One method is to apply a temperature gradient[1-4], to cause bubbles to move, because of the surface tension temperature dependency. Also, a floating bubble near[4] the surface of molten glass could be irradiated[6] with a specific laser beam.

1.  Thermal‐Gradient‐Induced Migration of Bubbles in Molten Glass     https://www.researchgate.net/publication/229761235_Thermal-Gradient-Induced_Migration_of_Bubbles_in_Molten_Glass

2.  Lumped Models of Gas Bubbles in Thermal Gradients    https://www.researchgate.net/publication/229401711_Lumped_Models_of_Gas_Bubbles_in_Thermal_Gradients

3.  Thermogravimetric Analysis     https://www.mri.psu.edu/materials-characterization-lab/characterization-techniques/thermogravimetric-analysis

4.  Bubble Bursting in Molten Glass    http://www.maths-in-industry.org/miis/353/1/Bubble-bursting-in-molton-glass.pdf

5.  Method for removing bubbles from molten glass and process for producing glass

US 8109118 B2     http://www.google.com/patents/US8109118

Article Lumped Models of Gas Bubbles in Thermal Gradients

Article Thermal‐Gradient‐Induced Migration of Bubbles in Molten Glass

If I get you right, yor're melting the glass directly on top of the tin? How high is the process temperature?

There is quite a number of differences to the industrial process:

If gas bubbles are exclusively at the "interface side": could it be that you're significantly exceeding the "traditional" temperature range - eventually catching Sn vapor bubbles while the glass is still at its highest temperature?

You may have to consider the seemingly possible occurrence of galvanic corrosion between the molten tin and graphite, which, if I have correctly interpreted the enunciate, are in direct (electric) contact, while they also contact the molten glass (electrolyte). Henceforth, tin may possibly be slowly oxidized (to Sn2+ or Sn4+) and dissolved in the melt, while cathodic reduction processe(s) could perhaps occur at the (more noble) graphite surface contacting the melt. Melt heterogeneity, possibly enhanced by high viscosity, may favour the overall process. The molten tin could hence become anodic, what might conceivably induce the liberation of O2 gas bubbles from O2- ions, if they can be somehow contributed by the melt: 2O2- → O2 + 4e-. Other gases could possibly be released, depending on the actual melt composition. Fast cooling may contribute for coarser bubbles containing low-pressure gas.  

Thank you all for your advices!

I did not expect any answer, so I have been too lazy to append some pictures. Now I appended to the question some illustrations of the bubbles.

@Ioannis Samaras

the solubility of gases in tin is falling with falling temperature. Thererfore my problem are bubbles, which on cooling arise on the tin surface, hang there and which leave hollowings in the bottom of the glass. Bubbles inside the glass are almost seldom. Therefore my actual goal is to avoid the degassing of the tin or any possible reaction between glass and tin. If I have done that, I will come back to your advice, it will be useful for removing the remaining bubbles.

@U.Dreher

may be that the higher temperature is the reason. The temperature is up to 1300°C (I use a ceramic kiln). I tried to lower the temperature, but can only estimate it to be about 1100°C. There are still more but smaller bubbles. See the pictures I appended. You mention Sn-bubbles. But I think, that on cooling the pressure of possible tin gas falls while the glass is still fluid. So the bubbles would collapse before solidification of the glass. That is plausible, because during cooling we go through the same temperature range as in ordinary float glass production, where  tin gas is no problem, Tin has a low vapor pressure.

@Carlos Araújo Queiroz

On the appended pictures you can see, that tin, glass and graphite are in contact with each other. Do I understand right: some oxygen may come from the glass itself? The bottom of the tin bath of float glass production lines is normally made of ceramics. But the liners on the sides of the bath are made of graphite and are in contact with tin and glass.

I could not find information about solubility of argon in tin. I don’t want to make the effort of building a device for melting under vacuum, which could be useless, if the bubbles come from oxygen stemming from the glass (if that is chemically possible). My next hope is to analyse the industrial process and find a solution there. But up to now, I could not get in contact with someone familiar with the industrial float glass process.

These are significant bubbles!

Did I get it right: when having an 1100 °C glass melt, this results in smaller, but more numerous bubbles than when starting at 1300 °C ? Wouldn't contradict any outgassing theory. Assuming that a higher start temperature results in a higher overall gas volume, smaller bubbles might "unite" - especially as long as the glass viscosity is lower. And obviously the glass's viscosity is then high enough as not to fill the voids then. Instead, tin is driven in. By some combination of pressure decrease and the weight of the glass itself.

What might play some role as well: different from the industrial process, I assume the temperature of your Sn bath is significantly higher. Considering the heat conduction in metals, the Sn bath temperature in the industrial process will be significantly lower than the glass temperature at the "hot end". I expect one of the challenges of the industrial process to achieve temperature stability: heating up due to the glass meot vs. conductive/radiative cooling vs. the ambient.

