films used for insulating one layer from another. In some cases, these films are less than 50 atoms thick. There often are multiple films being applied at the lower metal layers. Some of those materials need to be removed through heat. Others need to remain in place. And all of it has to be done with incredible precision.

In the past, with thicker films and larger distances between various components, this wasn’t an issue for most companies. But now, as transistor density increases at each new node, the problem isn’t just being able to apply these films accurately. It’s having them withstand the temperatures in various manufacturing processes and to either be removed through etching, or to dissolve completely so that not even a nanometer of material remains.

And that’s just one component in a manufacturing process that could include hundreds or thousands of steps. Polishing wafers once was as simple as grinding them flat. As wafers and dies are thinned out, and as they are bonded to various substrates and put into complex packages, sometimes these wafers warp. Still, it’s not necessarily the warpage that’s the problem. It’s the ability to polish the wafers down to the nanometer level with a complex material in which certain particles are perfectly distributed.

It doesn’t get much easier with the photoresist material, which typically is a liquid with particles suspended in very precisely manufactured goop. Most of that resist material is wasted. In fact, one of the major efforts underway is to limit the amount of material that is wasted during the litho process. But the real challenge is distribution of different particles within that goop, and the ability to remove all of it when necessary. A nanometer-sized particle that remains on the chip can cause havoc over its lifetime, which in the case of an automotive AI chip could be decades.

Add to that list the substrate. In the early days of semiconductor manufacturing, the substrate was little more than an inexpensive platform that largely didn’t interact with the electrical signals passing through the wires on a chip, other than to dissipate heat generated by current leakage. Today, tolerances are so tight that an imperfection in the substrate can create noise that can disrupt those signals. There are more materials, including SiC and GaN, as well as silicon on insulator and various other hybrid combinations. Imperfections at the atomic level can have big repercussions at the signal level.

All of this ultimately is physics, but chemistry is one of the best ways to limit the impact of physics. As the physics become more demanding, the chemistry likewise becomes more challenging. This has largely happened out of sight for most of the semiconductor industry in the past, but at each new node from here, chemistry is going to play a much more prominent and visible role — even if it’s at a scale that no one can actually see.