Heliogen Is Bill Gates’ Latest Venture That Is Only Good For Oil & Gas

Bill Gates seems to love to invest in things that aren’t going to make much of a difference to climate change but that are good for the fossil fuel industry. The latest is Heliogen, a company which uses machine learning to make solar ovens hotter and more reliable. The problems are rife with the technology and it’s not competitive except in niches that likely aren’t actually climate friendly. While I’m a big fan of both machine learning for clean technology and solar power, this instance proves that they don’t necessarily produce useful results.

Previously, I’ve gone deep and wide on Carbon Engineering, which from my analysis and its sole contract is only useful for putting on tapped-out oil fields to burn otherwise unmarketable natural gas in order to push CO2 underground to get typically 2-3 times as much CO2’s worth of unrecoverable oil above ground. That company received at least one early round of funding from Gates. He’s also funding experiments in solar geoengineering, something which is good to have in our back pockets, but if used will simply perpetuate the burning of fossil fuels.

Heliogen concentrating solar array

Heliogen concentrating solar array courtesy Heliogen press release

Let’s start with the good stuff. Heliogen actually has improved the performance of concentrated solar power. It uses four cameras pointed down at the array of mirrors to assess the light scatter around their edges to keep them tightly focused on the hotspot over the course of the day. This is a good example of something machine learning makes possible today that wasn’t easily possible a few years ago. This goes into CleanTechnica’s list of interesting uses of neural nets for clean technologies. And I’m sure that the method is useful in actually useful technical spaces too.

And there’s no reason to doubt that it works. Constant hyperfocusing of an array of mirrors to create a parabolic concentrator using this approach would create higher heat. It’s the same thing as a kid finding the focal point with a magnifying glass to fry an ant or make a classmate squeal. The ability to do this automatically and constantly means more rapid startup and a lot less downtime for calibration annually.

But that’s where it stops being particularly exciting too. This technology can create a single very hot spot, one that’s a long way off the ground. And it can only create it when the sun is highest. While it will generate lower levels of heat in the morning or late afternoon, it will only be able to generate its maximum temperatures for a few hours around solar noon on cloudless days.

There are five problems:

  • It’s a single, poorly scalable disk of high temperature
  • Tight coupling of energy production to energy consumption
  • Industrial facilities cast big shadows
  • Very high heat doesn’t store or transmit easily
  • We have solutions already that are much easier to integrate into industrial plants

The first two are chunky enough by themselves, so will be covered in this article. The remaining three will be covered in Part 2. Also in Part 2, we’ll define where the characteristics of the technology would actually be useful and identify what is likely to be its only potentially viable use case.

It’s a single, poorly scalable disk of high temperature

The released prototype image with no supporting additional material shows a tower with a concentrated point of high temperature roughly 55 feet or 17 meters off the ground. The mirror array is in the range of 16,000 square feet or 1,500 square meters meters to achieve about 1,800 degrees Fahrenheit or 1000 degrees Celsius. That point of high temperature from images appears to be in the range of a foot in diameter or about 30 centimeters across. So we have a single, relatively small spot of high temperature a few dozen feet off of the ground. The current disk is in the range of about 0.4 square feet or 0.03 square meters.

That means that Heliogen has achieved a very high temperatures but not sustained heat throughout a large volume. And it’s heat, not temperature, that’s the requirement for industrial scale processes.

Want to double the area of the disk and hence the amount of heat available? That requires doubling the area of mirrors as well. That would give you a disk of high temperature about a foot and half or 45 centimeters across. It would require roughly 32,000 square feet or 3,000 square meters of mirrors.

The Ivanpah concentrated solar power facility, for example, had a capacity just under 400 MW with an area of mirrors roughly 3,500 acres or 1,400 hectares. That’s 5.4 square miles or 14 square kilometers, an area about a quarter the size of Manhattan. Its annual production in 2018 was around 700 GWh. And let’s be clear, Heliogen isn’t magic. The amount of solar energy that it can concentrate hasn’t changed and isn’t higher than what Ivanpah achieved, the technology can just focus it on a smaller spot more consistently and with less calibration downtime. Where Ivanpah achieved about 85% of its design parameters in 2018, Heliogen is easily closer to 100% based on the technology it is employing. That’s useful, if this is useful technology at all.

Let’s do some comparisons. The company talks about a few use cases, so let’s look at examples of these pieces of equipment to get a sense of the challenge here. Let’s start with cement.

One of the things that mentioned in various places is decalcination of limestone into quicklime. That does require a lot of heat and is responsible for about half of cement’s emissions, which have estimates of CO2e emissions ranging from 5% to 12% of global CO2 annually, depending on what you look at.

