Rare oxygen isotope detected
nature.com
nature.com
> All naturally occurring tungsten isotopes are expected to alpha decay into hafnium, but with extremely long lifetimes. Since the decay energies for all these decays are in the same energy range as beta and gamma backgrounds from the natural decay chains, their observation is a difficult task. Yet with cryogenic scintillator experiments, these backgrounds can be discriminated from the alpha signal, leading to a basically background free measurement of such alpha decays, see figure 13. Hence, the natural decay of W-180 was observed unambiguously for the first time.
Update: Apparently it was not the only rare decay detected during a dark matter experiment, in fact dark matter searches are a major source of rare decay detection. Previously in 2003, Bismuth-209's radioactivity was also detected as a bonus result of a dark matter search, with a half life of 10^19 years. [1] In 2019, the XENON1T experiment detected the radioactivity of Xenon-124 (again, because Xe was used inside the detector), with a half life of 10^22 years. By far it's the rarest radioactive decay ever directly observed by physicists [2].
[0] https://en.wikipedia.org/wiki/Cryogenic_Rare_Event_Search_wi...
[1] https://physicsworld.com/a/bismuth-breaks-half-life-record-f...
[2] https://en.wikipedia.org/wiki/Isotopes_of_xenon#Xenon-124
If I have that right, it does seem mind blowing that they were able to detect it.
There are extensions to the Standard Model being used to search for dark matter, but until one of those is actually found experimentally, you can't say that it makes the Standard Model incomplete. It's sound within its domain -- aggravatingly so, since that makes it really hard to figure out what's next.
But similar to your example, neutron decay is a thing (a free neutron has a halflife of about 10 minutes), and yet many nuclei are stable for much longer.
Also to nitpick - “half life” is not applicable to subatomic particles
And that in turn might get us one step closer to a UFT.
However, there are many good reasons to expect that physics beyond the standard model do not conserve B or L (or B-L, which is what's actually conserved in the SM). In those models one expects proton decays for sure.
However, even in those models, electric charge is conserved. Unless there are gauge-variant dynamics at very high energy electric charge will always be conserved and electrons (and positrons), being the lightest electrically charged particles, will be absolutely stable.
The Tungsten isotope page lists two alpha decays per year per gram, that must have been quite the mass of Tungsten if they got a usable signal out of that. Amazing result, if you think about it: your measurement is so accurate that you can measure you measuring gear falling apart.
If the theories are incomplete or wrong, how could we accurately simulate things we don’t yet understand? It doesn’t matter how powerful the computer is.
We trust our physical understanding by experimentation, not simulation.
You can use simulations in areas that are fully understood to run calculation on new arrangements of those those things, without having to make the physical object. But it only works when you already understand the thing, you can't gain that original understanding from the simulation.
So sure run a simulation at whatever level of granularity you want that doesn’t mean it’s correct.
On the plus side, now that they have an empirical result, they can tweak the model such that it continues to accurately describe what it currently describes, and describes a short lived O28. Once they have those tweaks, they can find another experiment to see if their updated model accurately predicts what the experiment would produce. If it does, they gain more confidence in the model, if it does not, they go back to tweaking the model.
This is the essential core of scientific research, for science to be believable it needs to predict things that will happen given conditions, and then experimentalists establish those conditions and look for confirmation of the prediction. It is the only way to know if what we think we know is in fact worth knowing!
So imagine you have a simulation, and you get an answer out. Yay.
How do you know it is correct? You don't. You must compare against reality. Reality always wins.
This is not a "single atom," you might as well say "a single person." Each one of those protons is composed of two up quarks and one down quark. Each one of the neutrons is composed of two down quarks and one up quark. Each nucleon is therefore three quarks, held together by the exchange of virtual quarks. The nucleons themselves interact via a stepped-down approximation of that called the strong nuclear force. And you're not allowed to forget the electromagnetic force, either. And then there's self-interaction ...
There's a lot going inside of a nucleus.
Simulations are only useful for testing your models.
https://i0.wp.com/profmattstrassler.com/wp-content/uploads/2...
“Fig. 3: A more realistic, though still imperfect, image of protons and neutrons as full of quarks, anti-quarks and gluons, moving around at high speed. More precisely, a proton consists of two up quarks and a down quark plus many gluons (g) plus many quark/anti-quark pairs (u, d, s stand for up, down and strange quarks; anti-quarks are marked with a bar.) The edge of a proton or neutron is not sharp. Ignore the color-coding for now; it will become clearer in future articles.”
