Weekend Editor

Retired physicist and statistician, now a blogger.


Sorted by New

Wiki Contributions


For energy storage, you might consider the Ambri liquid metal batteries. They're being designed for this exact purpose: coupling intermittent renewable generating capacity to steady loads.

Also, they're cheap, rugged, and don't seem to lose capacity over multiple charge/discharge cycles.

JN Crossley, et al., What Is Mathematical Logic?

A 96-page intro to the basics of predicate calculus, model theory, and Gödel incompleteness. I've used it in the (distant) past a couple times when a student had trouble getting a practical grip on logic.

If you feel the need to do something in response to the advent of fusion power and high-capacity batteries, you might want to think about doing it sooner rather than later.

Fusion: I'm beginning to think this is nearer-term than most of us believe. Last September, Commonwealth Fusion Systems demonstrated a 20 Tesla superconducting magnet with a bore large enough for their tokamak.

  • It's a REBCO magnet (rare-earth boron copper oxide) which superconducts at nitrogen temperature, but they're going at 10-20K to have some running room for high field strength without quenching the magnet.
  • The volume of the tokamak scales as the inverse cube of the magnetic field. ITER runs about 9 Tesla, so at 20 Tesla CFS has more than a factor of 2 increase in field, hence 1/8th the volume. Tokamak costs scale roughly by volume, so there's potential cost savings of a factor of 8. (Potential, not yet demonstrated.)
  • Their reactor is a high-beta design, i.e., high plasma pressure. They're building a demonstrator reactor now, expected to have a Q (power out/power in) of about 11. Target completion date is in 2025, only 3 years from now. 140megawatts of power, delivered in 10 second bursts.
  • If the demo reactor works, they predict building full-scale commercially useful reactors by 2030.

High-Capacity Batteries: Lithium-ion batteries are wonderful for portable applications, but… they tend to degrade after a lot of discharge cycles, and in high-power, high-density situations they have thermal runaway problems. "Thermal runaway" is excessively polite language for "halt, catch fire, sometimes explode". The thermal management equipment and software are pretty gnarly.

Check out the liquid metal batteries from Ambri. They're basically a highly insulated box with 3 layers of molten antimony, calcium chloride, and molten calcium. Discharge it, and the Ca atoms give up a couple electrons, the ions migrate through the salt layer, and form CaSb at the other end. Charge it, and the reverse happens.

  • Yes, its molten metal. But about 1 charge/discharge cycle every day (say, when coupled to a solar array) is enough to keep it heated, given adequate insulation. Unlike lithium-ion, it likes to be hot.
  • The materials are cheap. Don Sadoway, an MIT prof who founded Ambri with one of his students, tells the same joke in every talk he gives (and I mean every talk!): "If you want it to be dirt cheap, make it out of dirt. Preferably local dirt, so nobody can cut off your supply." (He has a number of very engaging talks on YouTube.)
  • It appears to have no measurable degradation after hundreds of charge/discharge cycles. Obviously you can't form dendrites in liquid metal.
  • The round-trip efficiency (power out / power in) is about 80% (with the losses going to ohmic heating to keep the metals molten). Pumped hydro storage, for commercial comparison, is about 70% or so. So the efficiency is very much on point.
  • Ok, they're heavy. And full of molten metal. So they're not going in your car. But for power plant applications, that's just fine. Unlike lithium ion, they can't catch fire or explode, and when frozen at room temperature for shipping they're completely inert.
  • They'd be great for peak shaving: you have generation capacity for the average case, and use the batteries to store energy during low demand periods and supply energy during high demand periods.
  • They also couple ideally with solar arrays and wind farms, whose generation capacity is variable.

So there you go: 2 commercial interests in fusion and batteries, each with at least some chance of success. There are many others; it is very likely some of them will succeed within 10 years.

The evidence on vaccine efficacy waning is somewhat confusing to me!

On the one hand: Some of the initial data on waning from Israel (using Pfizer) was hopelessly confounded with age, leading to a huge Simpson's paradox effect. Once you (a) figure out the Bayes error (they calculated Pr(vax | hospitalization) when they really wanted Pr(hospitalization | vax)), (b) properly stratify by age, and (c) calculate confidence limits on vaccine efficacy, the effect goes away.

On the other hand: Later Israeli data presented at the Moderna booster hearing cleaned that up and showed there was a waning effect. Moderna did something similar with their vaccine, comparing the people in the treatment arm of their clinical trial vs those in the control arm who got the vaccine 6 months later when it read out, showing a waning effect between those 2 carefully matched groups with known distributions of age, race, gender, etc.