You have to bring large masses to high temperatures, not just one part of them. Modern lime kilns require about 6–8 million Btu per ton of quicklime produced. A Btu is the heat required to raise the temperature of one pound of water one degree Fahrenheit. Alternatively, that’s about 6.3 to 8.4 gigajoules. To put it into more familiar terms, that’s about 1.8 to 2.3 MWh per ton of quicklime.

Lime kilns for large scale production of quicklime are rotary drums up to 13.5 ft or 4 meters in diameter and 400 feet or 122 meters long.  Ignition of natural gas occurs inside the body of the rotating drum with high heat flames roughly 3 times the length of the diameter of the interior of the kiln, so in the case of the bigger kilns, that’s a jet of flame roughly 40 feet or 12 meters in length and about half the diameter of the kiln in width. The volume of the flame is roughly 1,300 cubic feet or 38 cubic meters. The interior of the kiln is maintained in the range of 600 degrees F or 315 degrees C for the entire day. Product takes 1-4 hours to transit the drum. Kilns at this scale can produce about 450 tons per day, so that’s the equivalent of 810 to 1035 MWh of electricity required.

So Heliogen can create a 1 foot or 20 cm disk of very high temperature for several hours, but processing cement at industrial scales requires a jet of flame 7 feet across and 40 feet long running all day long.

Let’s look at the energy requirements for quicklime vs the Ivanpah example, which produced sufficient heat for this process regardless, about 550 degrees Celsius in the receiver. Ivanpah was running about 29% boiler efficiency turning heat into useful electricity. If this could be made to work, the heat would be used more directly and efficiently in decalcination, so let’s double that to give this the benefit of the doubt. For round numbers, we’ll assume 400 MW capacity for Ivanpah operating over 8 hours, giving about 3,200 MWh of electricity. Doubling that for useful heat for the quicklime process gives about 6,400 MWh of heat.

That suggests that the Ivanpah facility of 5.4 square miles would be producing sufficient heat to produce about 6-8 times as much quicklime as a single plant that takes a tenth of an acre. Turning that around, a single lime kiln powered by Ivanpah technology would take up about 0.8 square miles, about 5,000 times as much space. Assuming higher efficiencies from calibration, maybe Heliogen’s approach would only take up 4,000 times as much space.

So if concentrated solar power was useful for cement, then Ivanpah’s technology produced more than high enough temperatures, so there’s no specific advantage for Heliogen there. And the area required to generate the necessary heat is vastly larger than today for scaled industrial processes.

There’s a big mismatch here, and that’s problem one. The same math applies to pretty much all the rest of the industrial processes that the company talks about. We’ll talk about steel and hydrogen in another section.

Tight vs loose coupling

So what about tight coupling? Well, when I write about the future of energy being electricity, one of the points I make is that loose coupling is key to successful technical innovations. What that means is that the more you can allow major components to operate independently, the better the overall solution is, all else being equal.

There’s nothing in Bill Gross’, Heliogen’s founder, background that suggests he would have any particular reason to have this insight, but it’s straight out of software engineering, and Bill Gates should know this one inside and out, even if object- and component-oriented design came after his hands-on days. And Heliogen is a software-enabled innovation. It undoubtedly has people on staff who internalized this principal a couple of decades ago just as I did.

What this means for energy generation and industrial processes is that requiring them to be put together is tight coupling, and severely constrains the solution.

Consider an alternative. Instead of using concentrated solar, imagine you have a grid of electricity powered by renewables with a bunch of storage spread around the continent, and the lime kiln uses electricity to power its operations. That’s loose coupling, and it’s very effective. You don’t care what creates the electrons, you just care that you have MWhs of them coming in through the wires to your tenth of an acre facility. Instead of buying or leasing 4,000 times the land for partial power through the day, you have your tenth of an acre facility with wires coming into it and quicklime flowing out 24/7 if you want to.

There’s another point here. Transporting tons of durable goods is expensive, but transporting electrons is cheap and efficient. Land costs for a 4,000x area cement plan eat into profitability a lot more than the cost of electricity would. After all, a MWh for industrial users can cost as little as $45 USD, as it does for major customers in BC. That covers a lot when the alternative is most of a square mile of land somewhere.

For the solar solution for cement to be viable, it has to be in the middle of nowhere for land prices not to significantly change economic costs and that means you are transporting a lot more physical material a lot further at a cost per ton mile which starts to get excessive.

Heliogen claims that roughly half of all industrial sites globally have sufficient room for their solution, but that’s a gross overstatement as far as I can tell. If they need a fairly small amount of high quality heat for a few hours a day, that might be true, but industrial processes need not just high temperatures but lots of heat. They are too tightly coupled for this to be useful for most existing sites and it doesn’t seem that reasonable for economically viable future sites compared to just using electricity, with a bit more on that later.

