I How can visible gas bubbles remain intact inside frozen water?

[shakhfenix]

TL;DR Summary

Froze a sealed plastic water bottle. Inside, visible gas bubbles remained perfectly shaped, untouched.
No rupture. No deformation.
According to classical thermodynamics and fluid behavior, is this expected?

I performed a simple test:

First, I briefly opened and closed a plastic mineral water bottle to allow air exchange.
Then I left it untouched at room temperature for about half an hour, until multiple gas bubbles formed inside.
After that, I placed the bottle in the freezer.

What caught my attention was that the air/gas bubbles trapped inside did not collapse, deform, or rupture — even after full ice solidification.
• The plastic bottle remained intact (no signs of deformation or pressure stress).
• The gas bubbles retained their spherical shape.
• There was no visible shift, compression, or expansion rupture.
• I observed the bubbles under a microscope before and after freezing. No collapse.

My question is:
How is it physically possible for these gas bubbles to remain spatially fixed and structurally intact while surrounded by ice?
Shouldn’t the freezing process — and water’s expansion — either displace, compress, or destroy them?

I’m aware that gas can become trapped in ice. However, this case seems fundamentally incompatible with what’s expected from the expansion behavior of water.
These were not microscopic gas inclusions — they were clearly visible, well-formed, perfectly spherical bubbles suspended in liquid water prior to freezing.

I had previously lost a 100% copper bottle to ice pressure. So the fact that not only the plastic bottle showed no deformation — but even the gas bubbles inside remained untouched — left me genuinely perplexed.

Under the microscope, using only refracted light, the “before” showed internal structures within the bubbles, incompatible with what we’d expect from simple gas-filled spheres.
After freezing, those internal structures appeared damaged or broken.

I’m not seeking speculative answers. I’m trying to understand this within the known framework of classical physics.
What is the expected behavior in this situation?

 

[Dale]

TL;DR Summary: Froze a sealed plastic water bottle. Inside, visible gas bubbles remained perfectly shaped, untouched.
No rupture. No deformation.
According to classical thermodynamics and fluid behavior, is this expected?

I’m not seeking speculative answers. I’m trying to understand this within the known framework of classical physics.
What is the expected behavior in this situation?

This is a much better start than your previous thread.

Here is a pretty recent paper on this topic:

“On the shape of air bubbles trapped in ice” by Thievenaz et al. https://doi.org/10.1073/pnas.2415027122

This is an interesting new paper as it not only describes the existence of such air bubbles in ice, but also derives the shape of the bubbles as they form. This process is well understood.

What is more interesting to me is not the fact that liquid water with bubbles will form ice with bubbles, but that even water without any bubbles will form ice with bubbles.

As water freezes, any gas that is dissolved in the liquid must be expelled. If it has nowhere else to go, then it must form bubbles. Even if no bubbles previously existed in the water.

 

[shakhfenix]

Thank you again for the reference and engagement.
As promised, I’m following up with the visual evidence that led me to raise this question in the first place.

The experiment, as described, is simple. But the results are difficult to reconcile with classical thermodynamics.

Attached are three image sets:
1. Direct visual setup:
A sealed 1.5L plastic water bottle, with bubbles clearly visible — and a microscope live-feeding the exact same bubbles onto a tablet.
This is not speculative. This is direct.
2. Before vs. After #1:
In the liquid state, the “gas” bubbles are spherical, smooth, structured.
After freezing, they collapse — but not uniformly. They break apart, showing internal fragmentations.
3. Before vs. After #2:
Once again, the “gas” bubbles appear hollow and coherent before.
After freezing, internal structures rupture — not what we would expect from inert trapped gas.
These were all viewed using only refracted light.
No dyes. No additives. No speculation.

And now, my question stands louder:

• If these are just gas bubbles, why do they appear structured under refracted light?
• Why do they fracture like biological cells, not compress like gas voids?
• Why does freezing not deform the bottle — but breaks what’s inside the bubbles?

The paper you referenced explains formation of new bubbles during freezing.
But it says nothing about pre-existing large gas structures showing post-freezing fragmentation.

I’m not here to speculate. I’m here to ask:
How does classical physics explain this?

 

[Dale]

But the results are difficult to reconcile with classical thermodynamics.

Why do you think that? The paper I cited used classical thermodynamics.

I had thought this thread would be better than the previous one. But let’s be very clear for both this question and future questions:

Just because you don’t know how some observation is explained using standard theories doesn’t in any way imply that the observation is incompatible with said theories. Until you have studied a topic at the graduate level you simply are not qualified to make such an assessment. Asking questions to learn something is good, but assuming that you know there is no answer is premature.

By the way, I have not studied thermodynamics at the graduate level. So even if someone like me doesn’t know the standard thermodynamics answer, it is still premature to claim that there isn’t one.

Why does freezing not deform the bottle — but breaks what’s inside the bubbles?

It does deform the bottle too. But what you are seeing in the bubbles is deformation and stress of the ice itself. The gas just compresses; it doesn’t break.

Ice forms many different crystalline structures. See here: https://en.m.wikipedia.org/wiki/Phases_of_ice with Ih II and III all being in typical ranges for sealed vessels in ordinary freezers.

And even without forming different crystalline structures, the same crystalline structure can be oriented differently with optically visible fissures between crystals. This is probably primarily what you are seeing.

 

[shakhfenix]

Why do you think that? The paper I cited used classical thermodynamics.

I had thought this thread would be better than the previous one. But let’s be very clear for both this question and future questions:

Just because you don’t know how some observation is explained using standard theories doesn’t in any way imply that the observation is incompatible with said theories. Until you have studied a topic at the graduate level you simply are not qualified to make such an assessment. Asking questions to learn something is good, but assuming that you know there is no answer is premature.

