good research projects

=projects =research =startups =projects

 

 

Here are some projects that I think are good choices for research, investment, and/or new commercialization attempts. These are all topics I've thought about, so if you want to work on one of them there's a chance I'd be able to help.

 

 

cheap flow batteries

Aqueous flow batteries are the most plausibly economically viable system for large-scale grid energy storage.

There's a significant amount of flow battery research going on, but the vast majority of it is on systems where the active materials are too expensive or too unstable to even potentially be practical for grid energy storage. There are also various designs with organic solvent where the required solvent is too expensive.

Some examples of potentially cheap flow batteries are:

A) iron-chromium flow batteries with some system for regenerating electrolyte as the ions cross to the wrong side and as hydrogen is slowly generated
B) iron-iron batteries using cheap and stable organic ligands to make the sides have different potentials

 

I particularly like (B). However, most chemists doing lab research at universities are not very good at estimating what the costs of new chemicals would be with large-scale production, and when trying to publish papers, they're incentivized to pretend their systems could be practical, so there's a lot of noise.

 

The directors who make the big decisions and say smart phrases like "emerging technologies such as AI, cloud, and blockchain" think that LiFePO4 will solve grid energy storage, because the price will go down to $50/kWh and they'll last for thousands of cycles. The new battery requirements for new buildings in Californa were probably designed for LiFePO4. However, the electrolyte degrades at the negative electrode of Li-ion batteries, which is the same in LiFePO4, so while the cycle lifetime is relatively high, there's still degradation over time. Also, the BloombergNEF surveyed prices are heavily weighted towards subsidized Chinese batteries that can't be purchased at those prices outside China.

 

 

GaN

Gallium nitride is the future of power electronics. It's better than silicon carbide and already somewhat competitive. That's despite GaN crystals currently being very expensive to make.

Silicon can be melted and crystallized, but GaN decomposes (making nitrogen gas) before melting at any reasonable pressure. There are a few different ways of growing GaN crystals: the main current approaches are physical vapor deposition (PVD), ammonothermal, and sodium flux. Usable PVD crystals are narrower than their seeds because they're bad around the edges, and the other 2 approaches are extremely slow, so my understanding is that some combination is used now.

 

 

replacing carbon black

Release of carbon black from tires as they wear away on roads is a significant source of particulate pollution. Calcium carbonate particles are much less hazardous, and it's possible to get the same performance as carbon black (or even slightly better) if you precipitate it in particular ways.

 

 

methanol fermentation

Metabolic engineering of microorganisms for synthetic methylotrophy is very possible. This is more difficult than using sugar, so it's only really worthwhile when the lower cost per energy of methanol vs sugar is significant, but I still expect it to be a foundational piece of future chemical production. Even if you want renewable chemicals, it's cheaper to gasify biomass and make methanol than to use sugar.

 

 

electron beam welding

Electron beam welding is the best type of welding, but requires a vacuum chamber, which makes things much more expensive but is still done industrially. With plasma windows, it's possible to do it without a vacuum chamber. So, I think plasma window electron beam welding should be developed and used commercially.

 

 

truck platooning

Fully self-driving vehicles are a difficult problem. A much easier problem is making automated vehicles closely follow a leader vehicle driven by a human.

It's also possible to reduce aerodynamic drag by having several trucks follow closely behind each other.

 

 

remote driving

Truck driving is an unpleasant job overall, but people pay money to play Euro Truck Simulator. With wireless internet now widely available on highways, it might be feasible to have remotely driven vehicles. This is already done for large mining trucks.

The main issues I'd be concerned with are dropped connections and accident rates. The crash rate for large UAVs is certainly much higher than that of piloted aircraft. Like automated driving, remote control is less of an issue on well-defined and segregated areas. A combination of automated driving with remote control would probably have better performance than just automated driving.

 

 

package loading

Loading and unloading things in trucks is a lot of work, and that work is sufficiently unpleasant to be a worthy target for automation. There are various companies working on doing that automatically, such as AncraSystems and Honeywell. Current systems are expensive or problematic, but I expect automated loading and unloading of boxes from trucks at warehouses to become standard.

 

 

reference counting

Reference counting as a garbage collection system for programs has long been considered to have much worse performance than mark-and-sweep generational collectors, but some relatively recent variations have shown similar performance and could potentially be superior for parallel systems. I wrote a bit about this.

 

 

hiring systems

Getting hired at a large company these days typically involves the following steps:

- A hiring manager sends a request to HR.
- HR makes a posting for a job opening.
- Someone searches onlike and sees the posting.
- Hundreds of people submit a resume using some online form system.
- HR reviews the resume to see if it matches the posting.
- A small % of submitted resumes go to the hiring manager.
- A small % of those lead to a phone interview.
- Someone (maybe) gets hired.

 

This system does not work very well. People sending out hundreds of resumes to get a job is not good. HR has no domain knowledge and can't write postings or evaluate resumes very well. Surely it's possible to design a system that works better.

Perhaps a starting point would be to find some companies with unusual hiring methods and do studies evaluating how well they work.


 

 


back to index