Regarding the fact that the bubbles are filled with tin during cooling: can you assess some remaining "void" volume or do they seem to filled completely? (The latter might really indicate Sn vapor that condensed again during cooling.)

How do you control the Sn temperature? Any way to increase the Sn mass/volume?

If you reduce the (glass) mass mg (to say : mg/5 or less), then the bubbles' volume will be reduced, under the same (thermal) conditions.

@Ioannis Samaras,

I don't understand the idea, why should the bubbles volume be reduced?

@ All,

I appended a further picture-file to the question. Because of the question of U.Dreher I checked the residual volume between glass-bubbles and tin. I cooked  the sample (glass on tin, once heated up to about 1300°C), and threw it into cold ink. On one side the glass-bubbles became blue. On top of some of them a gas bubble can be seen(not blue). The interface area between the bubbles is not blue. That is, because the distance there is very small, so that the ink can not be seen. Conversely it means, that the blue bubbles indicate  a residual volume in the bubbles.

On the other side of the sample nothing happened. My Interpretation: the surfaces are connected very tight, so no ink could penetrate.

Tin vapor should fill the entire bubble on cool down. This result therefore favors gas bubbles.

I made experiments with tin 20mm thick and 5mm thick. I do not rmember a significant difference in the bubbles number or size. That contradicts degassing of Argon out of the tin.

What kind of gas could it be? O2 or CO2 or Argon?

U.Dreher’s first question: yes, as you see in the pictures, lower temperature results in smaller (and more) bubbles. That is in agreement with degassing. The bubbles are fixed to the interface between glass and tin. Therefore I guess, that the bubbles come from below, means degassing of the tin. Or from some chemical reaction at the surface, which Carlos proposed. 

I do these experiments in a ceramic oven. I have no temperature mesurement up to now. I need some advice, because I have no money to waste.

Clever trick with the ink!

The images are a bit difficult to interpret, thus I have to resort mostly to your description.

Re tin vapor: this would condensate during cooling - at the cooler side (which I expect to be the tin surface). This might or might not result in some residual vacuum. (I'm not sure whether tin would be sucked in to completely fill the void or whether some residual vacuum would be left). So your inking experiment seems to indicate that the bubbles are filled with some gas at lower-than-ambient pressure, which can be expected due to the cooling down. (Because of the residual bubbles after inking.) Thinking again about this: could the ink itself release some gases?

Could it be that the completely uninked area is where originally some part of the glass was broken off? (I see something I interprest as a crack.) If so: the ink might not penetrate where there is no significant underpressure (which might have gone during the breaking).

I'm not sure about the "mechanical" setup: the tin is simply put into the graphite crucible, then molten with the glass on top? Or ...? I once again checked tin properties: an oxide layer on the surface exposed to air is to be expected - making some oxygen available for reactions. An Ar/H2 atmosphere might be able to reduce this layer to prevent the oxygen from reacting with "something else".

Regards

Thanks for your effort, to help me reasoning.

Tin is chosen in float glass production, because it has a rather  low vapor pressure at high temperatures. There should be no big temperature difference between the top of the glass and the bottom of the tin, because it is all in a covered graphite crucial which is in a stainless steel box in the center of the kiln. It is cooling to room temperature in about 8 hours. During cooling from 1300°C the crucible crosses 1100°C, that is the temperature, where float glass is produced. At this temperature glass is viscous like honey and the vapor pressure of tin is small enough, so that float glass production works. So I think, that bubbles full of vaporized tin must have collapsed at 1100°C.

The crack in the glass resulted from the change from hot water to cold ink.

The ink could have released some gas. My opinion is, that the flat area between the bubbles is not colored and the bubbles are blue, indicates, that there is some more space for the ink in the bubbles, what means, that at solidification temperature of the tin (235°C)  there was some gas in the cavities.

About

1. I have no equipment to measure the temperature, I have only the specification of the kiln and some  Pyrometric cones which indicate temperature in steps 1060°C 1200°C, 1250°C.

2. I append pictures to this question with two samples from the same heating to about 1300°C. One is the later inked sample, which is twice as high as the other sample. The other sample has a clear surface. The samples, who have much glass, so that they touch the crucible, have a wrinkled surface. This problem will be addressed later. Decide yourself, if the bubbles of both samples are significant different.

3. I tried to blow through the tin with Argon+H2 during heating without glass, but ended in a great mess. I must retry that.

I put small tin ingots into the crucial of graphite and on top a piece of float glass. Then I close the crucial with a graphite cover. This stands in a Steelbox which is slowly flooded with AR+H2 gas. May be I have some source of dirt (air) in my tubes?

