Links #14

Make new waterfalls, Haber-Bosch is here to stay, and more.

1.

Tourist discovers waterfall in China supplied through pipes. When I see this, I think “hell yeah”.

Desalination and solar can make water extremely cheap, why shouldn’t we juice up our waterfalls? Heck, why not add some new waterfalls while we’re at it? And geysers. And springs. And creeks. Cheap water means that we can create lush, managed ecosystems anywhere in the world.

By adding water, we can restore habitats to the way they were before climate change. But I think people should drop the pretense and make these ecosystems even more beautiful than they were before.

2.

Orca Sciences has a lot of interesting articles related to electrifying chemical and materials industries. If you only read one, check out Checking some prejudices on materials decarbonization.

Their key question is: what chemicals and materials production processes will be changed by cheap renewables? The old way of making stuff is to use high heat and pressure (thermochemistry) to make stuff. Think of the Haber-Bosch process which requires high heat and pressure to make ammonia, or chemical cracking which breaks down heavy hydrocarbons. The new and sexy way to make stuff is with *electrochemistry*.

I’ve talked a lot about using cheap solar electrons to make stuff, but when is it viable? In some cases, we will have to stick with the old processes and produce the heat and pressure with renewables.

This chart is a great way to visualize the different chemicals in our supply chain. The vertical axis is the value of each chemical divided by how much energy it takes to produce. The red line is the typical price of electricity at $100/MWh. If electricity is more expensive than your chemical, you shouldn’t use electricity to produce it!

The horizontal axis is the minimum voltage required for each reaction. The black line shows the equivalent “voltage” of reduction with carbon at 1000 C. Carbon sourced from coal is used to reduce all sorts of metals today.

The dots on the upper right of the chart need to be made with electricity, carbon won’t cut it and they already produce more value than the electricity they use. Magnesium is interesting here, it can be produced from seawater and has excellent properties, but so far it hasn’t been economical to produce. Cheap solar can change that!

The bottom right is the opposite: they’re too cheap to justify using electricity for their production. This is a bit disappointing, because some of the most important chemicals are down there. Ammonia feeds the world, methane powers industry, and ethylene is the basis of almost all plastics. It seems like solar can find use in these processes as a cheap energy source, but it’s not going to revolutionize them.

I just have one nitpick here, I think in the future the red line will be one tick lower, at $10/MWh (at least if you’re generating solar on-site) which could make a difference for some of these chemicals. Methane would still be below this line, presenting a challenge to companies like Terraform Industries.

There’s much more in the piece, such as discussing when to use biology versus synthetic chemistry.

3.

I want to zoom in on ammonia for a second. In past posts, I’ve often highlighted new ways to make ammonia using renewables. But from the previous section, it looks like it’s too cheap to justify leaving the Haber-Bosch process.

This is reinforced by a piece from Orca on Fertilizer from air plasma. They find that it’s probably not cost effective relative to Haber-Bosch (HB). Worse, this paper finds that using metals like lithium or calcium in an electrochemical process would use 10 times more energy than traditional HB, other metals could be more energy efficient but you probably have to heat them up a lot for it to work. At that point, have we really improved on HB?

All of this has pushed me to the realization that we should stick with HB and just get our hydrogen, heat, and pressure from renewables. That sort of looks like what the company Ammobia is doing, and they’re focusing on lowering capital costs so that the intermittency of renewables isn’t a problem.

That being said, I’m not going to stop talking about cool ways to make ammonia, like this Microwave catalytic synthesis of ammonia from methane and nitrogen (H/T ToughSF). Microwaves are somewhat interesting because we might be able to convert sunlight into microwaves using photonics, saving some of the energy losses from photovoltaics. But don’t bet on beating HB.

Everything else

I’ve bolded the things I think are particularly good.

Some other highlights from Orca:

Beamed power for electric aviation could potentially make eVTOL more viable. This would bring energy costs to near zero and make flying ubiquitous. Though there are infrastructure and legal challenges to deal with first.

Orca supports the company Savor making fats from carbon in the air.

Food without agriculture: Food from CO₂, biomass and hydrocarbons to secure humanity's food supply against global catastrophe

What I Won’t Eat. This pretty much accords with my thinking on the matter.

Also by Georgia Ray: This list of Broad-spectrum biotechnologies and their implications for synthetic biosecurity from 2016 highlights RNA vaccines and gene drives far before they became more well-known. I had never heard of DRACO before, but it seems like an important technology to develop.

Mosquito-Borne Disease: Environmental Justice, Geopolitics, and Biotechnology. A nice summary of why we should fight mosquito-borne diseases and promising technologies for that. Because of the evolutionary environment of insect-borne diseases, they tend to be much nastier than other pathogens. Since mosquitos are so deadly and not particularly critical to their ecosystems, I think there’s a case for targeted eradication of the species that infect humans.

Scientists Eliminate Cancer Using Evolutionary Principles. Very cool, attacks tumors by adding genetically-modified cancer cells resistant to a certain therapy. Then they apply the therapy, leaving only the modified cells and naturally resistant cancer cells. Then the modified cells burst, killing themselves and their neighbors. Feels like nanotechnology out of Stephenson’s Diamond Age. I wonder if other virulence management techniques would be useful in fighting cancer.

M2 Automation’s microdispensing systems are impressive. Feels like a basic capability needed to automate lots of lab work. Could something like this revive Theranos’ failed dream to do many tests on a single drop of blood?

Open-TeleVision is a robot teleoperation system uses only an Apple Vision Pro headset. How long before we start collecting a bunch of training data from people going about their daily lives wearing a headset like this? Oh wait, that’s already happening.

A New Way of Creating Space. Peter Hague profiles a the company Skyeports that is building a machine to make glass spheres out of lunar regolith for habitable structures. See my post on A glass O'Neill cylinder in LEO for a related proposal.

Wonderful video on the complexity of semiconductor device fabrication.

A Trustworthy, Free (Libre), Linux Capable, Self-Hosting 64bit RISC-V Computer. This feels like the path to getting much more secure hardware. It won’t really replace your PC, but using systems like this for security-critical applications like banking could be valuable.

I enjoyed the Distilling Singular Learning Theory series, particularly parts 1 and 2. SLT is a mathematical theory of how “singular models” (e.g. neural networks) learn in an idealized setting. The main takeaways are things we already knew about neural networks such as “neural networks generalize well even when they have lots of parameters”, “neural network training prefers simpler hypotheses”, but the fact that such a clear model with connections to other fields can show this is interesting. I think there’s a path to bringing the setting much closer to real-world training and using it to inform how we train AI models.