deltaH
Dangerous source of unreferenced speculation
Posts: 1663
Registered: 30-9-2013
Location: South Africa
Member Is Offline
Mood: Heavily protonated
|
|
Odd observation about azeotropes possibly leading to a new phenomenological law?
I've been mucking around with phase calculations for azeotropes using the UNIFAC additive model for activity coefficients in free software called
ChemSep Lite.
I started considering the Txy phase diagrams of azeotropic mixtures that have lower boiling points than either compound A and B (trivial case).
Now I expanded this to ternary systems (three component), seeing if I could find a compound C that shows a ternary azeotrope, that is to say, some
combination of A, B, C that have still lower an azeotropic boiling point.
This is analogous to ternary eutectics and higher, where having many solids mixed together lowers the eutectic still lower.
EXCEPT, in the case of liquid-vapour azeotropes, I could never find a component C that does this. In all cases, I could find lower boiling azeotropes
by introducing a C, but in all cases, either A or B would have to go to zero in the composition for the lowest possible boiling azeotrope to be found.
Put another way, on the 3-d surface plots of the boiling point and composition of A, B, C, the lowest azeotropic minimum would always be sitting on
one of the sides and not somewhere internally on the surface.
I was surprised by this because for solid-liquid phase equilibria, one can find ternary combinations that lower the eutectic point still lower.
Have I stumbled upon a new law concerning azeotropes or is this simply an artifact of UNIFAC or chance (i.e. exceptions exist which I missed).
For those who love maths, maybe there is a proof of this. Obviously if someone knows of an exception... then this is not true, so please do tell and I
can then check it in the program to see if this is an artifact of using UNIFAC (doubtful, but nevertheless).
EDIT: Oh and no, I don't particularly love maths 
[Edited on 7-12-2014 by deltaH]
|
|
|
forgottenpassword
Hazard to Others
Posts: 374
Registered: 12-12-2013
Member Is Offline
Mood: No Mood
|
|
Maybe you've discovered a bug in the software, rather than a new phenomenological law?! It's good that you're thinking big, though!
[Edited on 7-12-2014 by forgottenpassword]
|
|
|
Magpie
lab constructor
Posts: 5939
Registered: 1-11-2003
Location: USA
Member Is Offline
Mood: Chemistry: the subtle science.
|
|
Have you looked at this:
http://en.wikipedia.org/wiki/Azeotrope_%28data%29
It seems there are many examples.
The single most important condition for a successful synthesis is good mixing - Nicodem
|
|
|
deltaH
Dangerous source of unreferenced speculation
Posts: 1663
Registered: 30-9-2013
Location: South Africa
Member Is Offline
Mood: Heavily protonated
|
|
Thanks forgottenpassword, it's doubtful though as ChemSep is used heavily in chemical engineering circles and has been extensively
evaluated.
I'm been giving this some thought oven eating a McFlurry (spam?) just now
[Rubbish?]
Let's consider a specific case I was playing around with. Acetaldehyde (b.p. = 20C) is a highly polar molecule. As such, I suspected it's polarity
results in strong acetaldehyde-acetaldehyde electrostatic interactions with the effect of raising it's boiling point higher than what it would be
without those interactions. So going with that train of though, I'll introduced an alkane to break the interactions and hopefully form an azeotrope of
lower boiling point than either. This worked and some trial and error showed isopentane (b.p. = 28C) to be the most effective, resulting in an
azeotrope of a little over 12C according to the UNIFAC model.
The reason I think that searching for a third component to depress the acetaldehyde-isopentane azeotrope even further failed because as per my
rationalisation of why it formed in the first place, you needed something to break the acetaldehyde-acetaldehyde interactions and this could be done
with a strongly non-polar component. While it works with other alkanes as well, isopentane worked best, that is to say, anything that breaks those
interactions simply does it better or not, so would negate the third component.
There is no rationalisation I can think of that makes, say a n-pentane for example, better at breaking the polarity of acetaldehyde that would not be
best when used on its own (as is the case with isopentane). Anyhow, I'm not so sure I explained that well enough, but hopefully that makes sense.
Okay, that is a lot of blah and yet still not a proof  [/Rubish?]
****
Oh wow, Magpie, now all this thinking was for nothing 
Thanks, I'll try to plug those in...
Nice McFlurry though lol
[Edited on 7-12-2014 by deltaH]
|
|
|
Chemosynthesis
International Hazard
Posts: 1071
Registered: 26-9-2013
Member Is Offline
Mood: No Mood
|
|
I am going to say something in keeping with the already posted, but be very careful with software or computer models; when you are extrapolating data,
which isn't always well noted on software for QSPR, you are dealing with mathematical abstractions that are no longer grounded in empirical
observation. Using the math to look for a "new phenomenological law" is inherently troublesome. This is very, very different from using data science
to crunch empirical data statistics and I cannot express this enough.
Modeling can be useful; I have "played around" with it at work to point in the direction of discoveries for certain, which I then had to
experimentally or correlatively verify with measurement... but always remember a model is only as good as the computational simplifications made, and
the data added in. There is a thing as too much fitting of a curve, where fewer data points get more universally applicable equations or calculating
too good of a fit, where small errors in estimation begin to propagate to the point where more calculation gives more error, not more accuracy with
reality.