-- https://profmattstrassler.com/articles-and-posts/particle-ph...
And we know this, how? Using magic?
Has anyone ever seen a quark? We could barely detect atoms, now we're detecting something even smaller?
Atom detection has been ... quite a while. The parts of the atom: electron, proton, neutron (all somethings even smaller) started with the electron in 1897. Neutrons lagged until the 1930s. Quarks were hypothesized in 1964. Now, you'll never find free quarks (due to something called color confinement) but we started detecting that the nucleons (protons and neutrons) must have something even smaller inside via scattering experiments around 1968. We were producing charm quarks in 1974. 1977 we observed the bottom quark, and in 1995 we got the heaviest of the bunch, the top quark.
The current year is 2023.
But: it's a theory and it may well be displaced by something else at some point, but that something else would have to be even better at describing reality as observed than quarks are. Maybe a unified field theory will do away with the 'zoo' of subatomic particles but that would in itself be a very surprising result. But it could definitely happen.
Just to be clear, simulating the atomic nucleus isotope stability (like here) is something entirely completely different than simulating the quantum mechanics of electrons in one or more atoms (like we do in DFT), or simulating molecules (like we do e.g. in molecular dynamics). The latter two are comparably much easier.
Threw that into Google Scholar and the only hit had a link to the pdf of the paper
For that matter we can't simulate a single proton either. See: https://www.quantamagazine.org/inside-the-proton-the-most-co...
(Unrelated but this is why I don't believe singularities exist in the universe - we don't know enough about quark degeneracy pressure to know if it's actually possible for a star to collapse - it's possible the quark pressure keeps the matter from compressing.)
Why are you more comfortable with infinite pressure forces than infinite densities?
[0]: https://en.m.wikipedia.org/wiki/Kugelblitz_(astrophysics)
[1]: https://arxiv.org/abs/1408.2778
[2]: https://arxiv.org/abs/0805.3880
[3]: https://arxiv.org/abs/1105.5898 (building on [2])
Although I have a side question: Imagine three streams of light, each 1/3 the density needed to make a black hole, traveling at a slight angle from each other, and then meeting.
The moment they meet they are a black hole. How fast is that black hole moving afterward in order to concerve both momentum and energy? You'll find the answer is: The speed of light.
There are clearly unsolved issues with Kugelblitze.
When I do a naive version of the calculation I find it is slightly below the speed of light, with the amount below depending on the angles between the beams. The full calculation is beyond my skills.
But you have another issue: Even if you are just below the speed of light, most of the mass would become relativistic mass (and relativistic momentum), with almost no rest mass.
But there's a postulate that only rest mass can make a black hole, and relativistic mass doesn't count. (Because otherwise you could travel fast and see inside the black hole.)
So we are left with a contradiction.
As for the effect of moving masses, that kind of "drags" space-time along with it. And rotating masses do as well. In the case of rotating masses that's called https://en.wikipedia.org/wiki/Lense%E2%80%93Thirring_precess... and was what Gravity Probe B was measuring.
As for the light, you can solve the mystery as follows. Go to the reference frame in which the light is just all pouring into one spot. In that spot photons are meeting photons and creating a sea of pairs of particles - and now it is obvious that a black hole could be made.
So at that meeting spot they would go right through each other.
When a particle meets its antiparticle, you get 2 energetic photons. All such interactions are perfectly reversible. So when 2 energetic photons meet, you can get a particle and antiparticle. Both energy and momentum are conserved in this process.
But here we explicitly placed the photons to be traveling in the same direction, so there is some net momentum that can not be carried by the particles.
i.e. particle anti-particle annihilation will never produce two photons that are both traveling straight forward. So the event is not reversible.
You can read more about this at https://en.wikipedia.org/wiki/Two-photon_physics.
I read about the two photon physics - and the two photons can scatter off of each other, but they can't produce particles except in very high electromagnetic fields, which isn't the case here.
The process γ γ′ → e+ e− is a classic calculation in QED, and although it's never been observed directly for two single/isolated photons[0], there is no good reason to believe QED should be wrong in this particular case and, indeed, all evidence we have is pointing towards this process being perfectly possible[1].
[0]: https://en.m.wikipedia.org/wiki/Breit%E2%80%93Wheeler_proces...
[1]: https://www.energy.gov/science/np/articles/making-matter-col...
That's news to me. Could you provide a reference? "Relativistic mass" is just rest mass + kinetic energy, and so yes, it should influence the energy-momentum tensor like everything else.