By the way, I have not studied thermodynamics at the graduate level. So even if someone like me doesn’t know the standard thermodynamics answer, it is still premature to claim that there isn’t one.


It does deform the bottle too. But what you are seeing in the bubbles is deformation and stress of the ice itself. The gas just compresses; it doesn’t break.

Ice forms many different crystalline structures. See here: https://en.m.wikipedia.org/wiki/Phases_of_ice with Ih II and III all being in typical ranges for sealed vessels in ordinary freezers.

And even without forming different crystalline structures, the same crystalline structure can be oriented differently with optically visible fissures between crystals. This is probably primarily what you are seeing.

Hi Dale, thank you for your thoughtful reply and for keeping the discussion focused. First, I sincerely apologize if anything I wrote came across as presumptuous or offensive. That was never my intention. I’m here genuinely seeking clarity, and I deeply value your expertise and time.

To clarify, the experiment I referred to was designed in collaboration with the most advanced version of ChatGPT (GPT-4), using Deep Research mode. We actually plan to revisit this with an even more detailed AI-assisted review shortly. Every step — from the choice of container to the freezing conditions — followed ChatGPT’s own scientific expectations, based entirely on official thermodynamic principles.

So, when I said that the results are difficult to reconcile with classical thermodynamics, I was not speaking from personal authority — I’m not a thermodynamics graduate. What I meant is that even ChatGPT, despite operating within the framework of classical physics, predicted something quite different from what was observed. The discrepancy, therefore, is not due to my misunderstanding, but rather the inability of classical models — as interpreted by an advanced AI — to fully account for the outcome.

And yet, the data are clear: photos, videos, and live demonstrations were all captured through my smartphone camera — and precisely under the instructions given by the AI. The evidence is real, objective, and reproducible.

One important point I’d like to raise concerns the microscopy images. If you’ve had a chance to look at them: do the bubbles seen before freezing — perfectly spherical and symmetric — really look like empty gas pockets? They were visibly present before any phase change occurred. This is a crucial distinction.

The paper cited does not mention macroscopically visible, pre-existing spherical bubbles prior to freezing, let alone their preservation as flawless spheres through the solidification process. That’s what we’re seeing here, and it’s worth emphasizing.

Lastly, I’d like to ask one more question that may help clarify the phenomenon:

If the internal pressure is sufficient to rupture copper in certain setups, how do we explain the fact that in a glass bottle, these same “gas” bubbles vanish completely before the bottle itself breaks — while in plastic bottles, the bubbles remain intact, despite minimal or imperceptible deformation of the plastic? We’re preparing to record and share this comparative experiment, as we believe it may shed further light on the issue.

Thank you again for your time and openness. I truly hope we can keep exploring this mystery together — not to challenge the scientific method, but to advance it with honest curiosity and empirical rigor.

 

[Nugatory]

What I meant is that even ChatGPT, despite operating within the framework of classical physics, predicted something quite different from what was observed.

Which shows that ChatGPT and other LLMs are thoroughly unreliable on any but the most basic physics questions. This is why the forum rules do not allow them as references.

 

[renormalize]

The discrepancy, therefore, is not due to my misunderstanding, but rather the inability of classical models — as interpreted by an advanced AI — to fully account for the outcome.

There's your problem right there. How do can you possibly know that the AI interpretation is correct without checking it against credible models that are published by human researchers? There's a reason that one of the Physics Forum rules explicitly states:

• Answering a science or math question with AI-generated text, even with attribution, is not allowed. AI-generated text apps like ChatGPT are not valid sources. (emphasis added)

 

[Dale]

To clarify, the experiment I referred to was designed in collaboration with the most advanced version of ChatGPT (GPT-4), using Deep Research mode.

ChatGPT in deep research mode has also not studied this material at the graduate level. Nor has it published a single peer reviewed paper in this field. Nor has it conducted any notable experiments. It is not a reliable source.

Our policy is to not recognize AI as a valid source at this time. That policy is continually being re-evaluated, but I can tell you quite frankly that your two threads are examples supporting the current policy, rather than challenging it. These two topics are both addressable at the undergraduate thermodynamics level (my thermo level). For the AI to claim otherwise is simply a hallucination.

Where you might find value is in using it to collect references on a topic. Then you could read the original papers yourself and understand the topic.

The discrepancy, therefore, is not due to my misunderstanding, but rather the inability of classical models — as interpreted by an advanced AI — to fully account for the outcome.

ChatGPT is a large language model. It contains no fact models, nor any physics models. At best, when they don’t hallucinate, current LLM’s give you words and phrases that are typically associated with online descriptions of physics models.

They can fail by using crackpot or confused sources. They can also fail by producing hallucinations. Usually in this context this involves taking correct phrases from correct references but combining them in such a way that the resulting overall work is wrong.

The evidence is real, objective, and reproducible.

And not at all in contradiction with standard physics. As described above.

If you are interested in learning about a topic, that is great. But again, you are not qualified to claim that something is outside of standard thermodynamics (nor am I). The addition of an AI tool doesn’t change that fact.

 

[Dale]

After evaluation by the mentors we think that the ChatGPT tangent to this thread is too disruptive for it to continue. So we will close this thread.

However, the topic itself remains open. We encourage the OP to post questions about their experiments to a new thread with the following understanding:

1) Neither TikTok nor ChatGPT are valid sources here. (I recommend against using them to learn physics on your own, but, whatever you choose on your own time, they are certainly not acceptable for use here)

2) Neither you nor I are qualified to claim that a particular experimental result is incompatible with standard thermodynamic theory.

With those understandings we can have a productive discussion focused on the data and the theory.