As you can see in the last pictures, the free tin is white = oxidized, the tin covered under glas is clear. One interpretation: The glass reduces the oxygen away from the tin. The other interpretation: The inert Gas is too slow, so that on cooling down some air is sucked into the steelbox. I could not see a difference between experiments with Ar+H2 and pure Ar. But without Gas the crucible is burning away, therefore it has a protective effect.

@Friedrich Otto Richard Körner, "about your Q :  I don't understand the idea, why should the bubbles volume be reduced?"

The idea is that :

the ratio (volume / area) of the liquid glass decreases[1] with the mass of the glass and the path (length) for bubbles' out gas (from the smooth top surface) becomes shorter.

1.  For your containers' scheme : "U".

 @Ioannis Samaras: may be I understand: outgassing is proportional to the glass surface while bubble volume produced in the glass volume is proportional to this. I guess, that the bubbles grow at the interface.  If I find a way to release the bubbles from the tin-glass interface, it is easier to perish, if the glass layer is thin, because the path is shorter. 

Also, the thermal‐gradient[1] in the molten glass is higher, in favor of bubbles' migration, any bubbles' origin.

1.  Thermal‐Gradient‐Induced Migration of Bubbles in Molten Glass    https://www.researchgate.net/publication/229761235_Thermal-Gradient-Induced_Migration_of_Bubbles_in_Molten_Glass

Article Thermal‐Gradient‐Induced Migration of Bubbles in Molten Glass

It took me some time ...

Ever so slowly your setup becomes clear(er):

Thanks for the 3rd pdf - it gives me additional information. Some comments:

What I'd do (did similar things before - even if under totally different circumstances):

An alternative might be to get the ingots molten by placing the crucible on some hot plate. This step need not exceed the tin's melting temperature significantly but might help to minimize air getting trapped. In this case, have a second heating phase under protective/reducing atmosphere to get rid of the SnO2 stemming from the ingots and oxidation during melting.

Hope the approach is clear: make sure no gas is trapped in the tin prior melting the glass, remove the tin oxide by reduction and prevent further formation.

Regards

I append a new file experimental_setup.pdf.

There are 3 sources of gas:

O2 from eventually leaks or the reflow during cooling →  Re-engineering the box/gas supply

O2 from the Glass →  some electrochemical measures, other crucible or sacrificial anode

Tin bubbles →  will collapse below 1100°C

Ar from the forming gas → vacuum?

@ Carlos Araújo Queiroz  

What would happen, If  I change the crucible from graphite to ceramics (Al2O3)? I can do that with little effort. Main problem is then, that the glass will stick to the ceramics for ever. I could spray or paint a thin layer of graphite on the borders of the crucible, where the glass touches the crucibal. Would the assumed graphite problem remain?

@Ioannis Samaras

I must avoid temperature gradients in the glass, because I want to produce tension-free flat glass.

@U.Dreher

Yes, that is the setup.

I have read, that both mixtures Ar/H2, N2/H2 can be used in Float glass lines. I could not get N2/H2.

I flushed the box, maybe less than 10 times. I also had every time one crucible with lid and one without (because I burnt one under oxygen, to check the effect of Argon) - with the same result. The lid was there to protect from dirt.

I use a needle valve with bubble counter from a CO2-supply (aquarium), see experimental_setup.pdf. Four bubbles per second, volume of the steel-box about 2L. May be, that is too low: Eventually some air is sucked into the box when cooling down?! For flushing I used much more gas.

In liquid tin there are no gas bubbles, because it has a low viscosity. At higher temperatures more gas is dissolved in tin. If the tin cools down, the solubility of the gases decreases and therefore bubbles rise up.  On the tin there floats some dirt, probably Sn-oxide, I can easily remove it mechanically at 250°C.  

To follow your proposal for clean tin, I have first to re-engineer the box and the gas supply. I have to use a better temperature control, and let the temperature not exceed 1100°C. That will take some time. 

Some comments:

https://produkte.linde-gase.de/gase_zum_thermischen_trennen for some variants.

Regards

Residual stresses could be relieved, later, by a next (2nd) step, following a post annealing technique, optimized for your glass.

@ U.Dreher

The viscosity of tin at 1000°C is similar to water at room temperature (~1mPas): http://www.qucosa.de/fileadmin/data/qucosa/documents/5897/data/dissertation_mkehr.pdf,  page 54.  On page 55 there is a useful hint:  “It is possible that the liquid tin …  at the boundary surface reacts with the quartz glass to tin oxide. Such a behavior is known from the production of float glass …”  

@ Ioannis Samaras

Yes, I think that an annealing phase is a good idea after using a temperature gradient.

My plan now:

While keeping all under Ar/H2: Remove the dross mechanically from the liquid tin. Heating the tin up to 1000°C, using some temperature control. Pouring the glass on it. So the glass has a shorter time to react with the tin, and it shifts residual tin-oxide /dross besides.