> (Because otherwise you could travel fast and see inside the black hole.)
How so?
To see that this is a necessity, even for a single particle moving at relativistic speeds, consider energy conservation: If you accelerate a particle, you have to expend energy and that energy needs to come from somewhere (say nuclear fission). As a consequence, this energy will already show up in the energy-momentum tensor before acceleration (in the fission case: as rest mass of your isotopes) and, by local energy-momentum conservation (= 4-divergence of the tensor being zero), it then must also show up in the tensor afterwards.
Whether or not that energy can lead to a black hole is a different question. For a single massive object flying at relativistic speeds in an otherwise empty universe: Probably not, because one can simply transform that "relativistic mass" away by switching coordinates, as you say. For several objects moving relative to each other at high speeds, or even a single particle moving relative to the "rest of the universe", it's not so simple, though – you won't find a coordinate system in which all terms in the energy-momentum tensor are small (let alone zero) everywhere.
I don't see how that follows. You suggested that Pauli/degeneracy pressure is not well understood enough for us to conclude with certainty that gravitational collapse works and leads to black holes.
There is plenty of observational evidence for dark compact objects, though. How do you explain those?
As was already mentioned in another comment, adding two non-collinear null vectors won't give you a null vector but a timelike vector. Adding to the latter a third null vector will just give you another timelike vector, so not quite the speed of light.
Anyhow, even if the black hole did move at/close to the speed of light:
The speed of light limit holds only locally where you're roughly Minkowskian and there is no curvature. However, gravitational disturbances (= curvature disturbances) are not bound by the speed of light. They can indeed propagate at the speed of light. Consider, e.g., gravitational waves or comic expansion.
Of course, the difference from a physical POV is that a black hole has a nonzero mass whereas it's hard to define one for gravitational waves. So I get that a black hole moving at the speed of light would be concerning. But, again, the actual velocity would not quite be c, so all is good.
But if it's true, that's why you can do your hydrogen atom in Quantum 101, and why this is not merely O(28^2) harder (or do you need the electrons too?).
They were just apparently confused about what exponential means.
Edit: hint: focus on “if that’s true” and “merely”.
Read it again, armed with the following fact. What they claim to have been told is correct. It is exponentially hard in N.
Your theory is that they meant to say quadratic. This requires:
1. They heard exponential and understood it to be quadratic.
2. While writing down what they heard, they accidentally wrote exponential instead of quadratic.
3. While explaining their question, they failed to catch the second error.
Now consider my theory. That they are confused about the difference between exponential and quadratic. This requires:
1. Like many people I've met, that they are confused about the difference between exponential and quadratic.
2. They made no other mistakes. They remembered what they were told correctly. They repeated it correctly. And then asked a reasonable question based on their understanding.
In general a whole chain of coincidental errors is far less likely than a single error. Doubly so when it is a single error that is somewhat common. Most people will run across exponential vs quadratic in high school, forget it, then never use that information again. There's an exponent, it's exponential, right?
Therefore I think it is far more likely that they misunderstood the term, rather than that they understood the term and then made a whole series of other mistakes to produce the statement that we saw.
If it was classical mechanical 28 "balls" with pairwise interactions, complexity would be "merely O(28^2)". But it is not, which is what I said.
I mean, there is some irony in accusing people of not understanding the most basic of mathematical terms, while not understanding (repeated!) explanations of the comment.
Note: it's not currently proven that it's impossible for classical algorithms to simulate quantum systems in polynomial time, but it is strongly believed to be the case.
The interactions inside a nucleus are completely different from regular quantum mechanics with electrons etc. like in a quantum computer.
https://www.nature.com/articles/s41598-023-39263-7 shows them winning on simulating a nucleus.
Note that "if"--it's a big, unsolved problem right now.
The problem right now is that once you start adding qubits the noise in the system grows faster than your signal.
Then there is the uncertainty principle to deal with which may preclude one or more parameters from being known exactly in the first place. And so on. In the end you find that no matter how much computing power you throw at it even just a simplest atom is beyond your capability of simulation for as much as a tiny fraction of a second.
What we do in almost every simulation is to take a shortcut: instead of simulating the individuals atoms we simulate their observed properties and usually in larger numbers. This allows for useful work to be done in a timespan not measured in aeons. But it's an approximation at best, never a simulation accurate enough to make definitive statements about how any individual atom behaves and what its future state will be given some set of initial conditions with any accuracy.