On the gripping hand: Both of those show the onset of "waning" coincident with the onset of the Delta variant, i.e., last summer. So was it really waning, or was it Delta? I can't tell.

They show a waning effect with respect to initial infection, but continued robust protection against hospitalization (still 85-90%). That could be normal:

  • Antibodies do decrease with time. You're not carrying huge blood levels of antibodies for every virus you've ever encountered thus far in your life.
  • But T-cells and memory B-cells are still there. When an antigen from a previous infection is presented to the relevant memory B-cells, they trigger production of antibodies which then stop the infection. That way you can be technically infected for a couple days while that happens, but be asymptomatic or mildly symptomatic, and quickly clear the infection.

So it might be that we're just seeing antibodies fade, but which rapidly come back upon re-challenge with the virus.

On the (unnamed) fourth hand: I looked at a recent study by Townsend et al. at the Yale School of Public Health that analyzed 6 coronaviruses known to infect humans and related to SARS-CoV2, predicting that there's a 50% chance of reinfection at 18 months. So maybe periodic boosters will be required because of waning immunity?

JP Townsend et al., "The durability of immunity against reinfection by SARS-CoV-2: a comparative evolutionary study", The Lancet Microbe, 2021-Oct-01. DOI: 10.1016/S2666-5247(21)00219-6.

So... "I notice that I am confused." But since all this is happening frighteningly quickly and we're just learning how it all works, confusion is the normal state of affairs at the leading edge of science. Thus my confusion might be normal.

Or maybe I'm just plain old confused and need somebody to straighten me out. That's also regrettably normal, at least for me.

Until then, it seems that 2 things are worth remembering:

  1. Boosters do seem to work in terms of boosting ab levels, and don't seem to cause any worse side effects than the first doses.
  2. If there is waning, boosters are an effective strategy; if there is no waning, then because of the non-terrible side-effects, boosters are a safe strategy, i.e., unlikely to do much harm.

So I got a booster, and my spouse gets one in a few days when eligible.

Yes, absolutely.

The comment on study power is why I was pretty surprised at the FDA and CDC approvals, based on this study which itself says it's underpowered. I think they were mostly motivated by the safety results that boosters were at least as safe as the primers. So if it might do some good and probably does no harm, they can get to an EUA from there.

This is not the way they usually behave, but then again, these are not usual times.

“But something has really become clear: The mixing really is most impactful when you have a DNA/adenovirus vaccine first followed by the mRNA vaccine,” Gandhi said. WaPo

This is definitely true in terms of antibody fold induction: JnJ followed by either Pfizer or Moderna have the highest fold induction ratios.

However, they're starting from a lower baseline, since JnJ doesn't induce such high ab levels to begin with. (Though it might be better at training T cells and memory B cells, and have longer persistence? It's kind of frustratingly complicated, to me.)

If you look at absolute ab levels, JnJ followed by Moderna looks best.

(Data source: FDA presentation by Lyke, cited in previous comment, slide 22.)

Have a look at the presentation to the FDA's VRBPAC on 2021-Oct-15:

K Lyke, et al., "DMID 21-0012 - Heterologous Platform Boost Study Mix and Match", FDA VRBPAC 2021-Oct-15 Materials, retrieved 2021-Oct-15.

(If you're curious about the how the whole meeting went down, I wrote a little summary.)

Slide 22 (page 23, because the FDA tacks on a header page) should help. It's a 3x3 array of the 3 original vaccines x 3 possible booster vaccines. It shows the factor by which antibody levels are boosted by each primer/booster combination (geometric mean titer fold induction).

For your case of a Moderna primer, the middle column is the one to look at: Moderna booster gave 10x boost, JnJ booster gave 6.2x, and Pfizer booster gave 11x. Given the error bars on that slide, Moderna and Pfizer are more or less indistinguishable (10x vs 11x), but both are likely better than JnJ.

In terms of absolute ab levels (the number shown in blue), a Moderna booster is probably a touch higher than Pfizer (3727 vs 2801), though I haven't used the error bars to calculate the statistical significance of that. Both are pretty good.

For me, the bottom line is that it looks like either Pfizer or Moderna are fine for somebody who's already gotten 2 doses of Moderna. The important thing is to get the booster if you need it; the choice of Pfizer or Moderna is over-optimization.