If you agree with this procedure, I will disappear now for a while in the laboratory (workshop), and will report the results or construction plans later. More Ideas are welcome. Thanks for all your help.

I didn't expect the tin to react with the glass. Not sure whether your results until now showed tin oxide underneath the glass. On the other hand, some thin layer of tin oxide on top would prevent further oxidation. How this influences the "float process": difficult to assess.

What's still a mystery then: where are the bubbles from? We'll see...

Good luck with your experiments!

Ups, you are right. I must check if there is an equilibrium between the tin-oxides and tin + O2 . I try to find it out before experiment. Under the glass the tin is clear in contrast to the free tin surface, which is oxidized. I appended a new picture "Glas auf Zinn mit Kante_03.jpg". Is the  free tin oxidized by impurities in the protective gas, or comes it through the liquid tin originally from the glass? 

Tin oxidizes at the surface to air - similar to Al and Ti. Once there is a dense oxide layer, further oxidation should not take place.

Appearance of SnO2 is either white (thicker layers, bulk) or some "matte" appearance (easily distinguished from the "metallic glare" of non-oxidized metallic surfaces). Thus it is possible that the tin underneath your glass is oxidized too, but only in a very thin layer (if it has some matte look). I'm not sure whether it is feasible to assess this: eventually some resistance measurements might prove one or the other. But making sure that an eventual oxide layer is left undisturbed when probing - at the same time establishing some reliable contact - can be challenging. A "destructive" method would be to use some "electronics grade" rosin (or "flux") and heat the spot to inspect (preferably only up to slightly (10 .. 20 °C) below the melting temperature of tin). Rosin (flux) is capable to reduce tin oxide at the same time establishing some protective film (if applied in excess). This film can be removed with alcohols (ethanol, iso-propanol) when desired.

You can get bubbles at a metal/melt interface due to reduction of the "water" (hydroxyl) in the glass to release hydrogen, which, if the permeability of the glass is too low, with collect in bubbles. We found that simply reducing the hydroxyl content, or using glasses with higher hydrogen permeabilities, eliminated the problem. Float glass, unfortunately, has a pretty low hydrogen permeability, so that is a problem. I think there was a paper by Robert Eagan about this (at a silicon/glass interface) in the J. Am. Ceramic Soc back in the 1980's.

Thank you for your proposal. The bubbles grow with rising temperature only at the tin-glass interface. They do not evolve in the bulk of the glass.

I try to get Robert Eagan's paper.

Which kind of glass has a higher hydrogen permeability?

That is exactly what we saw. The gas is generated at the interface, so bubbles only form there. They will grow with temperature (and become fewer in number) by merger of smaller bubbles and expansion of the bubbles as their internal pressure rises with temperature and the melt becomes more fluid. The highest hydrogen permeabilities are found in vitreous silica and in many borosilicate glasses, esp. those with low alkali contents like the old GE sealing glasses and classic Pyrex (CGW-7740) or CGW-7070. The borosilicate glasses with high alkali/alkaline earth contents are just as bad as standard plate and bottle glasses. The other way to stop this would by using glass with a very low hydroxyl content. While it is not probable, it is possible that old float glass would be better than newer stuff since the newer glass contains more hydroxyl due to melting in oxygen/fuel furnaces as opposed to air/fuel ones. Lowering the alkali/alkaline earth content, or glasses made with potassium instead of sodium would also be going in the right direction.

Because I try to make optical mirrors, these glasses Pyrex or Silica are also the right direction to smaller thermal expansion coefficients. So I'll try to handle higher temperatures.

As atmosphere I use forming gas with 95%Ar and 5%H2. It is also used in float glass plants to protect the tin from oxydizing. Could the H2 content be a problem? I do not see less bubbles when I use pure Argon.

Where can I find classic Pyrex , Ebay?

In industrial tin baths they use 90% N2 and 10% H2, is there a reason you have chosen argon?

Thanks for your hint. I had contact to specialists and told them, but they did not notice. I took this gas, because it is easily available in the welding store and the H2 concentration is below the explosion threshold. Meanwhile I could make small samlpes (diameter about 6cm) with a better surface by pouring the glas on tin. I had some flat graphite crucibles made in China with 20cm diameter. But lack of money and time stopped the project for now. Some problems I could solve, and I could identify the next problems:

1. In a floatglas line there is at every point for several years the same temperature. Therefore there is no thermal stress on the lid. In my device I will heat up and cool down, which leads to thermal expansion/contraction of the lid and falling (graphite) dust. Which material for the lid could be promising?

2. Pouring the glass from a second crucible above the flat crucible yields a dirty dip in the middle of the sample. I stems from the bottom of the upper crucible.