That’s problem number 2, tightly coupling the primary source of energy to the industrial process instead of loosely coupling them to allow lots more sources of energy to provide the facility as needed.

In Part 2 we’ll cover the other the remaining three issues, the characteristics of a use case where this would be useful and the sole use case I can identify.

Bill Gates seems to love to invest in things that aren’t going to make much of a difference to climate change but that are good for the fossil fuel industry. The latest is Heliogen, a company which uses machine learning to make solar ovens hotter and more reliable. The problems are rife with the technology and it’s not competitive except in niches that likely aren’t actually climate friendly.

Bill Gates

This is the second of two articles assessing this technology. It introduced five problems with the technology as an industrial component:

  • It’s a single, poorly scalable disk of high temperature
  • Tight coupling of energy production to energy consumption
  • Industrial facilities cast big shadows
  • Very high heat doesn’t store or transmit easily
  • We have solutions already that are much easier to integrate into industrial plants

The first article covered the first two points, showing that tying this to, for example, a single reasonably sized cement plant would require roughly 4,000 times the space, and that decoupling energy creation from demand would provide substantially more flexibility and higher value. Now we’ll step through the remaining three problems.

Industrial facilities cast big shadows

You’ll note from the Heliogen demo site that the tower is casting shadows across the array of mirrors. As noted, it’s about 55 feet tall based on the image, not an unusual height for industrial facilities for those of us who have been near them.

But industrial facilities are a lot more solid and wider than the Heliogen tower. They are going to cast a much bigger shadow. That requires the concentrating mirrors to be further away from the facility or a lot more loss of useful generation of heat during hours of the day when the sun is lower in the sky, which is to say further away from noonday summer sunlight. Basically it just adds to the engineering and land use costs to find some niche where this is actually as opposed to hypothetically useful.

This doesn’t particularly get better as concentrating solar power scales, as the concentration point typically rises further and further above the ground, both increasing shadows from infrastructure and requiring the heat to be moved further with attendant quality degradation. Ivanpah’s tower, for example, is about 460 feet or 144 meters in height.

That’s problem number 3, that the shadows diminish further the economically viable use cases. It’s not insurmountable, just drives this further into niche categories.

Very high heat doesn’t store or transmit easily

So this is the next problem. As stated, maximum temperatures are only going to be achievable for a portion of sunlight hours, with diminishing heat available in shoulder hours. There are certainly stores for heat, but the higher the temperatures, the greater the engineering challenges.

When we get up into the 1,000-1,500 degrees Celsius range, normal pumps and pipes just get soft and deform, which is a significant limitation. Only recently was a low volume 1,400 degree ceramic pump demonstrated in a lab that didn’t shatter due to its brittle nature. The machine was incredibly high tolerance, hasn’t been proven to scale and screams ‘expensive’.

And as soon as you drop the temperature down to the range that metals are comfortable with, you lose a lot of the benefits, and you certainly lose the 1,000 degree + mark. Molten salts run from 150 degrees C to 600 degrees. It’s possible to combine them to get higher temperatures, but then you start running into the same challenges.

Heat 55 feet off the ground and on one side only isn’t that useful to most industrial processes. Typically heat is useful below and in the middle of industrial plants. That means you have to transmit it for it to be useful. That disk of extremely hot temperature isn’t directly useful in most applications, so it’s going to be heating up some heat exchange medium which will transmit the heat downward and inward in a facility, losing its temperature along the way.

And heat is only useful on the site where it’s generated for a relatively short period of time before it degrades further below the level of usefulness. Heat escapes and can’t be transmitted long distances.

By comparison, it’s trivial to deliver electricity from hundreds of miles away and have it power heat sources in directly the location that they are required. In fact, it’s probably more efficient than transmitting heat 100 meters end to end.

So that’s problem number 4 with Heliogen’s hypothetical industrial component. It doesn’t put heat where it’s useful, the very high temperature is very hard to transmit even a few dozen meters to where it is useful, and it’s currently incredibly difficult to store anything at that high a temperature. It’s certainly possible to engineer an industrial solution around this, but it’s hard to imagine how without increasing the shadow problem mentioned earlier. Basically, every industrial plant would have to be designed from the ground up to work with Heliogen technology, or they could just bolt in electrical heat sources in convenient locations.

We have solutions already that are much easier to integrate into industrial plants

Is Heliogen uniquely capable of providing high temperatures? No, not at all. Obviously, we’re doing this today in multiple ways. A lot of heat is provided by fossil fuels, but an awful lot is provided by electricity too.

