Please note, I've found sources in the literature saying reactor was going supercritical before pushing the SCRAM button, while other sources claim after SCRAM button was pushed it caused the reactor going supercritical. In my video, Ive assumed the first is true. It's a historical detail (or advanced Nuclear reactor detail) I'm not sure which is correct.
The correct answer is it went supercritical AFTER the button was pushed, the discrepancy is entirely fabricated to downplay just how bad the old reactor was designed and how risky the experiment was that was being performed. While both Western and soviet historians and scientists long accepted it was possible it could have gone supercritical before SCRAM, lately, experiments and a better understanding of physics have shown the SCRAM procedure is needed to push the situation over the edge into supercriticality.
"Ridiculously Badly Made Kettle" that should be on a T-shirt, it is brilliant! Fantastic video, I recall trying to model chain reactions for my video critical mass but gave up. Your is just great. And the runaway reaction was the cherry on top. Great video!
So, the visual simulation is interesting, but it runs the risk of having wildly different time-constants compared to the real-world system. In particular, Xenon burn-off has a time constant several orders-of-magnitude too slow to explain the rapid acceleration in Chernobyl-4's reaction rate. Xenon build-up played a key role in setting up the accident, by getting the operators to remove all but a handful of control rods, and by unbalancing the reaction so that it was proceeding entirely at the top and bottom of the reactor, and was almost totally stalled in the center of the reactor. But Xenon burn-off was NOT an important feedback factor in the reactor's rapid power rise. In addition to the slow time constants included in several simulation papers as well as the INSAG accident reports, it's relatively straight-forward to prove this from first-principles by observing that Xenon's absorption cross-section, of 2 million barns, or 2e-18 cm^2, times the neutron flux of an RBMK-1000, which was, IIRC, on the order of 10^14 /s/cm^2 at full power. So that tells us that at full power, your average Xenon-135 atom is going to take on the order of 500 seconds to absorb a neutron. Even during the transient, at 10X or 100X the normal reactor power, the transient was so rapid that only a small fraction of the Xenon atoms could be expected to burn off. It's a relatively slow feedback factor. Also, the accident can't be properly understood without understanding the role of the asymmetric power distribution prior to the accident. With the middle of the reactor was in the depths of a Xenon pit shut-down, there were really two separate, entirely decoupled reactions, one at the top of the reactor, and the other at the bottom. Thus pushing the control rods further into the middle of the reactor had little effect because the middle of the reactor was already dead, but pushing the graphite segments further down had a very significant effect. The SCRAM effectively shifted reactivity from the dead middle of the reactor to the over-active bottom of the reactor, further accelerating the reaction. Void reactivity was a known consideration to the reactor designers, and RBMK-1000 doesn't have a universally positive void coefficient in all scenarios. In particular the void coefficient is negative early in the fueling cycle and becomes positive later as more fuel is burned up and more control rods have to be removed to maintain reactivity. Chernobyl-4 was late in the fueling cycle, and the Xenon pit caused the operators to pull out even more control rods until the void coefficient became terrifyingly positive. That was a transient property of Chernobyl-4, not a universal condition of the RBMK-1000 in all conditions. One of the mitigations for void reactivity and other positive feedback factors is the even faster negative feedback from doppler broadening as the fuel temperature rises. That's one of the fastest feedback mechanisms and is supposed to help in stabilizing the reactor, but between putting the reactor into a configuration where it had a huge void coefficient, much larger than even a late-cycle RBMK-1000 normally would, and generating an asymmetric reaction that would be pushed over the cliff by the SCRAM event's modest reactivity spatial shift from the dead middle to the overheating bottom, the operators managed to thoroughly overwhelm doppler broadening's ability to limit a rapid transient. Regarding it "going supercritical", what you really mean is prompt thermal supercritical, as being delayed supercritical isn't all that exciting and merely implies that the reactor's power level is increasing, possibly very slowly. When the reaction needs to wait for the slower decay groups to achieve supercriticality, power rises very slowly, with time constants in the tens of seconds. When exceeding a dollar of supercriticality, reaction growth only needs to wait for the next neutron generation, and thing get out of hand very quickly, on the order of 10^-4 seconds, which is still wildly slower than prompt fast supercriticality in a bomb, 10^-8 seconds per generation; even in an out of control Chernobyl-4, the reaction still has to wait for graphite moderation between neutron generations, preventing it from significantly exceeding the level of energy release required to disassemble the reactor. For most of Chernobyl's power transient, it probably wasn't prompt thermal supercritical, and was probably growing at a rate throttled by of one of the fastest two delay groups, but when all the water flashed to steam, things got pretty crazy and none of the papers I've read can agree on the exact course of events. I'm still trying to understand the simulations myself and when exactly they may have reached prompt thermal supercriticality.