This is part of the reason you see modeling journals separate from experimental science journals. The burden of proof for publication is very
different.
All computations are a tradeoff of cycle expense (how good your computer is), speed, and accuracy. Pick two, as a modeling professor told me. It's
even harder when you consider what I said about data point picking and error compounding. Doesn't matter how popular or expensive the software you use
is. It might be better than someone else's at something, or easier to use... but it's still a model, and most commercial models are closed source and
hard to critique.
[Edited on 7-12-2014 by Chemosynthesis]
|
|
|
deltaH
Dangerous source of unreferenced speculation
Posts: 1663
Registered: 30-9-2013
Location: South Africa
Member Is Offline
Mood: Heavily protonated
|
|
Oh the flot plickens...
I've used Magpie treasure trove and considered the ternary azeotrope of Wiki's: methanol, chloroform, acetone (b.p of azeotrope =
57.5C.
BUT upon drawing the ternary diagram for that system, one observes a saddle point near 57.5C, but it's higher than the azeotrope on one side, that of
chloroform and acetone!
So my theorem still holds... will continue to try more...
***
Oh dratz, just tried the methanol, methyl acetate, n-hexane one, that one appears to have an internal minimum, however, its only very slightly lower
than n-hexane, methanol side curves and those side curves slow a slight double curve, typically indicative of liquid-liquid separation with this graph
would not show. (i.e. there might be 'lie' in the Txy diagram for this system because actually liquid-liquid equilibrium kicks in).
Will continue to hunt for a true minima where there isn't such liquid-liquid behaviour.
***
Ok tried water-ethanol-acetonitrile, similar behaviour to the previous, there appears a slightly lower minimum on the ternary plot, but it's close to
a liquid-liquid behaviour region.
I'm getting a bit tired of plugging these for now, but it seems that where a liquid-liquid separation would occur between that which is the very polar
component and the very non-polar component that breaks the interaction, the third component breaks the liquid-liquid problem thereby lowering the
azeotrope further... though these are to close to call on the phase diagrams I looking at that doesn't show me the liquid-liquid line (but I assume
this will be published somewhere).
Ok so conclussion for now, my theorem holds for systems that do not show liquid-liquid separation, for those that do, a third
component can lower it further, but presumably not a fourth?
[Edited on 7-12-2014 by deltaH]
|
|
|
Chemosynthesis
International Hazard
Posts: 1071
Registered: 26-9-2013
Member Is Offline
Mood: No Mood
|
|
Despite that suggesting more software doesn't test your hypothesis in an empirical manner (more models, less hard data), you may find it helpful to
compare with another algorithm.
http://vle-calc.com/azeotrope.html
If you wanted to publish a proposed theory, be sure to check the terms of service. Comparing calculations done Charm, MOPAC, GAMESS, Gaussian, etc. is
forbidden by terms of service last I checked.
|
|
|
deltaH
Dangerous source of unreferenced speculation
Posts: 1663
Registered: 30-9-2013
Location: South Africa
Member Is Offline
Mood: Heavily protonated
|
|
Thanks Chemosynthesis for the link and I did not know that (ban on comparison). If I ever were to publish it, then I think I would
need to do it preferably from the experimental data as opposed to models... while there's nothing wrong with models and in fact, as always, I am
impressed with how well UNIFAC works for even complex systems, I just think it will be easier to skirt all these doubts about accuracy or software
bugs.
Also, this is still pretty much just a phenomenological observation about phase behaviour. I suppose I compare data on as many known ternary systems
and show that they all have either just a saddle point higher up, or liquid-liquid region on the one side with a slightly lower point internally, it
becomes compelling enough... well until the day a four component lowest point azeotrope is found
By the way, UNIFAC is also perfectly capable of describing liquid-liquid phase equilibrium, it's just that ChemSep Lite doesn't allow the option of a
VLL calculation on its phase diagrams. The calculation simply entails equating the two liquid fugacities also using UNIFAC.
Though if I'm not mistaken, UNIFAC does not describe liquid-liquid equilibrium very accurately, instead, something like a NRTL model works much
better, but that is not predictive and requires parameter fitting to experimental data AFAIK.
[Edited on 7-12-2014 by deltaH]
|
|
|
Chemosynthesis
International Hazard
Posts: 1071
Registered: 26-9-2013
Member Is Offline
Mood: No Mood
|
|
You're welcome. I will be pleased if the link is helpful to you, but I am curious... is this really a phenomenonogical observation? QSPR doesn't work
that way, and the link I posted makes it very clear they are using models extrapolated from curve fits that don't necessarily take into account
outliers, and have an accepted average error for their chosen training dataset.
I won't believe it until I see it, and even if you do publish, I'll probably grumble criticisms unless and until I am forced to accept it  .
|
|
|
deltaH
Dangerous source of unreferenced speculation
Posts: 1663
Registered: 30-9-2013
Location: South Africa
Member Is Offline
Mood: Heavily protonated
|
|
I have published it already here on SM and you grumbled already
Anyhow, like I said, just an observation of what seems to be a trend, probably not a law just yet. But maybe now that it's been stated, people will
try to disprove it and hey presto... science at work!
|
|
|
|
![[*]](images/xpblue/default_icon.gif){kind=icon}