Let’s look at a couple of examples. Electric steel minimills are a good one. They use electric arcs to maintain temperatures of 1,800 degrees Celsius, well above the range Heliogen thinks that it might be able to achieve, and electric arcs can produce heat up to 3,000 degrees. Minimills currently provide about 70% of the steel used in North America without any problem at all. They’ve displaced most of North America’s steel manufacturing from iron ore with provision of high quality steel from scrap.

The same week that Heliogen made its announcement, Energywire carried a story about a new electric steel minimill being built in Sedalia, Missouri. What’s so special about it? Well, it’s going to run entirely off of renewable electricity. They’ve contracted for entirely renewable electricity, and more wind generation capacity is being added regionally to accommodate their demands.

Now that’s a model that makes sense. Put the plant where it is economically sensible to put the plant. Use incredibly well known, highly efficient technology for heat sources powered by electricity. Put the renewable generation where it makes sense to put the renewables. Run the electricity through existing transmission and distribution lines from generation to the plant. Don’t tightly couple them.

What other example springs to mind? Well, once again in the same week as Heliogen’s announcement, another steel innovation came to the forefront. Steel from scrap via electric minimills covers one part of the process, but doesn’t cover new steel from iron ore. That’s been proposed to be significantly decarbonized by using hydrogen in the process. Currently, coking coal is mostly used to iron ore reduction. When replaced with hydrogen, that eliminates CO2 from one major step and emits only water.

And how is the hydrogen to be created? From renewably generated electricity. Siemens has provided an industrial scale PEM hydrogen production machine for the facility, but the electricity is once again coming from wherever the renewable energy is generated over efficient transmission lines.

I reached out to Mark Z. Jacobson of Stanford, assuming that he and his team would have assessed concentrating solar power in some depth. He concurred.

“I don’t see the technology as filling any gap that was not filled previously, just another method of producing high temperature heat from renewable energy. Other methods of producing high-temperature heat from renewables include from electric resistance furnaces, electric induction furnaces, and electric arc furnaces, all powered by wind, water, or solar electricity. Whereas, Heliogen’s apparatus can produce temperatures up to 1,500C, an electric resistance furnace can produce temperatures of up to 3,000 C, so twice those of this new technology.”

All of which begs the question about Heliogen being useful for hydrogen production as well, another use case that it mentions. PEM hydrolysis is already 80% plus efficient and renewable generation to PEM device including transmission losses only drops that by a bit. If the proposed high temperature creation of hydrogen that Heliogen points to is useful, that heat can be supplied by electric arcs just as readily and without all of the other problems inherent in tight coupling of concentrating solar power to industrial facilities.

What does this all mean?

Well, it means that Heliogen has done some really cool machine learning to provide better alignment and capacity factors for a technology which continues not to prove that useful otherwise. I’m not downplaying the technical advancement. I’m sure that there are some industrial niches where Heliogen’s solution will be useful and economically viable, although I think it’s more likely that its core machine learning innovation regarding halo focusing will find a completely different niche outside of concentrated solar.

But the technology isn’t suitable for the vast majority of industrial facilities and their need for high quality heat. There are better alternatives, and systemically, electricity from renewables is far superior than direct use of the heat.

What would characterize a use case that this type of component would fit into? Well, it would have to be something where heat was required in the middle of nowhere where lots of land was cheaply available. It wouldn’t need incredibly high temperatures, just highly elevated ones. It wouldn’t require bulky above ground industrial facilities. It would have to be poorly served by grid connectivity, otherwise it would be simpler to just get electricity from widely spread renewables and use that with existing components that use electricity to provide high quality heat where required in the process. Even then, it’s going to need electricity to operate the neural net, the cameras and the drives the focus the mirrors, just not as much as heat would require.

What does that sound like to me? Steam assisted gravity drainage (SAGD) enhanced oil recovery for the oil sands. The oil sands in Alberta inject steam a couple of thousand feet underground through pipes with holes along their lengths to loosen up the oil trapped in the sand and allow it to be extracted through another hole nearby. The oil sands currently use co-generation gas units for this purpose, using the waste heat to create steam to inject underground, powering anything electrical with some of the electricity and selling the rest of the electricity to the utilities.

That’s right, once again, Gates has invested in a company whose only apparent natural market is for pumping more oil out of the ground that we can’t afford to burn. It might be a lower carbon way of getting at that oil, but we can’t afford to burn the very easily and cheaply recoverable oil, never mind most of what’s in oil sands right now.

And, once again, Bill Gates is spending his money and influence on dubious side bets instead of sensible technologies. Like Terrapower, Carbon Engineering, and solar geoengineering, Bill Gates is spending his money and influence in ways that aren’t beneficial to combating climate change. As I’ve said before, he needs new energy and climate advisors.


Michael Barnard