Back then (like, 30 years ago, and I stopped doing physics after this so my memory of this is fuzzy) we were looking at simulations of nuclei like O16. I say 'we' - nuclear theorists were _very_ thin on the ground, we were the only remaining group in the UK. Most particle physicists are of the kind looking at subatomic particles, not nuclei. Anyway, we were attempting to port the code to run on parallel processors (a 96 transputer rack at the time), and then diagonalise the matrices of the interactions to get out a spectrum of energy levels. IIRC the matrices worked out as ~20m x 20m, and the technique used was the https://en.wikipedia.org/wiki/Lanczos_algorithm ... the problem we had was that the state space explodes combinatorically with increasing numbers of nucleons; and the computation time scaled something like n^1.1 for n states, due to inter-processor communication.
In the end that was what killed the project - it became clear that with moores law we were about 10 years from having affordable access to a computer that could do the calculation for larger shells (including O28, which was well out of our range).
That was the state for _exact_ calculations, but there were alternative approaches - I recall VAMPYR being a German Monte-Carlo simulator for shell models that performed really well, and could extract properties even for quite heavy nuclei.
Looking back a lot of the problems were just a lack of memory, even more so than compute. The matrix elements weren't stored explicitly but recalculated on the fly because we lacked memory, this led to us not using off-the-shelf matrix code and the whole thing had numerical stability issues and used lanczos because we could fit that into the memory on board the processors. These days I use servers in AWS with ungodly amounts of memory and extremely fast cpus, I'm pretty sure they could simulate this for a couple of hundred bucks.
The numerical stability issues I complained about are inherent due to the accumulation of floating point errors in the Lanczos algorithm. You just end up with completely junk results. (so, you work with double precision all the time, chewing more memory). We do have some "initial conditions" in the calculation, in that the lanczos algorithm is repeatedly applying a matrix to a vector that represents state occupancies; but if this was numerically stable it wouldn't much matter what initial vector you chose (I can't remember if ours used (1,0,0...0) or 1/root(N)*(1,1,...,1))
AFAIK the code isn't available, that just wasn't how things worked back then, there were some ftp sites but open source was still nascent (Linux didn't exist yet). The architecture I was writing for was a Parsytec Supercluster https://en.wikipedia.org/wiki/Parsytec which doesn't exist as a thing any more, it was mostly straight K&R C tho with a harness written in Occam, controlled from a workstation running SunOS (not even Solaris).
If you happened to find the code it'd be completely worthless. Like I said above, we weren't able to use off the shelf, optimised matrix libraries because of the way we were generating coefficients on the fly, for what was a very sparse matrix. It's just not how you'd do it on modern hardware.
ok but what if the numbers do make sense to me tho, maybe i'm a computational physicist in some other area but not that one
> by the time you're ready to put your GPU goodies to work you'll probably find that path well trodden by others
yes that's what i was asking, what are the codes that others who have trodden that path are using
> Our GPU implementation shows a speedup up to 97.3 times over Matlab Implementation and 2.89 times over the Intel Math Kernel Library implementation on a Intel Core i7 920 Processor
Interesting it's only 3x as fast on GPU as MKL on CPU I wonder how it compares to OpenBLAS.
Oh but it's 2011 before GPU had really popped off, especially for larger precision. Probably in 2023 there is bigger GPU benefit vs. CPU.
In Dutch we have a proverb that says that one crazy person can ask more questions than 10 wise people can answer so if these are off-base feel free to ignore :)
"is the magnitude of the numbers such that this would be impractical..." is maybe the wrong way to look at our problem. Doubling memory to use double precision was an inconvenience, we were trying to squeeze out results with the hardware we had. The elephant in the room was the size of the matrix. Increasing the number of shells in the model - necessary to model heavier nuclei - massively expanded the size of the matrix, and needed a generation for the kind of computer a small research grant can afford to be capable of doing the next set of nuclei. And then it'll be another generation or two before the next set again.
And you can ask, but it's a lifetime ago, it's like asking if Napoleon could in some way have avoided defeat at Waterloo - just idle speculation on my part now. I have to laugh at the question in one of the other replies if this had been converted to run on GPU, man, those were 20 years away and some of the machines I coded for didn't have an FPU.
It's interesting how you now have the equivalent of a supercomputer from back then in your pocket. Talk about wasted potential...
I read the Parsytec page, I never knew about them, but very interesting machines.
Even though 28O is doubly magic, the unequal number of protons and neutrons creates computational difficulties for even ground-state computations, never mind decays.
Doubly magic: https://en.wikipedia.org/wiki/Magic_number_(physics)