The thing also overlooked here (in the simulation) are the graphite followers of the control and protective rods. If they'd been extended all the way to the bottom of the active core, then triggering EPS (aka SCRAM), would have simply let graphite go off from bottom as absorbing elements were inserted. This wasn't the case though, due to building size restrictions (remember the core height of ~7m). Thus, initially water was displaced by graphite at the bottom of the core by falling control rods, leading to pisitive scram effect, I cannot say if this started the power excursion, but at least accelerated. Then further down the few seconds, the positive void coefficient kicks in, coolant tubes rupture and mechanically jam control rods...
It's not a "visual simulation", it's a visual representation of ∑ solutions that stem from physical syntactic observations. Pardon the semantics, because I like the rest of your response.
@@defeatSpace Well, your model is not a solution of the true physically representative partial differential equations. Visually pleasing, yes, nowhere close to even rudimentary understanding of the physical phenomena in play. And if you'd like to split hair, you wouldn't refer to total power or flux as 'reactivity'. Reactivity is a unitless measure that is involved e.g. in the term void coefficient you refer to.
I think once the water flash-boiled, the xenon "burped" out of solution and escaped as gas (under high pressures), by simply venting to atmosphere. The graphite-tips accelerated the rates of thermal neutron production so fast, and the positive void coefficient meant once the steam bubble fraction was high enough, it just over pressurized the vessel and went boom. Then the xenon escaped and it really went boom and melted down. Just a pure, shit design with dumb operators and cockiness about safety.
@@nicklockard the Xenon and other fission gasses didn't mostly escape. Because they were within the fuel-matrix. During power excursion some did escape first to gas gap of the fuel rods, later as the fuel cladding shattered to coolant and out to atmosphere. However, this was insignificant on how the power excursion went. Things had unfolded by that time
Thank you! I used Manim with a Verlet integration for the physics solver. It's stitched together by me and sadly not available. Mostly because right now the code is confusing and poorly structured. It could be really fun to release of others found it useful though.
I think ddopson's comment below gives some of the most accurate and authoritative feedback needed to bring this neat little simulation closer to realism and correct some misconceptions. One I noted was that from the start it was assumed that it was the direct bombardment of moderator molecules that was (is ) responsible for energy transfer to that medium. Very likely this effect is the faster of many, but includes only about 5% of the fission energy (i.e., fission neutrons only carry away about that amount of the nuclear energy release). Roughly 90% is in the fission products (pre-decay), and so in the fuel rod. It is actually an important detail how fast the feedback process of fuel heating acts in conjunction with the "prompt" heating of neutrons and likely results in some of disparities between results of close-to-real-world simulations that ddopson alludes to. As well, once moderator starts disappearing, hardening of the neutron spectrum means that the much more prevalent fissionable-only fuel (U238) starts to participate in larger proportion in providing fission energy (that was previously disregarded), at least in that most interesting period before things are quenched by "rapid disassembly" of the core. This is often taken as a competing effect for doppler broadening for convenience in practical calculation. In any case show
I'm far from an expert on the incident, but from my understanding, the control rods jumping is pure fabrication by an individual who didn't understand how the reactor worked. The person who said the control rods were jumping also could not have been in that area and survived the explosion.
Can you add the effects of gravity-driven, internal circulation and also thermal diffusion? I know it gets hideously computationally intensive due to the turbulent flows that would arise, thus making it difficult to model, but even a dumbed-down model ignoring turbulence could show the cold spots and hot spots and how they affect the scenario, i think.
If I had to add another think thermal diffusion is probably the most important one to add. And yeah as you mention a "dumbed-down model" is probably sufficient. I thought about this, but I figured it wasn't absolutely critical for understanding the accident but now I regret not adding it. Maybe I will do a follow up.